mohitmishra786/interrupts-and-exceptions-baremetal

Interrupts and Exceptions (Bare-Metal)

Purpose

Guide agents through bare-metal interrupt handling on ARM Cortex-M: NVIC configuration, ISR writing rules, exception handlers (HardFault, BusFault), priority grouping, nesting, tail-chaining, and latency considerations.

When to Use

• Configuring peripheral IRQ priorities
• Writing ISRs that must not block
• Debugging HardFault after enabling interrupts
• Sharing data between ISR and main loop
• Optimizing interrupt latency

Workflow

1. NVIC overview (Cortex-M)

Exception / IRQ flow
├── NVIC receives IRQ (priority compare with BASEPRI/PRIMask)
├── Stacking: automatic save r0-r3, r12, lr, pc, psr
├── Branch to handler from vector table
├── Handler runs (should be short)
└── Unstack and return — tail-chain if another IRQ pending

2. Enable and prioritize an IRQ

#include "stm32f4xx.h"  /* CMSIS device header */

void uart_irq_init(void) {
    NVIC_SetPriority(USART2_IRQn, NVIC_EncodePriority(NVIC_GetPriorityGrouping(), 2, 0));
    NVIC_EnableIRQ(USART2_IRQn);
}

Priority: lower numeric value = higher urgency (on most Cortex-M implementations). Check vendor docs for grouping bits.

3. ISR template

void USART2_IRQHandler(void) {
    if (USART2->SR & USART_SR_RXNE) {
        uint8_t b = (uint8_t)USART2->DR;  /* read clears RXNE */
        ringbuf_push(b);
    }
    if (USART2->SR & USART_SR_ORE) {
        (void)USART2->DR;  /* clear overrun */
    }
}

ISR rules:

• No blocking calls (printf, malloc, long loops)
• Minimize work — defer to main via flag/ring buffer
• Clear interrupt flags per datasheet (read-to-clear vs write-1-clear)

4. Critical sections

uint32_t primask = __get_PRIMASK();
__disable_irq();
/* atomic section */
__set_PRIMASK(primask);

Or raise BASEPRI to mask lower-priority IRQs only.

5. HardFault handler

void HardFault_Handler(void) {
    __asm volatile(
        "tst lr, #4\n"
        "ite eq\n"
        "mrseq r0, msp\n"
        "mrsne r0, psp\n"
        "b hard_fault_c\n"
    );
}

void hard_fault_c(uint32_t *stack) {
    uint32_t r0  = stack[0];
    uint32_t pc  = stack[6];
    uint32_t psr = stack[7];
    /* log pc — GDB: info registers, bt */
    while (1);
}

Decode CFSR/HFSR registers for fault cause:

volatile uint32_t cfsr = SCB->CFSR;
volatile uint32_t hfsr = SCB->HFSR;
volatile uint32_t bfar = SCB->BFAR;

6. Latency and tail-chaining

| Factor | Impact | |--------|--------| | Higher priority IRQ | Preempts lower | | Tail-chaining | Back-to-back IRQs skip unstack/restack | | FPU context | Lazy stacking adds latency on M4F/M7 | | Long ISRs | Starves other IRQs and main |

Measure with GPIO toggle + scope, or DWT cycle counter (DWT->CYCCNT).

7. Agent usage examples

/interrupts-and-exceptions-baremetal Configure NVIC priority for UART vs SysTick
/interrupts-and-exceptions-baremetal Decode HardFault stacked PC with GDB

Common Problems

| Symptom | Cause | Fix | |---------|-------|-----| | IRQ never fires | NVIC not enabled or IRQ masked | NVIC_EnableIRQ; check PRIMASK | | Spurious re-entry | Flag not cleared | Clear per RM (ORE needs DR read) | | HardFault in ISR | Stack overflow | Increase _estack; check ISR stack | | Lost bytes | ISR too slow | Ring buffer + higher IRQ priority | | Priority inversion | Long critical section | Shorten __disable_irq window |

Related Skills

• skills/baremetal/baremetal-startup — vector table entries
• skills/baremetal/uart-serial-baremetal — UART IRQ handlers
• skills/embedded/openocd-jtag — GDB breakpoint in ISR
• skills/debuggers/gdb — examine fault stack frame
• skills/low-level-programming/assembly-arm — fault handler asm