Issues in Firmware Package "STCubeFW_F4_V1.25.0" for USB HID Code Generation

[BPaik]

Using Firmware Package "STCubeFW_F4_V1.25.0" in CubeMX appears to cause various issues with generating code for USB HID mode. I was originally able to get a USB HID project working by following the USB HID Custom Device Lab tutorial (found here https://www.youtube.com/watch?v=3JGRt3BFYrM&t=325s) using firmware package 1.24.1 with the STM32F405. USB functionality was broken after migrating my project to V1.25.0, which causes my USB PC Host to report "Transfer error on interrupt endpoint: Operation timed out". In order to verify the problem was in fact part of the V1.25.0 package, I repeated the ST tutorial linked above with both versions of the firmware package and verified that the project is able to communicate with the HID_Terminal .exe application in V1.24.1 but crashes the application with the V1.25.0 version.

Some other issues I noticed using CubeMX include the following:

• Setting the "CUSTOM_HID_FS_BINTERVAL" parameter in CubeMX does not have any effect on the code. The corresponding definition in "usbd_customhid.h" does not change regardless of what the value is set to.
• After following the tutorial to get my baseline project working, CubeMX will revert all the changes every time I regenerate my code to configure other peripherals. This is extremely frustrating and tedious to fix, and so far I haven't found a way to prevent this from happening. I would have expected the number of bytes transmitted to be configurable in CubeMX rather than fixed at 2 bytes. At the very least the default generated code should match the tutorial - as buffer pointer and 64 byte transfers are much more usable than the "state" and "event_idx" used in the "CUSTOM_HID_OutEvent_FS" routines.

[HLee.21]

This seems to be a regression. I'm having the exact same issue on an STM32L073CBTx with Firmware Package STM32Cube FW_L0 V1.12.1. After adapting the generated code like in the youtube video: HID_terminal.exe works two times (sends and recieves data by clicking on Send-Button) but fails after the secound time with an error "Operation timed out".

The failure is reproducible on Linux with a similar app.

@ST: After one and a half year, please recreate the issue and fix it.

[BPaik]

I've only tried this on the G4 so far, but here are a few things that got USB HID working after generating CubeMX code (some of this will be a repeat of information in the tutorial):

1. Change CUSTOM_HID_EPIN_SIZE and CUSTOM_HID_EPOUT_SIZE to 0x40 in usbd_customhid.h
2. Change the "OutEvent" function pointer in usbd_customhid.h to only have the state pointer as an input: int8_t (* OutEvent)(uint8_t *state)
3. Update the OutEvent function calls in usbd_customhid.c to only pass the buffer pointer
4. Make edits to the CUSTOM_HID_ReportDesc_FS in usb_hid_api.h as described in the tutorial video
5. Update CUSTOM_HID_OutEvent_FS in usb_hid_api.h to only have the state pointer as an input
6. Replace the contents of CUSTOM_HID_OutEvent_FS with the code below (g_usb_rx_bytes and g_usb_tx_bytes are 64 byte arrays for RX and TX data respectively).

 memcpy(g_usb_rx_bytes, state, 64);

 USBD_CUSTOM_HID_SendReport(&hUsbDeviceFS, g_usb_tx_bytes, 64);

 USBD_CUSTOM_HID_ReceivePacket(&hUsbDeviceFS);

 return (USBD_OK);

[HLee.21]

USBD_CUSTOM_HID_ReceivePacket(&hUsbDeviceFS); <- this isn't in the tutorial, does that make the difference?

[BPaik]

It did for me on the G4. I'm not sure if it will work for you but let me know if it does.

[HLee.21]

Looks a bit odd, but I'll test it tomorrow and get back to you. Thanks for your quick input!

[HLee.21]

In L0 there is no USBD_CUSTOM_HID_ReceivePacket(). Firmware Version V1.12.0 from year 2020 does work, ... so I guess I have to stick to that. But it would be nice if ST could fix the regression.

[BPaik]

Open the file "stm32l0xx_hal_pcd.c" around line 1873 and change the if statement to the following:

    /* 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)))

Let me know if that works.

[HLee.21]

--- a/Drivers/STM32L0xx_HAL_Driver/Src/stm32l0xx_hal_pcd.c
+++ b/Drivers/STM32L0xx_HAL_Driver/Src/stm32l0xx_hal_pcd.c
@@ -1851,28 +1851,6 @@ static HAL_StatusTypeDef PCD_EP_ISR_Handler(PCD_HandleTypeDef *hpcd)
         /* clear int flag */
         PCD_CLEAR_TX_EP_CTR(hpcd->Instance, epindex);
 
-        if (ep->type != EP_TYPE_BULK)
-        {
-          ep->xfer_len = 0U;
-
-          if ((wEPVal & USB_EP_DTOG_TX) != 0U)
-          {
-            PCD_SET_EP_DBUF0_CNT(hpcd->Instance, ep->num, ep->is_in, 0U);
-          }
-          else
-          {
-            PCD_SET_EP_DBUF1_CNT(hpcd->Instance, ep->num, ep->is_in, 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
-        /* Manage Bulk Single Buffer Transaction */
         if ((ep->type != EP_TYPE_BULK) ||
                              ((ep->type == EP_TYPE_BULK) && ((wEPVal & USB_EP_KIND) == 0U)))
         {

I had to apply the above patch to get it working - but I have 0 idea why it does! - an answer from ST would be great.

[HLee.21]

the ST code seems obviously wrong: They double test for ep->type != EP_TYPE_BULK:

if (a) {
...
} else if (a || b) {
 ...
}

... @ST: any comment???