drm/amd/display: powergate fe of reused pipes to reset ttu

When we exit MPO, disconnected pipes cannot be immediately powergated
because registers are double buffered, and actual disconnection does
not happen until VUPDATE. So it is differred for many flips.
However in the case of exiting full screen, the transition from MPO
to grph only back to MPO is very fast and also involves increasing of
watermarks. Since the underlay pipe is never powergated in this
scenario, it keeps its old TTU counter, which causes allowPstateSwitch
signal to be de-asserted when compared to the new increased watermark.
Since the new pipe is not enabled yet, the signal will be continously
de-asserted and hangs SMU, who's waiting for the signal to do pstate
switching.

Signed-off-by: Eric Yang <Eric.Yang2@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Harry Wentland <Harry.Wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c

@@ -1577,6 +1577,14 @@ static void dcn10_apply_ctx_for_surface(
 		struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i];
 		struct pipe_ctx *old_pipe_ctx =
 				&dc->current_context->res_ctx.pipe_ctx[i];
+		/*
+		 * Powergate reused pipes that are not powergated
+		 * fairly hacky right now, using opp_id as indicator
+		 */
+		if (pipe_ctx->surface && !old_pipe_ctx->surface) {
+			if (pipe_ctx->mpcc->opp_id != 0xf)
+				dcn10_power_down_fe(dc, pipe_ctx->pipe_idx);
+		}
 
 		if ((!pipe_ctx->surface && old_pipe_ctx->surface)
 				|| (!pipe_ctx->stream && old_pipe_ctx->stream)) {
@@ -1588,6 +1596,7 @@ static void dcn10_apply_ctx_for_surface(
 				continue;
 			}
 
+			/* reset mpc */
 			mpcc_cfg.opp_id = 0xf;
 			mpcc_cfg.top_dpp_id = 0xf;
 			mpcc_cfg.bot_mpcc_id = 0xf;