On Wed, Aug 07, 2019 at 01:38:08PM +0100, Mark Rutland wrote: > > I *believe* that there are not alias mappings (that I don't control > > myself) for pages coming from > > shmem_file_setup()/shmem_read_mapping_page().. > > AFAICT, that's regular anonymous memory, so there will be a cacheable > alias in the linear/direct map.
Yes. Although shmem is in no way special in that regard. Even with the normal dma_alloc_coherent implementation on arm and arm64 we keep the cacheable alias in the direct mapping and just create a new non-cacheable one. The only exception are CMA allocations on 32-bit arm, which do get remapped to uncachable in place.
