TMS570LC4357: mibSPI slave with several transfer groups connected/controlled by the same CS pin (HALCoGen + FreeRTOS)

Tool/software:

Hello,

I am trying to emulate an (mib)SPI slave with a rather cumbersome protocol.

Currently I am able to mimic the protocol on the SPI layer but not on the transport layer by using one transfer group on 16B (request 4B - response 12B) sent in the following manner:

MOSI:  --- | x1 | x2 | x3 | x4 |      -------- 15us ------- | NA | NA | NA | NA | NA | NA | NA | NA | NA |NA | NA | NA |  -----  | a1 | a2 | a3 | a4 |      -------- 15us ------- | NA | NA | NA | NA | NA | NA | NA | NA | NA |NA | NA | NA |  ---
MISO:  --- | NA1 | NA2 | NA3 | NA4 |  -------  15us ------- | y1 | y2 | y3 | y4 | y1 | y2 | y3 | y4 | y1 |y2 | y3 | y4 |  -----  | NA1 | NA2 | NA3 | NA4 |  -------  15us ------- | x1 | x2 | x3 | x4 | x1 | x2 | x3 | x4 | x1 | x2 | x3 | x4 | ---
           |                                                        TGx                                                |  -----  |                                                              TGx                                           |                                                           |  

NOTE: The response (a 4 x copy of the request in order to keep it simple) is delayed by one transmission of the transfer group TGx.

The problem however is that I need  the response immediately, i.e. ~15us after receiving the request (i.e. something like this):

MOSI:  --- | x1 | x2 | x3 | x4 |      -------- 15us ------- | NA | NA | NA | NA | NA | NA | NA | NA | NA |NA | NA | NA |   -----  | a1 | a2 | a3 | a4 |      -------- 15us -------  | NA | NA | NA | NA | NA | NA | NA | NA | NA |NA | NA | NA |  ---
MISO:  --- | NA1 | NA2 | NA3 | NA4 |  -------  15us -------  | x1 | x2 | x3 | x4 | x1 | x2 | x3 | x4 | x1 |x2 | x3 | x4 |  -----  | NA1 | NA2 | NA3 | NA4 |  -------  15us -------  | a1 | a2 | a3 | a4 | a1 | a2 | a3 | a4 | a1 | a2 | a3 | a4 | ---  
           |        TGx            |  -------  15us -------  |                          TGy                             |  -----  |         TGx           |  -------  15us -------  |                         TGy                               |

I have made an attempt to control the transfer group and the data in TGy in the notification function "mibspiGroupNotification".

Basically by doing the following (pseudo code):

void mibspiGroupNotification(mibspiBASE_t *mibspi, uint32 group)
{
    if (mibspi == mibspiREG1)
    {
        if (group == TGx)
        {
            /* Get TGx request data */
            mibspiGetData(mibspiREG1, TGx, &g_TGx_rx_data[0]);
            
            /* Copy request data to TGy response/tx data buffer g_TGy_tx_data[0] */
            ... // memcpy(...);
            
            /* Set TGy data */
            mibspiSetData(mibspiREG1, TGy, &g_TGy_tx_data[0]);
            
            /* Clear enable of TGx */
            mibspiREG1->TGCTRL[TGx] &= ~0x80000000U;
            
            /* Enable TGy */
            mibspiREG1->TGCTRL[TGy] |= 0x80000000U; // mibspiTransfer(mibspiREG1, TGy);
        } else {
            /* Clear TGy */
            mibspiREG1->TGCTRL[TGy] &= ~0x80000000U;
            /* Re-enable TGx */
            mibspiREG1->TGCTRL[TGx] |= 0x80000000U; // mibspiTransfer(mibspiREG1, TGx);
        }
    }
}

This approach seems to almost work, the problem is that I seem to send 1B of old/corrupted data before sending the intended data and by doing that disturbs the data content/order of the TGy buffer/response (as well as TGx). 

Hence I wonder if my intended use case is possible?

Any ideas? Does it seems like something that should be possible to accomplish? 

DISCLAIMER: I do not have a workspace/project that I can share at the moment, hence the reason for sharing only the necessary details needed in order to describe the issue at hand.