issue with STM32WB55 USB Custom Hid using V1.11.1 firmware pack.

Hi @MBuen.1​ ,

You are far from being an USB expert, the same for me :) . But I can help you by reporting the issue you highlighted to our experts.

I thank you for sharing the solution you found, I'll share it as well to confirm if it is all you have to do or not.

-Amel

Hi @MBuen.1

Just wanted to note, I also ran into the exact same issue with my application with STM32G474RE.

Thank you for the fix. I can now continuously receive data after the first send. However, my device does not appear to send anymore data afterwards.

What were the exact arguments that you passed into the HAL_PCD_EP_Transmit function?

It appears that there is not proper management for interrupt endpoint TX requests.

I've found a solution that works to be this (at least for single packet transfers)

It seems that the code labeled "manage bulk single buffer transaction" is really meant for "manage single buffer non-control transaction". (This also implies it does NOT service isochronous transactions as those are always double buffered). The change in question is on line 9 of the code snippet, which is on line 1854 of the stm32g4xx_hal_pcd.c file.

I added a check for interrupt endpoint, as those are never double buffered. It seems when the library was refactored, interrupt endpoints were overlooked as everyone typically uses CDC, Audio, or Mass Storage, which makes use of bulk or isochronous endpoints.

if ((wEPVal & USB_EP_CTR_TX) != 0U)
 {
 ep = &hpcd->IN_ep[epindex];
 
 /* clear int flag */
 PCD_CLEAR_TX_EP_CTR(hpcd->Instance, epindex);
 
 /* Manage Single Buffer Transaction */
 if (((ep->type == EP_TYPE_BULK) && ((wEPVal & USB_EP_KIND) == 0U)) || (ep->type == EP_TYPE_INTR))
 {
 /* multi-packet on the NON control IN endpoint */
 TxByteNbre = (uint16_t)PCD_GET_EP_TX_CNT(hpcd->Instance, ep->num);
 
 if (ep->xfer_len > TxByteNbre)
 {
 ep->xfer_len -= TxByteNbre;
 }
 else
 {
 ep->xfer_len = 0U;
 }
 
 /* Zero Length Packet? */
 if (ep->xfer_len == 0U)
 {
 /* TX COMPLETE */
#if (USE_HAL_PCD_REGISTER_CALLBACKS == 1U)
 hpcd->DataInStageCallback(hpcd, ep->num);
#else
 HAL_PCD_DataInStageCallback(hpcd, ep->num);
#endif /* USE_HAL_PCD_REGISTER_CALLBACKS */
 }
 else
 {
 /* Transfer is not yet Done */
 ep->xfer_buff += TxByteNbre;
 ep->xfer_count += TxByteNbre;
 (void)USB_EPStartXfer(hpcd->Instance, ep);
 }
 }
 /* Double Buffer Iso/bulk IN (bulk transfer Len > Ep_Mps) */
 else
 {
 (void)HAL_PCD_EP_DB_Transmit(hpcd, ep, wEPVal);
 }
 }

If you must use STM32CubeIDE or are constantly regenerating a CubeMX project, a really ugly hack to save yourself trouble is to uncheck "call HAL handler" in the NVIC settings, and copy the existing handler into a user function. Then have the interrupt target your new user handler. In the user handler, instead of calling "PCD_EP_ISR_Handler", copy this into another user function, and make the above change to that user function. Then modify the user copied HAL_PCD_IRQ_Handler to call the new user PCD_EP_ISR_Handler that has the changes. This is really ugly, and it would be much preferred if ST could push a hotfix for this in the firmware packages.

This is what the old HAL code looked like and and it can be seen why it worked, and also why they felt a need to refactor it. It's also clear how such a bug can be made. Overlooked the interrupt endpoint.

Reminder:

• control endpoint transactions are handled separately in the ISR.
• interrupt endpoints are never double buffered, due to their (relatively) low frequency, periodic nature.
• bulk endpoints are the optionally double buffed endpoints in question.
• isochronous endpoints are always double buffered.

/* Manage all non bulk transaction or Bulk Single Buffer Transaction */
 if ((ep->type != EP_TYPE_BULK) ||
 ((ep->type == EP_TYPE_BULK) && ((wEPVal & USB_EP_KIND) == 0U)))

Hi @Amel NASRI​ ​, if you could, can you add to the issue report that it has been experienced by multiple users? I appreciate the help you are providing.

Thank you.

-Kai Yin