On Tue, Apr 30, 2019 at 09:11:07PM +0000, Paul Burton wrote: > Right but dma_direct_alloc_pages() already checks for the PageHighMem > case & returns before ever calling arch_dma_prep_coherent(), no? True. And of course it can't be remapped into the uncached segment anyway. So yes, we should drop it. Eventually I want to add generic support for DMA_ATTR_NO_KERNEL_MAPPING, but that'll involved auditing all instances anyway.
