SoC Transaction Integrity via Endpoint CRC Over Standard NoC
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing solutions for ensuring end-to-end data integrity in System-on-Chip (SoC) communications are complex, expensive, and inefficient, particularly in scenarios where data is exchanged between CPU processors and external DRAM, and network-on-chip (NoC) interconnects lack built-in protection features.
Innovation Solution
Implementing a transaction integrity protection system using Transaction Checker Generator (TCG) and Transaction Checker Endpoint (TCEP) hardware circuits at the periphery of the interconnect to send extra protecting information transaction (PIT) messages with each regular transaction, utilizing CRC data to detect faults, and ensuring data integrity without modifying the underlying interconnect hardware.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If network-on-chip (NoC) interconnects include data integrity protection features, then data integrity is protected, but device complexity and implementation cost increase
Solution Approach 1:
The patent introduces an intermediary integrity checking mechanism that operates between the interconnect and the end devices. This mediator performs CRC calculations and comparisons without requiring modifications to the interconnect itself, thus protecting data integrity while keeping the interconnect simple and standard-compliant.
Solution Approach 2:
The integrity protection functionality is extracted from the interconnect and placed in separate end-device modules. This allows the interconnect to remain a simple, standard implementation while the complexity of integrity checking is isolated to specific endpoints that generate and verify protection information.
2Reliability
If lock-step NoC interconnect with discrepancy detection logic is implemented, then data integrity is protected, but silicon area doubles
Solution Approach 1:
Instead of implementing full duplicate logic across the entire interconnect, the patent applies integrity checking only at the endpoints where data is generated and consumed. This partial approach provides necessary protection without the excessive area cost of complete lock-step duplication throughout the interconnect fabric.
Solution Approach 2:
The patent uses CRC (Cyclic Redundancy Check) copying mechanisms where protection information is generated as a copy of the data and transmitted alongside it. This allows verification without requiring complete duplicate data paths, significantly reducing the area overhead compared to full lock-step implementation.
3Reliability
If extra protecting information transaction (PIT) messages are sent with each regular transaction, then end-to-end integrity protection is achieved, but bandwidth is consumed
Solution Approach 1:
The patent merges the protection information transmission with the existing data transmission protocol. CRC values and other integrity information are combined into the same transaction structures as the actual data, allowing both data and protection information to be transmitted efficiently without requiring separate dedicated channels that would double bandwidth consumption.
4Device complexity
If standard interconnect without built-in protection is used, then device complexity is reduced, but data integrity cannot be ensured
Solution Approach 1:
The patent segments the integrity protection function into separate end-device modules rather than embedding it in the interconnect. This allows the interconnect to remain simple and standard while the protection functionality is distributed to the endpoints that need it, achieving both simplicity and reliability through functional segmentation.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system-on-chip (SoC) method and apparatus are disclosed for checking end-to-end integrity of communications over an network interconnect, where the SoC includes an initiator subsystem connected over the network interconnect to a target subsystem, wherein a first integrity module is configured to compute a first integrity value based on regular transaction messages sent or received by the initiator subsystem and to send a protecting information transaction (PIT) message over the network interconnect to the target subsystem, wherein a second integrity module is configured to compute a second integrity value based on regular transaction messages sent or received by the destination subsystem and to send a PIT response message over the network interconnect to the initiator subsystem, and wherein a compatibility module compares the first and second integrity values to verify the end-to-end integrity of the regular transaction messages sent or received over the network interconnect.