stm32 - TIM2 DMA configuration for stm32h7

My problem: I can't configure DMA to working properly on Input Capture event. Data doesn't transfered and error occurs.

static void my_TIM2_initInputCaptureTimer(void) {

  // enable clock source for timer
  RCC->APB1LENR |= (0x1 << 0);

  // set prescaler to 200
  TIM2->PSC = 200 - 1;

  // choose TIM2_CH1 input
  TIM2->TISEL |= (0x0 << 0);
  // set channel 1 as input mapped on TI1
  TIM2->CCMR1 |= (0x1 << 0);
  // digital filter length (0)
  TIM2->CCMR1 |= (0x0 << 4);
  // rising & falling edge

  TIM2->CCER |= (0x1 << 1);
  TIM2->CCER |= (0x1 << 3);
  // prescaler to (0)
  TIM2->CCMR1 |= (0x0 << 2);
  // enable DMA interrupt & Capture/Compare interrupt
  TIM2->DIER |= (0x1 << 9) | (0x1 << 1);
  // capture enabled
  TIM2->CCER |= (0x1 << 0);
  // reset registers WARNING: need for preloading PSC  
  TIM2->EGR |= (0x1 << 0);


  // enable TIM3 timer
  TIM2->CR1 |= TIM_CR1_CEN;
  // enable interrupt request
  NVIC_EnableIRQ(TIM2_IRQn);
  // set priority
  NVIC_SetPriority(TIM2_IRQn, 1);

}

DMA configuration:

static void my_DMA_init(void) {
  // enable DMA1 clocking
  RCC->AHB1ENR |= (0x1 << 0);
  // clear EN bit to 0
  DMA1_Stream0->CR &= ~(0x1 << 0);
  // safeguard EN bit reset
  while (DMA1_Stream0->CR & 0x1);
  // check LISR HISR registers
  if ((DMA1->HISR == 0) && (DMA1->LISR == 0))

    printf("status registers is clear\r\n");
  else
    printf("status register is not clear -- DMA wont start\r\n");
  // set peripheral addres
  DMA1_Stream0->PAR = TIM2_CCR1_Address;
  // set memory addres
  DMA1_Stream0->M0AR = (unsigned int)buffer;
  // set total number of data items
  DMA1_Stream0->NDTR = 10;


  // NOTE: configurate TIM2_CH1 interrupt route
  // set DMAMUX to route request (TIM2_CH1)
  DMAMUX1_Channel0->CCR |= 18U;

  // set DMA priority (very high)
  DMA1_Stream0->CR |= (0x3 << 16);
  // set memory data size (32)
  DMA1_Stream0->CR |= (0x2 << 13);
  // set peripheral data size (32)
  DMA1_Stream0->CR |= (0x2 << 11);

  // set memory addres increment  (enable)
  DMA1_Stream0->CR |= (0x1 << 10);
  // set peripheral addres increment (disable)
  DMA1_Stream0->CR |= (0x0 << 9);
  // set circular buffer mode (enable)
  DMA1_Stream0->CR |= (0x1 << 8);
  // set data transfer direction (peripheral to memory)
  DMA1_Stream0->CR |= (0x0 << 6);
  // set transfer complete interrupt
  DMA1_Stream0->CR |= (0x1 << 4);

  // set transfer error interrupt
  DMA1_Stream0->CR |= (0x1 << 2);
  // enable DMA1
  DMA1_Stream0->CR |= (0x1 << 0);
  // enable IRQ
  NVIC_EnableIRQ(DMA1_Stream0_IRQn);
  printf("DMA1_Stream0 %u \r\n", (DMA1_Stream0->CR & 0x1));
}

Interrupt routines:

void TIM2_IRQHandler(void) {

  InCapTick = TIM2->CCR1;
  // reset interrupt flag
  TIM2->SR = 0;

  tick = systick_ms;
  flag++;
}


void DMA1_Stream0_IRQHandler(void) {

  printf("within\r\n");
  printf("LISR: %u \r\n", DMA1->LISR);
  // clear interrupt flag
  DMA1->LIFCR |= (0x1 << 3) | (0x1 << 5);

}

Full compiled project (gcc-arm\make) for thoose who interested: https://drive.google.com/open?id=1fwns5fKexGWJl64UDeYfDyFmSFA9BURZ

(stm32h743zi-nucleo board) expected behaviour: (when you press user button (the blue one) ->PG0 pin get SET (1) state -> so input capture pin (PA5, they should be connected via jumper) get low-high transition and capture first value (output it via com-port) when you free button, input capture pin get high-low transition and capture another value(output it via com-port), but none of this events do not kick start DMA request.

Data will be outputted via com port with 115200 speed.

Answer

DMA was initialized correctly. The problem was: buffer array wasn't initialized into any SRAM memory area, so when DMA trying to reach that memory via AHB (it's not possible) DMA crash with error and set TEI_flag (transfer error interrupt).

The easises way to make this work (and obviously it isn't correct way but for acquaintance purposes it works)

#define Destination_Address ((unsigned int)0x30000000)
  DMA1_Stream0->M0AR = Destination_Address;

0x30000000 -> is boundary of SRAM1 memory

and you need first enable SRAM1\2\3:

static void my_SRAM_init(void) {
  // activate SRAM1\2\3
  RCC->AHB2ENR |= (0x7 << 29);   
}

you can reach that area with:

  uint32_t *Ptr_Dest = (uint32_t *)Destination_Address;
  printf("buffer: %lu %lu \r\n", *Ptr_Dest, *(Ptr_Dest + 1));

For those this won't work or some UB happens (corrupt data, strange placing) possible reasons:

1. Cache (enabled cache could corrupt data)


2. Your addresses should be aligned with PSIZE and MSIZE

Example: if your MSIZE and PSIZE is 32bit address must end on 0x4

I hope this will help someone, because this was huge pain in my ass for straight 3 days, I hate it.