While stress testing a product that utilizes the MCC library (GitHub - toradex/mcc-kmod: Kernel module for Freescale/Timesys Multi-Core Communication), we saw critical failures in the M4 and the A5 where messages could no longer be sent from either the A5 or M4, cases where the M4 goes into a deadlock state within the mcc_cpu_to_cpu_isr function, as well as cases where the Linux kernel would freeze and become non-responsive (cannot use the debug UART or ssh to the machine). This occurs after stress testing the MCC channel used by our application so that many messages are being sent back and forth. Are there any clear guidelines on what the limitations of the MCC library are? Throttling the messages sent between cores seems to help but this is not discussed in NXP’s MCC User Guide document. Whether MCC library calls are thread safe is also not discussed.
This is a known issue and currently not planned (see Issue #13614). NXP replaced the MMC library with RPmsg & FreeRTOS which Toradex also backported for Colibri VF61.