USB superspeed peripherals Forum Discussions

Hello.

I've noticed that sometimes the USB endpoint communications locks up when a CYU3P_USBEP_SS_RESET_EVT occurs in the FX3 SDK 1.3.4.  My firmware handles the event by stalling the endpoint similar to the examples provided.  The only recovery I've found is to unplug/re-plug.  I tried resetting the endpoint, flushing memory, destroying DMA channels, etc.  Nothing could recover the FX3 endpoint.  I started troubleshooting to find the root cause. 

I tracked down the problem to somewhere in the libcyfxapi.a library in FX3 SDK 1.3.4.  (I am using the pre-built library in Cypress\EZ-USB FX3 SDK\1.3\fw_lib\1_3_4\fx3_release\libcyfxapi.a)  I then built that library myself from the 1.3.4 SDK source files.  When I build the library myself using the gcc tool suite (I do not have the RVCT tool suite),  the problem goes away with my gcc built library.

I tracked down the problem to the file cyu3usbpp.o in the pre-built library.  I was able to extract the cyu3usbpp.o file from the pre-built libcyfxapi.a (dated 5/23/2018) and substitute that cyu3usbpp.o into my gcc built library.  The problem tracks with the cyu3usbpp.o file from the pre-built library.

By carefully substituting functions from the pre-built library into my gcc build I was able to isolate the problem to the CyU3PUsbEpPrepare() routine in the pre-built library.  The CyU3PUsbEpPrepare() is called from CyU3PUsbResetEndpointMemories()  which happens upon a CYU3P_USBEP_SS_RESET_EVT event.

This is where it gets interesting.  I found that the CyU3PUsbEpPrepare() function in the pre-built libcyfxapi.a library does not call CyU3PUibSocketInit() upon entry (whereas the source code does).  Additionally, I found that if I remove the CyU3PUibSocketInit() call from my gcc built library, the problem occurs in my gcc built library as well.  Additionally, the problem occurs with the 1.3.3 SDK libraries which also are missing the CyU3PUibSocketInit() call.

Here are my questions:

1) Why is the function CyU3PUsbEpPrepare() different in the pre-built libcyfxapi.a library versus the supplied source files?

2) Since I do not have the source code for the function CyU3PUsbEpPrepare() that is in the pre-built libcyfxapi.a file, is that the only difference in the library versus the supplied source code?

3) Does it make sense that missing the CyU3PUibSocketInit() call upon entry to the CyU3PUsbEpPrepare() function, that the FX3 could never recover that endpoint communications?

4) Is it possible to get pre-built libraries that match the supplied source code in the SDK.

Frank

Hello,

I've still been working for an issue of the post.

Thank you for Infineon's supports for the post. 
The topic was closed while I was working on some another FPGA issues on my product but 
 I found that the reason of the issue was confliction between transfers of two end points.
It seems that a GPIF II -> USB Bulk IN transfer is interrupted by a UART -> USB Bulk IN transfer.

------------
My product uses several Bulk IN channels,

1) GPIF II -> USB (to transfer large amount of data from external FPGA with high transfer rate continuously) 
2) GPIF II -> USB (to transfer up to 32KB data from external FPGA occasionally) 
3) UART -> USB  (to transfer 32bytes fixed length short data from external MCU occasionally)
4) FX3 -> USB  (to transfer debug print messages generated in FX3 occasionally)

These channels are polled by PC application asynchronously.

I found that attempts of XferData() for UART -> USB channel always ended with timeout. 
The external MCU sends packets only occasionally and the FX3's buffer is empty most time.
No timeout is seen when a packet is sent from the MCU and FX3's buffer is filled.

I checked to call XferData() separately for each one of these 4 channels and
only UART -> USB channel had this issue.
(and  the issue of the post is solved by not using the UART -> USB channel and GPIF II->USB channel simultaneously)

I would like to confirm that

1) Is this behavior of UART->USB port normal?
Or should the FX3 response ZLP (or something) immediately when the buffer is empty?

2) Is it possible to use XferData() "by two or more threads" "asynchronously"
"for different endpoints" of "one FX3 device"?
I read KBA94607 and it's possible with my understanding, but it's still not clear for me.

------------

I attach a wireshark log of USB transfer attempts.
The address of my FX3 product is 1.14. The UART->USB channel is 1.14.3 (EP 0x83).
Most polling attempts end with timeout in the log, however, 
line 326 - 347 are intended data transfers and no problem is seen.

1) GPIF II -> USB,  4)FX3->USB channels are not activated to clarify the issue. 
Address 1.14.2 is the 2)GPIF II->USB channel. 

The UART channel setting is as shown below.

----------
UART Initialization

apiRetStatus = CyU3PUartInit();
if (apiRetStatus != CY_U3P_SUCCESS)
{
DebugPrint("UART_Initialize, CyU3PUartInit Failed, Error code = %d\n", apiRetStatus);
CommandErrorHandler(apiRetStatus);
}

//UART Settings
CyU3PMemSet ((uint8_t *)&uartConfig, 0, sizeof (uartConfig));
uartConfig.baudRate = CY_U3P_UART_BAUDRATE_115200;
uartConfig.stopBit = CY_U3P_UART_ONE_STOP_BIT;
uartConfig.parity = CY_U3P_UART_NO_PARITY;
uartConfig.txEnable = CyTrue;
uartConfig.rxEnable = CyTrue;
uartConfig.flowCtrl = CyFalse;
uartConfig.isDma = CyTrue;

apiRetStatus = CyU3PUartSetConfig (&uartConfig, NULL);

//Set the UART transfer to a really large value.
apiRetStatus = CyU3PUartTxSetBlockXfer (0xFFFFFFFF);

//TEST
CyU3PUartSetTimeout(1,1); // this did not affect the behavior of the channel

-------------
EP Initialization

CyU3PEpConfig_t epCfg;
CyU3PReturnStatus_t apiRetStatus = CY_U3P_SUCCESS;

CyU3PMemSet ((uint8_t *)&epCfg, 0, sizeof (epCfg));
epCfg.enable = CyTrue;
epCfg.epType = CY_U3P_USB_EP_BULK;
epCfg.burstLen = 0; //Burst length 1
epCfg.streams = 0;
epCfg.pcktSize = size;

apiRetStatus = CyU3PSetEpConfig(EP_COMMAND_CONSUMER, &epCfg);
if (apiRetStatus != CY_U3P_SUCCESS)

CyU3PUsbFlushEp(EP_COMMAND_CONSUMER);

--------------
DMA Initialization

CyU3PDmaChannelConfig_t dmaCfg;
CyU3PReturnStatus_t apiRetStatus = CY_U3P_SUCCESS;

dmaCfg.size = 32;
dmaCfg.count = 8;
dmaCfg.prodSckId = CY_U3P_LPP_SOCKET_UART_PROD;
dmaCfg.consSckId = EP_COMMAND_CONSUMER_SOCKET;
dmaCfg.dmaMode = CY_U3P_UIB_SOCKET_CONS_3;
dmaCfg.notification = 0;
dmaCfg.cb = NULL;
dmaCfg.prodHeader = 0;
dmaCfg.prodFooter = 0;
dmaCfg.consHeader = 0;
dmaCfg.prodAvailCount = 0;

apiRetStatus = CyU3PDmaChannelCreate (&glCommandIN_DMA_Handle,
CY_U3P_DMA_TYPE_AUTO, &dmaCfg);

apiRetStatus = CyU3PDmaChannelSetXfer (&glCommandIN_DMA_Handle, 0);

Hi infineon team,
we have a code that perfectly works on EZ-USB 1.3.1 version. now we try to build it using EZ-USB 1.3.4 version.
we faced memory related issues but we solved it and build was successful device enumerated as camera and HID as well . device is working fine.
the problem is when the HOST application is sending HID request to the device through the HID endpoint, the device didnt recogonise the request using HID callback function.
i have confirmed the request has been sent from the host application using protocol analyzer. But the request the was not detected by CX3 through HID call back function.
kindly help me to solve why CX3 is not receiving Request related to HID.

Note:
1) The HID call back was register successfully.we confirmed using CyU3PReturnStatus_t API.
2) Application is sending HID request.we confirmed using protocol analyzer.
3) The same code is working fine when build on 1.3.1 version.so i dont think there is logical mistake in code (like handling Callback functions).

Hi all,

I get CY_U3P_USB_EVENT_EP_UNDERRUN when getting data from two channels (audio, and video). It happens in two scenarios. 

1. If two channels are active for a long time 

2. if while both are active one is being stopped.

Seems like this is corrupting data and causes devices to stop responding.

I have checked the below threads for solution but the ones mentioned did not work for me.

1. In this thread I don't understand the usage of the "CyU3PUsbResetEndpointMemories()" to recover the corrupted state. Also found another API call "CyU3PUsbEnableEPPrefetch()" in FX3 1.3.4 version , which should fix the issue but I am not able to properly use it. Where should those guys be called?

https://community.infineon.com/t5/USB-superspeed-peripherals/What-triggers-CY-U3P-USB-EVENT-EP-UNDERRUN-on-EP-0x80/m-p/62387

2. In this thread the solution is pointing to the spec of GPIF but not exactly mentioned how to fix the GPIF state machine mirroring problem.

https://community.infineon.com/t5/USB-superspeed-peripherals/unexpected-values-of-CX3-machine-state/m-p/156093

Please suggest how to overcome this issue.

Thanks in advance.

Hi, I am already using ARM GCC and -O3, but I need my image to get even smaller to fit on EEPROM. My max file size is 118,528 bytes, and I'm slightly over that.

Is there anything I can strip out of the libraries? Please advise.

Hi all,

I have an Ascella developement kit with the CX3 chip. (https://www.e-consystems.com/Cypress-CX3-THine-ISP-13MP-RDK.asp)
I've successfully run firmware using the Cypress EZ USB Suite and UART for debugging.

Now I'm trying to get the JTAG debugger running. I did everything acording to the guide in the EZ USB suite.
However, it is not able to find a debugging device. I assume it has something to do with the drivers installed.

Could you tell me which exact drivers I need to install in order to get it running?

In the attachements I have screenshots from the currently installed drivers, my debug config and the error message.

Thanks for your help,
M.

Hello,

I would like to build an application where I have one USB->CPU and two CPU->peripheral dma channels. When a buffer from USB is received i have to decide to which of the two CPU->peripheral dma channels the buffer need to be forwarded.  I started from the FX3 cyfxbulklpmaninout example and it works. But in the example the incoming buffer is copied to the outgoing. To improve the performance I would like to directly send the incoming buffer to one of the outgoing channels. I cant figure out which API I have to use to send the received buffer to a second channel.

BR

I have implemented a GPIE state machine including 18 states. Two of them trigger an interrupt. In the callback  "CyFxApplnGPIFEventCB " I want to distinguish these two states using the "currentSmState" parameter. I expected numbers between 0 and 17, but I got 139. How do I have to interpret this number?