AM335x DDR to USB with DMA

I am trying to transfer bytes from DDR to a USB bulk endpoint at a fast data rate (>15MB/s).  I started with the usb_dev_bulk example project, running on a BeagleBoneBlack.  I defined DMA_MODE, included cppi4dma.h, initialized the DMA with the following code:

endpointInfo epInfo[] = {
{ USB_EP_TO_INDEX(USB_EP_1),
CPDMA_DIR_RX,
CPDMA_MODE_SET_TRANSPARENT, },
{ USB_EP_TO_INDEX(USB_EP_1),
CPDMA_DIR_TX,
CPDMA_MODE_SET_GRNDIS, } 
};

#define NUMBER_OF_ENDPOINTS 2 //Total number of send points(RX +TX) used in this USB configuration

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

tBulkInstance *psInst;

tUSBDBulkDevice *psDevice = (tUSBDBulkDevice *)&g_sBulkDevice;
// Get a pointer to our instance data.
psInst = psDevice->psPrivateBulkData;

Cppi41DmaInit(USB_INSTANCE, epInfo, NUMBER_OF_ENDPOINTS);

I am successfully able to transmit some bytes using the following two lines of code , but my program eventually crashes after a few megabytes.  

doDmaTxTransfer(USB_INSTANCE, 0x80000000,65536, psInst->ucINEndpoint);
enableCoreTxDMA(USB_INSTANCE, psInst->ucINEndpoint);

1.  Is it necessary to call cppiDmaAllocBuffer() if I'm transferring from DDR anyway?  0x80000000 is the start of DDR which contains my data.  I can verify this works with a few bytes.

2.  If I put those two lines of code in a loop to continuously send data, it crashes.  What do I need to watch for to make sure the transmit is complete before sending more?  The crash says "UndefInstHandler() 0x8000F508 (next frame is identical to an existing frame)".  How can I figure out what is causing the crash?  Am I not handling a callback or an interrupt?

3.  Is there some documentation for the "cppi41dma.h" API so I don't have to randomly guess how to call the proper functions?

Thanks!

-marc