@kayccc it would be really helpful to get a clear answer on this question. We also have a PCIe endpoint device from which we’re trying to fetch data from using Orin’s GPDMA facility an dmaengine_prep_dma_memcpy() targeting the device’s BAR. On other host SoC’s, this appears to work, but on the Orin, either we’re having some inconsistency in setting up the IOMMU mappings or this is not supported by GPDMA.
It would be very beneficial to both Nvidia and the community at large to provide a succinct example of DMAing from a PCIe endpoint device’s BAR, whether with GPDMA or the separate DMA engine residing inside Orin’s PCIe root port.