Skip to contentSkip to conversation

Two new drm crtc properties

PATCH Superseded [email protected] 4 participants 26 messages Mar 10, 2014 – Apr 20, 2014

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover Two new drm crtc properties

From: Akash Goel <[email protected]>

Added 2 new drm crtc properties. One property provides control to vary the PIPESRC value. With this the size of Pipe's output or Panel fitter input can be varied. The other property provides control to vary the size of horizontal & vertical borders. With this the size of the Panel fitter output or display window could be controlled.

Akash Goel (3):
drm/i915: Initialized 'set_property' fn pointer field of
intel_crtc_funcs structure
drm/i915: New drm crtc property for varying the Pipe Src size drm/i915: New drm crtc property for varying the size of borders

 drivers/gpu/drm/i915/i915_drv.h      |  13 ++++
 drivers/gpu/drm/i915/intel_display.c |  72 +++++++++++++++++
 drivers/gpu/drm/i915/intel_drv.h     |   4 +
 drivers/gpu/drm/i915/intel_panel.c   | 147 +++++++++++++++++++++++++++++++++++
 4 files changed, 236 insertions(+)
Signature · 1 line
1.8.5.2
#1 lore raw

PATCH 1/3 drm/i915: Initialized 'set_property' fn pointer field of intel_crtc_funcs structure

From: Akash Goel <[email protected]>

This patch defines a new function & assigns that to the 'set_property' function pointer field of the 'intel_crtc_funcs' structure.

---
 drivers/gpu/drm/i915/intel_display.c | 9 +++++++++
 1 file changed, 9 insertions(+)
drivers/gpu/drm/i915/intel_display.c +9 -0
index 8494194..5dfe156 100644
@@ -10431,6 +10431,14 @@ out_config:
 	return ret;
 }
 
+static int intel_crtc_set_property(struct drm_crtc *crtc,
+		struct drm_property *property, uint64_t val)
+{
+	int ret = -ENOENT;
+
+	return ret;
+}
+
 static const struct drm_crtc_funcs intel_crtc_funcs = {
 	.cursor_set = intel_crtc_cursor_set,
 	.cursor_move = intel_crtc_cursor_move,
@@ -10438,6 +10446,7 @@ static const struct drm_crtc_funcs intel_crtc_funcs = {
 	.set_config = intel_crtc_set_config,
 	.destroy = intel_crtc_destroy,
 	.page_flip = intel_crtc_page_flip,
+	.set_property = intel_crtc_set_property,
 };
 
 static void intel_cpu_pll_init(struct drm_device *dev)
Signature · 1 line
1.8.5.2
#2 lore raw

PATCH 2/3 drm/i915: New drm crtc property for varying the Pipe Src size

In reply to [email protected]

From: Akash Goel <[email protected]>

This patch adds a new drm crtc property for varying the Pipe Src size or the Panel fitter input size. Pipe Src controls the size that is scaled from.
This will allow to dynamically flip (without modeset) the frame buffers of different resolutions

---
 drivers/gpu/drm/i915/i915_drv.h      |  6 ++++++
 drivers/gpu/drm/i915/intel_display.c | 39 ++++++++++++++++++++++++++++++++++++
 2 files changed, 45 insertions(+)
drivers/gpu/drm/i915/i915_drv.h +6 -0
index a0d90ef..6f3af15 100644
@@ -1608,6 +1608,12 @@ typedef struct drm_i915_private {
 	struct drm_property *broadcast_rgb_property;
 	struct drm_property *force_audio_property;
 
+	/*
+	 * Property to dynamically vary the size of the
+	 * PIPESRC or Panel fitter input size
+	 */
+	struct drm_property *input_size_property;
+
 	uint32_t hw_context_size;
 	struct list_head context_list;
 
drivers/gpu/drm/i915/intel_display.c +39 -0
index 5dfe156..30374e9 100644
@@ -10434,8 +10434,38 @@ out_config:
 static int intel_crtc_set_property(struct drm_crtc *crtc,
 		struct drm_property *property, uint64_t val)
 {
+	struct drm_device *dev = crtc->dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
 	int ret = -ENOENT;
 
+	if (property == dev_priv->input_size_property) {
+		int new_width = (int)((val >> 16) & 0xffff);
+		int new_height = (int)(val & 0xffff);
+
+		if ((new_width == intel_crtc->config.pipe_src_w) &&
+		    (new_height == intel_crtc->config.pipe_src_h))
+			return 0;
+
+		intel_crtc->config.pipe_src_w = new_width;
+		intel_crtc->config.pipe_src_h = new_height;
+
+		intel_crtc->config.requested_mode.hdisplay = new_width;
+		intel_crtc->config.requested_mode.vdisplay = new_height;
+
+		crtc->mode.hdisplay = new_width;
+		crtc->mode.vdisplay = new_height;
+
+		/* pipesrc controls the size that is scaled from, which should
+		* always be the user's requested size.
+		*/
+		I915_WRITE(PIPESRC(intel_crtc->pipe),
+			((intel_crtc->config.pipe_src_w - 1) << 16) |
+			 (intel_crtc->config.pipe_src_h - 1));
+
+		return 0;
+	}
+
 	return ret;
 }
 
@@ -10586,6 +10616,15 @@ static void intel_crtc_init(struct drm_device *dev, int pipe)
 	dev_priv->pipe_to_crtc_mapping[intel_crtc->pipe] = &intel_crtc->base;
 
 	drm_crtc_helper_add(&intel_crtc->base, &intel_helper_funcs);
+
+	if (!dev_priv->input_size_property)
+		dev_priv->input_size_property =
+			drm_property_create_range(dev, 0, "input size", 0, 0xFFFFFFFF);
+
+	if (dev_priv->input_size_property)
+		drm_object_attach_property(&intel_crtc->base.base,
+					   dev_priv->input_size_property,
+					   0);
 }
 
 enum pipe intel_get_pipe_from_connector(struct intel_connector *connector)
Signature · 1 line
1.8.5.2
#3 lore raw

PATCH 3/3 drm/i915: New drm crtc property for varying the size of borders

In reply to [email protected]

From: Akash Goel <[email protected]>

This patch adds a new drm crtc property for varying the size of the horizontal & vertical borers of the output/display window. This will control the output of Panel fitter.

---
 drivers/gpu/drm/i915/i915_drv.h      |   7 ++
 drivers/gpu/drm/i915/intel_display.c |  28 ++++++-
 drivers/gpu/drm/i915/intel_drv.h     |   4 +
 drivers/gpu/drm/i915/intel_panel.c   | 147 +++++++++++++++++++++++++++++++++++
 4 files changed, 184 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +7 -0
index 6f3af15..eec32ed 100644
@@ -1614,6 +1614,13 @@ typedef struct drm_i915_private {
 	 */
 	struct drm_property *input_size_property;
 
+	/*
+	 * Property to dynamically vary the size of the
+	 * borders. This will indirectly control the size
+	 * of the display window i.e Panel fitter output
+	 */
+	struct drm_property *output_border_property;
+
 	uint32_t hw_context_size;
 	struct list_head context_list;
 
drivers/gpu/drm/i915/intel_display.c +26 -2
index 30374e9..2d213db 100644
@@ -10437,7 +10437,16 @@ static int intel_crtc_set_property(struct drm_crtc *crtc,
 	struct drm_device *dev = crtc->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
-	int ret = -ENOENT;
+
+	if (property == dev_priv->output_border_property) {
+		if (val == (uint64_t)intel_crtc->border_size)
+			return 0;
+
+		intel_crtc->border_size = (uint32_t)val;
+		intel_crtc->border_size_changed = 1;
+
+		goto done;
+	}
 
 	if (property == dev_priv->input_size_property) {
 		int new_width = (int)((val >> 16) & 0xffff);
@@ -10466,7 +10475,13 @@ static int intel_crtc_set_property(struct drm_crtc *crtc,
 		return 0;
 	}
 
-	return ret;
+	return -EINVAL;
+
+done:
+	if (crtc)
+		intel_crtc_restore_mode(crtc);
+
+	return 0;
 }
 
 static const struct drm_crtc_funcs intel_crtc_funcs = {
@@ -10625,6 +10640,15 @@ static void intel_crtc_init(struct drm_device *dev, int pipe)
 		drm_object_attach_property(&intel_crtc->base.base,
 					   dev_priv->input_size_property,
 					   0);
+
+	if (!dev_priv->output_border_property)
+		dev_priv->output_border_property =
+			drm_property_create_range(dev, 0, "border size", 0, 0xFFFFFFFF);
+
+	if (dev_priv->output_border_property)
+		drm_object_attach_property(&intel_crtc->base.base,
+					   dev_priv->output_border_property,
+					   0);
 }
 
 enum pipe intel_get_pipe_from_connector(struct intel_connector *connector)
drivers/gpu/drm/i915/intel_drv.h +4 -0
index 5360d16..68eec38 100644
@@ -387,6 +387,10 @@ struct intel_crtc {
 	bool cpu_fifo_underrun_disabled;
 	bool pch_fifo_underrun_disabled;
 
+	/* border info for the output/display window */
+	bool border_size_changed;
+	uint32_t border_size;
+
 	/* per-pipe watermark state */
 	struct {
 		/* watermarks currently being used  */
drivers/gpu/drm/i915/intel_panel.c +147 -0

171 lines not on this page · raw message

Signature · 1 line
1.8.5.2
#4 lore raw
In reply to [email protected]
Quoted 9 lines · On Mon, Mar 10, 2014 at 10:40:50AM +0530, [email protected] wrote:

From: Akash Goel <[email protected]>

Added 2 new drm crtc properties. One property provides control to vary the PIPESRC value. With this the size of Pipe's output or Panel fitter input can be varied. The other property provides control to vary the size of horizontal & vertical borders. With this the size of the Panel fitter output or display window could be controlled.

Quoted 5 lines

Akash Goel (3):
drm/i915: Initialized 'set_property' fn pointer field of
intel_crtc_funcs structure
drm/i915: New drm crtc property for varying the Pipe Src size drm/i915: New drm crtc property for varying the size of borders

Hm, where are the igt testcases for this feature? I expect some fun interactions at least with pageflips/modesets vs. setting this. The vblank timing code should be fine, since it always deals with adjusted timings.
-Daniel

Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
#5 lore raw

Sorry, will send across the corresponding IGT test case also shortly.

Best regards
Akash

-----Original Message-----
From: Daniel Vetter [mailto:[email protected]] On Behalf Of Daniel Vetter
Sent: Monday, March 10, 2014 10:44 AM
To: Goel, Akash
Cc: [email protected]
Subject: Re: [Intel-gfx] [PATCH 0/3] Two new drm crtc properties

Quoted 9 lines · On Mon, Mar 10, 2014 at 10:40:50AM +0530, [email protected] wrote:

From: Akash Goel <[email protected]>

Added 2 new drm crtc properties. One property provides control to vary the PIPESRC value. With this the size of Pipe's output or Panel fitter input can be varied.
The other property provides control to vary the size of horizontal & vertical borders. With this the size of the Panel fitter output or display window could be controlled.

Quoted 5 lines

Akash Goel (3):
drm/i915: Initialized 'set_property' fn pointer field of
intel_crtc_funcs structure
drm/i915: New drm crtc property for varying the Pipe Src size drm/i915: New drm crtc property for varying the size of borders

Hm, where are the igt testcases for this feature? I expect some fun interactions at least with pageflips/modesets vs. setting this. The vblank timing code should be fine, since it always deals with adjusted timings. -Daniel
--
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch

#6 lore raw

PATCH 1/3 drm/i915: Initialized 'set_property' fn pointer field of intel_crtc_funcs structure

In reply to Daniel Vetter

From: Akash Goel <[email protected]>

This patch defines a new function & assigns that to the 'set_property' function pointer field of the 'intel_crtc_funcs' structure.

---
 drivers/gpu/drm/i915/intel_display.c | 9 +++++++++
 1 file changed, 9 insertions(+)
drivers/gpu/drm/i915/intel_display.c +9 -0
index 8494194..5dfe156 100644
@@ -10431,6 +10431,14 @@ out_config:
 	return ret;
 }
 
+static int intel_crtc_set_property(struct drm_crtc *crtc,
+		struct drm_property *property, uint64_t val)
+{
+	int ret = -ENOENT;
+
+	return ret;
+}
+
 static const struct drm_crtc_funcs intel_crtc_funcs = {
 	.cursor_set = intel_crtc_cursor_set,
 	.cursor_move = intel_crtc_cursor_move,
@@ -10438,6 +10446,7 @@ static const struct drm_crtc_funcs intel_crtc_funcs = {
 	.set_config = intel_crtc_set_config,
 	.destroy = intel_crtc_destroy,
 	.page_flip = intel_crtc_page_flip,
+	.set_property = intel_crtc_set_property,
 };
 
 static void intel_cpu_pll_init(struct drm_device *dev)
Signature · 1 line
1.8.5.2
#7 lore raw

v2 cover Two new drm crtc properties

From: Akash Goel <[email protected]>

Added 2 new drm crtc properties. One property provides control to vary the PIPESRC value. With this the size of Pipe's output or Panel fitter input can be varied. The other property provides control to vary the size of horizontal & vertical borders. With this the size of the Panel fitter output or display window could be controlled.
v2: Added new checks and disabled the fb pitch mismatch check for VLV & HSW

Akash Goel (3):
drm/i915: Initialized 'set_property' fn pointer field of
intel_crtc_funcs structure
drm/i915: New drm crtc property for varying the Pipe Src size drm/i915: New drm crtc property for varying the size of borders

 drivers/gpu/drm/i915/i915_drv.h      |  13 +++
 drivers/gpu/drm/i915/intel_display.c |  93 +++++++++++++++++++-
 drivers/gpu/drm/i915/intel_drv.h     |   4 +
 drivers/gpu/drm/i915/intel_panel.c   | 159 +++++++++++++++++++++++++++++++++++
 4 files changed, 267 insertions(+), 2 deletions(-)
Signature · 1 line
1.8.5.2
#8 lore raw

Hi Ville,

Please could you review this patch & provide a feedback. This is as per your suggestions & our earlier discussions.

Regards,
Sourab

-----Original Message-----
From: Goel, Akash
Sent: Tuesday, March 11, 2014 6:24 PM
To: [email protected]
Cc: Purushothaman, Vijay A; G, Pallavi; Goel, Akash
Subject: [PATCH v2 0/3] Two new drm crtc properties

From: Akash Goel <[email protected]>

Added 2 new drm crtc properties. One property provides control to vary the PIPESRC value. With this the size of Pipe's output or Panel fitter input can be varied. The other property provides control to vary the size of horizontal & vertical borders. With this the size of the Panel fitter output or display window could be controlled.
v2: Added new checks and disabled the fb pitch mismatch check for VLV & HSW

Akash Goel (3):
drm/i915: Initialized 'set_property' fn pointer field of
intel_crtc_funcs structure
drm/i915: New drm crtc property for varying the Pipe Src size drm/i915: New drm crtc property for varying the size of borders

 drivers/gpu/drm/i915/i915_drv.h      |  13 +++
 drivers/gpu/drm/i915/intel_display.c |  93 +++++++++++++++++++-
 drivers/gpu/drm/i915/intel_drv.h     |   4 +
 drivers/gpu/drm/i915/intel_panel.c   | 159 +++++++++++++++++++++++++++++++++++
 4 files changed, 267 insertions(+), 2 deletions(-)

--
1.8.5.2

#9 lore raw

PATCH v2 2/3 drm/i915: New drm crtc property for varying the Pipe Src size

In reply to [email protected]

From: Akash Goel <[email protected]>

This patch adds a new drm crtc property for varying the Pipe Src size or the Panel fitter input size. Pipe Src controls the size that is scaled from.
This will allow to dynamically flip (without modeset) the frame buffers of different resolutions

v2: Added a check to fail the set property call if Panel fitter is disabled & new PIPESRC programming do not match with PIPE timings. Removed the pitch mismatch check on frame buffer, when being flipped. This is currently done only for VLV/HSW.

Testcase: igt/kms_panel_fitter_test

---
 drivers/gpu/drm/i915/i915_drv.h      |  6 ++++
 drivers/gpu/drm/i915/intel_display.c | 60 ++++++++++++++++++++++++++++++++++--
 2 files changed, 64 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +6 -0
index a0d90ef..6f3af15 100644
@@ -1608,6 +1608,12 @@ typedef struct drm_i915_private {
 	struct drm_property *broadcast_rgb_property;
 	struct drm_property *force_audio_property;
 
+	/*
+	 * Property to dynamically vary the size of the
+	 * PIPESRC or Panel fitter input size
+	 */
+	struct drm_property *input_size_property;
+
 	uint32_t hw_context_size;
 	struct list_head context_list;
 
drivers/gpu/drm/i915/intel_display.c +58 -2

80 lines not on this page · raw message

Signature · 1 line
1.8.5.2
#10 lore raw

PATCH v2 3/3 drm/i915: New drm crtc property for varying the size of borders

In reply to [email protected]

From: Akash Goel <[email protected]>

This patch adds a new drm crtc property for varying the size of the horizontal & vertical borers of the output/display window. This will control the output of Panel fitter.

v2: Added a new check for the invalid border size input

Testcase: igt/kms_panel_fitter_test

---
 drivers/gpu/drm/i915/i915_drv.h      |   7 ++
 drivers/gpu/drm/i915/intel_display.c |  28 +++++-
 drivers/gpu/drm/i915/intel_drv.h     |   4 +
 drivers/gpu/drm/i915/intel_panel.c   | 159 +++++++++++++++++++++++++++++++++++
 4 files changed, 196 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +7 -0
index 6f3af15..eec32ed 100644
@@ -1614,6 +1614,13 @@ typedef struct drm_i915_private {
 	 */
 	struct drm_property *input_size_property;
 
+	/*
+	 * Property to dynamically vary the size of the
+	 * borders. This will indirectly control the size
+	 * of the display window i.e Panel fitter output
+	 */
+	struct drm_property *output_border_property;
+
 	uint32_t hw_context_size;
 	struct list_head context_list;
 
drivers/gpu/drm/i915/intel_display.c +26 -2
index 8772390..68ec5b5 100644
@@ -10447,7 +10447,16 @@ static int intel_crtc_set_property(struct drm_crtc *crtc,
 	struct drm_device *dev = crtc->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
-	int ret = -ENOENT;
+
+	if (property == dev_priv->output_border_property) {
+		if (val == (uint64_t)intel_crtc->border_size)
+			return 0;
+
+		intel_crtc->border_size = (uint32_t)val;
+		intel_crtc->border_size_changed = 1;
+
+		goto done;
+	}
 
 	if (property == dev_priv->input_size_property) {
 		int new_width = (int)((val >> 16) & 0xffff);
@@ -10483,7 +10492,13 @@ static int intel_crtc_set_property(struct drm_crtc *crtc,
 		return 0;
 	}
 
-	return ret;
+	return -EINVAL;
+
+done:
+	if (crtc)
+		intel_crtc_restore_mode(crtc);
+
+	return 0;
 }
 
 static const struct drm_crtc_funcs intel_crtc_funcs = {
@@ -10642,6 +10657,15 @@ static void intel_crtc_init(struct drm_device *dev, int pipe)
 		drm_object_attach_property(&intel_crtc->base.base,
 					   dev_priv->input_size_property,
 					   0);
+
+	if (!dev_priv->output_border_property)
+		dev_priv->output_border_property =
+			drm_property_create_range(dev, 0, "border size", 0, 0xFFFFFFFF);
+
+	if (dev_priv->output_border_property)
+		drm_object_attach_property(&intel_crtc->base.base,
+					   dev_priv->output_border_property,
+					   0);
 }
 
 enum pipe intel_get_pipe_from_connector(struct intel_connector *connector)
drivers/gpu/drm/i915/intel_drv.h +4 -0
index 5360d16..68eec38 100644
@@ -387,6 +387,10 @@ struct intel_crtc {
 	bool cpu_fifo_underrun_disabled;
 	bool pch_fifo_underrun_disabled;
 
+	/* border info for the output/display window */
+	bool border_size_changed;
+	uint32_t border_size;
+
 	/* per-pipe watermark state */
 	struct {
 		/* watermarks currently being used  */
drivers/gpu/drm/i915/intel_panel.c +159 -0

183 lines not on this page · raw message

Signature · 1 line
1.8.5.2
#11 lore raw

PATCH v3 1/1 drm/i915: New drm crtc property for varying the size of borders

From: Akash Goel <[email protected]>

This patch adds a new drm crtc property for varying the size of the horizontal & vertical borers of the output/display window. This will control the output of Panel fitter.

v2: Added a new check for the invalid border size input

v3: Fixed bugs in output window calculation
Removed superfluous checks

Testcase: igt/kms_panel_fitter_test

---
 drivers/gpu/drm/i915/i915_drv.h      |   7 ++
 drivers/gpu/drm/i915/intel_display.c |  28 +++++-
 drivers/gpu/drm/i915/intel_drv.h     |   4 +
 drivers/gpu/drm/i915/intel_panel.c   | 180 ++++++++++++++++++++++++++++++++---
 4 files changed, 204 insertions(+), 15 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +7 -0
index 6f3af15..eec32ed 100644
@@ -1614,6 +1614,13 @@ typedef struct drm_i915_private {
 	 */
 	struct drm_property *input_size_property;
 
+	/*
+	 * Property to dynamically vary the size of the
+	 * borders. This will indirectly control the size
+	 * of the display window i.e Panel fitter output
+	 */
+	struct drm_property *output_border_property;
+
 	uint32_t hw_context_size;
 	struct list_head context_list;
 
drivers/gpu/drm/i915/intel_display.c +26 -2
index 8772390..68ec5b5 100644
@@ -10447,7 +10447,16 @@ static int intel_crtc_set_property(struct drm_crtc *crtc,
 	struct drm_device *dev = crtc->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
-	int ret = -ENOENT;
+
+	if (property == dev_priv->output_border_property) {
+		if (val == (uint64_t)intel_crtc->border_size)
+			return 0;
+
+		intel_crtc->border_size = (uint32_t)val;
+		intel_crtc->border_size_changed = 1;
+
+		goto done;
+	}
 
 	if (property == dev_priv->input_size_property) {
 		int new_width = (int)((val >> 16) & 0xffff);
@@ -10483,7 +10492,13 @@ static int intel_crtc_set_property(struct drm_crtc *crtc,
 		return 0;
 	}
 
-	return ret;
+	return -EINVAL;
+
+done:
+	if (crtc)
+		intel_crtc_restore_mode(crtc);
+
+	return 0;
 }
 
 static const struct drm_crtc_funcs intel_crtc_funcs = {
@@ -10642,6 +10657,15 @@ static void intel_crtc_init(struct drm_device *dev, int pipe)
 		drm_object_attach_property(&intel_crtc->base.base,
 					   dev_priv->input_size_property,
 					   0);
+
+	if (!dev_priv->output_border_property)
+		dev_priv->output_border_property =
+			drm_property_create_range(dev, 0, "border size", 0, 0xFFFFFFFF);
+
+	if (dev_priv->output_border_property)
+		drm_object_attach_property(&intel_crtc->base.base,
+					   dev_priv->output_border_property,
+					   0);
 }
 
 enum pipe intel_get_pipe_from_connector(struct intel_connector *connector)
drivers/gpu/drm/i915/intel_drv.h +4 -0
index 5360d16..68eec38 100644
@@ -387,6 +387,10 @@ struct intel_crtc {
 	bool cpu_fifo_underrun_disabled;
 	bool pch_fifo_underrun_disabled;
 
+	/* border info for the output/display window */
+	bool border_size_changed;
+	uint32_t border_size;
+
 	/* per-pipe watermark state */
 	struct {
 		/* watermarks currently being used  */
drivers/gpu/drm/i915/intel_panel.c +167 -13

220 lines not on this page · raw message

Signature · 1 line
1.8.5.2
#12 lore raw

PATCH v4 3/3 drm/i915: New drm crtc property for varying the size of borders

From: Akash Goel <[email protected]>

This patch adds a new drm crtc property for varying the size of the horizontal & vertical borers of the output/display window. This will control the output of Panel fitter.

v2: Added a new check for the invalid border size input

v3: Fixed bugs in output window calculation
Removed superfluous checks

v4: Added the capability to forecfully enable the Panel fitter. The property value is of 64 bits, first 32 bits are used for border dimensions. The 33rd bit can be used to forcefully enable the panel fitter. This is useful for Panels which do not override the User specified Pipe timings.

Testcase: igt/kms_panel_fitter_test

---
 drivers/gpu/drm/i915/i915_drv.h      |   7 ++
 drivers/gpu/drm/i915/intel_display.c |  39 +++++++-
 drivers/gpu/drm/i915/intel_drv.h     |   5 +
 drivers/gpu/drm/i915/intel_panel.c   | 176 ++++++++++++++++++++++++++++++++---
 4 files changed, 211 insertions(+), 16 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +7 -0
index 6f3af15..eec32ed 100644
@@ -1614,6 +1614,13 @@ typedef struct drm_i915_private {
 	 */
 	struct drm_property *input_size_property;
 
+	/*
+	 * Property to dynamically vary the size of the
+	 * borders. This will indirectly control the size
+	 * of the display window i.e Panel fitter output
+	 */
+	struct drm_property *output_border_property;
+
 	uint32_t hw_context_size;
 	struct list_head context_list;
 
drivers/gpu/drm/i915/intel_display.c +36 -3
index 7149123..a217f25 100644
@@ -10447,7 +10447,23 @@ static int intel_crtc_set_property(struct drm_crtc *crtc,
 	struct drm_device *dev = crtc->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
-	int ret = -ENOENT;
+
+	if (property == dev_priv->output_border_property) {
+		if ((val == (uint64_t)intel_crtc->border_size) &&
+		    (((val >> 32) & 0x1) == intel_crtc->pfit_enabled))
+			return 0;
+
+		intel_crtc->border_size = (uint32_t)val;
+		intel_crtc->pfit_enabled = (val >> 32) & 0x1;
+		if ((intel_crtc->border_size != 0) &&
+		    (!intel_crtc->pfit_enabled)) {
+			DRM_ERROR("Wrong setting, expect Pfit to be enabled when "
+				  "applying borders\n");
+			return -EINVAL;
+		}
+
+		goto done;
+	}
 
 	if (property == dev_priv->input_size_property) {
 		int new_width = (int)((val >> 16) & 0xffff);
@@ -10488,7 +10504,13 @@ static int intel_crtc_set_property(struct drm_crtc *crtc,
 		return 0;
 	}
 
-	return ret;
+	return -EINVAL;
+
+done:
+	if (crtc)
+		intel_crtc_restore_mode(crtc);
+
+	return 0;
 }
 
 static const struct drm_crtc_funcs intel_crtc_funcs = {
@@ -10641,12 +10663,23 @@ static void intel_crtc_init(struct drm_device *dev, int pipe)
 
 	if (!dev_priv->input_size_property)
 		dev_priv->input_size_property =
-			drm_property_create_range(dev, 0, "input size", 0, 0xFFFFFFFF);
+			drm_property_create_range(dev, 0, "input size",
+						0, 0xFFFFFFFF);
 
 	if (dev_priv->input_size_property)
 		drm_object_attach_property(&intel_crtc->base.base,
 					   dev_priv->input_size_property,
 					   0);
+
+	if (!dev_priv->output_border_property)
+		dev_priv->output_border_property =
+			drm_property_create_range(dev, 0, "border size",
+						0, (uint64_t)0x1FFFFFFFF);
+
+	if (dev_priv->output_border_property)
+		drm_object_attach_property(&intel_crtc->base.base,
+					   dev_priv->output_border_property,
+					   0);
 }
 
 enum pipe intel_get_pipe_from_connector(struct intel_connector *connector)
drivers/gpu/drm/i915/intel_drv.h +5 -0
index 5360d16..3cfc9da 100644
@@ -387,6 +387,11 @@ struct intel_crtc {
 	bool cpu_fifo_underrun_disabled;
 	bool pch_fifo_underrun_disabled;
 
+	/* for forceful enabling of panel fitter */
+	bool pfit_enabled;
+	/* border info for the output/display window */
+	uint32_t border_size;
+
 	/* per-pipe watermark state */
 	struct {
 		/* watermarks currently being used  */
drivers/gpu/drm/i915/intel_panel.c +163 -13

216 lines not on this page · raw message

Signature · 1 line
1.8.5.2
#13 lore raw

Re: [PATCH v4 3/3] drm/i915: New drm crtc property for varying the size of borders

Hi Ville,

Please could you review this patch.

Best Regards
Akash

-----Original Message-----
From: Goel, Akash
Sent: Wednesday, March 26, 2014 9:25 AM
To: [email protected]
Cc: Purushothaman, Vijay A; G, Pallavi; Goel, Akash
Subject: [PATCH v4 3/3] drm/i915: New drm crtc property for varying the size of borders

From: Akash Goel <[email protected]>

This patch adds a new drm crtc property for varying the size of the horizontal & vertical borers of the output/display window. This will control the output of Panel fitter.

v2: Added a new check for the invalid border size input

v3: Fixed bugs in output window calculation Removed superfluous checks

v4: Added the capability to forecfully enable the Panel fitter.
The property value is of 64 bits, first 32 bits are used for border dimensions. The 33rd bit can be used to forcefully enable the panel fitter. This is useful for Panels which do not override the User specified Pipe timings.

Testcase: igt/kms_panel_fitter_test

---
 drivers/gpu/drm/i915/i915_drv.h      |   7 ++
 drivers/gpu/drm/i915/intel_display.c |  39 +++++++-
 drivers/gpu/drm/i915/intel_drv.h     |   5 +
 drivers/gpu/drm/i915/intel_panel.c   | 176 ++++++++++++++++++++++++++++++++---
 4 files changed, 211 insertions(+), 16 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +7 -0
@@ -1614,6 +1614,13 @@ typedef struct drm_i915_private {
 	 */
 	struct drm_property *input_size_property;
 
+	/*
+	 * Property to dynamically vary the size of the
+	 * borders. This will indirectly control the size
+	 * of the display window i.e Panel fitter output
+	 */
+	struct drm_property *output_border_property;
+
 	uint32_t hw_context_size;
 	struct list_head context_list;
 
drivers/gpu/drm/i915/intel_display.c +17 -1
index 7149123..a217f25 100644
@@ -10447,7 +10447,23 @@ static int intel_crtc_set_property(struct drm_crtc *crtc,
 	struct drm_device *dev = crtc->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
-	int ret = -ENOENT;
+
+	if (property == dev_priv->output_border_property) {
+		if ((val == (uint64_t)intel_crtc->border_size) &&
+		    (((val >> 32) & 0x1) == intel_crtc->pfit_enabled))
+			return 0;
+
+		intel_crtc->border_size = (uint32_t)val;
+		intel_crtc->pfit_enabled = (val >> 32) & 0x1;
+		if ((intel_crtc->border_size != 0) &&
+		    (!intel_crtc->pfit_enabled)) {
+			DRM_ERROR("Wrong setting, expect Pfit to be enabled when "
+				  "applying borders\n");
+			return -EINVAL;
+		}
+
+		goto done;
+	}
 
 	if (property == dev_priv->input_size_property) {
 		int new_width = (int)((val >> 16) & 0xffff); @@ -10488,7 +10504,13 @@ static int intel_crtc_set_property(struct drm_crtc *crtc,

return 0;
}

- return ret;
+ return -EINVAL;
+
+done:
+ if (crtc)
+ intel_crtc_restore_mode(crtc);
+
+ return 0;
}

static const struct drm_crtc_funcs intel_crtc_funcs = { @@ -10641,12 +10663,23 @@ static void intel_crtc_init(struct drm_device *dev, int pipe)

if (!dev_priv->input_size_property)
dev_priv->input_size_property =
- drm_property_create_range(dev, 0, "input size", 0, 0xFFFFFFFF);
+ drm_property_create_range(dev, 0, "input size",
+ 0, 0xFFFFFFFF);

if (dev_priv->input_size_property)
drm_object_attach_property(&intel_crtc->base.base,
dev_priv->input_size_property, 0);
+
+ if (!dev_priv->output_border_property)
+ dev_priv->output_border_property =
+ drm_property_create_range(dev, 0, "border size",
+ 0, (uint64_t)0x1FFFFFFFF);
+
+ if (dev_priv->output_border_property)
+ drm_object_attach_property(&intel_crtc->base.base,
+ dev_priv->output_border_property,
+ 0);
}

enum pipe intel_get_pipe_from_connector(struct intel_connector *connector) diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h
index 5360d16..3cfc9da 100644

drivers/gpu/drm/i915/intel_drv.h +5 -0
@@ -387,6 +387,11 @@ struct intel_crtc {
 	bool cpu_fifo_underrun_disabled;
 	bool pch_fifo_underrun_disabled;
 
+	/* for forceful enabling of panel fitter */
+	bool pfit_enabled;
+	/* border info for the output/display window */
+	uint32_t border_size;
+
 	/* per-pipe watermark state */
 	struct {
 		/* watermarks currently being used  */ diff --git a/drivers/gpu/drm/i915/intel_panel.c b/drivers/gpu/drm/i915/intel_panel.c

index cb05840..86a486e 100644

drivers/gpu/drm/i915/intel_panel.c +13 -0
@@ -29,10 +29,25 @@
  */
 
 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+/* Max Downscale ratio of 1.125, expressed in 1.12 fixed point format 
+*/ #define MAX_DOWNSCALE_RATIO  (0x9 << 9)
 
 #include <linux/moduleparam.h>
 #include "intel_drv.h"
 
+static inline u32 panel_fitter_scaling(u32 source, u32 target) {
+	/*
+	 * Floating point operation is not supported. So the FACTOR
+	 * is defined, which can avoid the floating point computation
+	 * when calculating the panel ratio.
+	 */
+#define ACCURACY 12
+#define FACTOR (1 << ACCURACY)
+	u32 ratio = source * FACTOR / target;
+	return (FACTOR * ratio + FACTOR/2) / FACTOR; }
+
 void
 intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode,
 		       struct drm_display_mode *adjusted_mode) @@ -42,6 +57,60 @@ intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode,

drm_mode_set_crtcinfo(adjusted_mode, 0); }

+void
+intel_pch_manual_panel_fitting(struct intel_crtc *intel_crtc,
+ struct intel_crtc_config *pipe_config) {
+ struct drm_display_mode *adjusted_mode;
+ int x, y;
+ u32 pf_horizontal_ratio, pf_vertical_ratio;
+ u32 tot_width, tot_height;
+ u32 src_width, src_height; /* pipesrc.x, pipesrc.y */
+ u32 dst_width, dst_height;
+
+ adjusted_mode = &pipe_config->adjusted_mode;
+
+ src_width = pipe_config->pipe_src_w;
+ src_height = pipe_config->pipe_src_h;
+
+ tot_width = adjusted_mode->hdisplay;
+ tot_height = adjusted_mode->vdisplay;
+
+ /*
+ * Having non zero borders will reduce the size of 'HACTIVE/VACTIVE'
+ * region. So (HACTIVE - Left border - Right Border) *
+ * (VACTIVE - Top Border - Bottom border) will effectively be the
+ * output rectangle on screen
+ */
+ dst_width = tot_width -
+ (((intel_crtc->border_size >> 16) & 0xffff) * 2);
+ dst_height = tot_height -
+ ((intel_crtc->border_size & 0xffff) * 2);
+
+ if ((dst_width == 0) || (dst_height == 0)) {
+ DRM_ERROR("Invalid border size input\n");
+ return;
+ }
+
+ pf_horizontal_ratio = panel_fitter_scaling(src_width, dst_width);
+ pf_vertical_ratio = panel_fitter_scaling(src_height, dst_height);
+
+ if (pf_horizontal_ratio > MAX_DOWNSCALE_RATIO) {
+ DRM_ERROR("width is too small\n");
+ return;
+ } else if (pf_vertical_ratio > MAX_DOWNSCALE_RATIO) {
+ DRM_ERROR("height is too small\n");
+ return;
+ }
+
+ x = (adjusted_mode->hdisplay - dst_width + 1)/2;
+ y = (adjusted_mode->vdisplay - dst_height + 1)/2;
+
+ pipe_config->pch_pfit.pos = (x << 16) | y;
+ pipe_config->pch_pfit.size = (dst_width << 16) | dst_height;
+ pipe_config->pch_pfit.enabled = pipe_config->pch_pfit.size != 0; }
+
/* adjusted_mode has been preset to be the panel's fixed mode */ void intel_pch_panel_fitting(struct intel_crtc *intel_crtc, @@ -55,6 +124,13 @@ intel_pch_panel_fitting(struct intel_crtc *intel_crtc,

x = y = width = height = 0;

+ /* check if User wants to apply the borders, or wants to forcefully
+ enable the panel fitter, otherwise fall through the regular path */
+ if (intel_crtc->pfit_enabled ||
+ intel_crtc->border_size)
+ return intel_pch_manual_panel_fitting(intel_crtc,
+ pipe_config);
+
/* Native modes don't need fitting */
if (adjusted_mode->hdisplay == pipe_config->pipe_src_w &&
adjusted_mode->vdisplay == pipe_config->pipe_src_h) @@ -157,19 +233,6 @@ centre_vertically(struct drm_display_mode *mode,
mode->crtc_vsync_end = mode->crtc_vsync_start + sync_width; }

-static inline u32 panel_fitter_scaling(u32 source, u32 target) -{
- /*
- * Floating point operation is not supported. So the FACTOR
- * is defined, which can avoid the floating point computation
- * when calculating the panel ratio.
- */
-#define ACCURACY 12
-#define FACTOR (1 << ACCURACY)
- u32 ratio = source * FACTOR / target;
- return (FACTOR * ratio + FACTOR/2) / FACTOR;
-}
-
static void i965_scale_aspect(struct intel_crtc_config *pipe_config,
u32 *pfit_control)
{
@@ -247,6 +310,86 @@ static void i9xx_scale_aspect(struct intel_crtc_config *pipe_config,
}
}

+void intel_gmch_manual_panel_fitting(struct intel_crtc *intel_crtc,
+ struct intel_crtc_config *pipe_config) {
+ struct drm_device *dev = intel_crtc->base.dev;
+ u32 pfit_control = 0, border = 0;
+ u32 pf_horizontal_ratio, pf_vertical_ratio;
+ struct drm_display_mode *adjusted_mode;
+ u32 tot_width, tot_height;
+ u32 src_width, src_height; /* pipesrc.x, pipesrc.y */
+ u32 dst_width, dst_height;
+
+ adjusted_mode = &pipe_config->adjusted_mode;
+
+ src_width = pipe_config->pipe_src_w;
+ src_height = pipe_config->pipe_src_h;
+
+ tot_width = adjusted_mode->hdisplay;
+ tot_height = adjusted_mode->vdisplay;
+
+ /*
+ * Having non zero borders will reduce the size of 'HACTIVE/VACTIVE'
+ * region. So (HACTIVE - Left border - Right Border) *
+ * (VACTIVE - Top Border - Bottom border) will effectively be the
+ * output rectangle on screen
+ */
+ dst_width = tot_width -
+ (((intel_crtc->border_size >> 16) & 0xffff) * 2);
+ dst_height = tot_height -
+ ((intel_crtc->border_size & 0xffff) * 2);
+
+ if ((dst_width == 0) || (dst_height == 0)) {
+ DRM_ERROR("Invalid border size input\n");
+ return;
+ }
+
+ pf_horizontal_ratio = panel_fitter_scaling(src_width, dst_width);
+ pf_vertical_ratio = panel_fitter_scaling(src_height, dst_height);
+
+ if (pf_horizontal_ratio > MAX_DOWNSCALE_RATIO) {
+ DRM_ERROR("width is too small\n");
+ return;
+ } else if (pf_vertical_ratio > MAX_DOWNSCALE_RATIO) {
+ DRM_ERROR("height is too small\n");
+ return;
+ }
+
+ if (dst_width != tot_width)
+ centre_horizontally(adjusted_mode, dst_width);
+ if (dst_height != tot_height)
+ centre_vertically(adjusted_mode, dst_height);
+
+ /* No scaling needed now, but still enable the panel fitter,
+ as that will allow the User to subequently do the dynamic
+ flipping of fbs of different resolutions */
+ if (adjusted_mode->crtc_hdisplay == pipe_config->pipe_src_w &&
+ adjusted_mode->crtc_vdisplay == pipe_config->pipe_src_h) {
+ BUG_ON(!intel_crtc->pfit_enabled);
+ DRM_DEBUG_KMS("Forcefully enabling the Panel fitter\n");
+ }
+
+ border = LVDS_BORDER_ENABLE;
+
+ if (INTEL_INFO(dev)->gen >= 4) {
+ /* PFIT_SCALING_PROGRAMMED is de-featured on BYT */
+ pfit_control |= PFIT_ENABLE | PFIT_SCALING_AUTO;
+ pfit_control |= ((intel_crtc->pipe << PFIT_PIPE_SHIFT) | PFIT_FILTER_FUZZY);
+ } else {
+ pfit_control |= (PFIT_ENABLE |
+ VERT_AUTO_SCALE | HORIZ_AUTO_SCALE |
+ VERT_INTERP_BILINEAR | HORIZ_INTERP_BILINEAR);
+ }
+
+ /* Make sure pre-965 set dither correctly for 18bpp panels. */
+ if (INTEL_INFO(dev)->gen < 4 && pipe_config->pipe_bpp == 18)
+ pfit_control |= PANEL_8TO6_DITHER_ENABLE;
+
+ pipe_config->gmch_pfit.control = pfit_control;
+ pipe_config->gmch_pfit.lvds_border_bits = border; }
+
void intel_gmch_panel_fitting(struct intel_crtc *intel_crtc,
struct intel_crtc_config *pipe_config, int fitting_mode)
@@ -257,6 +400,13 @@ void intel_gmch_panel_fitting(struct intel_crtc *intel_crtc,

adjusted_mode = &pipe_config->adjusted_mode;

+ /* check if User wants to apply the borders, or wants to forcefully
+ enable the panel fitter, otherwise fall through the regular path */
+ if (intel_crtc->pfit_enabled ||
+ intel_crtc->border_size)
+ return intel_gmch_manual_panel_fitting(intel_crtc,
+ pipe_config);
+
/* Native modes don't need fitting */
if (adjusted_mode->hdisplay == pipe_config->pipe_src_w &&
adjusted_mode->vdisplay == pipe_config->pipe_src_h)
--
1.8.5.2

#14 lore raw

Re: [PATCH v4 3/3] drm/i915: New drm crtc property for varying the size of borders

Quoted 3 lines · Akash

Hi Ville,

Please could you review this patch.

You need a pass of checkpatch.pl in here. Not sure what happened with the coding style.

Signature · 1 line
Damien
#15 lore raw

Re: [PATCH v4 3/3] drm/i915: New drm crtc property for varying the size of borders

Quoted 7 lines · Damien Lespiau
Quoted 3 lines · Akash

Hi Ville,

Please could you review this patch.

You need a pass of checkpatch.pl in here. Not sure what happened with the coding style.

Ignore this, I was looking at the wrong thing.

Signature · 1 line
Damien
#16 lore raw

Re: [PATCH v4 3/3] drm/i915: New drm crtc property for varying the size of borders

In reply to [email protected]
Quoted 183 lines · On Wed, Mar 26, 2014 at 09:25:12AM +0530, [email protected] wrote:

From: Akash Goel <[email protected]>

This patch adds a new drm crtc property for varying the size of the horizontal & vertical borers of the output/display window. This will control the output of Panel fitter.

v2: Added a new check for the invalid border size input

v3: Fixed bugs in output window calculation
Removed superfluous checks

v4: Added the capability to forecfully enable the Panel fitter. The property value is of 64 bits, first 32 bits are used for border dimensions. The 33rd bit can be used to forcefully enable the panel fitter. This is useful for Panels which do not override the User specified Pipe timings.

Testcase: igt/kms_panel_fitter_test

---
 drivers/gpu/drm/i915/i915_drv.h      |   7 ++
 drivers/gpu/drm/i915/intel_display.c |  39 +++++++-
 drivers/gpu/drm/i915/intel_drv.h     |   5 +
 drivers/gpu/drm/i915/intel_panel.c   | 176 ++++++++++++++++++++++++++++++++---
 4 files changed, 211 insertions(+), 16 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +7 -0
index 6f3af15..eec32ed 100644
@@ -1614,6 +1614,13 @@ typedef struct drm_i915_private {
 	 */
 	struct drm_property *input_size_property;
 
+	/*
+	 * Property to dynamically vary the size of the
+	 * borders. This will indirectly control the size
+	 * of the display window i.e Panel fitter output
+	 */
+	struct drm_property *output_border_property;
+
 	uint32_t hw_context_size;
 	struct list_head context_list;
 
drivers/gpu/drm/i915/intel_display.c +36 -3
index 7149123..a217f25 100644
@@ -10447,7 +10447,23 @@ static int intel_crtc_set_property(struct drm_crtc *crtc,
 	struct drm_device *dev = crtc->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
-	int ret = -ENOENT;
+
+	if (property == dev_priv->output_border_property) {
+		if ((val == (uint64_t)intel_crtc->border_size) &&
+		    (((val >> 32) & 0x1) == intel_crtc->pfit_enabled))
+			return 0;
+
+		intel_crtc->border_size = (uint32_t)val;
+		intel_crtc->pfit_enabled = (val >> 32) & 0x1;
+		if ((intel_crtc->border_size != 0) &&
+		    (!intel_crtc->pfit_enabled)) {
+			DRM_ERROR("Wrong setting, expect Pfit to be enabled when "
+				  "applying borders\n");
+			return -EINVAL;
+		}
+
+		goto done;
+	}
 
 	if (property == dev_priv->input_size_property) {
 		int new_width = (int)((val >> 16) & 0xffff);
@@ -10488,7 +10504,13 @@ static int intel_crtc_set_property(struct drm_crtc *crtc,
 		return 0;
 	}
 
-	return ret;
+	return -EINVAL;
+
+done:
+	if (crtc)
+		intel_crtc_restore_mode(crtc);
+
+	return 0;
 }
 
 static const struct drm_crtc_funcs intel_crtc_funcs = {
@@ -10641,12 +10663,23 @@ static void intel_crtc_init(struct drm_device *dev, int pipe)
 
 	if (!dev_priv->input_size_property)
 		dev_priv->input_size_property =
-			drm_property_create_range(dev, 0, "input size", 0, 0xFFFFFFFF);
+			drm_property_create_range(dev, 0, "input size",
+						0, 0xFFFFFFFF);
 
 	if (dev_priv->input_size_property)
 		drm_object_attach_property(&intel_crtc->base.base,
 					   dev_priv->input_size_property,
 					   0);
+
+	if (!dev_priv->output_border_property)
+		dev_priv->output_border_property =
+			drm_property_create_range(dev, 0, "border size",
+						0, (uint64_t)0x1FFFFFFFF);
+
+	if (dev_priv->output_border_property)
+		drm_object_attach_property(&intel_crtc->base.base,
+					   dev_priv->output_border_property,
+					   0);
 }
 
 enum pipe intel_get_pipe_from_connector(struct intel_connector *connector)
drivers/gpu/drm/i915/intel_drv.h +5 -0
index 5360d16..3cfc9da 100644
@@ -387,6 +387,11 @@ struct intel_crtc {
 	bool cpu_fifo_underrun_disabled;
 	bool pch_fifo_underrun_disabled;
 
+	/* for forceful enabling of panel fitter */
+	bool pfit_enabled;
+	/* border info for the output/display window */
+	uint32_t border_size;
+
 	/* per-pipe watermark state */
 	struct {
 		/* watermarks currently being used  */
drivers/gpu/drm/i915/intel_panel.c +34 -0
index cb05840..86a486e 100644
@@ -29,10 +29,25 @@
  */
 
 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+/* Max Downscale ratio of 1.125, expressed in 1.12 fixed point format */
+#define MAX_DOWNSCALE_RATIO  (0x9 << 9)
 
 #include <linux/moduleparam.h>
 #include "intel_drv.h"
 
+static inline u32 panel_fitter_scaling(u32 source, u32 target)
+{
+	/*
+	 * Floating point operation is not supported. So the FACTOR
+	 * is defined, which can avoid the floating point computation
+	 * when calculating the panel ratio.
+	 */
+#define ACCURACY 12
+#define FACTOR (1 << ACCURACY)
+	u32 ratio = source * FACTOR / target;
+	return (FACTOR * ratio + FACTOR/2) / FACTOR;
+}
+
 void
 intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode,
 		       struct drm_display_mode *adjusted_mode)
@@ -42,6 +57,60 @@ intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode,
 	drm_mode_set_crtcinfo(adjusted_mode, 0);
 }
 
+void
+intel_pch_manual_panel_fitting(struct intel_crtc *intel_crtc,
+			struct intel_crtc_config *pipe_config)
+{
+	struct drm_display_mode *adjusted_mode;
+	int x, y;
+	u32 pf_horizontal_ratio, pf_vertical_ratio;
+	u32 tot_width, tot_height;
+	u32 src_width, src_height; /* pipesrc.x, pipesrc.y */
+	u32 dst_width, dst_height;
+
+	adjusted_mode = &pipe_config->adjusted_mode;
+
+	src_width = pipe_config->pipe_src_w;
+	src_height = pipe_config->pipe_src_h;
+
+	tot_width  = adjusted_mode->hdisplay;
+	tot_height = adjusted_mode->vdisplay;
+

drivers/gpu/drm/i915/intel_panel.c intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode, · line 88

Quoted 10 lines
drivers/gpu/drm/i915/intel_panel.c +10 -0
+	/*
+	 * Having non zero borders will reduce the size of 'HACTIVE/VACTIVE'
+	 * region. So (HACTIVE - Left border - Right Border) *
+	 * (VACTIVE  - Top Border  - Bottom border) will effectively be the
+	 * output rectangle on screen
+	 */
+	dst_width = tot_width -
+			(((intel_crtc->border_size >> 16) & 0xffff) * 2);
+	dst_height = tot_height -
+			((intel_crtc->border_size & 0xffff) * 2);

I'm thinking that we should allow the user to specify each border width individually rather than just assume left==right and top==bottom.

drivers/gpu/drm/i915/intel_panel.c intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode, · line 92

Quoted 4 lines
drivers/gpu/drm/i915/intel_panel.c +4 -0
+
+	if ((dst_width == 0) || (dst_height == 0)) {
+		DRM_ERROR("Invalid border size input\n");
+		return;

This is clear sign here that we should do all this stuff in the compute config stage so that we can fail gracefully and tell userspace that things didn't work out.

Quoted 166 lines
drivers/gpu/drm/i915/intel_panel.c +115 -13
+	}
+
+	pf_horizontal_ratio = panel_fitter_scaling(src_width, dst_width);
+	pf_vertical_ratio   = panel_fitter_scaling(src_height, dst_height);
+
+	if (pf_horizontal_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("width is too small\n");
+		return;
+	} else if (pf_vertical_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("height is too small\n");
+		return;
+	}
+
+	x = (adjusted_mode->hdisplay - dst_width + 1)/2;
+	y = (adjusted_mode->vdisplay - dst_height + 1)/2;
+
+	pipe_config->pch_pfit.pos = (x << 16) | y;
+	pipe_config->pch_pfit.size = (dst_width << 16) | dst_height;
+	pipe_config->pch_pfit.enabled = pipe_config->pch_pfit.size != 0;
+}
+
 /* adjusted_mode has been preset to be the panel's fixed mode */
 void
 intel_pch_panel_fitting(struct intel_crtc *intel_crtc,
@@ -55,6 +124,13 @@ intel_pch_panel_fitting(struct intel_crtc *intel_crtc,
 
 	x = y = width = height = 0;
 
+	/* check if User wants to apply the borders, or wants to forcefully
+	   enable the panel fitter, otherwise fall through the regular path */
+	if (intel_crtc->pfit_enabled ||
+			intel_crtc->border_size)
+		return intel_pch_manual_panel_fitting(intel_crtc,
+						      pipe_config);
+
 	/* Native modes don't need fitting */
 	if (adjusted_mode->hdisplay == pipe_config->pipe_src_w &&
 	    adjusted_mode->vdisplay == pipe_config->pipe_src_h)
@@ -157,19 +233,6 @@ centre_vertically(struct drm_display_mode *mode,
 	mode->crtc_vsync_end = mode->crtc_vsync_start + sync_width;
 }
 
-static inline u32 panel_fitter_scaling(u32 source, u32 target)
-{
-	/*
-	 * Floating point operation is not supported. So the FACTOR
-	 * is defined, which can avoid the floating point computation
-	 * when calculating the panel ratio.
-	 */
-#define ACCURACY 12
-#define FACTOR (1 << ACCURACY)
-	u32 ratio = source * FACTOR / target;
-	return (FACTOR * ratio + FACTOR/2) / FACTOR;
-}
-
 static void i965_scale_aspect(struct intel_crtc_config *pipe_config,
 			      u32 *pfit_control)
 {
@@ -247,6 +310,86 @@ static void i9xx_scale_aspect(struct intel_crtc_config *pipe_config,
 	}
 }
 
+void intel_gmch_manual_panel_fitting(struct intel_crtc *intel_crtc,
+				     struct intel_crtc_config *pipe_config)
+{
+	struct drm_device *dev = intel_crtc->base.dev;
+	u32 pfit_control = 0, border = 0;
+	u32 pf_horizontal_ratio, pf_vertical_ratio;
+	struct drm_display_mode *adjusted_mode;
+	u32 tot_width, tot_height;
+	u32 src_width, src_height; /* pipesrc.x, pipesrc.y */
+	u32 dst_width, dst_height;
+
+	adjusted_mode = &pipe_config->adjusted_mode;
+
+	src_width = pipe_config->pipe_src_w;
+	src_height = pipe_config->pipe_src_h;
+
+	tot_width = adjusted_mode->hdisplay;
+	tot_height = adjusted_mode->vdisplay;
+
+	/*
+	 * Having non zero borders will reduce the size of 'HACTIVE/VACTIVE'
+	 * region. So (HACTIVE - Left border - Right Border) *
+	 * (VACTIVE  - Top Border  - Bottom border) will effectively be the
+	 * output rectangle on screen
+	 */
+	dst_width = tot_width -
+			(((intel_crtc->border_size >> 16) & 0xffff) * 2);
+	dst_height = tot_height -
+			((intel_crtc->border_size & 0xffff) * 2);
+
+	if ((dst_width == 0) || (dst_height == 0)) {
+		DRM_ERROR("Invalid border size input\n");
+		return;
+	}
+
+	pf_horizontal_ratio = panel_fitter_scaling(src_width, dst_width);
+	pf_vertical_ratio   = panel_fitter_scaling(src_height, dst_height);
+
+	if (pf_horizontal_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("width is too small\n");
+		return;
+	} else if (pf_vertical_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("height is too small\n");
+		return;
+	}
+
+	if (dst_width != tot_width)
+		centre_horizontally(adjusted_mode, dst_width);
+	if (dst_height != tot_height)
+		centre_vertically(adjusted_mode, dst_height);
+
+	/* No scaling needed now, but still enable the panel fitter,
+	   as that will allow the User to subequently do the dynamic
+	   flipping of fbs of different resolutions */
+	if (adjusted_mode->crtc_hdisplay == pipe_config->pipe_src_w &&
+	    adjusted_mode->crtc_vdisplay == pipe_config->pipe_src_h) {
+		BUG_ON(!intel_crtc->pfit_enabled);
+		DRM_DEBUG_KMS("Forcefully enabling the Panel fitter\n");
+	}
+
+	border = LVDS_BORDER_ENABLE;
+
+	if (INTEL_INFO(dev)->gen >= 4) {
+		/* PFIT_SCALING_PROGRAMMED is de-featured on BYT */
+		pfit_control |= PFIT_ENABLE | PFIT_SCALING_AUTO;
+		pfit_control |= ((intel_crtc->pipe << PFIT_PIPE_SHIFT) | PFIT_FILTER_FUZZY);
+	} else {
+		pfit_control |= (PFIT_ENABLE |
+				 VERT_AUTO_SCALE | HORIZ_AUTO_SCALE |
+				 VERT_INTERP_BILINEAR | HORIZ_INTERP_BILINEAR);
+	}
+
+	/* Make sure pre-965 set dither correctly for 18bpp panels. */
+	if (INTEL_INFO(dev)->gen < 4 && pipe_config->pipe_bpp == 18)
+		pfit_control |= PANEL_8TO6_DITHER_ENABLE;
+
+	pipe_config->gmch_pfit.control = pfit_control;
+	pipe_config->gmch_pfit.lvds_border_bits = border;
+}
+
 void intel_gmch_panel_fitting(struct intel_crtc *intel_crtc,
 			      struct intel_crtc_config *pipe_config,
 			      int fitting_mode)
@@ -257,6 +400,13 @@ void intel_gmch_panel_fitting(struct intel_crtc *intel_crtc,
 
 	adjusted_mode = &pipe_config->adjusted_mode;
 
+	/* check if User wants to apply the borders, or wants to forcefully
+	   enable the panel fitter, otherwise fall through the regular path */
+	if (intel_crtc->pfit_enabled ||
+			intel_crtc->border_size)
+		return intel_gmch_manual_panel_fitting(intel_crtc,
+						       pipe_config);
+
 	/* Native modes don't need fitting */
 	if (adjusted_mode->hdisplay == pipe_config->pipe_src_w &&
 	    adjusted_mode->vdisplay == pipe_config->pipe_src_h)
Signature · 6 lines
1.8.5.2

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 2 lines
Ville Syrjälä
Intel OTC
author
#17 lore raw

Re: [PATCH v4 3/3] drm/i915: New drm crtc property for varying the size of borders

Quoted 189 lines · Ville Syrjälä
Quoted 188 lines · On Wed, Mar 26, 2014 at 09:25:12AM +0530, [email protected] wrote:

From: Akash Goel <[email protected]>

This patch adds a new drm crtc property for varying the size of the horizontal & vertical borers of the output/display window. This will control the output of Panel fitter.

v2: Added a new check for the invalid border size input

v3: Fixed bugs in output window calculation
Removed superfluous checks

v4: Added the capability to forecfully enable the Panel fitter. The property value is of 64 bits, first 32 bits are used for border dimensions. The 33rd bit can be used to forcefully enable the panel fitter. This is useful for Panels which do not override the User specified Pipe timings.

Testcase: igt/kms_panel_fitter_test

---
 drivers/gpu/drm/i915/i915_drv.h      |   7 ++
 drivers/gpu/drm/i915/intel_display.c |  39 +++++++-
 drivers/gpu/drm/i915/intel_drv.h     |   5 +
 drivers/gpu/drm/i915/intel_panel.c   | 176 ++++++++++++++++++++++++++++++++---
 4 files changed, 211 insertions(+), 16 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +7 -0
index 6f3af15..eec32ed 100644
@@ -1614,6 +1614,13 @@ typedef struct drm_i915_private {
 	 */
 	struct drm_property *input_size_property;
 
+	/*
+	 * Property to dynamically vary the size of the
+	 * borders. This will indirectly control the size
+	 * of the display window i.e Panel fitter output
+	 */
+	struct drm_property *output_border_property;
+
 	uint32_t hw_context_size;
 	struct list_head context_list;
 
drivers/gpu/drm/i915/intel_display.c +36 -3
index 7149123..a217f25 100644
@@ -10447,7 +10447,23 @@ static int intel_crtc_set_property(struct drm_crtc *crtc,
 	struct drm_device *dev = crtc->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_crtc *intel_crtc = to_intel_crtc(crtc);
-	int ret = -ENOENT;
+
+	if (property == dev_priv->output_border_property) {
+		if ((val == (uint64_t)intel_crtc->border_size) &&
+		    (((val >> 32) & 0x1) == intel_crtc->pfit_enabled))
+			return 0;
+
+		intel_crtc->border_size = (uint32_t)val;
+		intel_crtc->pfit_enabled = (val >> 32) & 0x1;
+		if ((intel_crtc->border_size != 0) &&
+		    (!intel_crtc->pfit_enabled)) {
+			DRM_ERROR("Wrong setting, expect Pfit to be enabled when "
+				  "applying borders\n");
+			return -EINVAL;
+		}
+
+		goto done;
+	}
 
 	if (property == dev_priv->input_size_property) {
 		int new_width = (int)((val >> 16) & 0xffff);
@@ -10488,7 +10504,13 @@ static int intel_crtc_set_property(struct drm_crtc *crtc,
 		return 0;
 	}
 
-	return ret;
+	return -EINVAL;
+
+done:
+	if (crtc)
+		intel_crtc_restore_mode(crtc);
+
+	return 0;
 }
 
 static const struct drm_crtc_funcs intel_crtc_funcs = {
@@ -10641,12 +10663,23 @@ static void intel_crtc_init(struct drm_device *dev, int pipe)
 
 	if (!dev_priv->input_size_property)
 		dev_priv->input_size_property =
-			drm_property_create_range(dev, 0, "input size", 0, 0xFFFFFFFF);
+			drm_property_create_range(dev, 0, "input size",
+						0, 0xFFFFFFFF);
 
 	if (dev_priv->input_size_property)
 		drm_object_attach_property(&intel_crtc->base.base,
 					   dev_priv->input_size_property,
 					   0);
+
+	if (!dev_priv->output_border_property)
+		dev_priv->output_border_property =
+			drm_property_create_range(dev, 0, "border size",
+						0, (uint64_t)0x1FFFFFFFF);
+
+	if (dev_priv->output_border_property)
+		drm_object_attach_property(&intel_crtc->base.base,
+					   dev_priv->output_border_property,
+					   0);
 }
 
 enum pipe intel_get_pipe_from_connector(struct intel_connector *connector)
drivers/gpu/drm/i915/intel_drv.h +5 -0
index 5360d16..3cfc9da 100644
@@ -387,6 +387,11 @@ struct intel_crtc {
 	bool cpu_fifo_underrun_disabled;
 	bool pch_fifo_underrun_disabled;
 
+	/* for forceful enabling of panel fitter */
+	bool pfit_enabled;
+	/* border info for the output/display window */
+	uint32_t border_size;
+
 	/* per-pipe watermark state */
 	struct {
 		/* watermarks currently being used  */
drivers/gpu/drm/i915/intel_panel.c +39 -0
index cb05840..86a486e 100644
@@ -29,10 +29,25 @@
  */
 
 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+/* Max Downscale ratio of 1.125, expressed in 1.12 fixed point format */
+#define MAX_DOWNSCALE_RATIO  (0x9 << 9)
 
 #include <linux/moduleparam.h>
 #include "intel_drv.h"
 
+static inline u32 panel_fitter_scaling(u32 source, u32 target)
+{
+	/*
+	 * Floating point operation is not supported. So the FACTOR
+	 * is defined, which can avoid the floating point computation
+	 * when calculating the panel ratio.
+	 */
+#define ACCURACY 12
+#define FACTOR (1 << ACCURACY)
+	u32 ratio = source * FACTOR / target;
+	return (FACTOR * ratio + FACTOR/2) / FACTOR;
+}
+
 void
 intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode,
 		       struct drm_display_mode *adjusted_mode)
@@ -42,6 +57,60 @@ intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode,
 	drm_mode_set_crtcinfo(adjusted_mode, 0);
 }
 
+void
+intel_pch_manual_panel_fitting(struct intel_crtc *intel_crtc,
+			struct intel_crtc_config *pipe_config)
+{
+	struct drm_display_mode *adjusted_mode;
+	int x, y;
+	u32 pf_horizontal_ratio, pf_vertical_ratio;
+	u32 tot_width, tot_height;
+	u32 src_width, src_height; /* pipesrc.x, pipesrc.y */
+	u32 dst_width, dst_height;
+
+	adjusted_mode = &pipe_config->adjusted_mode;
+
+	src_width = pipe_config->pipe_src_w;
+	src_height = pipe_config->pipe_src_h;
+
+	tot_width  = adjusted_mode->hdisplay;
+	tot_height = adjusted_mode->vdisplay;
+
+	/*
+	 * Having non zero borders will reduce the size of 'HACTIVE/VACTIVE'
+	 * region. So (HACTIVE - Left border - Right Border) *
+	 * (VACTIVE  - Top Border  - Bottom border) will effectively be the
+	 * output rectangle on screen
Quoted 9 lines

drivers/gpu/drm/i915/intel_panel.c intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode, · line 88

Quoted 5 lines
drivers/gpu/drm/i915/intel_panel.c +5 -0
+	 */
+	dst_width = tot_width -
+			(((intel_crtc->border_size >> 16) & 0xffff) * 2);
+	dst_height = tot_height -
+			((intel_crtc->border_size & 0xffff) * 2);

I'm thinking that we should allow the user to specify each border width individually rather than just assume left==right and top==bottom.

Sorry I thought that the Top/Bottom & left/Right borders would be symmetric only.
Tried setting the borders on EDP & HDMI panels by manipulating the Pipe timings (through the logic used in 'centre_horizontally' & 'centre_vertically' functions), but it didn't work.
Is this logic effective for the LVDS panel only ?

Quoted 9 lines

drivers/gpu/drm/i915/intel_panel.c intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode, · line 92

Quoted 4 lines
drivers/gpu/drm/i915/intel_panel.c +4 -0
+
+	if ((dst_width == 0) || (dst_height == 0)) {
+		DRM_ERROR("Invalid border size input\n");
+		return;

This is clear sign here that we should do all this stuff in the compute config stage so that we can fail gracefully and tell userspace that things didn't work out.

Actually this call to decide the panel fitter config, is being made in the context of 'compute config' only. But it's originating from the 'crtc set property' & not from the modeset.
Do you mean to say that if done from the 'modeset' call, we can report back the error to User space for an invalid border setting. Actually currently the return type is 'void' only for the 2 existing panel fitter config functions.
Also we don't have a check in place for the max supported upscaling ratio.

Quoted 162 lines
Quoted 162 lines
drivers/gpu/drm/i915/intel_panel.c +115 -13
+	}
+
+	pf_horizontal_ratio = panel_fitter_scaling(src_width, dst_width);
+	pf_vertical_ratio   = panel_fitter_scaling(src_height, dst_height);
+
+	if (pf_horizontal_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("width is too small\n");
+		return;
+	} else if (pf_vertical_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("height is too small\n");
+		return;
+	}
+
+	x = (adjusted_mode->hdisplay - dst_width + 1)/2;
+	y = (adjusted_mode->vdisplay - dst_height + 1)/2;
+
+	pipe_config->pch_pfit.pos = (x << 16) | y;
+	pipe_config->pch_pfit.size = (dst_width << 16) | dst_height;
+	pipe_config->pch_pfit.enabled = pipe_config->pch_pfit.size != 0;
+}
+
 /* adjusted_mode has been preset to be the panel's fixed mode */
 void
 intel_pch_panel_fitting(struct intel_crtc *intel_crtc,
@@ -55,6 +124,13 @@ intel_pch_panel_fitting(struct intel_crtc *intel_crtc,
 
 	x = y = width = height = 0;
 
+	/* check if User wants to apply the borders, or wants to forcefully
+	   enable the panel fitter, otherwise fall through the regular path */
+	if (intel_crtc->pfit_enabled ||
+			intel_crtc->border_size)
+		return intel_pch_manual_panel_fitting(intel_crtc,
+						      pipe_config);
+
 	/* Native modes don't need fitting */
 	if (adjusted_mode->hdisplay == pipe_config->pipe_src_w &&
 	    adjusted_mode->vdisplay == pipe_config->pipe_src_h)
@@ -157,19 +233,6 @@ centre_vertically(struct drm_display_mode *mode,
 	mode->crtc_vsync_end = mode->crtc_vsync_start + sync_width;
 }
 
-static inline u32 panel_fitter_scaling(u32 source, u32 target)
-{
-	/*
-	 * Floating point operation is not supported. So the FACTOR
-	 * is defined, which can avoid the floating point computation
-	 * when calculating the panel ratio.
-	 */
-#define ACCURACY 12
-#define FACTOR (1 << ACCURACY)
-	u32 ratio = source * FACTOR / target;
-	return (FACTOR * ratio + FACTOR/2) / FACTOR;
-}
-
 static void i965_scale_aspect(struct intel_crtc_config *pipe_config,
 			      u32 *pfit_control)
 {
@@ -247,6 +310,86 @@ static void i9xx_scale_aspect(struct intel_crtc_config *pipe_config,
 	}
 }
 
+void intel_gmch_manual_panel_fitting(struct intel_crtc *intel_crtc,
+				     struct intel_crtc_config *pipe_config)
+{
+	struct drm_device *dev = intel_crtc->base.dev;
+	u32 pfit_control = 0, border = 0;
+	u32 pf_horizontal_ratio, pf_vertical_ratio;
+	struct drm_display_mode *adjusted_mode;
+	u32 tot_width, tot_height;
+	u32 src_width, src_height; /* pipesrc.x, pipesrc.y */
+	u32 dst_width, dst_height;
+
+	adjusted_mode = &pipe_config->adjusted_mode;
+
+	src_width = pipe_config->pipe_src_w;
+	src_height = pipe_config->pipe_src_h;
+
+	tot_width = adjusted_mode->hdisplay;
+	tot_height = adjusted_mode->vdisplay;
+
+	/*
+	 * Having non zero borders will reduce the size of 'HACTIVE/VACTIVE'
+	 * region. So (HACTIVE - Left border - Right Border) *
+	 * (VACTIVE  - Top Border  - Bottom border) will effectively be the
+	 * output rectangle on screen
+	 */
+	dst_width = tot_width -
+			(((intel_crtc->border_size >> 16) & 0xffff) * 2);
+	dst_height = tot_height -
+			((intel_crtc->border_size & 0xffff) * 2);
+
+	if ((dst_width == 0) || (dst_height == 0)) {
+		DRM_ERROR("Invalid border size input\n");
+		return;
+	}
+
+	pf_horizontal_ratio = panel_fitter_scaling(src_width, dst_width);
+	pf_vertical_ratio   = panel_fitter_scaling(src_height, dst_height);
+
+	if (pf_horizontal_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("width is too small\n");
+		return;
+	} else if (pf_vertical_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("height is too small\n");
+		return;
+	}
+
+	if (dst_width != tot_width)
+		centre_horizontally(adjusted_mode, dst_width);
+	if (dst_height != tot_height)
+		centre_vertically(adjusted_mode, dst_height);
+
+	/* No scaling needed now, but still enable the panel fitter,
+	   as that will allow the User to subequently do the dynamic
+	   flipping of fbs of different resolutions */
+	if (adjusted_mode->crtc_hdisplay == pipe_config->pipe_src_w &&
+	    adjusted_mode->crtc_vdisplay == pipe_config->pipe_src_h) {
+		BUG_ON(!intel_crtc->pfit_enabled);
+		DRM_DEBUG_KMS("Forcefully enabling the Panel fitter\n");
+	}
+
+	border = LVDS_BORDER_ENABLE;
+
+	if (INTEL_INFO(dev)->gen >= 4) {
+		/* PFIT_SCALING_PROGRAMMED is de-featured on BYT */
+		pfit_control |= PFIT_ENABLE | PFIT_SCALING_AUTO;
+		pfit_control |= ((intel_crtc->pipe << PFIT_PIPE_SHIFT) | PFIT_FILTER_FUZZY);
+	} else {
+		pfit_control |= (PFIT_ENABLE |
+				 VERT_AUTO_SCALE | HORIZ_AUTO_SCALE |
+				 VERT_INTERP_BILINEAR | HORIZ_INTERP_BILINEAR);
+	}
+
+	/* Make sure pre-965 set dither correctly for 18bpp panels. */
+	if (INTEL_INFO(dev)->gen < 4 && pipe_config->pipe_bpp == 18)
+		pfit_control |= PANEL_8TO6_DITHER_ENABLE;
+
+	pipe_config->gmch_pfit.control = pfit_control;
+	pipe_config->gmch_pfit.lvds_border_bits = border;
+}
+
 void intel_gmch_panel_fitting(struct intel_crtc *intel_crtc,
 			      struct intel_crtc_config *pipe_config,
 			      int fitting_mode)
@@ -257,6 +400,13 @@ void intel_gmch_panel_fitting(struct intel_crtc *intel_crtc,
 
 	adjusted_mode = &pipe_config->adjusted_mode;
 
+	/* check if User wants to apply the borders, or wants to forcefully
+	   enable the panel fitter, otherwise fall through the regular path */
+	if (intel_crtc->pfit_enabled ||
+			intel_crtc->border_size)
+		return intel_gmch_manual_panel_fitting(intel_crtc,
+						       pipe_config);
+
 	/* Native modes don't need fitting */
 	if (adjusted_mode->hdisplay == pipe_config->pipe_src_w &&
 	    adjusted_mode->vdisplay == pipe_config->pipe_src_h)
Signature · 1 line
1.8.5.2
Quoted 5 lines
Quoted 4 lines

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#18 lore raw

Re: [PATCH v4 3/3] drm/i915: New drm crtc property for varying the size of borders

Quoted 29 lines · Akash Goel
Quoted 28 lines · Ville Syrjälä
Quoted 27 lines · On Wed, Mar 26, 2014 at 09:25:12AM +0530, [email protected] wrote:

From: Akash Goel <[email protected]>

This patch adds a new drm crtc property for varying the size of the horizontal & vertical borers of the output/display window. This will control the output of Panel fitter.

v2: Added a new check for the invalid border size input

v3: Fixed bugs in output window calculation
Removed superfluous checks

v4: Added the capability to forecfully enable the Panel fitter. The property value is of 64 bits, first 32 bits are used for border dimensions. The 33rd bit can be used to forcefully enable the panel fitter. This is useful for Panels which do not override the User specified Pipe timings.

Testcase: igt/kms_panel_fitter_test

---
 drivers/gpu/drm/i915/i915_drv.h      |   7 ++
 drivers/gpu/drm/i915/intel_display.c |  39 +++++++-
 drivers/gpu/drm/i915/intel_drv.h     |   5 +
 drivers/gpu/drm/i915/intel_panel.c   | 176 ++++++++++++++++++++++++++++++++---
 4 files changed, 211 insertions(+), 16 deletions(-)

<snip>

Quoted 38 lines
Quoted 28 lines
Quoted 28 lines
drivers/gpu/drm/i915/intel_panel.c +24 -0
@@ -42,6 +57,60 @@ intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode,
 	drm_mode_set_crtcinfo(adjusted_mode, 0);
 }
 
+void
+intel_pch_manual_panel_fitting(struct intel_crtc *intel_crtc,
+			struct intel_crtc_config *pipe_config)
+{
+	struct drm_display_mode *adjusted_mode;
+	int x, y;
+	u32 pf_horizontal_ratio, pf_vertical_ratio;
+	u32 tot_width, tot_height;
+	u32 src_width, src_height; /* pipesrc.x, pipesrc.y */
+	u32 dst_width, dst_height;
+
+	adjusted_mode = &pipe_config->adjusted_mode;
+
+	src_width = pipe_config->pipe_src_w;
+	src_height = pipe_config->pipe_src_h;
+
+	tot_width  = adjusted_mode->hdisplay;
+	tot_height = adjusted_mode->vdisplay;
+
+	/*
+	 * Having non zero borders will reduce the size of 'HACTIVE/VACTIVE'
+	 * region. So (HACTIVE - Left border - Right Border) *
+	 * (VACTIVE  - Top Border  - Bottom border) will effectively be the
+	 * output rectangle on screen
Quoted 9 lines

drivers/gpu/drm/i915/intel_panel.c intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode, · line 88

Quoted 5 lines
drivers/gpu/drm/i915/intel_panel.c +5 -0
+	 */
+	dst_width = tot_width -
+			(((intel_crtc->border_size >> 16) & 0xffff) * 2);
+	dst_height = tot_height -
+			((intel_crtc->border_size & 0xffff) * 2);

I'm thinking that we should allow the user to specify each border width individually rather than just assume left==right and top==bottom.

Quoted 2 lines

Sorry I thought that the Top/Bottom & left/Right borders would be symmetric only.

I don't see a reason to limit ourselves here.

Quoted 4 lines

Tried setting the borders on EDP & HDMI panels by manipulating the Pipe timings (through the logic used in 'centre_horizontally' & 'centre_vertically' functions), but it didn't work.
Is this logic effective for the LVDS panel only ?

Could be. Certainly the border enable bit is there only for LVDS. The gmch panel fitter isn't very flexible so it's possible we can't actually make it do many of the things the pch pfit can do.

What happens if we set up the pfit to use manual scaling ratios but configure both scaling ratios so that scaled image won't fill the active video region in either direction? Does it position the scaled image at coordinates 0,0 and simply scan black the rest of the time after it's run out of source pixel data? Or does it automagically center the image and scan black on both sides? Or does it fail in some way?

Quoted 11 lines
Quoted 10 lines

drivers/gpu/drm/i915/intel_panel.c intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode, · line 92

Quoted 5 lines
drivers/gpu/drm/i915/intel_panel.c +4 -0
+
+	if ((dst_width == 0) || (dst_height == 0)) {
+		DRM_ERROR("Invalid border size input\n");
+		return;

This is clear sign here that we should do all this stuff in the compute config stage so that we can fail gracefully and tell userspace that things didn't work out.

Quoted 3 lines

Actually this call to decide the panel fitter config, is being made in the context of 'compute config' only. But it's originating from the 'crtc set property' & not from the modeset.

We need to check things in both cases, and return an error if things can't work out.

Quoted 6 lines

Do you mean to say that if done from the 'modeset' call, we can report back the error to User space for an invalid border setting. Actually currently the return type is 'void' only for the 2 existing panel fitter config functions.
Also we don't have a check in place for the max supported upscaling ratio.

Is there an upscaling limit? I know there's a downscaling limit of IIRC 1.125x or something close to that. I don't think we check that either.

Quoted 169 lines
Quoted 168 lines
Quoted 167 lines
drivers/gpu/drm/i915/intel_panel.c +115 -13
+	}
+
+	pf_horizontal_ratio = panel_fitter_scaling(src_width, dst_width);
+	pf_vertical_ratio   = panel_fitter_scaling(src_height, dst_height);
+
+	if (pf_horizontal_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("width is too small\n");
+		return;
+	} else if (pf_vertical_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("height is too small\n");
+		return;
+	}
+
+	x = (adjusted_mode->hdisplay - dst_width + 1)/2;
+	y = (adjusted_mode->vdisplay - dst_height + 1)/2;
+
+	pipe_config->pch_pfit.pos = (x << 16) | y;
+	pipe_config->pch_pfit.size = (dst_width << 16) | dst_height;
+	pipe_config->pch_pfit.enabled = pipe_config->pch_pfit.size != 0;
+}
+
 /* adjusted_mode has been preset to be the panel's fixed mode */
 void
 intel_pch_panel_fitting(struct intel_crtc *intel_crtc,
@@ -55,6 +124,13 @@ intel_pch_panel_fitting(struct intel_crtc *intel_crtc,
 
 	x = y = width = height = 0;
 
+	/* check if User wants to apply the borders, or wants to forcefully
+	   enable the panel fitter, otherwise fall through the regular path */
+	if (intel_crtc->pfit_enabled ||
+			intel_crtc->border_size)
+		return intel_pch_manual_panel_fitting(intel_crtc,
+						      pipe_config);
+
 	/* Native modes don't need fitting */
 	if (adjusted_mode->hdisplay == pipe_config->pipe_src_w &&
 	    adjusted_mode->vdisplay == pipe_config->pipe_src_h)
@@ -157,19 +233,6 @@ centre_vertically(struct drm_display_mode *mode,
 	mode->crtc_vsync_end = mode->crtc_vsync_start + sync_width;
 }
 
-static inline u32 panel_fitter_scaling(u32 source, u32 target)
-{
-	/*
-	 * Floating point operation is not supported. So the FACTOR
-	 * is defined, which can avoid the floating point computation
-	 * when calculating the panel ratio.
-	 */
-#define ACCURACY 12
-#define FACTOR (1 << ACCURACY)
-	u32 ratio = source * FACTOR / target;
-	return (FACTOR * ratio + FACTOR/2) / FACTOR;
-}
-
 static void i965_scale_aspect(struct intel_crtc_config *pipe_config,
 			      u32 *pfit_control)
 {
@@ -247,6 +310,86 @@ static void i9xx_scale_aspect(struct intel_crtc_config *pipe_config,
 	}
 }
 
+void intel_gmch_manual_panel_fitting(struct intel_crtc *intel_crtc,
+				     struct intel_crtc_config *pipe_config)
+{
+	struct drm_device *dev = intel_crtc->base.dev;
+	u32 pfit_control = 0, border = 0;
+	u32 pf_horizontal_ratio, pf_vertical_ratio;
+	struct drm_display_mode *adjusted_mode;
+	u32 tot_width, tot_height;
+	u32 src_width, src_height; /* pipesrc.x, pipesrc.y */
+	u32 dst_width, dst_height;
+
+	adjusted_mode = &pipe_config->adjusted_mode;
+
+	src_width = pipe_config->pipe_src_w;
+	src_height = pipe_config->pipe_src_h;
+
+	tot_width = adjusted_mode->hdisplay;
+	tot_height = adjusted_mode->vdisplay;
+
+	/*
+	 * Having non zero borders will reduce the size of 'HACTIVE/VACTIVE'
+	 * region. So (HACTIVE - Left border - Right Border) *
+	 * (VACTIVE  - Top Border  - Bottom border) will effectively be the
+	 * output rectangle on screen
+	 */
+	dst_width = tot_width -
+			(((intel_crtc->border_size >> 16) & 0xffff) * 2);
+	dst_height = tot_height -
+			((intel_crtc->border_size & 0xffff) * 2);
+
+	if ((dst_width == 0) || (dst_height == 0)) {
+		DRM_ERROR("Invalid border size input\n");
+		return;
+	}
+
+	pf_horizontal_ratio = panel_fitter_scaling(src_width, dst_width);
+	pf_vertical_ratio   = panel_fitter_scaling(src_height, dst_height);
+
+	if (pf_horizontal_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("width is too small\n");
+		return;
+	} else if (pf_vertical_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("height is too small\n");
+		return;
+	}
+
+	if (dst_width != tot_width)
+		centre_horizontally(adjusted_mode, dst_width);
+	if (dst_height != tot_height)
+		centre_vertically(adjusted_mode, dst_height);
+
+	/* No scaling needed now, but still enable the panel fitter,
+	   as that will allow the User to subequently do the dynamic
+	   flipping of fbs of different resolutions */
+	if (adjusted_mode->crtc_hdisplay == pipe_config->pipe_src_w &&
+	    adjusted_mode->crtc_vdisplay == pipe_config->pipe_src_h) {
+		BUG_ON(!intel_crtc->pfit_enabled);
+		DRM_DEBUG_KMS("Forcefully enabling the Panel fitter\n");
+	}
+
+	border = LVDS_BORDER_ENABLE;
+
+	if (INTEL_INFO(dev)->gen >= 4) {
+		/* PFIT_SCALING_PROGRAMMED is de-featured on BYT */
+		pfit_control |= PFIT_ENABLE | PFIT_SCALING_AUTO;
+		pfit_control |= ((intel_crtc->pipe << PFIT_PIPE_SHIFT) | PFIT_FILTER_FUZZY);
+	} else {
+		pfit_control |= (PFIT_ENABLE |
+				 VERT_AUTO_SCALE | HORIZ_AUTO_SCALE |
+				 VERT_INTERP_BILINEAR | HORIZ_INTERP_BILINEAR);
+	}
+
+	/* Make sure pre-965 set dither correctly for 18bpp panels. */
+	if (INTEL_INFO(dev)->gen < 4 && pipe_config->pipe_bpp == 18)
+		pfit_control |= PANEL_8TO6_DITHER_ENABLE;
+
+	pipe_config->gmch_pfit.control = pfit_control;
+	pipe_config->gmch_pfit.lvds_border_bits = border;
+}
+
 void intel_gmch_panel_fitting(struct intel_crtc *intel_crtc,
 			      struct intel_crtc_config *pipe_config,
 			      int fitting_mode)
@@ -257,6 +400,13 @@ void intel_gmch_panel_fitting(struct intel_crtc *intel_crtc,
 
 	adjusted_mode = &pipe_config->adjusted_mode;
 
+	/* check if User wants to apply the borders, or wants to forcefully
+	   enable the panel fitter, otherwise fall through the regular path */
+	if (intel_crtc->pfit_enabled ||
+			intel_crtc->border_size)
+		return intel_gmch_manual_panel_fitting(intel_crtc,
+						       pipe_config);
+
 	/* Native modes don't need fitting */
 	if (adjusted_mode->hdisplay == pipe_config->pipe_src_w &&
 	    adjusted_mode->vdisplay == pipe_config->pipe_src_h)
Signature · 6 lines
1.8.5.2

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 2 lines
Ville Syrjälä
Intel OTC
author
#19 lore raw

Re: [PATCH v4 3/3] drm/i915: New drm crtc property for varying the size of borders

Quoted 69 lines · Ville Syrjälä
Quoted 29 lines · Akash Goel
Quoted 28 lines · Ville Syrjälä
Quoted 27 lines · On Wed, Mar 26, 2014 at 09:25:12AM +0530, [email protected] wrote:

From: Akash Goel <[email protected]>

This patch adds a new drm crtc property for varying the size of the horizontal & vertical borers of the output/display window. This will control the output of Panel fitter.

v2: Added a new check for the invalid border size input

v3: Fixed bugs in output window calculation
Removed superfluous checks

v4: Added the capability to forecfully enable the Panel fitter. The property value is of 64 bits, first 32 bits are used for border dimensions. The 33rd bit can be used to forcefully enable the panel fitter. This is useful for Panels which do not override the User specified Pipe timings.

Testcase: igt/kms_panel_fitter_test

---
 drivers/gpu/drm/i915/i915_drv.h      |   7 ++
 drivers/gpu/drm/i915/intel_display.c |  39 +++++++-
 drivers/gpu/drm/i915/intel_drv.h     |   5 +
 drivers/gpu/drm/i915/intel_panel.c   | 176 ++++++++++++++++++++++++++++++++---
 4 files changed, 211 insertions(+), 16 deletions(-)

<snip>

Quoted 38 lines
Quoted 37 lines
Quoted 23 lines
drivers/gpu/drm/i915/intel_panel.c +19 -0
@@ -42,6 +57,60 @@ intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode,
 	drm_mode_set_crtcinfo(adjusted_mode, 0);
 }
 
+void
+intel_pch_manual_panel_fitting(struct intel_crtc *intel_crtc,
+			struct intel_crtc_config *pipe_config)
+{
+	struct drm_display_mode *adjusted_mode;
+	int x, y;
+	u32 pf_horizontal_ratio, pf_vertical_ratio;
+	u32 tot_width, tot_height;
+	u32 src_width, src_height; /* pipesrc.x, pipesrc.y */
+	u32 dst_width, dst_height;
+
+	adjusted_mode = &pipe_config->adjusted_mode;
+
+	src_width = pipe_config->pipe_src_w;
+	src_height = pipe_config->pipe_src_h;
+
+	tot_width  = adjusted_mode->hdisplay;
+	tot_height = adjusted_mode->vdisplay;
+

drivers/gpu/drm/i915/intel_panel.c intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode, · line 88

Quoted 10 lines
drivers/gpu/drm/i915/intel_panel.c +10 -0
+	/*
+	 * Having non zero borders will reduce the size of 'HACTIVE/VACTIVE'
+	 * region. So (HACTIVE - Left border - Right Border) *
+	 * (VACTIVE  - Top Border  - Bottom border) will effectively be the
+	 * output rectangle on screen
+	 */
+	dst_width = tot_width -
+			(((intel_crtc->border_size >> 16) & 0xffff) * 2);
+	dst_height = tot_height -
+			((intel_crtc->border_size & 0xffff) * 2);

I'm thinking that we should allow the user to specify each border width individually rather than just assume left==right and top==bottom.

Quoted 5 lines
Quoted 2 lines

Sorry I thought that the Top/Bottom & left/Right borders would be symmetric only.

I don't see a reason to limit ourselves here.

Fine, will extend this property to set each border separately. Can I use the 12 bits for each border value, as that shall be sufficient.

Quoted 9 lines
Quoted 4 lines

Tried setting the borders on EDP & HDMI panels by manipulating the Pipe timings (through the logic used in 'centre_horizontally' & 'centre_vertically' functions), but it didn't work.
Is this logic effective for the LVDS panel only ?

Could be. Certainly the border enable bit is there only for LVDS. The gmch panel fitter isn't very flexible so it's possible we can't actually make it do many of the things the pch pfit can do.

Yes the GMCH panel fitter function is not equally capable as the PCH counterpart. Here except the LVDS panel, it seems that borders cannot be realized on any other panel, just via the manipulation of Pipe timings (the way it can be done in PCH one).
For the same reason the 'Center' Panel fitting mode of "scaling mode" property is not working on VLV (at least for HDMI/EDP panels).

Quoted 7 lines

What happens if we set up the pfit to use manual scaling ratios but configure both scaling ratios so that scaled image won't fill the active video region in either direction? Does it position the scaled image at coordinates 0,0 and simply scan black the rest of the time after it's run out of source pixel data? Or does it automagically center the image and scan black on both sides? Or does it fail in some way?

Already tried that, but in vain. As per the VLV spec, the support for Manual scaling ratio mode itself has been de-featured, so it didn't work at all. So Auto/LetterBox/PillarBox modes are being supported.

As a next step tried to manipulate the Pipe timings, so as to produce the borders on 4 sides of the panel. Similar to the PCH panel fitting logic, kept the HSYNC/VSYNC pulse width same as well as the size of HBLANK/VBLANK intervals, just manipulated the HYSNC/VSYNC and HBLANK/VBLANK start & end points to create borders around the active region.

Tried to add Left/Right borders of 32 columns and Top/bottom borders of 30 lines.
For that
HACTIVE=1920, HBLANK START=1920, HSYNC START=1968, HYSNC END=2000, HBLANK END=2080, HTOTAL=2080.
was changed to
HACTIVE=1856, HBLANK START=1888, HSYNC START=1952, HYSNC END=1984,
HBLANK END=2048, HTOTAL=2080.

And Similarly
VACTIVE=1200, VBLANK START=1200, VSYNC START=1203, VYSNC END=1209, VBLANK END=1235, VTOTAL=1235.
was changed to
VACTIVE=1140, VBLANK START=1170, VSYNC START=1185, VYSNC END=1191, VBLANK END=1205, VTOTAL=1235.

After this manipulation, saw that the HMDI panel turned blank and showed a "No Signal" message.
But for the same experiment, observed a different behavior with EDP panel, that the display was active but the image was being shown in the Top left part of the screen only, with the area outside active rectangle having all junk data.

Quoted 11 lines
Quoted 11 lines
Quoted 10 lines

drivers/gpu/drm/i915/intel_panel.c intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode, · line 92

Quoted 5 lines
drivers/gpu/drm/i915/intel_panel.c +4 -0
+
+	if ((dst_width == 0) || (dst_height == 0)) {
+		DRM_ERROR("Invalid border size input\n");
+		return;

This is clear sign here that we should do all this stuff in the compute config stage so that we can fail gracefully and tell userspace that things didn't work out.

Quoted 7 lines
Quoted 3 lines

Actually this call to decide the panel fitter config, is being made in the context of 'compute config' only. But it's originating from the 'crtc set property' & not from the modeset.

We need to check things in both cases, and return an error if things can't work out.

Can this be done in a next patch set i.e. adding return types to the functions so that if there is an error in panel fitter configuration, it can be communicated back to User space.

Quoted 10 lines
Quoted 6 lines

Do you mean to say that if done from the 'modeset' call, we can report back the error to User space for an invalid border setting. Actually currently the return type is 'void' only for the 2 existing panel fitter config functions.
Also we don't have a check in place for the max supported upscaling ratio.

Is there an upscaling limit? I know there's a downscaling limit of IIRC 1.125x or something close to that. I don't think we check that either.

Sorry my mistake, it's only the downscaling ratio which has a max value of 1.125 i.e. PIPESRC cannot be downscaled by more than a factor of 1.125.

Quoted 163 lines
Quoted 163 lines
Quoted 163 lines
Quoted 163 lines
drivers/gpu/drm/i915/intel_panel.c +115 -13
+	}
+
+	pf_horizontal_ratio = panel_fitter_scaling(src_width, dst_width);
+	pf_vertical_ratio   = panel_fitter_scaling(src_height, dst_height);
+
+	if (pf_horizontal_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("width is too small\n");
+		return;
+	} else if (pf_vertical_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("height is too small\n");
+		return;
+	}
+
+	x = (adjusted_mode->hdisplay - dst_width + 1)/2;
+	y = (adjusted_mode->vdisplay - dst_height + 1)/2;
+
+	pipe_config->pch_pfit.pos = (x << 16) | y;
+	pipe_config->pch_pfit.size = (dst_width << 16) | dst_height;
+	pipe_config->pch_pfit.enabled = pipe_config->pch_pfit.size != 0;
+}
+
 /* adjusted_mode has been preset to be the panel's fixed mode */
 void
 intel_pch_panel_fitting(struct intel_crtc *intel_crtc,
@@ -55,6 +124,13 @@ intel_pch_panel_fitting(struct intel_crtc *intel_crtc,
 
 	x = y = width = height = 0;
 
+	/* check if User wants to apply the borders, or wants to forcefully
+	   enable the panel fitter, otherwise fall through the regular path */
+	if (intel_crtc->pfit_enabled ||
+			intel_crtc->border_size)
+		return intel_pch_manual_panel_fitting(intel_crtc,
+						      pipe_config);
+
 	/* Native modes don't need fitting */
 	if (adjusted_mode->hdisplay == pipe_config->pipe_src_w &&
 	    adjusted_mode->vdisplay == pipe_config->pipe_src_h)
@@ -157,19 +233,6 @@ centre_vertically(struct drm_display_mode *mode,
 	mode->crtc_vsync_end = mode->crtc_vsync_start + sync_width;
 }
 
-static inline u32 panel_fitter_scaling(u32 source, u32 target)
-{
-	/*
-	 * Floating point operation is not supported. So the FACTOR
-	 * is defined, which can avoid the floating point computation
-	 * when calculating the panel ratio.
-	 */
-#define ACCURACY 12
-#define FACTOR (1 << ACCURACY)
-	u32 ratio = source * FACTOR / target;
-	return (FACTOR * ratio + FACTOR/2) / FACTOR;
-}
-
 static void i965_scale_aspect(struct intel_crtc_config *pipe_config,
 			      u32 *pfit_control)
 {
@@ -247,6 +310,86 @@ static void i9xx_scale_aspect(struct intel_crtc_config *pipe_config,
 	}
 }
 
+void intel_gmch_manual_panel_fitting(struct intel_crtc *intel_crtc,
+				     struct intel_crtc_config *pipe_config)
+{
+	struct drm_device *dev = intel_crtc->base.dev;
+	u32 pfit_control = 0, border = 0;
+	u32 pf_horizontal_ratio, pf_vertical_ratio;
+	struct drm_display_mode *adjusted_mode;
+	u32 tot_width, tot_height;
+	u32 src_width, src_height; /* pipesrc.x, pipesrc.y */
+	u32 dst_width, dst_height;
+
+	adjusted_mode = &pipe_config->adjusted_mode;
+
+	src_width = pipe_config->pipe_src_w;
+	src_height = pipe_config->pipe_src_h;
+
+	tot_width = adjusted_mode->hdisplay;
+	tot_height = adjusted_mode->vdisplay;
+
+	/*
+	 * Having non zero borders will reduce the size of 'HACTIVE/VACTIVE'
+	 * region. So (HACTIVE - Left border - Right Border) *
+	 * (VACTIVE  - Top Border  - Bottom border) will effectively be the
+	 * output rectangle on screen
+	 */
+	dst_width = tot_width -
+			(((intel_crtc->border_size >> 16) & 0xffff) * 2);
+	dst_height = tot_height -
+			((intel_crtc->border_size & 0xffff) * 2);
+
+	if ((dst_width == 0) || (dst_height == 0)) {
+		DRM_ERROR("Invalid border size input\n");
+		return;
+	}
+
+	pf_horizontal_ratio = panel_fitter_scaling(src_width, dst_width);
+	pf_vertical_ratio   = panel_fitter_scaling(src_height, dst_height);
+
+	if (pf_horizontal_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("width is too small\n");
+		return;
+	} else if (pf_vertical_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("height is too small\n");
+		return;
+	}
+
+	if (dst_width != tot_width)
+		centre_horizontally(adjusted_mode, dst_width);
+	if (dst_height != tot_height)
+		centre_vertically(adjusted_mode, dst_height);
+
+	/* No scaling needed now, but still enable the panel fitter,
+	   as that will allow the User to subequently do the dynamic
+	   flipping of fbs of different resolutions */
+	if (adjusted_mode->crtc_hdisplay == pipe_config->pipe_src_w &&
+	    adjusted_mode->crtc_vdisplay == pipe_config->pipe_src_h) {
+		BUG_ON(!intel_crtc->pfit_enabled);
+		DRM_DEBUG_KMS("Forcefully enabling the Panel fitter\n");
+	}
+
+	border = LVDS_BORDER_ENABLE;
+
+	if (INTEL_INFO(dev)->gen >= 4) {
+		/* PFIT_SCALING_PROGRAMMED is de-featured on BYT */
+		pfit_control |= PFIT_ENABLE | PFIT_SCALING_AUTO;
+		pfit_control |= ((intel_crtc->pipe << PFIT_PIPE_SHIFT) | PFIT_FILTER_FUZZY);
+	} else {
+		pfit_control |= (PFIT_ENABLE |
+				 VERT_AUTO_SCALE | HORIZ_AUTO_SCALE |
+				 VERT_INTERP_BILINEAR | HORIZ_INTERP_BILINEAR);
+	}
+
+	/* Make sure pre-965 set dither correctly for 18bpp panels. */
+	if (INTEL_INFO(dev)->gen < 4 && pipe_config->pipe_bpp == 18)
+		pfit_control |= PANEL_8TO6_DITHER_ENABLE;
+
+	pipe_config->gmch_pfit.control = pfit_control;
+	pipe_config->gmch_pfit.lvds_border_bits = border;
+}
+
 void intel_gmch_panel_fitting(struct intel_crtc *intel_crtc,
 			      struct intel_crtc_config *pipe_config,
 			      int fitting_mode)
@@ -257,6 +400,13 @@ void intel_gmch_panel_fitting(struct intel_crtc *intel_crtc,
 
 	adjusted_mode = &pipe_config->adjusted_mode;
 
+	/* check if User wants to apply the borders, or wants to forcefully
+	   enable the panel fitter, otherwise fall through the regular path */
+	if (intel_crtc->pfit_enabled ||
+			intel_crtc->border_size)
+		return intel_gmch_manual_panel_fitting(intel_crtc,
+						       pipe_config);
+
 	/* Native modes don't need fitting */
 	if (adjusted_mode->hdisplay == pipe_config->pipe_src_w &&
 	    adjusted_mode->vdisplay == pipe_config->pipe_src_h)
Signature · 1 line
1.8.5.2
Quoted 7 lines
Quoted 6 lines
Quoted 5 lines
Quoted 4 lines

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#20 lore raw

Re: [PATCH v4 3/3] drm/i915: New drm crtc property for varying the size of borders

Quoted 73 lines · Akash Goel
Quoted 72 lines · Ville Syrjälä
Quoted 29 lines · Akash Goel
Quoted 28 lines · Ville Syrjälä
Quoted 27 lines · On Wed, Mar 26, 2014 at 09:25:12AM +0530, [email protected] wrote:

From: Akash Goel <[email protected]>

This patch adds a new drm crtc property for varying the size of the horizontal & vertical borers of the output/display window. This will control the output of Panel fitter.

v2: Added a new check for the invalid border size input

v3: Fixed bugs in output window calculation
Removed superfluous checks

v4: Added the capability to forecfully enable the Panel fitter. The property value is of 64 bits, first 32 bits are used for border dimensions. The 33rd bit can be used to forcefully enable the panel fitter. This is useful for Panels which do not override the User specified Pipe timings.

Testcase: igt/kms_panel_fitter_test

---
 drivers/gpu/drm/i915/i915_drv.h      |   7 ++
 drivers/gpu/drm/i915/intel_display.c |  39 +++++++-
 drivers/gpu/drm/i915/intel_drv.h     |   5 +
 drivers/gpu/drm/i915/intel_panel.c   | 176 ++++++++++++++++++++++++++++++++---
 4 files changed, 211 insertions(+), 16 deletions(-)

<snip>

Quoted 40 lines
Quoted 28 lines
Quoted 28 lines
drivers/gpu/drm/i915/intel_panel.c +24 -0
@@ -42,6 +57,60 @@ intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode,
 	drm_mode_set_crtcinfo(adjusted_mode, 0);
 }
 
+void
+intel_pch_manual_panel_fitting(struct intel_crtc *intel_crtc,
+			struct intel_crtc_config *pipe_config)
+{
+	struct drm_display_mode *adjusted_mode;
+	int x, y;
+	u32 pf_horizontal_ratio, pf_vertical_ratio;
+	u32 tot_width, tot_height;
+	u32 src_width, src_height; /* pipesrc.x, pipesrc.y */
+	u32 dst_width, dst_height;
+
+	adjusted_mode = &pipe_config->adjusted_mode;
+
+	src_width = pipe_config->pipe_src_w;
+	src_height = pipe_config->pipe_src_h;
+
+	tot_width  = adjusted_mode->hdisplay;
+	tot_height = adjusted_mode->vdisplay;
+
+	/*
+	 * Having non zero borders will reduce the size of 'HACTIVE/VACTIVE'
+	 * region. So (HACTIVE - Left border - Right Border) *
+	 * (VACTIVE  - Top Border  - Bottom border) will effectively be the
+	 * output rectangle on screen
Quoted 9 lines

drivers/gpu/drm/i915/intel_panel.c intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode, · line 88

Quoted 5 lines
drivers/gpu/drm/i915/intel_panel.c +5 -0
+	 */
+	dst_width = tot_width -
+			(((intel_crtc->border_size >> 16) & 0xffff) * 2);
+	dst_height = tot_height -
+			((intel_crtc->border_size & 0xffff) * 2);

I'm thinking that we should allow the user to specify each border width individually rather than just assume left==right and top==bottom.

Sorry I thought that the Top/Bottom & left/Right borders would be symmetric only.

Quoted 6 lines
Quoted 2 lines

I don't see a reason to limit ourselves here.

Fine, will extend this property to set each border separately. Can I use the 12 bits for each border value, as that shall be sufficient.

Maybe just add separate properties for each border value. We already have similar properties for TV outputs.

Quoted 26 lines
Quoted 10 lines
Quoted 5 lines

Tried setting the borders on EDP & HDMI panels by manipulating the Pipe timings (through the logic used in 'centre_horizontally' & 'centre_vertically' functions), but it didn't work.
Is this logic effective for the LVDS panel only ?

Could be. Certainly the border enable bit is there only for LVDS. The gmch panel fitter isn't very flexible so it's possible we can't actually make it do many of the things the pch pfit can do.

Yes the GMCH panel fitter function is not equally capable as the PCH counterpart. Here except the LVDS panel, it seems that borders cannot be realized on any other panel, just via the manipulation of Pipe timings (the way it can be done in PCH one).
For the same reason the 'Center' Panel fitting mode of "scaling mode" property is not working on VLV (at least for HDMI/EDP panels).

Quoted 7 lines

What happens if we set up the pfit to use manual scaling ratios but configure both scaling ratios so that scaled image won't fill the active video region in either direction? Does it position the scaled image at coordinates 0,0 and simply scan black the rest of the time after it's run out of source pixel data? Or does it automagically center the image and scan black on both sides? Or does it fail in some way?

Quoted 3 lines

Already tried that, but in vain. As per the VLV spec, the support for Manual scaling ratio mode itself has been de-featured, so it didn't work at all. So Auto/LetterBox/PillarBox modes are being supported.

I see.

Quoted 17 lines

As a next step tried to manipulate the Pipe timings, so as to produce the borders on 4 sides of the panel. Similar to the PCH panel fitting logic, kept the HSYNC/VSYNC pulse width same as well as the size of HBLANK/VBLANK intervals, just manipulated the HYSNC/VSYNC and HBLANK/VBLANK start & end points to create borders around the active region.

Tried to add Left/Right borders of 32 columns and Top/bottom borders of 30 lines.
For that
HACTIVE=1920, HBLANK START=1920, HSYNC START=1968, HYSNC END=2000, HBLANK END=2080, HTOTAL=2080.
was changed to
HACTIVE=1856, HBLANK START=1888, HSYNC START=1952, HYSNC END=1984,
HBLANK END=2048, HTOTAL=2080.

Quoted 6 lines

And Similarly
VACTIVE=1200, VBLANK START=1200, VSYNC START=1203, VYSNC END=1209, VBLANK END=1235, VTOTAL=1235.
was changed to
VACTIVE=1140, VBLANK START=1170, VSYNC START=1185, VYSNC END=1191, VBLANK END=1205, VTOTAL=1235.

Your sync values seem to be a bit off. AFAICS they should be:
hsync_start = 1936, hsync_end = 1968
vsync_start = 1173, vsync_end = 1179

Quoted 3 lines

After this manipulation, saw that the HMDI panel turned blank and showed a "No Signal" message.

Might be good to repeat with fixed sync positions.

There's also some kind of border enable bit in the sdvo/hdmi control register. No idea if that does anything useful.

Quoted 4 lines

But for the same experiment, observed a different behavior with EDP panel, that the display was active but the image was being shown in the Top left part of the screen only, with the area outside active rectangle having all junk data.

:(

I suppose it's still likely that things won't actually work even w/ the sync pulses in the correct position. And that would mean we can't support borders on non-LVDS outputs on gmch platforms.

And that would also mean we already have a bug with the current scaling_mode property on VLV eDP.

Quoted 16 lines
Quoted 16 lines
Quoted 15 lines
Quoted 11 lines

drivers/gpu/drm/i915/intel_panel.c intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode, · line 92

Quoted 6 lines
drivers/gpu/drm/i915/intel_panel.c +4 -0
+
+	if ((dst_width == 0) || (dst_height == 0)) {
+		DRM_ERROR("Invalid border size input\n");
+		return;

This is clear sign here that we should do all this stuff in the compute config stage so that we can fail gracefully and tell userspace that things didn't work out.

Actually this call to decide the panel fitter config, is being made in the context of 'compute config' only. But it's originating from the 'crtc set property' & not from the modeset.

Quoted 7 lines
Quoted 3 lines

We need to check things in both cases, and return an error if things can't work out.

Can this be done in a next patch set i.e. adding return types to the functions so that if there is an error in panel fitter configuration, it can be communicated back to User space.

I don't see a reason for delaying it.

Quoted 186 lines
Quoted 11 lines
Quoted 7 lines

Do you mean to say that if done from the 'modeset' call, we can report back the error to User space for an invalid border setting. Actually currently the return type is 'void' only for the 2 existing panel fitter config functions.
Also we don't have a check in place for the max supported upscaling ratio.

Is there an upscaling limit? I know there's a downscaling limit of IIRC 1.125x or something close to that. I don't think we check that either.

Sorry my mistake, it's only the downscaling ratio which has a max value of 1.125 i.e. PIPESRC cannot be downscaled by more than a factor of 1.125.

Quoted 170 lines
Quoted 169 lines
Quoted 168 lines
Quoted 167 lines
drivers/gpu/drm/i915/intel_panel.c +115 -13
+	}
+
+	pf_horizontal_ratio = panel_fitter_scaling(src_width, dst_width);
+	pf_vertical_ratio   = panel_fitter_scaling(src_height, dst_height);
+
+	if (pf_horizontal_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("width is too small\n");
+		return;
+	} else if (pf_vertical_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("height is too small\n");
+		return;
+	}
+
+	x = (adjusted_mode->hdisplay - dst_width + 1)/2;
+	y = (adjusted_mode->vdisplay - dst_height + 1)/2;
+
+	pipe_config->pch_pfit.pos = (x << 16) | y;
+	pipe_config->pch_pfit.size = (dst_width << 16) | dst_height;
+	pipe_config->pch_pfit.enabled = pipe_config->pch_pfit.size != 0;
+}
+
 /* adjusted_mode has been preset to be the panel's fixed mode */
 void
 intel_pch_panel_fitting(struct intel_crtc *intel_crtc,
@@ -55,6 +124,13 @@ intel_pch_panel_fitting(struct intel_crtc *intel_crtc,
 
 	x = y = width = height = 0;
 
+	/* check if User wants to apply the borders, or wants to forcefully
+	   enable the panel fitter, otherwise fall through the regular path */
+	if (intel_crtc->pfit_enabled ||
+			intel_crtc->border_size)
+		return intel_pch_manual_panel_fitting(intel_crtc,
+						      pipe_config);
+
 	/* Native modes don't need fitting */
 	if (adjusted_mode->hdisplay == pipe_config->pipe_src_w &&
 	    adjusted_mode->vdisplay == pipe_config->pipe_src_h)
@@ -157,19 +233,6 @@ centre_vertically(struct drm_display_mode *mode,
 	mode->crtc_vsync_end = mode->crtc_vsync_start + sync_width;
 }
 
-static inline u32 panel_fitter_scaling(u32 source, u32 target)
-{
-	/*
-	 * Floating point operation is not supported. So the FACTOR
-	 * is defined, which can avoid the floating point computation
-	 * when calculating the panel ratio.
-	 */
-#define ACCURACY 12
-#define FACTOR (1 << ACCURACY)
-	u32 ratio = source * FACTOR / target;
-	return (FACTOR * ratio + FACTOR/2) / FACTOR;
-}
-
 static void i965_scale_aspect(struct intel_crtc_config *pipe_config,
 			      u32 *pfit_control)
 {
@@ -247,6 +310,86 @@ static void i9xx_scale_aspect(struct intel_crtc_config *pipe_config,
 	}
 }
 
+void intel_gmch_manual_panel_fitting(struct intel_crtc *intel_crtc,
+				     struct intel_crtc_config *pipe_config)
+{
+	struct drm_device *dev = intel_crtc->base.dev;
+	u32 pfit_control = 0, border = 0;
+	u32 pf_horizontal_ratio, pf_vertical_ratio;
+	struct drm_display_mode *adjusted_mode;
+	u32 tot_width, tot_height;
+	u32 src_width, src_height; /* pipesrc.x, pipesrc.y */
+	u32 dst_width, dst_height;
+
+	adjusted_mode = &pipe_config->adjusted_mode;
+
+	src_width = pipe_config->pipe_src_w;
+	src_height = pipe_config->pipe_src_h;
+
+	tot_width = adjusted_mode->hdisplay;
+	tot_height = adjusted_mode->vdisplay;
+
+	/*
+	 * Having non zero borders will reduce the size of 'HACTIVE/VACTIVE'
+	 * region. So (HACTIVE - Left border - Right Border) *
+	 * (VACTIVE  - Top Border  - Bottom border) will effectively be the
+	 * output rectangle on screen
+	 */
+	dst_width = tot_width -
+			(((intel_crtc->border_size >> 16) & 0xffff) * 2);
+	dst_height = tot_height -
+			((intel_crtc->border_size & 0xffff) * 2);
+
+	if ((dst_width == 0) || (dst_height == 0)) {
+		DRM_ERROR("Invalid border size input\n");
+		return;
+	}
+
+	pf_horizontal_ratio = panel_fitter_scaling(src_width, dst_width);
+	pf_vertical_ratio   = panel_fitter_scaling(src_height, dst_height);
+
+	if (pf_horizontal_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("width is too small\n");
+		return;
+	} else if (pf_vertical_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("height is too small\n");
+		return;
+	}
+
+	if (dst_width != tot_width)
+		centre_horizontally(adjusted_mode, dst_width);
+	if (dst_height != tot_height)
+		centre_vertically(adjusted_mode, dst_height);
+
+	/* No scaling needed now, but still enable the panel fitter,
+	   as that will allow the User to subequently do the dynamic
+	   flipping of fbs of different resolutions */
+	if (adjusted_mode->crtc_hdisplay == pipe_config->pipe_src_w &&
+	    adjusted_mode->crtc_vdisplay == pipe_config->pipe_src_h) {
+		BUG_ON(!intel_crtc->pfit_enabled);
+		DRM_DEBUG_KMS("Forcefully enabling the Panel fitter\n");
+	}
+
+	border = LVDS_BORDER_ENABLE;
+
+	if (INTEL_INFO(dev)->gen >= 4) {
+		/* PFIT_SCALING_PROGRAMMED is de-featured on BYT */
+		pfit_control |= PFIT_ENABLE | PFIT_SCALING_AUTO;
+		pfit_control |= ((intel_crtc->pipe << PFIT_PIPE_SHIFT) | PFIT_FILTER_FUZZY);
+	} else {
+		pfit_control |= (PFIT_ENABLE |
+				 VERT_AUTO_SCALE | HORIZ_AUTO_SCALE |
+				 VERT_INTERP_BILINEAR | HORIZ_INTERP_BILINEAR);
+	}
+
+	/* Make sure pre-965 set dither correctly for 18bpp panels. */
+	if (INTEL_INFO(dev)->gen < 4 && pipe_config->pipe_bpp == 18)
+		pfit_control |= PANEL_8TO6_DITHER_ENABLE;
+
+	pipe_config->gmch_pfit.control = pfit_control;
+	pipe_config->gmch_pfit.lvds_border_bits = border;
+}
+
 void intel_gmch_panel_fitting(struct intel_crtc *intel_crtc,
 			      struct intel_crtc_config *pipe_config,
 			      int fitting_mode)
@@ -257,6 +400,13 @@ void intel_gmch_panel_fitting(struct intel_crtc *intel_crtc,
 
 	adjusted_mode = &pipe_config->adjusted_mode;
 
+	/* check if User wants to apply the borders, or wants to forcefully
+	   enable the panel fitter, otherwise fall through the regular path */
+	if (intel_crtc->pfit_enabled ||
+			intel_crtc->border_size)
+		return intel_gmch_manual_panel_fitting(intel_crtc,
+						       pipe_config);
+
 	/* Native modes don't need fitting */
 	if (adjusted_mode->hdisplay == pipe_config->pipe_src_w &&
 	    adjusted_mode->vdisplay == pipe_config->pipe_src_h)
Signature · 6 lines
1.8.5.2

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 2 lines
Ville Syrjälä
Intel OTC
author
#21 lore raw

Re: [PATCH v4 3/3] drm/i915: New drm crtc property for varying the size of borders

Quoted 77 lines · Ville Syrjälä
Quoted 76 lines · Akash Goel
Quoted 74 lines · Ville Syrjälä
Quoted 29 lines · Akash Goel
Quoted 28 lines · Ville Syrjälä
Quoted 27 lines · On Wed, Mar 26, 2014 at 09:25:12AM +0530, [email protected] wrote:

From: Akash Goel <[email protected]>

This patch adds a new drm crtc property for varying the size of the horizontal & vertical borers of the output/display window. This will control the output of Panel fitter.

v2: Added a new check for the invalid border size input

v3: Fixed bugs in output window calculation
Removed superfluous checks

v4: Added the capability to forecfully enable the Panel fitter. The property value is of 64 bits, first 32 bits are used for border dimensions. The 33rd bit can be used to forcefully enable the panel fitter. This is useful for Panels which do not override the User specified Pipe timings.

Testcase: igt/kms_panel_fitter_test

---
 drivers/gpu/drm/i915/i915_drv.h      |   7 ++
 drivers/gpu/drm/i915/intel_display.c |  39 +++++++-
 drivers/gpu/drm/i915/intel_drv.h     |   5 +
 drivers/gpu/drm/i915/intel_panel.c   | 176 ++++++++++++++++++++++++++++++++---
 4 files changed, 211 insertions(+), 16 deletions(-)

<snip>

Quoted 38 lines
Quoted 28 lines
Quoted 28 lines
drivers/gpu/drm/i915/intel_panel.c +24 -0
@@ -42,6 +57,60 @@ intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode,
 	drm_mode_set_crtcinfo(adjusted_mode, 0);
 }
 
+void
+intel_pch_manual_panel_fitting(struct intel_crtc *intel_crtc,
+			struct intel_crtc_config *pipe_config)
+{
+	struct drm_display_mode *adjusted_mode;
+	int x, y;
+	u32 pf_horizontal_ratio, pf_vertical_ratio;
+	u32 tot_width, tot_height;
+	u32 src_width, src_height; /* pipesrc.x, pipesrc.y */
+	u32 dst_width, dst_height;
+
+	adjusted_mode = &pipe_config->adjusted_mode;
+
+	src_width = pipe_config->pipe_src_w;
+	src_height = pipe_config->pipe_src_h;
+
+	tot_width  = adjusted_mode->hdisplay;
+	tot_height = adjusted_mode->vdisplay;
+
+	/*
+	 * Having non zero borders will reduce the size of 'HACTIVE/VACTIVE'
+	 * region. So (HACTIVE - Left border - Right Border) *
+	 * (VACTIVE  - Top Border  - Bottom border) will effectively be the
+	 * output rectangle on screen
Quoted 9 lines

drivers/gpu/drm/i915/intel_panel.c intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode, · line 88

Quoted 5 lines
drivers/gpu/drm/i915/intel_panel.c +5 -0
+	 */
+	dst_width = tot_width -
+			(((intel_crtc->border_size >> 16) & 0xffff) * 2);
+	dst_height = tot_height -
+			((intel_crtc->border_size & 0xffff) * 2);

I'm thinking that we should allow the user to specify each border width individually rather than just assume left==right and top==bottom.

Quoted 2 lines

Sorry I thought that the Top/Bottom & left/Right borders would be symmetric only.

I don't see a reason to limit ourselves here.

Quoted 7 lines
Quoted 3 lines

Fine, will extend this property to set each border separately. Can I use the 12 bits for each border value, as that shall be sufficient.

Maybe just add separate properties for each border value. We already have similar properties for TV outputs.

ok so should define 4 new properties like "left margin", "right margin", "top margin", "bottom margin" already defined for TV.

Quoted 56 lines
Quoted 26 lines
Quoted 10 lines
Quoted 5 lines

Tried setting the borders on EDP & HDMI panels by manipulating the Pipe timings (through the logic used in 'centre_horizontally' & 'centre_vertically' functions), but it didn't work.
Is this logic effective for the LVDS panel only ?

Could be. Certainly the border enable bit is there only for LVDS. The gmch panel fitter isn't very flexible so it's possible we can't actually make it do many of the things the pch pfit can do.

Yes the GMCH panel fitter function is not equally capable as the PCH counterpart. Here except the LVDS panel, it seems that borders cannot be realized on any other panel, just via the manipulation of Pipe timings (the way it can be done in PCH one).
For the same reason the 'Center' Panel fitting mode of "scaling mode" property is not working on VLV (at least for HDMI/EDP panels).

Quoted 7 lines

What happens if we set up the pfit to use manual scaling ratios but configure both scaling ratios so that scaled image won't fill the active video region in either direction? Does it position the scaled image at coordinates 0,0 and simply scan black the rest of the time after it's run out of source pixel data? Or does it automagically center the image and scan black on both sides? Or does it fail in some way?

Quoted 3 lines

Already tried that, but in vain. As per the VLV spec, the support for Manual scaling ratio mode itself has been de-featured, so it didn't work at all. So Auto/LetterBox/PillarBox modes are being supported.

I see.

Quoted 17 lines

As a next step tried to manipulate the Pipe timings, so as to produce the borders on 4 sides of the panel. Similar to the PCH panel fitting logic, kept the HSYNC/VSYNC pulse width same as well as the size of HBLANK/VBLANK intervals, just manipulated the HYSNC/VSYNC and HBLANK/VBLANK start & end points to create borders around the active region.

Tried to add Left/Right borders of 32 columns and Top/bottom borders of 30 lines.
For that
HACTIVE=1920, HBLANK START=1920, HSYNC START=1968, HYSNC END=2000, HBLANK END=2080, HTOTAL=2080.
was changed to
HACTIVE=1856, HBLANK START=1888, HSYNC START=1952, HYSNC END=1984,
HBLANK END=2048, HTOTAL=2080.

Quoted 6 lines

And Similarly
VACTIVE=1200, VBLANK START=1200, VSYNC START=1203, VYSNC END=1209, VBLANK END=1235, VTOTAL=1235.
was changed to
VACTIVE=1140, VBLANK START=1170, VSYNC START=1185, VYSNC END=1191, VBLANK END=1205, VTOTAL=1235.

Quoted 4 lines

Your sync values seem to be a bit off. AFAICS they should be:
hsync_start = 1936, hsync_end = 1968
vsync_start = 1173, vsync_end = 1179

Actually these are the values outputted by 'centre_horizontally' & 'centre_vertically' functions, where the logic is to position the hsyc/vsync pulses in the middle of the corresponding hblank/vblank intervals.

Thus the HSYNC START=1952, HYSNC END=1984 (pulse of 32 pixels) is lying in the middle of HBLANK START=1888, HBLANK END=2048 interval (160 pixels).

Quoted 14 lines
Quoted 3 lines

After this manipulation, saw that the HMDI panel turned blank and showed a "No Signal" message.

Might be good to repeat with fixed sync positions.

There's also some kind of border enable bit in the sdvo/hdmi control register. No idea if that does anything useful.

Quoted 4 lines

But for the same experiment, observed a different behavior with EDP panel, that the display was active but the image was being shown in the Top left part of the screen only, with the area outside active rectangle having all junk data.

Quoted 9 lines

:(

I suppose it's still likely that things won't actually work even w/ the sync pulses in the correct position. And that would mean we can't support borders on non-LVDS outputs on gmch platforms.

And that would also mean we already have a bug with the current scaling_mode property on VLV eDP.

Yes, the 'Center' Panel fitting mode of "scaling mode" property is not working on VLV (for HDMI/EDP panels).

Quoted 20 lines
Quoted 20 lines
Quoted 19 lines
Quoted 12 lines
Quoted 11 lines

drivers/gpu/drm/i915/intel_panel.c intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode, · line 92

Quoted 6 lines
drivers/gpu/drm/i915/intel_panel.c +4 -0
+
+	if ((dst_width == 0) || (dst_height == 0)) {
+		DRM_ERROR("Invalid border size input\n");
+		return;

This is clear sign here that we should do all this stuff in the compute config stage so that we can fail gracefully and tell userspace that things didn't work out.

Quoted 3 lines

Actually this call to decide the panel fitter config, is being made in the context of 'compute config' only. But it's originating from the 'crtc set property' & not from the modeset.

We need to check things in both cases, and return an error if things can't work out.

Quoted 193 lines
Quoted 3 lines

Can this be done in a next patch set i.e. adding return types to the functions so that if there is an error in panel fitter configuration, it can be communicated back to User space.

I don't see a reason for delaying it.

Quoted 186 lines
Quoted 11 lines
Quoted 7 lines

Do you mean to say that if done from the 'modeset' call, we can report back the error to User space for an invalid border setting. Actually currently the return type is 'void' only for the 2 existing panel fitter config functions.
Also we don't have a check in place for the max supported upscaling ratio.

Is there an upscaling limit? I know there's a downscaling limit of IIRC 1.125x or something close to that. I don't think we check that either.

Sorry my mistake, it's only the downscaling ratio which has a max value of 1.125 i.e. PIPESRC cannot be downscaled by more than a factor of 1.125.

Quoted 170 lines
Quoted 169 lines
Quoted 168 lines
Quoted 167 lines
drivers/gpu/drm/i915/intel_panel.c +115 -13
+	}
+
+	pf_horizontal_ratio = panel_fitter_scaling(src_width, dst_width);
+	pf_vertical_ratio   = panel_fitter_scaling(src_height, dst_height);
+
+	if (pf_horizontal_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("width is too small\n");
+		return;
+	} else if (pf_vertical_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("height is too small\n");
+		return;
+	}
+
+	x = (adjusted_mode->hdisplay - dst_width + 1)/2;
+	y = (adjusted_mode->vdisplay - dst_height + 1)/2;
+
+	pipe_config->pch_pfit.pos = (x << 16) | y;
+	pipe_config->pch_pfit.size = (dst_width << 16) | dst_height;
+	pipe_config->pch_pfit.enabled = pipe_config->pch_pfit.size != 0;
+}
+
 /* adjusted_mode has been preset to be the panel's fixed mode */
 void
 intel_pch_panel_fitting(struct intel_crtc *intel_crtc,
@@ -55,6 +124,13 @@ intel_pch_panel_fitting(struct intel_crtc *intel_crtc,
 
 	x = y = width = height = 0;
 
+	/* check if User wants to apply the borders, or wants to forcefully
+	   enable the panel fitter, otherwise fall through the regular path */
+	if (intel_crtc->pfit_enabled ||
+			intel_crtc->border_size)
+		return intel_pch_manual_panel_fitting(intel_crtc,
+						      pipe_config);
+
 	/* Native modes don't need fitting */
 	if (adjusted_mode->hdisplay == pipe_config->pipe_src_w &&
 	    adjusted_mode->vdisplay == pipe_config->pipe_src_h)
@@ -157,19 +233,6 @@ centre_vertically(struct drm_display_mode *mode,
 	mode->crtc_vsync_end = mode->crtc_vsync_start + sync_width;
 }
 
-static inline u32 panel_fitter_scaling(u32 source, u32 target)
-{
-	/*
-	 * Floating point operation is not supported. So the FACTOR
-	 * is defined, which can avoid the floating point computation
-	 * when calculating the panel ratio.
-	 */
-#define ACCURACY 12
-#define FACTOR (1 << ACCURACY)
-	u32 ratio = source * FACTOR / target;
-	return (FACTOR * ratio + FACTOR/2) / FACTOR;
-}
-
 static void i965_scale_aspect(struct intel_crtc_config *pipe_config,
 			      u32 *pfit_control)
 {
@@ -247,6 +310,86 @@ static void i9xx_scale_aspect(struct intel_crtc_config *pipe_config,
 	}
 }
 
+void intel_gmch_manual_panel_fitting(struct intel_crtc *intel_crtc,
+				     struct intel_crtc_config *pipe_config)
+{
+	struct drm_device *dev = intel_crtc->base.dev;
+	u32 pfit_control = 0, border = 0;
+	u32 pf_horizontal_ratio, pf_vertical_ratio;
+	struct drm_display_mode *adjusted_mode;
+	u32 tot_width, tot_height;
+	u32 src_width, src_height; /* pipesrc.x, pipesrc.y */
+	u32 dst_width, dst_height;
+
+	adjusted_mode = &pipe_config->adjusted_mode;
+
+	src_width = pipe_config->pipe_src_w;
+	src_height = pipe_config->pipe_src_h;
+
+	tot_width = adjusted_mode->hdisplay;
+	tot_height = adjusted_mode->vdisplay;
+
+	/*
+	 * Having non zero borders will reduce the size of 'HACTIVE/VACTIVE'
+	 * region. So (HACTIVE - Left border - Right Border) *
+	 * (VACTIVE  - Top Border  - Bottom border) will effectively be the
+	 * output rectangle on screen
+	 */
+	dst_width = tot_width -
+			(((intel_crtc->border_size >> 16) & 0xffff) * 2);
+	dst_height = tot_height -
+			((intel_crtc->border_size & 0xffff) * 2);
+
+	if ((dst_width == 0) || (dst_height == 0)) {
+		DRM_ERROR("Invalid border size input\n");
+		return;
+	}
+
+	pf_horizontal_ratio = panel_fitter_scaling(src_width, dst_width);
+	pf_vertical_ratio   = panel_fitter_scaling(src_height, dst_height);
+
+	if (pf_horizontal_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("width is too small\n");
+		return;
+	} else if (pf_vertical_ratio > MAX_DOWNSCALE_RATIO) {
+		DRM_ERROR("height is too small\n");
+		return;
+	}
+
+	if (dst_width != tot_width)
+		centre_horizontally(adjusted_mode, dst_width);
+	if (dst_height != tot_height)
+		centre_vertically(adjusted_mode, dst_height);
+
+	/* No scaling needed now, but still enable the panel fitter,
+	   as that will allow the User to subequently do the dynamic
+	   flipping of fbs of different resolutions */
+	if (adjusted_mode->crtc_hdisplay == pipe_config->pipe_src_w &&
+	    adjusted_mode->crtc_vdisplay == pipe_config->pipe_src_h) {
+		BUG_ON(!intel_crtc->pfit_enabled);
+		DRM_DEBUG_KMS("Forcefully enabling the Panel fitter\n");
+	}
+
+	border = LVDS_BORDER_ENABLE;
+
+	if (INTEL_INFO(dev)->gen >= 4) {
+		/* PFIT_SCALING_PROGRAMMED is de-featured on BYT */
+		pfit_control |= PFIT_ENABLE | PFIT_SCALING_AUTO;
+		pfit_control |= ((intel_crtc->pipe << PFIT_PIPE_SHIFT) | PFIT_FILTER_FUZZY);
+	} else {
+		pfit_control |= (PFIT_ENABLE |
+				 VERT_AUTO_SCALE | HORIZ_AUTO_SCALE |
+				 VERT_INTERP_BILINEAR | HORIZ_INTERP_BILINEAR);
+	}
+
+	/* Make sure pre-965 set dither correctly for 18bpp panels. */
+	if (INTEL_INFO(dev)->gen < 4 && pipe_config->pipe_bpp == 18)
+		pfit_control |= PANEL_8TO6_DITHER_ENABLE;
+
+	pipe_config->gmch_pfit.control = pfit_control;
+	pipe_config->gmch_pfit.lvds_border_bits = border;
+}
+
 void intel_gmch_panel_fitting(struct intel_crtc *intel_crtc,
 			      struct intel_crtc_config *pipe_config,
 			      int fitting_mode)
@@ -257,6 +400,13 @@ void intel_gmch_panel_fitting(struct intel_crtc *intel_crtc,
 
 	adjusted_mode = &pipe_config->adjusted_mode;
 
+	/* check if User wants to apply the borders, or wants to forcefully
+	   enable the panel fitter, otherwise fall through the regular path */
+	if (intel_crtc->pfit_enabled ||
+			intel_crtc->border_size)
+		return intel_gmch_manual_panel_fitting(intel_crtc,
+						       pipe_config);
+
 	/* Native modes don't need fitting */
 	if (adjusted_mode->hdisplay == pipe_config->pipe_src_w &&
 	    adjusted_mode->vdisplay == pipe_config->pipe_src_h)
Signature · 6 lines
1.8.5.2

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#22 lore raw

Re: [PATCH v4 3/3] drm/i915: New drm crtc property for varying the size of borders

Quoted 2 lines · Akash Goel

Yes, the 'Center' Panel fitting mode of "scaling mode" property is not working on VLV (for HDMI/EDP panels).

That sounds like a really good reason for a stab at a CRC based testcase for all this - we've blown up the different upscaling modes way too often already.
-Daniel

Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
#23 lore raw

PATCH v5 3/4 drm/i915: New drm crtc property for varying the size of borders

In reply to Ville Syrjälä

From: Akash Goel <[email protected]>

This patch adds a new drm crtc property for varying the size of the horizontal & vertical borers of the output/display window. This will control the output of Panel fitter.

v2: Added a new check for the invalid border size input

v3: Fixed bugs in output window calculation
Removed superfluous checks

v4: Added the capability to forecfully enable the Panel fitter. The property value is of 64 bits, first 32 bits are used for border dimensions. The 33rd bit can be used to forcefully enable the panel fitter. This is useful for Panels which do not override the User specified Pipe timings.

v5: Splitted the single border property into 4 separate properties so as to allow a control on the size of each border left/top/bottom/right (Ville)
Reverted the changes done in v4 of the patch, which provided forceful enabling of panel fitter.
Also refactored based on latest codebase.

Testcase: igt/kms_panel_fitter_test

---
 drivers/gpu/drm/i915/i915_drv.h      |  10 ++
 drivers/gpu/drm/i915/intel_display.c | 107 ++++++++++++++++-
 drivers/gpu/drm/i915/intel_drv.h     |  12 ++
 drivers/gpu/drm/i915/intel_panel.c   | 217 ++++++++++++++++++++++++++++++++---
 4 files changed, 331 insertions(+), 15 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +10 -0
index 104a232..dfe3e03 100644
@@ -1447,6 +1447,16 @@ struct drm_i915_private {
 	 */
 	struct drm_property *input_size_property;
 
+	/*
+	 * Properties to dynamically vary the size of the
+	 * borders. This will indirectly control the size
+	 * of the display window i.e Panel fitter output
+	 */
+	struct drm_property *left_border_property;
+	struct drm_property *right_border_property;
+	struct drm_property *top_border_property;
+	struct drm_property *bottom_border_property;
+
 	uint32_t hw_context_size;
 	struct list_head context_list;
 
drivers/gpu/drm/i915/intel_display.c +105 -2

134 lines not on this page · raw message

drivers/gpu/drm/i915/intel_drv.h +12 -0
index c551472..f00d075 100644
@@ -109,6 +109,14 @@
 #define INTEL_DSI_COMMAND_MODE	0
 #define INTEL_DSI_VIDEO_MODE	1
 
+#define MAX_BORDER_VALUE 256
+enum panel_border {
+	PANEL_BORDER_LEFT = 0,
+	PANEL_BORDER_RIGHT,
+	PANEL_BORDER_TOP,
+	PANEL_BORDER_BOTTOM,
+};
+
 struct intel_framebuffer {
 	struct drm_framebuffer base;
 	struct drm_i915_gem_object *obj;
@@ -395,6 +403,10 @@ struct intel_crtc {
 	bool cpu_fifo_underrun_disabled;
 	bool pch_fifo_underrun_disabled;
 
+	/* border info for the output/display window */
+	uint32_t border[4];
+	bool apply_borders;
+
 	/* per-pipe watermark state */
 	struct {
 		/* watermarks currently being used  */
drivers/gpu/drm/i915/intel_panel.c +204 -13

263 lines not on this page · raw message

Signature · 1 line
1.9.2
#24 lore raw
In reply to Daniel Vetter
Quoted 16 lines · Daniel Vetter
Quoted 9 lines · On Mon, Mar 10, 2014 at 10:40:50AM +0530, [email protected] wrote:

From: Akash Goel <[email protected]>

Added 2 new drm crtc properties. One property provides control to vary the PIPESRC value. With this the size of Pipe's output or Panel fitter input can be varied. The other property provides control to vary the size of horizontal & vertical borders. With this the size of the Panel fitter output or display window could be controlled.

Quoted 5 lines

Akash Goel (3):
drm/i915: Initialized 'set_property' fn pointer field of
intel_crtc_funcs structure
drm/i915: New drm crtc property for varying the Pipe Src size drm/i915: New drm crtc property for varying the size of borders

Quoted 4 lines

Hm, where are the igt testcases for this feature? I expect some fun interactions at least with pageflips/modesets vs. setting this. The vblank timing code should be fine, since it always deals with adjusted timings.

Follow-up from my side ... could we just expose this through primary planes, or is the damn panel fitter at the wrong spot in the pipeline? -Daniel

Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
#25 lore raw

Actually since the Panel fitter is at the Pipe level and also the PIPESRC register, this has been exposed through crtc property. But with crtc also, it is kind of associated with primary planes only.

Best Regards
Akash

-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of Daniel Vetter
Sent: Saturday, March 22, 2014 1:24 AM
To: Goel, Akash
Cc: intel-gfx
Subject: Re: [Intel-gfx] [PATCH 0/3] Two new drm crtc properties

Quoted 16 lines · Daniel Vetter
Quoted 9 lines · On Mon, Mar 10, 2014 at 10:40:50AM +0530, [email protected] wrote:

From: Akash Goel <[email protected]>

Added 2 new drm crtc properties. One property provides control to vary the PIPESRC value. With this the size of Pipe's output or Panel fitter input can be varied.
The other property provides control to vary the size of horizontal & vertical borders. With this the size of the Panel fitter output or display window could be controlled.

Quoted 5 lines

Akash Goel (3):
drm/i915: Initialized 'set_property' fn pointer field of
intel_crtc_funcs structure
drm/i915: New drm crtc property for varying the Pipe Src size drm/i915: New drm crtc property for varying the size of borders

Quoted 4 lines

Hm, where are the igt testcases for this feature? I expect some fun interactions at least with pageflips/modesets vs. setting this. The vblank timing code should be fine, since it always deals with adjusted timings.

Follow-up from my side ... could we just expose this through primary planes, or is the damn panel fitter at the wrong spot in the pipeline? -Daniel
--
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch