gpu: nvgpu: Update Linux side VM code for API solidification

Update the Linux specific code to match the MM API docs in the
previous patch. The user passed page size is plumbed through
the Linux VM mapping calls but is ultimately ignored once the
core VM code is called. This will be handled in the next
patch.

This also adds some code to make the CDE page size picking
happen semi-intelligently. In many cases the CDE buffers can
be mapped with large pages.

Bug 2011640

Change-Id: I20e78e7d5a841e410864b474179e71da1c2482f4
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1740610
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

1 files changed, 28 insertions, 1 deletions

diff --git a/drivers/gpu/nvgpu/os/linux/cde.c b/drivers/gpu/nvgpu/os/linux/cde.c
index 66a80403..052a1d21 100644
--- a/drivers/gpu/nvgpu/os/linux/cde.c
+++ b/drivers/gpu/nvgpu/os/linux/cde.c
@@ -975,6 +975,30 @@ static struct gk20a_cde_ctx *gk20a_cde_allocate_context(struct nvgpu_os_linux *l
         return cde_ctx;
 }
 
+static u32 gk20a_cde_mapping_page_size(struct vm_gk20a *vm,
+                                       u32 map_offset, u32 map_size)
+{
+        struct gk20a *g = gk20a_from_vm(vm);
+
+        /*
+         * To be simple we will just make the map size depend on the
+         * iommu'ability of the driver. If there's an IOMMU we can rely on
+         * buffers being contiguous. If not, then we'll use 4k pages since we
+         * know that will work for any buffer.
+         */
+        if (!nvgpu_iommuable(g))
+                return SZ_4K;
+
+        /*
+         * If map size or offset is not 64K aligned then use small pages.
+         */
+        if (map_size & (vm->big_page_size - 1) ||
+            map_offset & (vm->big_page_size - 1))
+                return SZ_4K;
+
+        return vm->big_page_size;
+}
+
 int gk20a_cde_convert(struct nvgpu_os_linux *l,
                       struct dma_buf *compbits_scatter_buf,
                       u64 compbits_byte_offset,
@@ -1071,7 +1095,10 @@ __releases(&l->cde_app->mutex)
         err = nvgpu_vm_map_linux(cde_ctx->vm, compbits_scatter_buf, 0,
                                  NVGPU_VM_MAP_CACHEABLE |
                                  NVGPU_VM_MAP_DIRECT_KIND_CTRL,
-                                 NVGPU_KIND_INVALID,
+                                 gk20a_cde_mapping_page_size(cde_ctx->vm,
+                                                             map_offset,
+                                                             map_size),
+                                 NV_KIND_INVALID,
                                  compbits_kind, /* incompressible kind */
                                  gk20a_mem_flag_none,
                                  map_offset, map_size,