Issue with FFTC and Packet DMA in TMS320TC16​618

Hi Srinivas,

Welcome to the TI E2E forum. I hope you will find many good answers here and in the TI.com documents and in the TI Wiki Pages (for processor issues). Be sure to search those for helpful information and to browse for the questions others may have asked on similar topics (e2e.ti.com).

1. Have you observed this TI provided examples? 

We are working with experts to answer your post. Thank you for your patience.

Hi Rajasekaran,

Similar issue is raised by others in this forum.

Restarting an SRIO Messaging application on C6670 without the use of a System Reset

http://e2e.ti.com/support/dsp/c6000_multi-core_dsps/f/639/t/297491.aspx

Overall issue is same: How to reset the "Multicore Navigator" (to bring into the known state) using DSP core at restart? My observation is similar to what others are reported. Soft reset is not brining "Multicore Navigator" to the known state, where as "Hard reset" works fine.

I used the following code to reinitialize the QMSS (Thanks to Lance S) (picked up from the following thread)
http://e2e.ti.com/support/dsp/c6000_multi-core_dsps/f/639/t/297491.aspx
Executed the below code on Core#1  (not on core#0)
// reset the queue manager by clearing the memory regions and pushing a NULL
// pointer on each queue (this should destroy the descriptors of each queue).
void ReinitializeQmssDevice()
{
    uint32 u32MemoryRegion;
    uint32 u32QueueNum;
    uint16 u16Count;
    CSL_Qm_descriptor_region_configRegs *poConfigReg = (CSL_Qm_descriptor_region_configRegs *)CSL_QM_SS_CFG_DESCRIPTION_REGS;

    // Zero out the memory regions. The C6670 has 20 separate memory regions for descriptor sets.
    for(u32MemoryRegion = 0; u32MemoryRegion < 20; u32MemoryRegion++)
    {
        poConfigReg->MEMORY_REGION_BASE_ADDRESS_GROUP[u32MemoryRegion].MEMORY_REGION_BASE_ADDRESS_REG = 0;
        poConfigReg->MEMORY_REGION_BASE_ADDRESS_GROUP[u32MemoryRegion].MEMORY_REGION_DESCRIPTOR_SETUP_REG = 0;
        poConfigReg->MEMORY_REGION_BASE_ADDRESS_GROUP[u32MemoryRegion].MEMORY_REGION_START_INDEX_REG = 0;
    }

    // Push a NULL pointer on each queue. This will loop through all the queues
    // and ensure each is in an empty state. The Qmss_queueEmpty() will write a
    // NULL pointer to the Queue Register D.
    for(u32QueueNum = 0; u32QueueNum < 8192; u32QueueNum++)
    {
        u16Count = Qmss_getQueueEntryCount(u32QueueNum);
        if(u16Count)
        {
            Qmss_queueEmpty((Qmss_QueueHnd) u32QueueNum);
        }
    }
}

I executed the above code on Core#1 at restart initialization routine. This doesn't help. I got the same error in executing FFTC (as listed below):
0X21F0014  : FFTC_ERROR_INTERRUPT_RAW_STATUS_REGISTER 0x00000004
==> RX buffer starvation error occurred in queue n.
0x21F0014  : FFTC_ERROR_INTERRUPT_SET_REGISTER 0x00000004
==> RX buffer starvation error occurred in queue n.

What I have not tried yet.
1. Packet DMA teardown using CSL functions (as handles are lost during retrain).
Please list the code to do this.
2. Setting QMSS registers to default value.
Is it available in TI docs? (I am planning to put a break point, capture the QMSS default values and hardccode them in the code, bruteforce approach).

Thanks & Regards,
Srinivas

Hi Rajashekaran,

I tried to run the following routines at powerup initialization along with resetting the QMSSdevice( ) (as listed above).

targetFreeQs();
hwQmTeardown();

ReinitializeQmssDevice();

I see the same behaviour. Code reset (soft) still gives the Rx buffer starvation error (as listed above), where as Hard Reset (using a button on TI EVM 6670) works fine. I am assuming hwQMTearDown() will also do the packet DMA teardown. If not, please specify how to do the Packet DMA teardown using the CSL functions.

Please specify the default settings to be forced on QMSS registers (like the way Hard reset button).

Please help me. I am struck. Not able to proceed as I am still developing and debugging the LTE Phy. Hard reset is not a proper approach as LTE Phy do go for retrain.

Thanks & Regards,

Srinivas

Srinivas,

If navigator is not locked up, you can disable each CPPI/PKTDMA TX/RX DMA channel, reset each PDSP, write 0 into each queue management D register to throw away all descriptors in the queue, write 0 onto the base address of each descriptor region.  This puts it back into the initial state.  This should not be done unless all the cores are restarted else this will cause navigator code on those cores to fail.
 
You can also look at the PSC to see if there is reset control for the navigator, without power down control (its always on).