Enabling IP DMA on USB when enabling DCACHE on STM32H7A(B)

[Claydonkey]

I have referred to this document but it gives no solution.

https://community.st.com/s/article/how-to-enable-dma-in-usb-with-stm32h7-devices?t=1659888985140

I have tried putting all the USB variables and arrays into a noncacheable region of AXI1 RAM but the USB is not firing up. I will provide more details of what is happening after I debug...

[Claydonkey]

Solved problem...

Sorry this was not true after I reset the device. As Piranha says below it is only the pointer....

I was using CDC and it's buffer was allocated inside the init function:

USBD_CDC_Init(USBD_HandleTypeDef *pdev, uint8_t cfgidx)
{
...
USBD_CDC_HandleTypeDef   *hcdc = (USBD_CDC_HandleTypeDef *)USBD_malloc(sizeof(USBD_CDC_HandleTypeDef));
...
}

#define _NO_CACHE_USB_ __attribute__((section (".DATA_RAM_NOCACHE"))) 
 
USBD_CDC_HandleTypeDef  _NO_CACHE_USB_ *hcdc;static uint8_t 
 
USBD_CDC_Init(USBD_HandleTypeDef *pdev, uint8_t cfgidx)
{
...
hcdc = (USBD_CDC_HandleTypeDef *)USBD_malloc(sizeof(USBD_CDC_HandleTypeDef));
...
}

After moving it out and assigning it a non cacheable region the USB works as expected

[Piranha]

You have moved only the pointer to the allocated memory, which doesn't solve anything...

https://community.st.com/s/question/0D50X00009XkZZeSAN/cubemx-static-usbd-setting

https://community.st.com/s/question/0D50X00009Xkft5SAB/does-usb-device-code-for-stm32-really-allocate-dynamic-memory-from-interrupts

https://community.st.com/s/question/0D50X0000AAHSlfSQH/why-does-the-hal-usb-cdc-driver-call-malloc-in-an-isr-context

[Claydonkey]

Indeed it was a red herring.

There is a strange phenomena though... It works immediately after flashing a DCACHE'ed build but only once and it borks afte a reset... wierd.

I also neglected to say that I am using the MPU to set the area of AXI1 RAM to non catchable. (probably obvious).

So, as Piranha rightly pointed out - I only moved the pointer. Therefore I moved hcdc to the stack "hcdc [1]". Still no dice. As far as I can see I have moved all the variables used by HAL_USB to the non cachable region - alas I am shooting in the dark and I suppose a more through understanding of the IP DMA is required on my part