search

STMicroelectronics / stm32h7xx-hal-driver

Provides the STM32Cube MCU Component "hal_driver" of the STM32H7 series.
BSD 3-Clause "New" or "Revised" License
98 stars 43 forks source link

SDMMC stuck with HAL_SD_ERROR_DATA_TIMEOUT #30

Closed daniel-starke closed 5 months ago

daniel-starke commented 2 years ago

Describe the set-up SD card socket connected without external pull-ups (only internal ones) to a NUCLEO-H723ZG board. This made the data lines instable. Adding an external logic analyzer for example brought the lines to a working level. Switching from the non-working variant after HAL_SD_ERROR_DATA_TIMEOUT online to the a working variant failed while being stuck with HAL_SD_ERROR_DATA_TIMEOUT.

Describe the bug I have observed cases like above where HAL_SD_ERROR_DATA_TIMEOUT was set via SDMMC_FLAG_DTIMEOUT and HAL_SD_Abort() + HAL_SD_Init() could not recover from this.

How To Reproduce

  1. Provoke a HAL_SD_ERROR_DATA_TIMEOUT error.

  2. Correct the hardware setup online (i.e. by adding pull-ups without a system reset).

  3. Use HAL_SD_Abort(), HAL_SD_DeInit() and HAL_SD_Init() to re-initialize the device.

  4. Initialization fails and the error remains.

Additional context I suggest to call SDMMC_CmdStopTransfer() unconditionally within HAL_SD_Abort() to allow proper error recovery. This issue may be also present in other STM32 series HAL like STM32F7.

KRASTM commented 5 months ago

Hello @daniel-starke,

Sorry for the very late response, I hope you are well. I want to know if you still got this problem even after the update of the firmware during this time.

With regards.

daniel-starke commented 5 months ago

I do not have the test setup anymore so it is hard for me to try out at the point. Would you like me to do a code review instead?

KRASTM commented 5 months ago

Hello @daniel-starke,

Since you do not have the setup anymore, I would like to share some information related to the issue.

Usually, it's recommended to use external Pull up to get a better and stable operation, already other board equipped by SD Card contain Pull ups on the lines. I shared your proposal with our development team, they don't accept it because we use HAL_SD_Abort() only when something is ongoing, and the state is busy. So, the call of SDMMC_CmdStopTransfer() must be conditionally within HAL_SD_Abort().

Finally, please allow me to close this issue, thank you for your comprehension.

With regards.

daniel-starke commented 5 months ago

Hello @KRASTM,

The issue is that in case of a HAL_SD_ERROR_DATA_TIMEOUT error (the missing pull-up was just an easy example on how to reproduce it) the HAL layer goes out of sync with the PHY. HAL falsely assumes that the PHY has finished. As is has not and HAL performs no appropriate reset the PHY remains in the error state until system reset. Feel free to use another mechanism to recover from that state. Calling SDMMC_CmdStopTransfer() unconditionally within HAL_SD_Abort() was an easy workaround which worked for me.

KRASTM commented 5 months ago

Hello @daniel-starke,

Thank you for the explanation and the report, at least we have a trace on the topic. I will share this with our team in order to enhance or to find a solution for that case (maybe your workaround in another way).

With regards.