Why check length at the end of ux_device_class_storage_inquiry? It seems not compliant with protocol.

[xuzihan351]

code place: ux_device_class_storage_inquiry.c:+249

/* Check length.  */
if (storage -> ux_slave_class_storage_host_length != inquiry_length)
{
    storage -> ux_slave_class_storage_csw_residue = storage -> ux_slave_class_storage_host_length - inquiry_length;
    _ux_device_stack_endpoint_stall(endpoint_in);
}

When deal with inquiry cmd, the transfer length may not equal with dCBWDataTransferLength, the code check it and stall the endpoint if it is not match.

Considered code in the same file, ux_device_class_storage_inquiry.c:+194

     /* Send a data payload with the inquiry response buffer.  */
    if (inquiry_length > 24)
        inquiry_length = 24;

the inquiry_length's value is changed, but storage -> ux_slave_class_storage_host_length is not. So the endpoint stall will occur once the code run into line 194.

[xiaocq2001]

According to Section 6.7 The Thirteen Cases USB mass storage class bulk-only transport spec, the way to operate on device side in case host requested length not equal to device sending length is:

The case matrix:

Case (5) and case (7) is the reference:

Your description is Hi > Di (case 5), the actual data back is not as expected and we are not padding data, so there should be a STALL.

Note the behavior here is checked by USB CV test tool.

[xuzihan351]

Thanks for the explanation, it is indeed correct. Some other usb stacks do not make endpoint into stall status in the case, so that is their fault.