LP-AM263: receive of data from FSI-interface (loopback-mode, DMA) in debug mode does never finish

Hi

The ticket is assigned to expert

Thanks & regards

Sri Vidya

Hi Jo,

I cannot reproduce your reported issue with the following steps:

1. Initialize the AM263x CC. The UART will display the following:

2. Load the program. 

3. Set the breakpoint after the semaphores

4. Run the program and reach the breakpoint

5 Continue to run the program.

The program finishes properly.

Your reported issue only happens when I re-load and run the program immediately after the first run without initializing the AM263x CC. I think it may be caused by the resource (like EDMA) may not be released completely.

You may need to flash the SBL_NULL into the QSPI flash to initialize the AM263x CC properly before running the program again.

Please refer to the URL for details on how to initialize the AM263x CC: AM263x MCU+ SDK: Additional Details

Best regards,

Ming

Hello Ming,

I have some further questions:

- what SDK-version did you use?
- did you test also with LP-board? I have no CC-board, I have only a LP-board

Best Regards,

Jo Scho

Hello Ming,

here a some supplements for this problem:

I tried with the SBL-NULL configuration, as you mentioned in your last reply.
(Code Composer Studio 12.1, LP-Board, SDK 8.5.0.24 --> we never reach the second breakpoint here, the firmware waits endless for the RX semaphore (only in debug mode)

I tried now the new SDK version 8.6.0.34 with CCS 12.1 and CCS 12.2 --> the problem is not existing with SDK 8.6.0.34 for both CCS versions

We would like to understand the problem, to be save in further developement.

Best regards

Jo Scho

Hello Ming,

thank you very much four your answer. I just have one question for understanding.

You said in your last answer, that you have changed the Hwi trigger from pulse to level mode.
I tried to understand the modification in the SDK from version 8.5 to 8.6. I recognized the change in the file "source/drivers/edma/v0/edma.c"

The flag "isPulse is set to 1 here, now. In SDK version 8.5 the flag is 0.

The comment of the flag variable indicates another behaviour. (file HwiP.h,  0 == level interrupt, 1== pulse interrupt)

The register description of AM263x would also suggest that setting a 1 means pulse mode.

Are my assumptions wrong? Or is the pulse mode the correct mode, to be used for correct behaviour?

Best regards,

Jo Scho

Hi Ming,

your answer solved the problem, but I discussed the problem and the solution with my collegues yesterday.

We do not really understand, the change of level to pulse or in detail, the effect of the change and why it functions with pulse mode now.
In SDK 8.5 there was level mode for Hwi trigger, in SDK 8.6 this is changed to pulse mode.

In the technical reference manual spruj42c.pdf there is the description of the interrupt type register, this is the register, where the pulse/level information is configured to.

The description says, that the value, set to this register has no effect on how the VIM hardware function. Due to that description, we tried to understand 
that with pulse mode the debugging of the FSI example functions properly and with level mode, debugging fails.

In our understanding, the register is only a buffer, whether the firmware should make a level or pulse evaluation and has no hardware effect. Is this correct?

In this context, we searched for all places in the source code where the function HwiP_isPulse() is called.
The function is used in the Modul HwiP_arm7vr_handlers_nortos.c in the functions HwiP_fiq_handler() and HwiP_irq_handler_c().

EDMA does not use the FIQ-Handler, it uses the normal IRQ-Handler, we think.

Our problem at the moment is, that we do not really understand the source code of the function HwiP_irq_handler_c:

In case of pulse mode, the pending interrupt is cleared before the service routine is executed, in level mode, the clear of a pending IRQ is after the ISR-execution. But in our understanding (VIM hardware is, due to the reference manual, not influenced by the pulse or level mode configuraion) this change no makes no difference for the FSI example.

Perhaps you can give us some more information or explain us, why it functions in pulse mode, but not in level mode.

Best regards

Jo Scho