V4l2dec and v4l2enc in dmabuf-importmode

Hi All,


I am working on a v4l2 decoder and encoder which works on dmabuf-import
mode on both capture and output interface.



But it is evident from the log that v4l2encoder throws error  if downstream
plugin does not provide a pool



example pipeline : gst-launch-1.0 videotestsrc ! v4l2h264enc
capture-io-mode=5 output-io-mode=5 ! fakesink



*gst_v4l2_object_decide_allocation:<v4l2h264enc0> error: When importing
DMABUF or USERPTR, we need a pool to import from*







Similarly in case of decoder also if upstream plugin pushes non-fd buffer
then pipeline throws error dmabuf import failed.



example pipeline : gst-launch-1.0 filesrc location=/1.mp4 ! qtdemux !
h264parse ! v4l2h264dec capture-io-mode=5 output-io-mode=5 ! fakesink



*gst_v4l2_allocator_import_dmabuf:<v4l2h264dec0:pool0:sink:allocator>
Memory 0 is not of DMABUF*

*v4l2bufferpool
gstv4l2bufferpool.c:396:gst_v4l2_buffer_pool_import_dmabuf:<v4l2h264dec0:pool0:sink>
failed to import dmabuf*



What is the recommendation to handle such cases?



1. Implement a two separate gstreamer plugins which generates dma buffers,
for example predmabufgen and postdmabufgen in the following examples



Example Pipeline:



gst-launch-1.0 filesrc location=/1.mp4 ! qtdemux ! h264parse !
*predmabufgen*  v4l2h264dec capture-io-mode=5 output-io-mode=5 !
*postdmabufgen* ! fakesink

gst-launch-1.0 videotestsrc ! *predmabufgen* ! v4l2h264enc
capture-io-mode=5 output-io-mode=5 ! *postdmabufgen* ! fakesink



2. Allocate dmabuf in the v4l2dec/v4l2enc plugin in case of dmabuf-import
if downstream is not providing a pool or upstream plugin is snot sending fd
buffer?


Thanks

Arindam