RM48L952: Interrupting FEE write causes a reset

Will Treharne

Part Number: RM48L952
Other Parts Discussed in Thread: HALCOGEN

Hello,

We are using the HALCoGen FEE driver in our project to store changes to calibratable parameters. All seemed to be working well until we recently implemented nested interrupts as described in the TI application note. It appears interrupting the write operation causes the microcontroller to crash. As a workaround, we have disabled all interrupts while writing to the FEE which is called from a 1 second task, but this is not desirable as the write operation takes around 50ms to complete and we have a 10ms task which will be significantly delayed.

We are using the asynchronous write command as follows:

/* Disable interrupts during FEE write */
_disable_IRQ();

/* Write the block into FEE sector Asynchronously. When status is IDLE the write operation has completed */
TI_Fee_WriteAsync(ctBlockNumber, (uint8_t *)ptrRamAddr);

do {

TI_Fee_MainFunction();

stFeeStatus = TI_Fee_GetStatus(GET_LAST_ACTION_STATUS);

}
while (stFeeStatus != IDLE);

/* Re-enable interrupts after writing to FEE is complete */
_enable_IRQ();

Is anyone able to tell me why interrupting the write sequence would cause a problem? Are we able to move the interrupt disable to cover only the TI_Fee_MainFunction() so we do not delay our higher priority timed tasks?

Many thanks

over 8 years ago

0 QJ Wang over 8 years ago

TI__Guru**** 198766 points

Hello Will,

I have forwarded this question to our SW expert, and he will come back to you soon.

0 Vishwanath Reddy over 8 years ago in reply to QJ Wang

TI__Mastermind 19395 points

Hello Will,

Can you describe in more detail about crash? Is it a watchdog reset? Have you configured TI_Fee_MainFunction() to be called from any other task which can interrupt the TI_Fee_WriteAsync API?

0 Will Treharne over 8 years ago in reply to Vishwanath Reddy

Intellectual 270 points

Hi Vishwanath,

Thanks for getting back to me. It does appear to be a watchdog reset. I tried disabling the watchdog yesterday, but must have made a mistake as it was still resetting. When I tried again today I can see what is happening.

With the watchdog disabled, my code is stuck in the dabort handler after the FEE write starts. It looks like the error is occurring at the following line in the C_irqDispatch function in irqDispatch_c.c:

if (0U == u32IrqIndex)

which has the following disassebly:

00016160: E59DC000 ldr r12, [sp]
00016164: E35C0000 cmp r12, #0
00016168: 1A000000 bne $C$L2

The error occurs on the first instruction. SP is holding the value 0x07FFFF48 so it seems the user stack has overflowed. After increasing the stack size, my program no longer crashes.

Am I right in thinking the TI nested interrupt handler uses only the user stack rather than IRQ stack and so the IRQ stack size can now be set to zero? Are there any other potential issues with interrupting an FEE write, provided that the interrupting task doesn't contain any FEE operations?

Many thanks