On 23.04.2021 15:51, Lars-Peter Clausen wrote:

quoted

On 23-04-21, 11:17, Lars-Peter Clausen wrote:

quoted

It seems to me what we are missing from the DMAengine API is the equivalent
of device_prep_dma_memcpy() that is able to take SG lists. There is already
a memset_sg, it should be possible to add something similar for memcpy.

You mean something like dmaengine_prep_dma_sg() which was removed?

Ah, that's why I could have sworn we already had this!

Yes, after that API function was removed, the Xilinx DMA driver effectively
ceased to support memory-to-memory SG transfers. My assumption was if it wasn't
ever reimplemented through the callback you mentioned before, is because there
isn't truly much interest in this device.

However we do need this functionality in our system. At the moment, and for our
particular use of it, we can always guarantee that either the source or
destination will be one contiguous chunk of memory, so just one scatterlist
pointer array was enough to fully program an operation. I think this would fit
alright into memset_sg.  What do you think?

Also, at the moment we're using it for transfers between main memory and
CPU-mapped PCI memory, which cannot be allocated with kmalloc. Wouldn't this
fall into the use case for device_prep_slave_sg?

Adrian

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel