RCCU, FCCU, and Lockstep Diagnostic Data Flow on S32G3 Dear NXP, Based on my understanding, RCCU detects Lockstep mismatches and reports errors to FCCU, which handles fault management and may provide diagnostic information through the Linux kernel log (dmesg). However, the BSP documentation describes this process only at a high level. Could you please provide more specific implementation guidance for the S32G378A with Linux BSP 48.0.0? Available diagnostic data: What types of data can be collected from RCCU and FCCU, such as error status, fault source, error type, or fault context? What information is actually available in hardware, and what can be reported through Linux dmesg? End-to-end data flow: What is the exact flow from a Lockstep mismatch being detected by RCCU, through FCCU fault handling, to Linux kernel reporting? Which registers, interrupts, drivers, or software components are involved? Source code and implementation: Which source files, device drivers, device-tree entries, or configuration files need to be modified or enabled in BSP 48.0.0 to collect and report this information? Does NXP provide an existing GitHub repository, reference implementation, or sample code demonstrating the complete flow? Lockstep input and output: What does the Lockstep mechanism actually compare as its input? Is the input generated internally by the redundant A53 execution paths, or must software configure or load any input data? Where is the relevant configuration performed, and what diagnostic output can software retrieve when a mismatch occurs? Please provide the relevant Reference Manual sections, source-code references, and a concrete implementation workflow, as the current BSP documentation does not explain these details sufficiently. Re: RCCU, FCCU, and Lockstep Diagnostic Data Flow on S32G3 Thank you for your reply. I would like to ask one more question. The Linux BSP 48.0.0 User Manual describes FCCU configuration using Device Tree properties such as `nxp,ncf_fault_list` and `nxp,ncf_actions`. Could you please clarify which source file we should modify to apply this configuration? After saving the changes, will they take effect through the Device Tree, or do we also need to modify the FCCU driver? If possible, please provide a simple example for S32G378A-EVB3. Best regards, Toan Re: RCCU, FCCU, and Lockstep Diagnostic Data Flow on S32G3 Hello, @Toan
Thanks for the post
1. The FCCU provides a hardware interface to collect faults when a failure is detected in the device, it is hardware behavior, different fault number would reflect different failure, you may refer the fault map file attached in S32G3 RM. The Linux FCCU driver in BSP48 allows updating the selected list of Non-Critical Faults handled in FCCU and selecting one of the supported NCF actions using device tree, if the reaction set is an alarm, then the alarm information is printed in the kernel’s log if the corresponding failure occurred.
2. When RCCU detected the issue, the FCCU will record it and then execute the actions set by the user, for example, if alarm is set as the action, then the information could be found in the kernel log, the Linux FCCU driver, DTS, FCCU alarm interrupt handler would be involved.
3. The corresponding information could be obtained from the Chapter 6 of BSP48 UM.
4. In the lockstep mode, one cluster is considered the main cluster and the second cluster is considered the checker cluster. From a software perspective, the two clusters operate as a single cluster in the lockstep mode. Cortex-A53 Cluster0 is the main execution cluster, where Cortex-A53 Cluster1 follows execution of Cortex-A53 Cluster0. Equivalent operation of replicated resources is supervised by comparators (RCCU) on all functional signals leaving the cluster for the rest of the MCU. Any operational deviations between the supervised signals causes the FCCU to be notified of the discrepancy.
In S32G3 A53 Lockstep mode, software does not need to provide or load input data for the Lockstep comparison. The comparison is performed internally by the hardware during normal A53 execution and is transparent to application software.
Software mainly configures the Lockstep operating mode, rather than the comparison data. Software can set the corresponding action for the fault. If a mismatch is detected, it is reported through the RCCU/FCCU safety fault mechanism and the relevant actions would be executed.
BR
Chenyin
View full article