gpu: nvgpu: vgpu: add new GMMU map interface

The server now exposes a new GMMU map interface that
can accept a scatter-gather list. This is needed to support
SMMU-bypass configurations.

Bug 1677153
JIRA VFND-689

Change-Id: I7b5af145db57dcebe2c9125ec90c689798d7e69e
Signed-off-by: Aingara Paramakuru <aparamakuru@nvidia.com>
Reviewed-on: http://git-master/r/792558
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Richard Zhao <rizhao@nvidia.com>
Reviewed-by: Vladislav Buzov <vbuzov@nvidia.com>

1 files changed, 22 insertions, 1 deletions

diff --git a/include/linux/tegra_vgpu.h b/include/linux/tegra_vgpu.h
index 2b327d09..6fc298e0 100644
--- a/include/linux/tegra_vgpu.h
+++ b/include/linux/tegra_vgpu.h
@@ -64,7 +64,8 @@ enum {
         TEGRA_VGPU_CMD_SUBMIT_RUNLIST,
         TEGRA_VGPU_CMD_GET_ZCULL_INFO,
         TEGRA_VGPU_CMD_ZBC_SET_TABLE,
-        TEGRA_VGPU_CMD_ZBC_QUERY_TABLE
+        TEGRA_VGPU_CMD_ZBC_QUERY_TABLE,
+        TEGRA_VGPU_CMD_AS_MAP_EX
 };
 
 struct tegra_vgpu_connect_params {
@@ -136,6 +137,25 @@ struct tegra_vgpu_as_map_params {
         u32 ctag_offset;
 };
 
+struct tegra_vgpu_as_map_ex_params {
+        u64 handle;
+        u64 gpu_va;
+        u64 size;
+        u32 mem_desc_count;
+        u8 pgsz_idx;
+        u8 iova;
+        u8 kind;
+        u8 cacheable;
+        u8 clear_ctags;
+        u8 prot;
+        u32 ctag_offset;
+};
+
+struct tegra_vgpu_mem_desc {
+        u64 addr;
+        u64 length;
+};
+
 struct tegra_vgpu_as_invalidate_params {
         u64 handle;
 };
@@ -236,6 +256,7 @@ struct tegra_vgpu_cmd_msg {
                 struct tegra_vgpu_as_share_params as_share;
                 struct tegra_vgpu_as_bind_share_params as_bind_share;
                 struct tegra_vgpu_as_map_params as_map;
+                struct tegra_vgpu_as_map_ex_params as_map_ex;
                 struct tegra_vgpu_as_invalidate_params as_invalidate;
                 struct tegra_vgpu_channel_config_params channel_config;
                 struct tegra_vgpu_ramfc_params ramfc;