On Tuesday 16 August 2011, Marek Szyprowski wrote:
On Friday, August 12, 2011 5:01 PM Arnd Bergmann wrote:
How about something like
if (arch_is_coherent() || nommu()) ret = alloc_simple_buffer(); else if (arch_is_v4_v5()) ret = alloc_remap(); else if (gfp & GFP_ATOMIC) ret = alloc_from_pool(); else ret = alloc_from_contiguous();
This also allows a natural conversion to dma_map_ops when we get there.
Ok. Is it ok to enable CMA permanently for ARMv6+? If CMA is left conditional the dma pool code will be much more complicated, because it will need to support both CMA and non-CMA cases.
I think that is ok, yes.
Arnd