[PATCH 1/1] usb: request only 8 bytes for USB_SPEED_FULL bMaxPacketSize0 discovery

[Marek Vasut]

On 8/29/22 08:31, Janne Grunau wrote:
> Fixes probing of various keyboards with DWC3 as integrated into Apple
> silicon SoCs. The problem appears to be requesting more data than the
> devices's bMaxPacketSize0 of 8. Older Logitech unifying receivers
> (bcdDevice 12.03 or 12.10) are for eaxample affected.
> 
> Signed-off-by: Janne Grunau <j at jannau.net>
> Tested-by: Thomas Glanzmann <thomas at glanzmann.de>
> ---
>   common/usb.c | 8 +++++---
>   1 file changed, 5 insertions(+), 3 deletions(-)
> 
> diff --git a/common/usb.c b/common/usb.c
> index 6fcf1e8428e9..48a310e8599d 100644
> --- a/common/usb.c
> +++ b/common/usb.c
> @@ -993,10 +993,12 @@ static int usb_setup_descriptor(struct usb_device *dev, bool do_read)
>   		 *
>   		 * At least the DWC2 controller needs to be programmed with
>   		 * the number of packets in addition to the number of bytes.
> -		 * A request for 64 bytes of data with the maxpacket guessed
> -		 * as 64 (above) yields a request for 1 packet.
> +		 * Requesting more than 8 bytes causes probing errors on the
> +		 * DWC3 controller integrated into Apple silicon SoCs with
> +		 * devices with bMaxPacketSize0 of 8. So limit the read request
> +		 * to the used size of 8 bytes.
>   		 */
> -		err = get_descriptor_len(dev, 64, 8);
> +		err = get_descriptor_len(dev, 8, 8);
>   		if (err)
>   			return err;
>   	}

Sorry for the delay.

Can you be more specific about those logitech receivers ? I might have 
one of those devices, and I have DWC3 in i.MX8MP and i.MX8MQ, as well as 
ZynqMP, so I should be able to try and trigger the problem. Can you 
share the reproducer test case for this problem ?

If the problem is specific to the Apple instance of the controller, 
maybe we need some sort of quirk instead ?

Thanks