gpu: nvgpu: remove usage of nvgpu_fence

Remove the usage of nvgpu_fence splattered across nvgpu,
and replace with a struct defined in common code.
The usage is still inside Linux, but this helps the
subsequent unification efforts, e.g. to unify the submit
path.

VQRM-3465

Change-Id: Ic3737450123dfc5e1c40ca5b6b8d8f6b3070aa0d
Signed-off-by: Sourab Gupta <sourabg@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1691977
Reviewed-by: Konsta Holtta <kholtta@nvidia.com>
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>

drivers/gpu/nvgpu/common/linux/cde.c

@@ -748,7 +748,7 @@ deinit_image:
 }
 
 static int gk20a_cde_execute_buffer(struct gk20a_cde_ctx *cde_ctx,
-				    u32 op, struct nvgpu_fence *fence,
+				    u32 op, struct nvgpu_channel_fence *fence,
 				    u32 flags, struct gk20a_fence **fence_out)
 {
 	struct nvgpu_os_linux *l = cde_ctx->l;
@@ -975,7 +975,7 @@ int gk20a_cde_convert(struct nvgpu_os_linux *l,
 		      struct dma_buf *compbits_scatter_buf,
 		      u64 compbits_byte_offset,
 		      u64 scatterbuffer_byte_offset,
-		      struct nvgpu_fence *fence,
+		      struct nvgpu_channel_fence *fence,
 		      u32 __flags, struct gk20a_cde_param *params,
 		      int num_params, struct gk20a_fence **fence_out)
 __acquires(&l->cde_app->mutex)
@@ -1482,7 +1482,7 @@ static int gk20a_buffer_convert_gpu_to_cde_v1(
 		u64 offset, u64 compbits_hoffset, u64 compbits_voffset,
 		u64 scatterbuffer_offset,
 		u32 width, u32 height, u32 block_height_log2,
-		u32 submit_flags, struct nvgpu_fence *fence_in,
+		u32 submit_flags, struct nvgpu_channel_fence *fence_in,
 		struct gk20a_buffer_state *state)
 {
 	struct gk20a *g = &l->g;
@@ -1624,7 +1624,7 @@ static int gk20a_buffer_convert_gpu_to_cde(
 		u64 offset, u64 compbits_hoffset, u64 compbits_voffset,
 		u64 scatterbuffer_offset,
 		u32 width, u32 height, u32 block_height_log2,
-		u32 submit_flags, struct nvgpu_fence *fence_in,
+		u32 submit_flags, struct nvgpu_channel_fence *fence_in,
 		struct gk20a_buffer_state *state)
 {
 	struct gk20a *g = &l->g;
@@ -1656,7 +1656,7 @@ int gk20a_prepare_compressible_read(
 		u64 compbits_hoffset, u64 compbits_voffset,
 		u64 scatterbuffer_offset,
 		u32 width, u32 height, u32 block_height_log2,
-		u32 submit_flags, struct nvgpu_fence *fence,
+		u32 submit_flags, struct nvgpu_channel_fence *fence,
 		u32 *valid_compbits, u32 *zbc_color,
 		struct gk20a_fence **fence_out)
 {

drivers/gpu/nvgpu/common/linux/cde.h

@@ -44,7 +44,7 @@ struct device;
 struct nvgpu_os_linux;
 struct gk20a;
 struct gk20a_fence;
-struct nvgpu_fence;
+struct nvgpu_channel_fence;
 struct channel_gk20a;
 struct vm_gk20a;
 struct nvgpu_gpfifo;
@@ -306,7 +306,7 @@ int gk20a_cde_convert(struct nvgpu_os_linux *l,
 		struct dma_buf *compbits_buf,
 		u64 compbits_byte_offset,
 		u64 scatterbuffer_byte_offset,
-		struct nvgpu_fence *fence,
+		struct nvgpu_channel_fence *fence,
 		u32 __flags, struct gk20a_cde_param *params,
 		int num_params, struct gk20a_fence **fence_out);
 
@@ -315,7 +315,7 @@ int gk20a_prepare_compressible_read(
 		u64 compbits_hoffset, u64 compbits_voffset,
 		u64 scatterbuffer_offset,
 		u32 width, u32 height, u32 block_height_log2,
-		u32 submit_flags, struct nvgpu_fence *fence,
+		u32 submit_flags, struct nvgpu_channel_fence *fence,
 		u32 *valid_compbits, u32 *zbc_color,
 		struct gk20a_fence **fence_out);
 int gk20a_mark_compressible_write(

drivers/gpu/nvgpu/common/linux/ce2.c

@@ -56,7 +56,7 @@ int gk20a_ce_execute_ops(struct gk20a *g,
 	u32 cmd_buf_read_offset;
 	u32 dma_copy_class;
 	struct nvgpu_gpfifo gpfifo;
-	struct nvgpu_fence fence = {0,0};
+	struct nvgpu_channel_fence fence = {0, 0};
 	struct gk20a_fence *ce_cmd_buf_fence_out = NULL;
 
 	if (!ce_app->initialised ||ce_app->app_state != NVGPU_CE_ACTIVE)

drivers/gpu/nvgpu/common/linux/channel.c

@@ -416,7 +416,7 @@ static void trace_write_pushbuffer_range(struct channel_gk20a *c,
  * Handle the submit synchronization - pre-fences and post-fences.
  */
 static int gk20a_submit_prepare_syncs(struct channel_gk20a *c,
-				      struct nvgpu_fence *fence,
+				      struct nvgpu_channel_fence *fence,
 				      struct channel_gk20a_job *job,
 				      struct priv_cmd_entry **wait_cmd,
 				      struct priv_cmd_entry **incr_cmd,
@@ -663,7 +663,7 @@ int gk20a_submit_channel_gpfifo(struct channel_gk20a *c,
 				struct nvgpu_submit_gpfifo_args *args,
 				u32 num_entries,
 				u32 flags,
-				struct nvgpu_fence *fence,
+				struct nvgpu_channel_fence *fence,
 				struct gk20a_fence **fence_out,
 				bool force_need_sync_fence,
 				struct fifo_profile_gk20a *profile)

drivers/gpu/nvgpu/common/linux/channel.h

@@ -24,7 +24,7 @@
 struct channel_gk20a;
 struct nvgpu_gpfifo;
 struct nvgpu_submit_gpfifo_args;
-struct nvgpu_fence;
+struct nvgpu_channel_fence;
 struct gk20a_fence;
 struct fifo_profile_gk20a;
 struct nvgpu_os_linux;
@@ -75,7 +75,7 @@ int gk20a_submit_channel_gpfifo(struct channel_gk20a *c,
 				struct nvgpu_submit_gpfifo_args *args,
 				u32 num_entries,
 				u32 flags,
-				struct nvgpu_fence *fence,
+				struct nvgpu_channel_fence *fence,
 				struct gk20a_fence **fence_out,
 				bool force_need_sync_fence,
 				struct fifo_profile_gk20a *profile);

drivers/gpu/nvgpu/common/linux/ioctl_channel.c

@@ -614,6 +614,14 @@ static void nvgpu_get_gpfifo_args(
 							alloc_gpfifo_args->flags);
 }
 
+static void nvgpu_get_fence_args(
+		struct nvgpu_fence *fence_args_in,
+		struct nvgpu_channel_fence *fence_args_out)
+{
+	fence_args_out->id = fence_args_in->id;
+	fence_args_out->value = fence_args_in->value;
+}
+
 static int gk20a_channel_wait_semaphore(struct channel_gk20a *ch,
 					ulong id, u32 offset,
 					u32 payload, u32 timeout)
@@ -760,6 +768,7 @@ static int gk20a_ioctl_channel_submit_gpfifo(
 	struct channel_gk20a *ch,
 	struct nvgpu_submit_gpfifo_args *args)
 {
+	struct nvgpu_channel_fence fence;
 	struct gk20a_fence *fence_out;
 	struct fifo_profile_gk20a *profile = NULL;
 
@@ -779,8 +788,9 @@ static int gk20a_ioctl_channel_submit_gpfifo(
 		!capable(CAP_SYS_NICE))
 		return -EPERM;
 
+	nvgpu_get_fence_args(&args->fence, &fence);
 	ret = gk20a_submit_channel_gpfifo(ch, NULL, args, args->num_entries,
-					  args->flags, &args->fence,
+					  args->flags, &fence,
 					  &fence_out, false, profile);
 
 	if (ret)

drivers/gpu/nvgpu/common/linux/ioctl_ctrl.c

@@ -339,7 +339,7 @@ static int gk20a_ctrl_prepare_compressible_read(
 
 #ifdef CONFIG_NVGPU_SUPPORT_CDE
 	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
-	struct nvgpu_fence fence;
+	struct nvgpu_channel_fence fence;
 	struct gk20a_fence *fence_out = NULL;
 	int flags = args->submit_flags;
 

drivers/gpu/nvgpu/gk20a/channel_gk20a.h

@@ -47,6 +47,18 @@ struct fifo_profile_gk20a;
 #define NVGPU_GPFIFO_FLAGS_REPLAYABLE_FAULTS_ENABLE	(1 << 2)
 #define NVGPU_GPFIFO_FLAGS_USERMODE_SUPPORT		(1 << 3)
 
+/*
+ * The binary format of 'struct nvgpu_channel_fence' introduced here
+ * should match that of 'struct nvgpu_fence' defined in uapi header, since
+ * this struct is intended to be a mirror copy of the uapi struct. This is
+ * not a hard requirement though because of nvgpu_get_fence_args conversion
+ * function.
+ */
+struct nvgpu_channel_fence {
+	u32 id;
+	u32 value;
+};
+
 struct nvgpu_gpfifo_args {
 	u32 num_entries;
 	u32 num_inflight_jobs;