Hi NXP Team,
Platform details:
Problem statement:
In production we occasionally see the A35 complex crash or hard-hang (kernel panic / lockup). Today we have no persisted crash evidence and no autonomous recovery — the cluster stays dead until a manual power cycle, and field/DV occurrences cannot be debugged. We want to implement (a) panic-log persistence across a watchdog reset using pstore/ramoops, and (b) supervision of the A35 by the CM4_0 with the ability to reset only the A35 partition.
On i.MX8QXP C0, when the system watchdog (imx-sc-wdt, handled by SCFW) fires, what type of reset is performed — full SoC/board reset or A-cluster partition reset? Is this configurable via SCFW board file or sc_pm API?
Hello,
yes, you may use a virtual watchdog managed by the SCFW to reset the partitionSC_TIMER_WDOG_ACTION_PARTITION behavior is what you want it resets only the Linux/A35 partition.
Also, there is a way to achieve the same effect from CM4_0 firmware directly by calling sc_pm_reset_partition() on the A35 partition via the SCFW sc_pm API, which gives the CM4_0 full supervisory control independent of the watchdog mechanism.
You can find more information about this on the SCFW porting guide for the specific version that you are using.
Best regards/Saludos,
Aldo.