uDMA not working with SSI0 in slave mode

Hi,

I used the debugger to inspect the area of memory that the uDMA is using and I am getting the transfer but I am not getting the interrupt to let me know about the transfer.

Regards,

Susie Johnson

I finally got the interrupts going this morning.  I had the receive and transmit buffers sized at 512 bytes and was only sending 11 bytes from the Aardvark to the micro.  I was setting up the transfer:

     uDMAChannelTransferSet(UDMA_CHANNEL_SSI0RX | UDMA_PRI_SELECT,

                                UDMA_MODE_BASIC,

                                (void *)(SSI0_BASE + SSI_O_DR),

                                SomRxBuf, sizeof(SomRxBuf));

which has a size of 512 bytes.  This morning I changed the call to:

     uDMAChannelTransferSet(UDMA_CHANNEL_SSI0RX | UDMA_PRI_SELECT,

                                UDMA_MODE_BASIC,

                                (void *)(SSI0_BASE + SSI_O_DR),

                                SomRxBuf, 11);

And now I get the receive interrupt when the transfer is complete.  I am still getting some peculiar behavior on the transmit interrupts but am making progress.

I have gotten the receive side working perfectly but the transmit side is not working properly.  I have SSI0 connected to an Aardvark SPI tool running at 125kHz and using mode SSI_FRF_MOTO_MODE_1.  SSI0 is set up in slave mode with the Aardvark set to master mode.  I have loaded the transmit buffer on the LM4F232 micro with data to transmit to the Aardvark.  When I send data from the Aardvark to the micro, I do the data from the buffer on the micro but it is not transmitted in the correct order. 

 Data in buffer: 0x41, 0x42, 0x43,  0x44, 0x45, 0x46, 0x47, 0x48, 0x49, 0x4A, 0x4B, 0x4C, 0x4D, 0x4E, 0x4F, 0x50

 Data received on the Aardvark 1^st transmission:  0x4B, 0x4C, 0x4D, 0x4E, 0x4F, 0x50, 0x41, 0x42, 0x41, 0x42, 0x43,  0x44, 0x45, 0x46, 0x47, 0x48

Data received on the Aardvark 2^nd transmission:  0x4A, 0x4B, 0x4C, 0x4D, 0x4E, 0x4F, 0x50, 0x41, 0x42, 0x43,  0x44, 0x45, 0x46, 0x47, 0x48, 0x49

Data received on the Aardvark 3^rd transmission:  0x4B, 0x4C, 0x4D, 0x4E, 0x4F, 0x50, 0x41, 0x42, 0x43,  0x44, 0x45, 0x46, 0x47, 0x48, 0x49, 0x4A

Data received on the Aardvark 4^th transmission:  0x4C, 0x4D, 0x4E, 0x4F, 0x50, 0x41, 0x42, 0x43,  0x44, 0x45, 0x46, 0x47, 0x48, 0x49, 0x4A, 0x4B

 The first transmission is always totally messed up and then it starts the shifting byte thing.  I have attached my latest code.

  4810.SOMspi.c

I did notice something interesting this morning.  I am getting two interrupts per transfer.  When I step through the interrupts, the following is what I notice:

 In the first Interrupt the following code gets executed in the ISR:

 ulStatus = ROM_SSIIntStatus(SSI0_BASE, 1);

 ROM_SSIIntClear(SSI0_BASE, ulStatus);

 ulMode = ROM_uDMAChannelModeGet(UDMA_CHANNEL_SSI0RX | UDMA_PRI_SELECT);  (After this is executed ulMode is zero.)

     if(!ROM_uDMAChannelIsEnabled(UDMA_CHANNEL_SSI0RX))

   {

        uDMAChannelTransferSet(UDMA_CHANNEL_SSI0RX | UDMA_PRI_SELECT,

                                   UDMA_MODE_BASIC,

                                   (void *)(SSI0_BASE + SSI_O_DR),

                                   SomRxBuf, 16);

        ROM_uDMAChannelEnable(UDMA_CHANNEL_SSI0RX);

    }

    if(!ROM_uDMAChannelIsEnabled(UDMA_CHANNEL_SSI0TX))

    {

        TxBufCount++;

        uDMAChannelTransferSet(UDMA_CHANNEL_SSI0TX | UDMA_PRI_SELECT,

                                   UDMA_MODE_BASIC, SomTxBuf,

                                   (void *)(SSI0_BASE + SSI_O_DR),

                                   16);

        ROM_uDMAChannelEnable(UDMA_CHANNEL_SSI0TX);

    }

 In the second Interrupt the following code gets executed in the ISR:

 ulStatus = ROM_SSIIntStatus(SSI0_BASE, 1);

 ROM_SSIIntClear(SSI0_BASE, ulStatus);

 ulMode = ROM_uDMAChannelModeGet(UDMA_CHANNEL_SSI0RX | UDMA_PRI_SELECT);  (After this is executed ulMode is one.)    

 if(ulMode == UDMA_MODE_BASIC)

     {

        RxBufCount++;

   }

 I thought there was supposed to be only one completion interrupt generated by the uDMA channel. 

First - must applaud your efforts - detail and care in your writings.  My "non-mastery" of uDMA prevents better response.

I can offer the guidance that when an interrupt is cleared - too deep (or far down) w/in a service routine - re-entry to that service routine may occur.  You clear yours @ 2nd instruction - so that appears not to be your case.  (unless the code you presented here - is a small extract of the "real deal."

Sometimes writing test code in the simplest form possible - lends greater light on such issues.  Perhaps you could increment a variable w/in the troublesome service - and see if indeed it agrees/confirms your +2 assertion. 

Debuggers have been seen to sometimes cause - or erroneously report - such issues.  Would be beneficial to run the code and observe - w/ test variable as suggested...

Wish I knew and/or could assist further - perhaps one w/greater uDMA "history" w/Stellaris can help...

Thanks cb1_mobile for the suggestion.  I put an interrupt counter variable in the code and incremented it each time the interrupt was called.  I sent 3 transactions with no breakpoints to interrupt the running  code and my variable reports that it was indeed called 6 times.  So I am getting two interrupts per transaction.

hi Susi,

there is an unclear thing in your source code.

To clear the interrupt you use the following statements:

ulStatus = ROM_SSIIntStatus(SSI0_BASE, 1);

ROM_SSIIntClear(SSI0_BASE, ulStatus);

but while SSIIntStatus return a bit mask made by 4 bits, SSIIntClear take a bit mask made by only 2 bits as parameter.

according to me there is something wrong

best regards

mastupristi