On Wed, Aug 01, 2018 at 04:02:34PM -0700, Wesley Terpstra wrote: > FYI, I don't believe this solution is the one we intended to merge > upstream. There is a different approach used in the riscv-linux tree > from Palmer. This is the solution I as the dma-mapping maintainer want for upstream, and I'm trying to convince the PCI maintainers to agree with me. What is in the RISC-V trees is a horribly bad idea that should have never seen the light of day (copying dma_map_ops and patching them up).
