drm/amd/display: Add DC_FP helper to check FPU state

To fully isolate FPU operations in a single place, we must avoid
situations where compilers spill FP values to registers due to FP enable
in a specific C file. Note that even if we isolate all FPU functions in
a single file and call its interface from other files, the compiler
might enable the use of FPU before we call DC_FP_START. Nevertheless, it
is the programmer's responsibility to invoke DC_FP_START/END in the
correct place. To highlight situations where developers forgot to use
the FP protection before calling the DC FPU interface functions, we
introduce a helper that checks if the function is invoked under FP
protection. If not, it will trigger a kernel warning.

Changes cince V3:
- Rebase

Changes cince V2 (Christian):
- Do not use this_cpu_* between get/put_cpu_ptr().
- In the kernel documentation, better describe restrictions.
- Make dc_assert_fp_enabled trigger the ASSERT message.

Changes since V1:
- Remove fp_enable variables
- Rename dc_is_fp_enabled to dc_assert_fp_enabled
- Replace wrong variable type

Cc: Harry Wentland <[e-mail]>
Cc: Anson Jacob <[e-mail]>
Cc: Christian König <[e-mail]>
Cc: Hersen Wu <[e-mail]>
Cc: Aric Cyr <[e-mail]>
Cc: Jun Lei <[e-mail]>
Cc: Dmytro Laktyushkin <[e-mail]>
Cc: Qingqing Zhuo <[e-mail]>
Reported-by: kernel test robot <[e-mail]>
Signed-off-by: Rodrigo Siqueira <[e-mail]>
Reviewed-by: Christian König <[e-mail]>
Signed-off-by: Alex Deucher <[e-mail]>

drivers/gpu/drm/amd/display/amdgpu_dm/dc_fpu.c

···
 static DEFINE_PER_CPU(int, fpu_recursion_depth);
 
 /**
+ * dc_assert_fp_enabled - Check if FPU protection is enabled
+ *
+ * This function tells if the code is already under FPU protection or not. A
+ * function that works as an API for a set of FPU operations can use this
+ * function for checking if the caller invoked it after DC_FP_START(). For
+ * example, take a look at dcn2x.c file.
+ */
+inline void dc_assert_fp_enabled(void)
+{
+	int *pcpu, depth = 0;
+
+	pcpu = get_cpu_ptr(&fpu_recursion_depth);
+	depth = *pcpu;
+	put_cpu_ptr(&fpu_recursion_depth);
+
+	ASSERT(depth > 1);
+}
+
+/**
  * dc_fpu_begin - Enables FPU protection
  * @function_name: A string containing the function name for debug purposes
  *   (usually __func__)

drivers/gpu/drm/amd/display/amdgpu_dm/dc_fpu.h

···
 #ifndef __DC_FPU_H__
 #define __DC_FPU_H__
 
+void dc_assert_fp_enabled(void);
 void dc_fpu_begin(const char *function_name, const int line);
 void dc_fpu_end(const char *function_name, const int line);
 

drivers/gpu/drm/amd/display/dc/dcn20/dcn20_resource.c

···
 	}
 
 	/* populate writeback information */
+	DC_FP_START();
 	dc->res_pool->funcs->populate_dml_writeback_from_context(dc, res_ctx, pipes);
+	DC_FP_END();
 
 	return pipe_cnt;
 }

drivers/gpu/drm/amd/display/dc/dml/dcn2x/dcn2x.c

···
  *    that deals with FP register is contained within this call.
  * 3. All function that needs to be accessed outside this file requires a
  *    public interface that not uses any FPU reference.
+ * 4. Developers **must not** use DC_FP_START/END in this file, but they need
+ *    to ensure that the caller invokes it before access any function available
+ *    in this file. For this reason, public functions in this file must invoke
+ *    dc_assert_fp_enabled();
+ *
+ * Let's expand a little bit more the idea in the code pattern. To fully
+ * isolate FPU operations in a single place, we must avoid situations where
+ * compilers spill FP values to registers due to FP enable in a specific C
+ * file. Note that even if we isolate all FPU functions in a single file and
+ * call its interface from other files, the compiler might enable the use of
+ * FPU before we call DC_FP_START. Nevertheless, it is the programmer's
+ * responsibility to invoke DC_FP_START/END in the correct place. To highlight
+ * situations where developers forgot to use the FP protection before calling
+ * the DC FPU interface functions, we introduce a helper that checks if the
+ * function is invoked under FP protection. If not, it will trigger a kernel
+ * warning.
  */
 
 void dcn20_populate_dml_writeback_from_context(struct dc *dc,
···
 					       display_e2e_pipe_params_st *pipes)
 {
 	int pipe_cnt, i;
+
+	dc_assert_fp_enabled();
 
 	for (i = 0, pipe_cnt = 0; i < dc->res_pool->pipe_count; i++) {
 		struct dc_writeback_info *wb_info = &res_ctx->pipe_ctx[i].stream->writeback_info[0];