Merge branch 'perf/urgent' into perf/core, to pick up fixes and refresh the branch

Signed-off-by: Ingo Molnar <mingo@kernel.org>

5 files changed, 146 insertions, 32 deletions

diff --git a/drivers/gpu/drm/radeon/atombios_crtc.c b/drivers/gpu/drm/radeon/atombios_crtc.c
index 26c12a3..a03c734 100644
--- a/drivers/gpu/drm/radeon/atombios_crtc.c
+++ b/drivers/gpu/drm/radeon/atombios_crtc.c
@@ -1052,7 +1052,7 @@ static void atombios_crtc_set_pll(struct drm_crtc *crtc, struct drm_display_mode
 	int encoder_mode = atombios_get_encoder_mode(radeon_crtc->encoder);
 
 	/* pass the actual clock to atombios_crtc_program_pll for DCE5,6 for HDMI */
-	if (ASIC_IS_DCE5(rdev) && !ASIC_IS_DCE8(rdev) &&
+	if (ASIC_IS_DCE5(rdev) &&
 	    (encoder_mode == ATOM_ENCODER_MODE_HDMI) &&
 	    (radeon_crtc->bpc > 8))
 		clock = radeon_crtc->adjusted_clock;
@@ -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) {
@@ -1174,33 +1175,73 @@ static int dce4_crtc_do_set_base(struct drm_crtc *crtc,
 	radeon_bo_get_tiling_flags(rbo, &tiling_flags, NULL);
 	radeon_bo_unreserve(rbo);
 
-	switch (target_fb->bits_per_pixel) {
-	case 8:
+	switch (target_fb->pixel_format) {
+	case DRM_FORMAT_C8:
 		fb_format = (EVERGREEN_GRPH_DEPTH(EVERGREEN_GRPH_DEPTH_8BPP) |
 			     EVERGREEN_GRPH_FORMAT(EVERGREEN_GRPH_FORMAT_INDEXED));
 		break;
-	case 15:
+	case DRM_FORMAT_XRGB4444:
+	case DRM_FORMAT_ARGB4444:
+		fb_format = (EVERGREEN_GRPH_DEPTH(EVERGREEN_GRPH_DEPTH_16BPP) |
+			     EVERGREEN_GRPH_FORMAT(EVERGREEN_GRPH_FORMAT_ARGB4444));
+#ifdef __BIG_ENDIAN
+		fb_swap = EVERGREEN_GRPH_ENDIAN_SWAP(EVERGREEN_GRPH_ENDIAN_8IN16);
+#endif
+		break;
+	case DRM_FORMAT_XRGB1555:
+	case DRM_FORMAT_ARGB1555:
 		fb_format = (EVERGREEN_GRPH_DEPTH(EVERGREEN_GRPH_DEPTH_16BPP) |
 			     EVERGREEN_GRPH_FORMAT(EVERGREEN_GRPH_FORMAT_ARGB1555));
+#ifdef __BIG_ENDIAN
+		fb_swap = EVERGREEN_GRPH_ENDIAN_SWAP(EVERGREEN_GRPH_ENDIAN_8IN16);
+#endif
+		break;
+	case DRM_FORMAT_BGRX5551:
+	case DRM_FORMAT_BGRA5551:
+		fb_format = (EVERGREEN_GRPH_DEPTH(EVERGREEN_GRPH_DEPTH_16BPP) |
+			     EVERGREEN_GRPH_FORMAT(EVERGREEN_GRPH_FORMAT_BGRA5551));
+#ifdef __BIG_ENDIAN
+		fb_swap = EVERGREEN_GRPH_ENDIAN_SWAP(EVERGREEN_GRPH_ENDIAN_8IN16);
+#endif
 		break;
-	case 16:
+	case DRM_FORMAT_RGB565:
 		fb_format = (EVERGREEN_GRPH_DEPTH(EVERGREEN_GRPH_DEPTH_16BPP) |
 			     EVERGREEN_GRPH_FORMAT(EVERGREEN_GRPH_FORMAT_ARGB565));
 #ifdef __BIG_ENDIAN
 		fb_swap = EVERGREEN_GRPH_ENDIAN_SWAP(EVERGREEN_GRPH_ENDIAN_8IN16);
 #endif
 		break;
-	case 24:
-	case 32:
+	case DRM_FORMAT_XRGB8888:
+	case DRM_FORMAT_ARGB8888:
 		fb_format = (EVERGREEN_GRPH_DEPTH(EVERGREEN_GRPH_DEPTH_32BPP) |
 			     EVERGREEN_GRPH_FORMAT(EVERGREEN_GRPH_FORMAT_ARGB8888));
 #ifdef __BIG_ENDIAN
 		fb_swap = EVERGREEN_GRPH_ENDIAN_SWAP(EVERGREEN_GRPH_ENDIAN_8IN32);
 #endif
 		break;
+	case DRM_FORMAT_XRGB2101010:
+	case DRM_FORMAT_ARGB2101010:
+		fb_format = (EVERGREEN_GRPH_DEPTH(EVERGREEN_GRPH_DEPTH_32BPP) |
+			     EVERGREEN_GRPH_FORMAT(EVERGREEN_GRPH_FORMAT_ARGB2101010));
+#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:
+		fb_format = (EVERGREEN_GRPH_DEPTH(EVERGREEN_GRPH_DEPTH_32BPP) |
+			     EVERGREEN_GRPH_FORMAT(EVERGREEN_GRPH_FORMAT_BGRA1010102));
+#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 depth %d\n",
-			  target_fb->bits_per_pixel);
+		DRM_ERROR("Unsupported screen format %s\n",
+			  drm_get_format_name(target_fb->pixel_format));
 		return -EINVAL;
 	}
 
@@ -1329,6 +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);
@@ -1396,6 +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) {
@@ -1433,18 +1487,30 @@ static int avivo_crtc_do_set_base(struct drm_crtc *crtc,
 	radeon_bo_get_tiling_flags(rbo, &tiling_flags, NULL);
 	radeon_bo_unreserve(rbo);
 
-	switch (target_fb->bits_per_pixel) {
-	case 8:
+	switch (target_fb->pixel_format) {
+	case DRM_FORMAT_C8:
 		fb_format =
 		    AVIVO_D1GRPH_CONTROL_DEPTH_8BPP |
 		    AVIVO_D1GRPH_CONTROL_8BPP_INDEXED;
 		break;
-	case 15:
+	case DRM_FORMAT_XRGB4444:
+	case DRM_FORMAT_ARGB4444:
+		fb_format =
+		    AVIVO_D1GRPH_CONTROL_DEPTH_16BPP |
+		    AVIVO_D1GRPH_CONTROL_16BPP_ARGB4444;
+#ifdef __BIG_ENDIAN
+		fb_swap = R600_D1GRPH_SWAP_ENDIAN_16BIT;
+#endif
+		break;
+	case DRM_FORMAT_XRGB1555:
 		fb_format =
 		    AVIVO_D1GRPH_CONTROL_DEPTH_16BPP |
 		    AVIVO_D1GRPH_CONTROL_16BPP_ARGB1555;
+#ifdef __BIG_ENDIAN
+		fb_swap = R600_D1GRPH_SWAP_ENDIAN_16BIT;
+#endif
 		break;
-	case 16:
+	case DRM_FORMAT_RGB565:
 		fb_format =
 		    AVIVO_D1GRPH_CONTROL_DEPTH_16BPP |
 		    AVIVO_D1GRPH_CONTROL_16BPP_RGB565;
@@ -1452,8 +1518,8 @@ static int avivo_crtc_do_set_base(struct drm_crtc *crtc,
 		fb_swap = R600_D1GRPH_SWAP_ENDIAN_16BIT;
 #endif
 		break;
-	case 24:
-	case 32:
+	case DRM_FORMAT_XRGB8888:
+	case DRM_FORMAT_ARGB8888:
 		fb_format =
 		    AVIVO_D1GRPH_CONTROL_DEPTH_32BPP |
 		    AVIVO_D1GRPH_CONTROL_32BPP_ARGB8888;
@@ -1461,9 +1527,20 @@ static int avivo_crtc_do_set_base(struct drm_crtc *crtc,
 		fb_swap = R600_D1GRPH_SWAP_ENDIAN_32BIT;
 #endif
 		break;
+	case DRM_FORMAT_XRGB2101010:
+	case DRM_FORMAT_ARGB2101010:
+		fb_format =
+		    AVIVO_D1GRPH_CONTROL_DEPTH_32BPP |
+		    AVIVO_D1GRPH_CONTROL_32BPP_ARGB2101010;
+#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 depth %d\n",
-			  target_fb->bits_per_pixel);
+		DRM_ERROR("Unsupported screen format %s\n",
+			  drm_get_format_name(target_fb->pixel_format));
 		return -EINVAL;
 	}
 
@@ -1502,6 +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);
diff --git a/drivers/gpu/drm/radeon/evergreen_reg.h b/drivers/gpu/drm/radeon/evergreen_reg.h
index a0f63ff..333d143 100644
--- a/drivers/gpu/drm/radeon/evergreen_reg.h
+++ b/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
diff --git a/drivers/gpu/drm/radeon/r500_reg.h b/drivers/gpu/drm/radeon/r500_reg.h
index 1dd0d32..136b7bc 100644
--- a/drivers/gpu/drm/radeon/r500_reg.h
+++ b/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
diff --git a/drivers/gpu/drm/radeon/radeon_connectors.c b/drivers/gpu/drm/radeon/radeon_connectors.c
index 933c5c3..1b9177e 100644
--- a/drivers/gpu/drm/radeon/radeon_connectors.c
+++ b/drivers/gpu/drm/radeon/radeon_connectors.c
@@ -1288,17 +1288,15 @@ static int radeon_dvi_mode_valid(struct drm_connector *connector,
 		    (radeon_connector->connector_object_id == CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_D) ||
 		    (radeon_connector->connector_object_id == CONNECTOR_OBJECT_ID_HDMI_TYPE_B))
 			return MODE_OK;
-		else if (radeon_connector->connector_object_id == CONNECTOR_OBJECT_ID_HDMI_TYPE_A) {
-			if (ASIC_IS_DCE6(rdev)) {
-				/* HDMI 1.3+ supports max clock of 340 Mhz */
-				if (mode->clock > 340000)
-					return MODE_CLOCK_HIGH;
-				else
-					return MODE_OK;
-			} else
+		else if (ASIC_IS_DCE6(rdev) && drm_detect_hdmi_monitor(radeon_connector->edid)) {
+			/* HDMI 1.3+ supports max clock of 340 Mhz */
+			if (mode->clock > 340000)
 				return MODE_CLOCK_HIGH;
-		} else
+			else
+				return MODE_OK;
+		} else {
 			return MODE_CLOCK_HIGH;
+		}
 	}
 
 	/* check against the max pixel clock */
@@ -1549,6 +1547,8 @@ out:
 static int radeon_dp_mode_valid(struct drm_connector *connector,
 				  struct drm_display_mode *mode)
 {
+	struct drm_device *dev = connector->dev;
+	struct radeon_device *rdev = dev->dev_private;
 	struct radeon_connector *radeon_connector = to_radeon_connector(connector);
 	struct radeon_connector_atom_dig *radeon_dig_connector = radeon_connector->con_priv;
 
@@ -1579,14 +1579,23 @@ static int radeon_dp_mode_valid(struct drm_connector *connector,
 					return MODE_PANEL;
 			}
 		}
-		return MODE_OK;
 	} else {
 		if ((radeon_dig_connector->dp_sink_type == CONNECTOR_OBJECT_ID_DISPLAYPORT) ||
-		    (radeon_dig_connector->dp_sink_type == CONNECTOR_OBJECT_ID_eDP))
+		    (radeon_dig_connector->dp_sink_type == CONNECTOR_OBJECT_ID_eDP)) {
 			return radeon_dp_mode_valid_helper(connector, mode);
-		else
-			return MODE_OK;
+		} else {
+			if (ASIC_IS_DCE6(rdev) && drm_detect_hdmi_monitor(radeon_connector->edid)) {
+				/* HDMI 1.3+ supports max clock of 340 Mhz */
+				if (mode->clock > 340000)
+					return MODE_CLOCK_HIGH;
+			} else {
+				if (mode->clock > 165000)
+					return MODE_CLOCK_HIGH;
+			}
+		}
 	}
+
+	return MODE_OK;
 }
 
 static const struct drm_connector_helper_funcs radeon_dp_connector_helper_funcs = {
diff --git a/drivers/gpu/drm/radeon/radeon_display.c b/drivers/gpu/drm/radeon/radeon_display.c
index 5ed6170..8fc362a 100644
--- a/drivers/gpu/drm/radeon/radeon_display.c
+++ b/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)
@@ -357,8 +358,9 @@ void radeon_crtc_handle_flip(struct radeon_device *rdev, int crtc_id)
 
 	spin_unlock_irqrestore(&rdev->ddev->event_lock, flags);
 
+	drm_vblank_put(rdev->ddev, radeon_crtc->crtc_id);
 	radeon_fence_unref(&work->fence);
-	radeon_irq_kms_pflip_irq_get(rdev, work->crtc_id);
+	radeon_irq_kms_pflip_irq_put(rdev, work->crtc_id);
 	queue_work(radeon_crtc->flip_queue, &work->unpin_work);
 }
 
@@ -459,6 +461,12 @@ static void radeon_flip_work_func(struct work_struct *__work)
 		base &= ~7;
 	}
 
+	r = drm_vblank_get(crtc->dev, radeon_crtc->crtc_id);
+	if (r) {
+		DRM_ERROR("failed to get vblank before flip\n");
+		goto pflip_cleanup;
+	}
+
 	/* We borrow the event spin lock for protecting flip_work */
 	spin_lock_irqsave(&crtc->dev->event_lock, flags);
 
@@ -473,6 +481,16 @@ static void radeon_flip_work_func(struct work_struct *__work)
 
 	return;
 
+pflip_cleanup:
+	if (unlikely(radeon_bo_reserve(work->new_rbo, false) != 0)) {
+		DRM_ERROR("failed to reserve new rbo in error path\n");
+		goto cleanup;
+	}
+	if (unlikely(radeon_bo_unpin(work->new_rbo) != 0)) {
+		DRM_ERROR("failed to unpin new rbo in error path\n");
+	}
+	radeon_bo_unreserve(work->new_rbo);
+
 cleanup:
 	drm_gem_object_unreference_unlocked(&work->old_rbo->gem_base);
 	radeon_fence_unref(&work->fence);