how to implement video decode with buffer queue?

hello experts, i want to implement video decode with buffer queue just like the encoder in sample/frontend.
but always got error when decoder initiating, failed to get return value from dqEvent().
can you please have a look? thx a lot.

Failed to query video capabilities: Inappropriate ioctl for device
[DEBUG] (NvV4l2Element.cpp:70) :Opened, fd = 14
[DEBUG] (NvV4l2Element.cpp:228) :Successfully subscribed to event 5
[DEBUG] (NvV4l2Element.cpp:190) :Set controls
[DEBUG] (NvVideoDecoder.cpp:174) :Disabling decoder complete frame input buffer: success
NvMMLiteOpen : Block : BlockType = 261
TVMR: NvMMLiteTVMRDecBlockOpen: 7650: NvMMLiteBlockOpen
NvMMLiteBlockCreate : Block : BlockType = 261
[DEBUG] (NvV4l2ElementPlane.cpp:290) :Output Plane:VIDIOC_S_FMT at capture plane successful
NvMMLiteTVMRDecSetAttribute:: Error status reporting set to 1

here is my code section:

in main():
  pthread_create(&rtspClientHandle, &attr, rtspClientThread, NULL); // to feed decoder buffer queue 
...
  pthread_create(&decodeHandle, &attr, decoderThread, NULL);

in rtspClientThread()
in a loop:
    BufferInfo buf = rtspClinetBufferQueue.pop(); // get empty buffer from input queue
    void *wPtr = NULL;
    NvBufferMemMap(buf.fd, 0, NvBufferMem_Write, &wPtr);
    buf.size = rtspClientRead();
    success = rtspClinetOutBufferQueue.push(buf); // put filled buffer to output queue 

in decoderThread()
    VideoDecoder videoDecoder("dec");
    videoDecoder.initialize(); // 1.subscribeEvent(V4L2_EVENT_RESOLUTION_CHANGE, 0, 0);
                               // 2.disableCompleteFrameInputBuffer();
                               // 3.setOutputPlaneFormat(V4L2_PIX_FMT_H264, chunkSize);
                               // 4.output_plane.setupPlane(<b>V4L2_MEMORY_DMABUF</b>, 10, false, false);
                               // 5.output_plane.setStreamStatus(true);
                               // 6.capture_plane.setDQThreadCallback();
                               // 7.capture_plane.startDQThread
    while (output_plane.getNumQueuedBuffers() < output_plane.getNumBuffers())
    {
        BufferInfo buf = rtspClinetOutBufferQueue.pop(); // get filled buffer from another thread
        v4l2_buf.index = videoDecoder.m_VideoDecoder->output_plane.getNumQueuedBuffers();
        v4l2_buf.m.planes = planes;
        v4l2_buf.m.planes[0].m.fd = buf.fd;
        v4l2_buf.m.planes[0].bytesused = buf.size;
        videoDecoder.m_VideoDecoder->output_plane.qBuffer(v4l2_buf, NULL); // queue the buffer to output plane 
        videoDecoder.insertFdToDmaBufFdSet(buf.fd);
        resetDecoderResolution();  // one frame each time, no loop in this function
                                   // ret = dqEvent(ev, false);
                                   // if (ret == 0 && V4L2_EVENT_RESOLUTION_CHANGE == ev.type)
                                   //   capture_plane.getFormat(format);
                                   //   capture_plane.getCrop(crop);
                                   //   capture_plane.deinitPlane();
                                   //   setCapturePlaneFormat(<b>V4L2_PIX_FMT_NV12M</b>/*format.fmt.pix_mp.pixelformat*/, format.fmt.pix_mp.width, format.fmt.pix_mp.height);
                                   //   getMinimumCapturePlaneBuffers(min_dec_capture_buffers);
                                   //   capture_plane.setupPlane(V4L2_MEMORY_MMAP, min_dec_capture_buffers + 5, false, false);
                                   //   capture_plane.setStreamStatus(true);
                                   //   Enqueue all the empty capture plane buffers
                                   //   return true;
                                   // else
                                   //  return false;
      rtspClinetBufferQueue.push(buf); // return buf to rtspClientThread() thread
    }
  //process
  while (1)
  {
    BufferInfo buf = rtspClinetOutBufferQueue.pop();
    ret = videoDecoder.decodeFromFd(buf); // decode one frame 
    videoDecOutBufferQueue.push(buf); // put yuv data to post-process
  }

update:
seems no error during decoder creation and configuration, but fatal crashed (Segmentation fault) in dqEvent().

Hi,
https://devtalk.nvidia.com/default/topic/1030108/jetson-tx2/call-ioctl-video_fd-vidioc_qbuf-amp-queuebuffer-the-program-is-blocked/post/5241305/#5241305

We have the implementation open source. Please refer to it.

hi DaneLLL, thank you so much.

i can run dec now, but using MMAP instead of DMABuffer as outplane input.
my question is, the decoder can be running well with MMAP, but faild with DMA.
how to config the NvBuffer or V4L2 buffer beforing Qing it into decoder outplane??

currently:

//ret = m_VideoDecoder->output_plane.setupPlane(V4L2_MEMORY_DMABUF, MAX_QUEUED_BUFFERS, false, false);
  ret = m_VideoDecoder->output_plane.setupPlane(V4L2_MEMORY_MMAP, MAX_QUEUED_BUFFERS, true, false);
//....
    struct v4l2_buffer v4l2_buf;
    struct v4l2_plane planes[MAX_PLANES];
    NvBuffer *buffer;
    void *rPtr = NULL;

    BufferInfo buf = BufferQueue.pop();
    memset(&v4l2_buf, 0, sizeof(v4l2_buf));
    memset(planes, 0, sizeof(planes));

    v4l2_buf.index = m_VideoDecoder->output_plane.getNumQueuedBuffers();
    buffer = m_VideoDecoder->output_plane.getNthBuffer(v4l2_buf.index);

    v4l2_buf.m.planes = planes;

    //will be failed w/o the following section, meanwhile set DMA in outplane
    NvBufferMemMap(buf.fd, 0, NvBufferMem_Read, &rPtr);
    NvBufferMemSyncForCpu(buf.fd, 0, &rPtr);
    memcpy((char*) buffer->planes[0].data, (const char*) rPtr, buf.size);

    //the following section should be useless since MMAP working, but is this enough if DMA??
    buffer->planes[0].fd = buf.fd;
    v4l2_buf.m.planes[0].m.fd = buf.fd;

    NvBufferMemUnMap(buf.fd, 0, &rPtr);
    
    //for MMAP only
    v4l2_buf.m.planes[0].bytesused = buf.size;
    buffer->planes[0].bytesused = buf.size;

    m_VideoDecoder->output_plane.qBuffer(v4l2_buf, NULL);

Hi liuliang,
For configuring V4L2_MEMORY_DMABUF, you have to allocate buffers via NvBufferCreate() and queue to the plane.