MPC5744P IVOR1 Machine Check Handler not hit with uncorrectable FLASH ECC error on RTOS, baremetal S

[hunanpuxi]

Hi,I am using the FLASH_ECC_Error_Injection_MPC5744P demo project from SDK 2.1 to test FLASH ECC fault handling.

1. In the bare-metal SDK example:After calling Generate_noncorrectable_FLASH_ECC_error, the core properly traps the fault and enters Machine_check_handler (IVOR1 Handler) as expected.
2. After porting the identical ECC injection logic to my RTOS-based project:No machine check exception occurs, and the code never jumps into Machine_check_handler.

I have already completed the following porting & configuration steps:

• Fully ported and defined IVOR1_Handler in the startup .s assembly file, consistent with the bare-metal demo;
• Full FCCU alarm interrupt configuration is implemented, but no FCCU alarm interrupt is triggered either when injecting the uncorrectable FLASH ECC error.

I have 3 core questions to troubleshoot this issue:

1. What mandatory hardware/register configurations are required to enable entry to the Machine Check Handler (IVOR1) for FLASH uncorrectable ECC errors?
2. Are there special considerations or constraints when running ECC fault injection and machine check exception handling under an RTOS?
3. Could you provide a step-by-step debug/troubleshooting workflow to locate the root cause of this missing machine check exception?

[davidtosenovjan]

I don't know mentioned example, I suppose it is ported version on my appnote I did with using of GHS compiler.
https://www.nxp.com/docs/en/application-note/AN13179.pdf
https://www.nxp.com/docs/en/application-note-software/AN13179SW.zip

If RTOS affects the bahavior, then it is needed to investigate what is does with MSR register. Pay attention to chapter 5.

Also you may check section 8 to understand ECC handling.

[hunanpuxi]

HI,

I found during testing:

In my project, the IVOR1_Vector handler cannot be triggered when compiling with -O1 optimization level. The exception works normally under -O0.

However, the reference demo project works correctly with -O1 using the same optimization setting.

I would appreciate it if you could share some suggestions to help me locate the root cause.