USB superspeed peripherals Forum Discussions

Hello,

I'm trying to implement error handling for UVC controls on CX3. When a control is set to an invalid value, the device should respond with a STALL condition. This in turn causes the host driver to query the error code in a separate control transfer.

To do this, I need to read all of the OUT data of the control transfer, validate it, and decide whether to set the STALL condition. However the documentation for CyU3PUsbGetEP0Data has this:

    If the control request is to be failed with a STALL handshake, the stall
    call has to be made before all of the OUT data has been read. The request
    will be completed with a positive ACK as soon as all of the OUT data has
    been received by the device.

Is there any way around this limitation? I expected to have some way to keep EP0 in forced NAK until the firmware decides what to do, but nothing I have tried seemed to work.

Thanks!

Hello, I'm trying to achieve a simple direct GPIF to USB data transfer with a CYUSB3KIT-003. I started with the GpifToUSB example that can be found in C:\Program Files (x86)\Cypress\EZ-USB FX3 SDK\1.3\firmware\basic_examples\cyfxgpiftousb and that is mentioned in AN86947 Optimizing USB 3.0 Throughput with EZ-USB® FX3™

The problem is/was that I lose data when the DMA buffer switches. I learned that the solution to this is to implement two or more threads and switch between those. A modified GpifToUSB example implementing exactly that can be found in https://community.infineon.com/t5/USB-superspeed-peripherals/EZ-USB-FX3-Explorer-kit-as-16-channel-logic-analyzers-gt-dropped-samples-after/m-p/635888#M35636

I used the source code there and slightly changed it to use an external clock.

Unfortunately it does not solve my problem, I still lose data every ~16 KB (which is the configured DMA buffer size).

I've attached my source code. It's mostly the example from the forum post above, changed to an external clock.

Also, I'm not quite sure what the reasoning behind the size of the counter in the state machine is. Why is it 1/4th of the DMA buffer size and not 1/2?

Any help would be greatly appreciated!

Here's the UART debug log, from the point where I flash the .img into RAM. Afterwards I shortly start and stop a data transfer using the CollectData.exe from CYUSB3KIT\SuperSpeed Design Examples V1.2.1\PC Utilities\

debug initialized
Prod:0 , Cons:0
USB event: 11 0
About to connect to USB host
USB event: 0 1
CY_U3P_USB_EVENT_CONNECT detected
CyFxApplnInit complete
USB event: 8 0
USB event: 5 1
Using GPIF:cyfxgpif2config
Prod:6 , Cons:0
Prod:2 , Cons:0
Vendor Command Received: 0xA0 
Vendor Command Received: 0xA0 
Vendor Command Received: 0xA0 
Vendor Command Received: 0xB5 
Prod:2 , Cons:0
Prod:2 , Cons:0
Prod:2 , Cons:0
Prod:2 , Cons:0
Prod:2 , Cons:0
Prod:1 , Cons:0
Prod:2 , Cons:0
Prod:2 , Cons:0
Prod:1 , Cons:0
Prod:1 , Cons:0
Prod:2 , Cons:0
Prod:1 , Cons:0
Prod:2 , Cons:0
Prod:2 , Cons:0
Prod:1 , Cons:0
Prod:1 , Cons:0
...
[shortened]
...
Prod:2 , Cons:0
Prod:2 , Cons:0
Prod:1 , Cons:0
Prod:1 , Cons:0
Prod:2 , Cons:0
Prod:2 , Cons:0
Prod:2 , Cons:0
Prod:2 , Cons:0
Prod:2 , Cons:0
Prod:2 , Cons:0
Prod:1 , Cons:0
Vendor Command Received: 0xB5 
Prod:1 , Cons:0
Prod:2 , Cons:0
Prod:2 , Cons:0
Prod:1 , Cons:0
Prod:1 , Cons:0
Prod:2 , Cons:0

Here is my state machine:

Edit: I've attached a binary file of about 130 KB of transferred data. It can be easily viewed in a hex editor for example. The transferred data is a 7 bit monotonically rising number. On addresses 0x4000, 0x8000, 0xC000, ... (multiples of 16384 bytes, the configured DMA buffer size) you can see data loss. For convenience I also added a screenshot showing the loss:

And here is a trace of the incoming data (non-shown lines are kept at logic low). I've marked a random sampling edge to show that data is sampled on the rising edge: