• State Resolved
  • Locked Locked
  • Replies 2 replies
  • Subscribers 98 subscribers
  • Views 992 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Mapping C6678 PCIe MSI interrupt to EDMA event

Gordon Deane
Expert 1255 points

Hi,

I have a C6678 connected to an FPGA over PCIe, and am reading blocks of data from the FPGA, working in a ring buffer.  The FPGA will send an MSI interrupt whenever another block is ready.

We can catch the MSI interrupt on the cores.  I'd like to send the MSI directly to an EDMA channel controller (any controller).  I can also find lots of tables of interrupts, events, etc.  What I can't see is any EDMA input event that I can map in this way.  The MSI events don't seem to be a possible input for the EDMA directly or for the CIC2 controller indirectly.

Am I missing something, or is this not possible?

Thanks,

Gordon

  • 0
    TI__Genius 12065 points

    Gordon,

    As you can see the interrupt tables in the C6678 data manual (section 7.9), there is no direct connection between PCIe MSI interrupt to EDMA (CIC 2 or 3). 

    So you may have to use MSI to generate interrupt to CPU and let CPU trigger the EDMA transfer.

    Another way is, you can prepare the EDMA configuration (PaRAM) first and then whenever the block is ready, you can use send PCIe write packets from FPGA to write to the EDMA registers (such as ESR) to trigger the EDMA transfer.

    In this way, you do not need to use interrupt but you have to map some of the PCIe inbound translation region to EDMA registers region and be careful of the operation (it is configuration space instead of memory space).

    And you probably need to set MST_PRIV = 1 in PCIe PRIORITY register in order to grant PCIe the access to EDMA configuration space.

  • 0 in reply to Steven Ji
    Expert 1255 points

    Thanks, I think I'll go with the CPU interrupt route.  It is a little more intrusive on the CPU but has the nice advantage that it should stop automatically when the CPU is halted in the debugger.

    We could probably toggle a GPIO line from the FPGA to DSP if we really wanted to, asI believe the EDMA can use that as an event.