video: tegra: host: use dma APIs for gr ctx

video: tegra: host: use dma APIs for gr ctx

- use dma_alloc/free_attrs() APIs to allocate/free memory for gr_ctx
- pass DMA_ATTR_NO_KERNEL_MAPPING to above APIs to avoid kernel
  mapping during allocation
- use vmap()/unmap() to get kernel mapping whenever needed
- use gk20a_gmmu_map()/unmap() to update gmmu ptes
- define new API gk20a_get_sgtable_from_pages() which creates sg_table
  of the buffer from 'struct page **'

Bug 1397647

Change-Id: I9f8a9c9f8e7b309cf4d6e5d906747d200ba3dd29
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: http://git-master/r/352250
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Tested-by: Terje Bergstrom <tbergstrom@nvidia.com>