DWC2: Assert on replug of device after update from 0.16.0 to 0.17.0

[Olstyle]

Operating System

Windows 10

Board

Custom STM32U585 PCB

Firmware

tinyusb + FreeRTOS based project which unfortunately I can not share completely
With the following initialisation of an MSC and Vendor Class device

/* This variable uses the defines from usb_dwc2_cfg.h */
uint8_t const g_ConfigDesc[] =
{
  /* Config number, interface count, string index, total length, attribute, power in mA */
  TUD_CONFIG_DESCRIPTOR(1, ITF_NUM_TOTAL, 0, CONFIG_TOTAL_LEN, 0x00, 100),

  /* Interface number, string descriptor index, EP Out & EP In address, EP size */
  TUD_MSC_DESCRIPTOR(ITF_NUM_MSC, 4, EPNUM_MSC_OUT, EPNUM_MSC_IN, 64),

  /* Interface number, string descriptor index, EP Out & EP In address, EP size */
  TUD_VENDOR_DESCRIPTOR(ITF_NUM_VENDOR_WCID, 5, EPNUM_WCID_OUT, EPNUM_WCID_IN, 64)
};


const char g_VendorId[] = "Me";
const char g_ProductId[] = "My Prog. Dev 3";
const char g_ProductRev[] = "1.0";

static uint8_t s_Serial[] = "0000000000";
/* array of pointer to string descriptors */
char const* g_StringDesc_P[] =
{
  (char[]) { 0x09, 0x04 },        /* 0: is supported language is English (0x0409) */
  "Me",                          /* 1: Manufacturer */
  "My Prog. Dev 3",              /* 2: Product */
  (char*)&s_Serial,               /* 3: Serial */
  "My MSC",                      /* 4: MSC Interface */
  "My WCID",                     /* 5: WCID Interface */
};

/* Public Function Definition ***************************************/

/* USB Device Driver task
 * This top level thread process all usb events and invoke callbacks
 */
void usbHdl_deviceTask(void* Param)
{
    uint8_t SerialArray[4];
    uint32_t Serial;
    (void) Param;

    usbHdl_reOpenMSC();

    portALLOCATE_SECURE_CONTEXT( configMINIMAL_SECURE_STACK_SIZE * 2 );
    //read clock as reference for the USB driver
    SystemCoreClock = mcu_getSystemClock();

    /* Init MSC */
    usb_msc_init(g_VendorId, sizeof(g_VendorId),
                 g_ProductId, sizeof(g_ProductId),
                 g_ProductRev, sizeof(g_ProductRev));


    vendord_init();

    devinfo_getSerial(SerialArray);
    Serial = SerialArray[0] << 24 | SerialArray[1] << 16 | SerialArray[2] << 8 | SerialArray[0];
    convU32ToAsciiString(Serial, s_Serial, 10);

    /* Init module */
    usb_dwc2_init(g_ConfigDesc, g_StringDesc_P, sizeof(g_StringDesc_P)/sizeof(g_StringDesc_P[0]));

    /* set status once via pin read */
    usb_dwc2_setPlug( (dio_readChannel(PIN_USB_SENSE_E) == DIO_HIGH_E) );

    /* RTOS forever loop */
    while (1)
    {
        /* put this thread to waiting state until there is new events */
        usb_dwc2_task(WDGMAN_MAX_TRIG_MS);
    }
}

What happened ?

After switching from version 01.16.0 to 0.17.01, If I plug out the device and plug it in again, I get stuck at an assert in usbd.c line 1049

 // Failed if there is no supported drivers
   TU_ASSERT(drv_id < TOTAL_DRIVER_COUNT);
 }

(the device has a battery which keeps it running after disconnect)
Could this be due to bad descriptors or is this a bug in tinyusb?

How to reproduce ?

1. plug in device
2. plug out device
3. plug device in again

Debug Log as txt file (LOG/CFG_TUSB_DEBUG=2)

logComplete.txt
logFromSuspend.txt

Screenshots

No response

I have checked existing issues, dicussion and documentation

• I confirm I have checked existing issues, dicussion and documentation.

[HiFiPhile]

Please try #2820

[Olstyle]

#2820 fixes it, thank you. Though looking at the change in the PR and the changes from, 0.16.0 to 0.17.0, I ask myself the typical "How did that ever work?". The old code was a direct call to tu_fifo_clear, and tu_edpt_stream_clear is just an alias for that. Why is the additional close suddenly necessary?

[HiFiPhile]

Why is the additional close suddenly necessary?

It's not the fifo who cause the issue. In old version vendor class is erased on reset

tinyusb/src/class/vendor/vendor_device.c

Line 185 in 8d345cd

tu_memclr(p_itf, ITF_MEM_RESET_SIZE);

Since 0.17.0 transfer management has been migrated to stream api but forgot to erase endpoint on reset, the next class opening attempt will fail.