drm/radeon: Bypass hw lut's for > 8 bpc framebuffer scanout.

The hardware lut's only have 256 slots for indexing by a
8 bpc framebuffer. In 10 bpc scanout modes, framebuffer
color values would get truncated to their 8 msb's,
thereby losing the extra precision afforded by a 10 bpc
framebuffer.

To retain full precision, bypass the hw lut in 10 bpc
scanout mode.

Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/radeon/atombios_crtc.c

@@ -1136,6 +1136,7 @@ static int dce4_crtc_do_set_base(struct drm_crtc *crtc,
 	u32 fb_swap = EVERGREEN_GRPH_ENDIAN_SWAP(EVERGREEN_GRPH_ENDIAN_NONE);
 	u32 tmp, viewport_w, viewport_h;
 	int r;
+	bool bypass_lut = false;
 
 	/* no fb bound */
 	if (!atomic && !crtc->primary->fb) {
@@ -1225,6 +1226,8 @@ static int dce4_crtc_do_set_base(struct drm_crtc *crtc,
 #ifdef __BIG_ENDIAN
 		fb_swap = EVERGREEN_GRPH_ENDIAN_SWAP(EVERGREEN_GRPH_ENDIAN_8IN32);
 #endif
+		/* Greater 8 bpc fb needs to bypass hw-lut to retain precision */
+		bypass_lut = true;
 		break;
 	case DRM_FORMAT_BGRX1010102:
 	case DRM_FORMAT_BGRA1010102:
@@ -1233,6 +1236,8 @@ static int dce4_crtc_do_set_base(struct drm_crtc *crtc,
 #ifdef __BIG_ENDIAN
 		fb_swap = EVERGREEN_GRPH_ENDIAN_SWAP(EVERGREEN_GRPH_ENDIAN_8IN32);
 #endif
+		/* Greater 8 bpc fb needs to bypass hw-lut to retain precision */
+		bypass_lut = true;
 		break;
 	default:
 		DRM_ERROR("Unsupported screen format %s\n",
@@ -1365,6 +1370,18 @@ static int dce4_crtc_do_set_base(struct drm_crtc *crtc,
 	WREG32(EVERGREEN_GRPH_CONTROL + radeon_crtc->crtc_offset, fb_format);
 	WREG32(EVERGREEN_GRPH_SWAP_CONTROL + radeon_crtc->crtc_offset, fb_swap);
 
+	/*
+	 * The LUT only has 256 slots for indexing by a 8 bpc fb. Bypass the LUT
+	 * for > 8 bpc scanout to avoid truncation of fb indices to 8 msb's, to
+	 * retain the full precision throughout the pipeline.
+	 */
+	WREG32_P(EVERGREEN_GRPH_LUT_10BIT_BYPASS_CONTROL + radeon_crtc->crtc_offset,
+		 (bypass_lut ? EVERGREEN_LUT_10BIT_BYPASS_EN : 0),
+		 ~EVERGREEN_LUT_10BIT_BYPASS_EN);
+
+	if (bypass_lut)
+		DRM_DEBUG_KMS("Bypassing hardware LUT due to 10 bit fb scanout.\n");
+
 	WREG32(EVERGREEN_GRPH_SURFACE_OFFSET_X + radeon_crtc->crtc_offset, 0);
 	WREG32(EVERGREEN_GRPH_SURFACE_OFFSET_Y + radeon_crtc->crtc_offset, 0);
 	WREG32(EVERGREEN_GRPH_X_START + radeon_crtc->crtc_offset, 0);
@@ -1432,6 +1449,7 @@ static int avivo_crtc_do_set_base(struct drm_crtc *crtc,
 	u32 fb_swap = R600_D1GRPH_SWAP_ENDIAN_NONE;
 	u32 tmp, viewport_w, viewport_h;
 	int r;
+	bool bypass_lut = false;
 
 	/* no fb bound */
 	if (!atomic && !crtc->primary->fb) {
@@ -1517,6 +1535,8 @@ static int avivo_crtc_do_set_base(struct drm_crtc *crtc,
 #ifdef __BIG_ENDIAN
 		fb_swap = R600_D1GRPH_SWAP_ENDIAN_32BIT;
 #endif
+		/* Greater 8 bpc fb needs to bypass hw-lut to retain precision */
+		bypass_lut = true;
 		break;
 	default:
 		DRM_ERROR("Unsupported screen format %s\n",
@@ -1559,6 +1579,13 @@ static int avivo_crtc_do_set_base(struct drm_crtc *crtc,
 	if (rdev->family >= CHIP_R600)
 		WREG32(R600_D1GRPH_SWAP_CONTROL + radeon_crtc->crtc_offset, fb_swap);
 
+	/* LUT only has 256 slots for 8 bpc fb. Bypass for > 8 bpc scanout for precision */
+	WREG32_P(AVIVO_D1GRPH_LUT_SEL + radeon_crtc->crtc_offset,
+		 (bypass_lut ? AVIVO_LUT_10BIT_BYPASS_EN : 0), ~AVIVO_LUT_10BIT_BYPASS_EN);
+
+	if (bypass_lut)
+		DRM_DEBUG_KMS("Bypassing hardware LUT due to 10 bit fb scanout.\n");
+
 	WREG32(AVIVO_D1GRPH_SURFACE_OFFSET_X + radeon_crtc->crtc_offset, 0);
 	WREG32(AVIVO_D1GRPH_SURFACE_OFFSET_Y + radeon_crtc->crtc_offset, 0);
 	WREG32(AVIVO_D1GRPH_X_START + radeon_crtc->crtc_offset, 0);

drivers/gpu/drm/radeon/evergreen_reg.h

@@ -116,6 +116,8 @@
 #       define EVERGREEN_GRPH_ARRAY_LINEAR_ALIGNED      1
 #       define EVERGREEN_GRPH_ARRAY_1D_TILED_THIN1      2
 #       define EVERGREEN_GRPH_ARRAY_2D_TILED_THIN1      4
+#define EVERGREEN_GRPH_LUT_10BIT_BYPASS_CONTROL         0x6808
+#       define EVERGREEN_LUT_10BIT_BYPASS_EN            (1 << 8)
 #define EVERGREEN_GRPH_SWAP_CONTROL                     0x680c
 #       define EVERGREEN_GRPH_ENDIAN_SWAP(x)            (((x) & 0x3) << 0)
 #       define EVERGREEN_GRPH_ENDIAN_NONE               0

drivers/gpu/drm/radeon/r500_reg.h

@@ -402,6 +402,7 @@
  * block and vice versa.  This applies to GRPH, CUR, etc.
  */
 #define AVIVO_D1GRPH_LUT_SEL                                    0x6108
+#       define AVIVO_LUT_10BIT_BYPASS_EN                        (1 << 8)
 #define AVIVO_D1GRPH_PRIMARY_SURFACE_ADDRESS                    0x6110
 #define R700_D1GRPH_PRIMARY_SURFACE_ADDRESS_HIGH                0x6914
 #define R700_D2GRPH_PRIMARY_SURFACE_ADDRESS_HIGH                0x6114

drivers/gpu/drm/radeon/radeon_display.c

@@ -66,7 +66,8 @@ static void avivo_crtc_load_lut(struct drm_crtc *crtc)
 			     (radeon_crtc->lut_b[i] << 0));
 	}
 
-	WREG32(AVIVO_D1GRPH_LUT_SEL + radeon_crtc->crtc_offset, radeon_crtc->crtc_id);
+	/* Only change bit 0 of LUT_SEL, other bits are set elsewhere */
+	WREG32_P(AVIVO_D1GRPH_LUT_SEL + radeon_crtc->crtc_offset, radeon_crtc->crtc_id, ~1);
 }
 
 static void dce4_crtc_load_lut(struct drm_crtc *crtc)