I'm using a STM32H745 MCU, with two processors, one of which is an M7. With the M7 I have problems with the DMA, I have seen that it is a very discussed topic on the internet, where there are many solutions, but I still have some doubts: I would like to use the DMA with the UART, the first thing I tried to do is force the buffers into the RAM_D2 memory area:
__attribute__ ((section(".buffer"), used)) uint8_t TxBuf[1024];
and in flash.ld:
.buffer(NOLOAD) :
{
. = ALIGN (1);
} > RAM_D2
But it didn't work, Even just out of curiosity, know why?
Then in a post I found this other solution: Before every use of DMA, I use the function: SCB_InvalidateDCache_by_Addr, in this way:
SCB_InvalidateDCache_by_Addr((uint32_t*)(((uint32_t)rx_buffer) & ~(uint32_t)0x1F), SERIAL_BUFFER_SIZE+32);
HAL_UARTEx_ReceiveToIdle_DMA(&DATA_INTERFACE, rx_buffer, SERIAL_BUFFER_SIZE);
__HAL_DMA_DISABLE_IT(&hdma_usart3_rx, DMA_IT_HT);
or
SCB_CleanDCache_by_Addr((uint32_t*)(((uint32_t)tx_buffer) & ~(uint32_t)0x1F), SERIAL_BUFFER_SIZE+32);
HAL_UART_Transmit_DMA(&DATA_INTERFACE, tx_buffer, SERIAL_BUFFER_SIZE);
And it work! But I have some questions:
1 - Can I use this method every time I use the DMA, for example I would need it to receive data from the I2C, just call the SCB_InvalidateDCache_by_Addr before using the DMA and is everything ok?
2 - Is this method ok, can it be used safely? Are there any contraindications or things to consider?
3 - Are there better, safer or more suitable methods to do these operations with DMA?
The last one is more of a curiosity than a question, but because in the article I found the SCB_InvalidateDCache_by_Addr is called like this:
SCB_CleanDCache_by_Addr((uint32_t*)(((uint32_t)tx_buffer) & ~(uint32_t)0x1F), SERIAL_BUFFER_SIZE+32);
instead of just like this:
SCB_CleanDCache_by_Addr((uint32_t *)tx_buffer, SERIAL_BUFFER_SIZE);
