Bare-metal C, memory management, and hardware interrupts.
3 questions in Embedded Systems & IoT.
Stack and heap sizing on fixed RAM starts from the link map, worst-case call depth and interrupt nesting. Measure stack high-water marks with fill patterns, avoid unbounded heap allocation, and prefer fixed pools allocated at startup for predictable firmware behaviour.
Interrupt service routine rules are simple: acknowledge the hardware, capture the minimum safe state and return fast. Blocking calls, allocation, logging and long loops belong in deferred work because ISR runtime becomes worst-case latency for interrupts at the same or lower priority.
Watchdog recovery in firmware only works if the code that kicks the watchdog can detect task progress. Kick from one supervisor after checking heartbeats, record reset cause and diagnostics in retained memory, and treat watchdog resets as defects to investigate.