Contents

Platform services on the HPC safety island (MCU)

The safety island checks that critical software runs correctly, validates exchanged data and records faults so the system can respond safely.

What runs here

These services support real-time control and safety monitors on the HPC’s MCU cores. They detect execution and communication faults, support a defined reaction and preserve diagnostic evidence. AUTOSAR Classic is a common implementation approach; the exact allocation depends on the SoC and safety architecture.

Component Role
Watchdog Manager (WdgM) Supervises software execution through configured checkpoints.
End-to-end (E2E) protection Checks safety-relevant data exchanged with other software or nodes.
Diagnostic Event Manager (Dem) Processes reported diagnostic events and manages fault records.
Nonvolatile storage services Preserve selected records across resets, where suitable storage is available.
Project-specific fault reaction logic Coordinates degraded operation, output inhibition or another specified safe response.

WdgM provides alive supervision, deadline supervision and logical supervision: whether execution occurs often enough, whether checkpoint intervals meet limits and whether checkpoints follow allowed sequences. It coordinates watchdog triggering through Watchdog Interface (WdgIf) and watchdog drivers. AUTOSAR Watchdog Manager specification

E2E protection uses profile-specific CRCs, counters and data identification mechanisms. Receivers interpret check results together with communication timing requirements. E2E is a library or transformer-based mechanism near communication endpoints, although grouped here as a platform capability. AUTOSAR E2E Library specification

Dem manages diagnostic status and configured event data. Durable storage commonly uses NVRAM Manager (NvM), but an island without local nonvolatile memory needs an explicitly designed persistence path. Recording a fault and reacting to it are separate responsibilities. AUTOSAR Diagnostic Event Manager specification

Why it looks like this

The island supports bounded execution and fault response while application processors handle larger, more changeable workloads. Its services therefore favour static configuration, bounded resource use and carefully controlled changes. A small, focused service set conserves MCU memory and limits integration and verification effort.

Lockstep cores and ECC memory support hardware fault detection. They do not establish software correctness or make the whole HPC independent: clocks, power, interconnects and reset paths may still be shared. An ASIL D allocation requires evidence for the complete safety mechanism and its dependencies; choosing an MCU or AUTOSAR module does not provide that evidence automatically.

Supervision limits must accommodate valid scheduling variation while leaving enough time to detect a fault and complete the required reaction within the fault tolerant time interval. Fault storage must fit within those constraints. A flash write must never delay a necessary safety action.

How it connects

Above this layer, control applications report checkpoints and monitor results through RTE interfaces or configured service APIs. They receive data validity and supervision status for their decision logic. Fault reaction logic translates detections into actions defined by the safety concept.

Below it, the OS schedules service execution; watchdog interfaces and drivers reach hardware; memory services reach persistent storage. SoC-specific mechanisms may also expose hardware fault indications and reset controls.

Across the chip, the island exchanges protected data with application processors through an interprocessor channel, such as shared memory and mailboxes. Across the vehicle, configured communication stacks carry protected messages to other ECUs. Protection endpoints must cover the intended communication path, and receivers need explicit stale-data handling. Diagnostic access may be routed through a host or gateway; responsibility for records and their availability after host failure must be clear.

Compared with other node classes

  • HPC application processor: Platform services emphasize OTA, SOVD, intrusion detection and extensive logging; island services prioritize bounded safety supervision and reaction.
  • Zone controller: Similar MCU services coexist with network management, diagnostics and local power-control responsibilities.
  • MCU-based ECU: Often hosts a fuller local diagnostic, storage and security stack; an integrated island may rely on other HPC resources.
  • Smart node / LIN slave: Usually provides smaller configuration and diagnostic facilities, with more supervision responsibility assigned to an upstream controller.

Best practices

  • Trace each detection and reaction to a safety requirement, including timing budgets and ownership.
  • Place checkpoints around meaningful work; verify that stalled or skipped execution cannot still satisfy supervision.
  • Agree E2E profiles, identifiers, layouts, counter handling and startup behaviour at both endpoints.
  • Test host hangs, communication loss, shared-resource faults and reset-domain behaviour using fault injection.
  • Bound diagnostic processing and persistence costs; preserve useful reset evidence without blocking safety reactions.
  • Coordinate island and host updates with interface compatibility checks and regression verification.

Common pitfalls

  • Treating lockstep execution or physical core separation as proof of independence.
  • Feeding a watchdog from a periodic task that continues running after the supervised function fails.
  • Treating an E2E CRC as sender authentication, or accepting old data indefinitely after reception stops.
  • Assuming every fault requires a whole-HPC reset; the required response depends on the safety concept.

What it means for your discipline

  • Functional safety Core

    Defines detection coverage, reaction timing and safe or degraded behaviour. Reviews whether shared SoC resources can defeat the island’s monitoring or response.

  • Base software Core

    Configures WdgM, E2E integration, Dem and storage services. Verifies scheduling, mode transitions and driver dependencies against the safety requirements.

  • E/E architecture Partial

    Allocates supervision, reset authority and diagnostic ownership between the island, host and external nodes. Makes shared-resource dependencies explicit.

  • Diagnostics & lifecycle Partial

    Defines event records, persistence and retrieval paths. Ensures workshop evidence remains useful after resets or loss of the application processor.

  • Software update & OTA Partial

    Coordinates programming and restart modes with island watchdog supervision and fault reactions. Checks that expected update interruptions do not trigger unintended resets and that monitoring resumes correctly after activation.

No direct stake: Cybersecurity · HPC & application software · Hardware & silicon · Networks & communication

Where it sits in the software stack

Generic layered software of a central HPC (application processor with hypervisor, plus a safety island) and of a classic MCU-based ECU (AUTOSAR Classic). Highlighted modules matter for this topic. Click any module to open its page.

HPC · application processor (MPU) and safety island (MCU)
  • WdgM · E2E · Dem — Supervises execution with WdgM, checks exchanged data with E2E protection and manages diagnostic fault records through Dem.
MCU-based ECU · AUTOSAR Classic

Not directly involved on this node type.