1 files changed, 15 insertions, 1 deletions

diff --git a/drivers/gpu/nvgpu/common/mm/pd_cache.c b/drivers/gpu/nvgpu/common/mm/pd_cache.c
index f0abc03d..4c3e06ba 100644
--- a/drivers/gpu/nvgpu/common/mm/pd_cache.c
+++ b/drivers/gpu/nvgpu/common/mm/pd_cache.c
@@ -143,6 +143,7 @@ int __nvgpu_pd_cache_alloc_direct(struct gk20a *g,
                                   struct nvgpu_gmmu_pd *pd, u32 bytes)
 {
         int err;
+        unsigned long flags = 0;
 
         pd_dbg(g, "PD-Alloc [D] %u bytes", bytes);
 
@@ -152,7 +153,20 @@ int __nvgpu_pd_cache_alloc_direct(struct gk20a *g,
                 return -ENOMEM;
         }
 
-        err = nvgpu_dma_alloc(g, bytes, pd->mem);
+        /*
+         * If bytes == PAGE_SIZE then it's impossible to get a discontiguous DMA
+         * allocation. Some DMA implementations may, despite this fact, still
+         * use the contiguous pool for page sized allocations. As such only
+         * request explicitly contiguous allocs if the page directory is larger
+         * than the page size. Also, of course, this is all only revelant for
+         * GPUs not using an IOMMU. If there is an IOMMU DMA allocs are always
+         * going to be virtually contiguous and we don't have to force the
+         * underlying allocations to be physically contiguous as well.
+         */
+        if (!nvgpu_iommuable(g) && bytes > PAGE_SIZE)
+                flags = NVGPU_DMA_FORCE_CONTIGUOUS;
+
+        err = nvgpu_dma_alloc_flags(g, flags, bytes, pd->mem);
         if (err) {
                 nvgpu_err(g, "OOM allocating page directory!");
                 nvgpu_kfree(g, pd->mem);