Fix(host) enumeration with 8bitdo devices

[hathach]

Describe the PR

• fix(hcd_rp2040) assert/panic endpoint already active: when a device reset while having on-going control transfer
• featuer(host) always get language id, manufacturer, product and serial string. Which is required by some device such as 8bitdo. Should fix HID host issues with RetroFlag Classic USB Gamepad #2163
• add tuh_descriptor_get_string_langid() API

@tannewt please try to see if this works with your 8bitdo gamepad

PS: IAR build failed due to server maintanence

[HiFiPhile]

Since you are working on usbh, I found at least on NUCLEO-H7S3L8 speed detection is more stable by putting tusb_time_delay_ms_api(ENUM_DEBOUNCING_DELAY_MS); at the start of enum_new_device().

[hathach]

Since you are working on usbh, I found at least on NUCLEO-H7S3L8 speed detection is more stable by putting tusb_time_delay_ms_api(ENUM_DEBOUNCING_DELAY_MS); at the start of enum_new_device().

hmm that is weird, delay before port_reset() shouldn't make a difference since port got reset anyway, and the delay afterwards is rather generous ~450ms. Thank you for the hint, I will check this out next time I test with h7rs.

[HiFiPhile]

If I remember correctly dwc2 does speed detection at the end of reset. If DEBOUNCING_DELAY_MS is placed after the end of reset then it won't help speed detection since it's already done.

[hathach]

@HiFiPhile it is weird, there is ENUM_RESET_DELAY_MS which is 50ms per USB specs, it should be enough for dwc2 to detect the link speed, the ENUM_DEBOUNCING_DELAY_MS is mostly waiting for device to bootup/clock init before we start talking to it iirc. I will check again later.

    hcd_port_reset(_dev0.rhport);

    // Since we are in middle of rhport reset, frame number is not available yet.
    // need to depend on tusb_time_millis_api()
    tusb_time_delay_ms_api(ENUM_RESET_DELAY_MS);

    hcd_port_reset_end(_dev0.rhport);

    // wait until device connection is stable TODO non blocking
    tusb_time_delay_ms_api(ENUM_DEBOUNCING_DELAY_MS);

[HiFiPhile]

it is weird, there is ENUM_RESET_DELAY_MS which is 50ms per USB specs,

I think some devices need extra time to power on (DP is pulled high but not ready to communicate), in which case ENUM_RESET_DELAY_MS is not enough. Speed detection is done once hcd_port_reset_end(_dev0.rhport); is called but the device is not ready yet. Then we wait for ENUM_DEBOUNCING_DELAY_MS but since the detection is already done it's not very useful.