On Wed, Jul 27, 2016 at 11:28:35AM +0200, Stefan Roese wrote: > Hi, > > I'm currently thinking about adding SG support to the "old" Marvell > XOR engine DMA driver (mv_xor.c). My use-case is to transfer an area > between system memory (represented via a scatterlist) and memory on > a PCI device. Currently the XOR driver completely lags SG support. > Now I've started comparing the APIs of both DMA_SG and DMA_SLAVE to > see, which one is better matching my use-case. And because my > device address space is contiguous, the DMA_SLAVE API with only > one scatterlist seems to be more appropriate here. Even though its > not strictly speaking a "device" that I'm transferring data with. > > So my main question is, is it appropriate to add the DMA_SLAVE > infrastructure to the mv_xor driver, even though the XOR engine > is not really designed to be used on "devices" (with fixed > addresses, FIFOs etc) but more for mem2mem xfers? Or should this > SG support better be added by using the DMA_SG design? from the look at it, it should be DMA_SG. The memory is on another device, but DMA can transfer. Please do ensure you setup IOMMU properly -- ~Vinod -- To unsubscribe from this list: send the line "unsubscribe dmaengine" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
