SDIO Gets stuck

[drJonesy]

hi 

I am trying to get an sd card to work with my custom board, uses STM32F412RET6.

when i try to mount or do anything other than init in 1 bit mode the code gets stuck at this point. I have been able to read block size and number from the SD card but that is it.

 while(!__HAL_SD_GET_FLAG(hsd, SDIO_FLAG_RXOVERR | SDIO_FLAG_DCRCFAIL | SDIO_FLAG_DTIMEOUT))
  {
    if(__HAL_SD_GET_FLAG(hsd, SDIO_FLAG_RXDAVL))
    {
      *(tempscr + index) = SDIO_ReadFIFO(hsd->Instance);
      index++;
    }
    else if(!__HAL_SD_GET_FLAG(hsd, SDIO_FLAG_RXACT))
    {
      break;
    }

    if((HAL_GetTick() - tickstart) >=  SDMMC_SWDATATIMEOUT)
    {
      return HAL_SD_ERROR_TIMEOUT;
    }
  }

 

[TDK]

The SDIO peripheral has race conditions where data needs read out quickly enough otherwise it overflows. Use a higher clock prescaler to see if that solves the problem. If it does, the solution would be to use DMA or rewrite code to be faster. Using Release mode can be a bandaid.

[drJonesy]

DMA is enabled and prescaler is set to 250

 

[Saket_Om]

Hello ​@drJonesy 

The execution is not expected to stay stuck in this loop indefinitely, because the code contains an internal timeout protection.
If the SDIO transfer does not progress and none of the expected status flags are detected, the timeout condition is reached:

if((HAL_GetTick() - tickstart) >= SDMMC_SWDATATIMEOUT)
{
    return HAL_SD_ERROR_TIMEOUT;
}

 

[AScha.3]

Right, ​@Saket_Om  , and the timeout is : ?

 

afaik is on : SDMMC_ERROR_TIMEOUT ((uint32_t)0x80000000U) /*!< Timeout error ,

so at systick 1ms , → 24.8 days . 

But ok, just wait a little...

[TDK]

It’s actually hilariously worse than that.

https://github.com/STMicroelectronics/stm32f4xx-hal-driver/blob/1f6451c3e07728b4c830744de380e56bf5bc0026/Inc/stm32f4xx_ll_sdmmc.h#L308

#ifndef SDMMC_DATATIMEOUT /*Hardware Data Timeout (ms) */
#define SDMMC_DATATIMEOUT                  ((uint32_t)0xFFFFFFFFU)
#endif /* SDMMC_DATATIMEOUT */

#ifndef SDMMC_SWDATATIMEOUT /*Software Data Timeout (ms) */
#define SDMMC_SWDATATIMEOUT                SDMMC_DATATIMEOUT
#endif /* SDMMC_SWDATATIMEOUT */

So yes you can change it, but by default it’s 49.7 days. 🤣 And if your program is busy during that particular 1ms, you’ll have to wait another 49.7 days for the timeout to trigger. And if you’re not using the default 1ms tick, but 10ms or 100ms instead, it will never timeout.

[AScha.3]

Hi,

so you just have the command - communication, not the data in-out on D0 .

Typical a hardware problem, to long wires cpu to sd-card, or bad setting of pin speed.

• wires: keep all lines to sd-card at about same length and  shorter than 50mm 
• set all cpu pins to sd-card to medium speed + pullups ON 
• set the sd clock to < 50 MHz (...clock divide..), try at 20MHz and 1-bit mode
• set using no DMA , hardw.control.flow ON, give sdmmc a good priority interrupt (and enable it ! in NVIC)
• try only mount (…, 1) = mount NOW , or a read of file;  NO write or format !!!

 

[drJonesy]

Hi 

Thanks for the heads up about what the problem may be. I have set GPIO to medium speed and pull up but that doesn’t help.  I will focus on the circuit design and check for wire lengths. 

 

Just as I started writing this, I checked my circuit, for Errors and low and behold. The problem was on the eclamp pin 16. it is set dat0 in the schematic but Dat1 in data sheet. Very disappointed in myself for not checking.

 

Thanks for the heads up. I will modify my circuit and see if i get a fix.

[drJonesy]

Hi All

 

I am now reading the SD CARD. Problem as above. Thank you AScha.3 for your help with this issue. You can now close this issue.