SV#8 during concurrence of two adjacent ISR

[Cesare]

Good evening to everyone,
I am developing along STM32F407VGT on STM32F4-Discovery board.
I have one thread which continuously fills a double DMA buffer where the DMA moves data from memory to a GPIO peripheral at very high frequency. I have a DMA ISR every 570 uS.

I have another function, called from the shell, which uses chprintf() and writes characters to a stream managed by the SERIAL driver (SD3).
If I execute this function, very soon the OS gets stuck at SV#8 error. This happens during the DMA ISR at the state check __dbg_check_enter_isr().

Code: Select all

/**
 * @brief   DMA2 stream 1 shared interrupt handler.
 */
OSAL_IRQ_HANDLER(STM32_DMA2_CH1_HANDLER) {
  uint32_t flags;

  OSAL_IRQ_PROLOGUE();         <===== OS GET STUCK HERE =====

  flags = (DMA2->LISR >> 6U) & STM32_DMA_ISR_MASK;
  DMA2->LIFCR = flags << 6U;
  if (dma.streams[9].func)
    dma.streams[9].func(dma.streams[9].param, flags);

  OSAL_IRQ_EPILOGUE();
}

This is the stack trace:

Thread #1 (Suspended : Signal : SIGINT:Interrupt)
chSysHalt() at chsys.c:232 0x80015ae
__dbg_check_enter_isr() at chdebug.c:220 0x8001940
Vector124() at stm32_dma.c:329 0x8003c9e
<signal handler called>() at 0xfffffffd
chSchGoSleepS() at chschd.c:313 0x800231c
chSchGoSleepTimeoutS() at chschd.c:374 0x80023c2
chThdEnqueueTimeoutS() at chthreads.c:879 0x8002418
osalThreadEnqueueTimeoutS() at osal.h:895 0x800453e
oqPutTimeout() at hal_queues.c:546 0x800453e
_put() at hal_serial.c:64 0x80045cc
chvprintf() at chprintf.c:329 0x800279a
chprintf() at chprintf.c:371 0x800279a
render_dump() at eeprom_shell.c:65 0x8002f22
do_dump() at eeprom_shell.c:88 0x8002f22
argparse_getvalue() at argparse.c:149 0x8003172
parse_short_opt() at argparse.c:181 0x80031d2
ARGPARSE_Parse() at argparse.c:313 0x8003274
EEPROM_Shell() at eeprom_shell.c:126 0x8003460
cmdexec() at shell.c:106 0x80051ae
shellThread() at shell.c:405 0x80051ae
__port_thread_start() at chcoreasm.S:203 0x80002fe

I had to manage this error already several times, all those times when I forgot to use the I-class API. But this time I can't figure out what I am doing wrong.
If I get SV#8 it probably means I am missing some chSysLock() during the chprintf(), but I supposed this was already managed in the stream object.

Thanks in advance for any hint.

[Giovanni]

Hi,

It is also possible that you are using an invalid priority for that IRQ, use values 3...15. 0...2 could cause that error because those can preempt the kernel critical sections.

Giovanni

[Cesare]

Giovanni,
you are always the MCU 'Professor'

My issue was very simple, the DMA IRQ priority in the NVIC register was erroneously set to 0. Changing it in the range 3-15 made it working perfectly.

Thank you so much.