RISC-V LinuxBetaOPEN SILICON. SHARED KNOWLEDGE.

Reconfiguring System-Level Protection Through Bounded Delegation in RISC-V MCUs

AI Analysis Abstract Microcontroller-based embedded and cyber-physical systems increasingly demand dynamic protection policies that can adapt at runtime. Existing commercial off-the-shelf (COTS) MCU protection mechanisms are often static and fail to constrain the scope of policy…

AI Analysis

Abstract

Microcontroller-based embedded and cyber-physical systems increasingly demand dynamic protection policies that can adapt at runtime. Existing commercial off-the-shelf (COTS) MCU protection mechanisms are often static and fail to constrain the scope of policy changes an authorized domain can make, potentially leading to unauthorized expansion or interference. This work introduces an ownership-aware model for reconfiguring system-level protection, featuring explicit per-rule ownership, protection-state-derived authority bounds, and validate-before-commit updates. This model enables secure, dynamic policy evolution and recursive delegation of configuration authority. Implemented as an Ownership-Bounded World Checker (OB-WC) in an open-source RISC-V CVA6-based System-on-Chip, the solution demonstrates no degradation in access latency or operating frequency. Authors report a hardware overhead of 7.1% LUT and 1.2% FF for an eight-rule checker. A case study on over-the-air updates shows a significant reduction in policy-transition latency by 89.5% compared to hypervisor mediation and 94.6% over trusted firmware mediation, while maintaining established authority boundaries.

Key Highlights

  • Introduces bounded configuration authority as a missing architectural primitive for dynamic protection policies in MCUs.
  • Proposes an ISA-independent ownership-aware model combining explicit per-rule ownership, protection-state-derived authority bounds, and validate-before-commit updates.
  • Implements the model in an open-source RISC-V World Checker (OB-WC) integrated into a CVA6-based SoC.
  • Demonstrates no degradation in access latency or operating frequency, with modest hardware overhead (7.1% LUT and 1.2% FF for eight rules).
  • Achieves substantial reductions in policy-transition latency (89.5% to 94.6%) through direct delegation compared to software mediation in an OTA update scenario.

Technical Details

  • The proposed model defines three architectural semantics: explicit per-rule ownership, protection-state-derived authority bounds, and validate-before-commit reconfiguration.
  • Each protection rule is associated with an owner identifier (OWNER) and an ownership definition flag (ODEF), separating configuration authority from resource access permissions.
  • An update controller is inserted on the memory-mapped configuration path to authenticate requests, validate candidate ranges, and atomically commit accepted updates.
  • Configuration requests proceed through owner verification, pre-validation, optional sequential range validation (against enabled rules), and final commit or rejection.
  • Sequential range validation checks if a candidate update overlaps a rule not owned by the requester or if it is fully contained by an enabled requester-owned rule.
  • The OB-WC implementation extends a baseline RISC-V World Checker, preserving the normal access datapath and confining new logic to the configuration path.
  • The prototype supports eight Worlds and an eight-slot OB-WC, integrated with OpenSBI, Bao, and Zephyr for hierarchical delegation.

Implications

  • Enables secure, dynamic evolution of protection policies at runtime in MCU-based systems, addressing limitations of static COTS mechanisms.
  • Reduces policy-transition latency by allowing direct hardware-supported delegation of configuration authority, minimizing reliance on privileged software mediation.
  • Provides an architectural basis for system-level protection checkers in next-generation RISC-V MCUs and can inform ongoing RISC-V specifications.
  • Allows for hierarchical delegation of configuration authority, where domains can refine and further delegate authority within established boundaries.
  • The off-datapath validation microarchitecture ensures that bounded reconfiguration does not affect latency-critical resource access performance.

Limitations

  • The model targets a hierarchical trust model where preceding owners may remain part of the Trusted Computing Base (TCB), rather than confidential-computing-style protection from more privileged software.
  • The current design does not require delegators to relinquish access to delegated resources; supporting non-retentive ownership transfer would require additional hardware mechanisms.
  • Multi-rule policy transitions are not atomic; software must order updates to ensure intermediate committed states preserve intended security properties.
  • The prototype's validation scope is limited to the local rule set of a single checker; software must coordinate policy transitions involving resources protected by multiple checkers.
  • Synchronization of resource accesses during policy transitions is left to software, which must stop relevant initiators and wait for outstanding transactions to complete.
  • The sequential range validator's latency depends on the number of enabled rules; a parallel implementation could reduce this variable cost at the expense of additional hardware.

Read next

Explore all articles →