Skip to contentSkip to conversation

More Haswell patches

PATCH Applied Eugeni Dodonov 4 participants 44 messages Jun 28, 2012 – Jul 6, 2012

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

cover More Haswell patches

Hi,

Those are the patches which I had in my queue for past few weeks, which I am sending now for your bikeshedding pleasure :).

As major changes, this patch series adds support for DP, eDP and RC6 features for Haswell. Some of the DP-related patches were already posted here, but I am resending them as part of this larger series for giving them proper context.

Also, for DP and eDP, Paulo Zanoni and Shobhit Kumar are already working on a larger series of patches which should appear in the next few days as well. So while the patches in this series should work on most configurations, considerable amount of improvements are still to come.

(And yes, our entire intel_dp dungeon is getting too dangerous for mere mortals to enter, so I think a refactoring of its internals is highly needed.)

Moving to RC6, I've added the new forcewake sequence (which is mostly similar to gen6), and a new enable/disable set of routines. Most of the sequence is similar to what we had on SNB/IVB, but there are some changes in registers and values. So instead of adding another layer of if(IS_HASWELL()) checks all around splitting it into a separate set of functions seems cleaner. This also prepares the ground for additional patches which are yet to come.

I've also added haswell_init_clock_gating routine, which is almost identical to Ivy Bridge. The reason for this was the same as with RC6 - to avoid a large nesting IS_HASWELL chunks on top of older code.

For both those cases, I'd be interested in your opinion on which approach you prefer - reusing GEN6 stuff with IS_HASWELL() chunks; or splitting the functionality into Haswell-specific routines. I can see pros and contras with both approaches, so I am open to suggestions.

Finally, there are some small cleanup patches for LPT debug/error messages a small fix for bogus FBC update messages, and a fix for a broken PIPE_WM_LINETIME register from Paulo which we all overlooked somehow.

So in overall, with those patches, the status of Haswell support should be more or less on-par with Ivy Bridge kernel-wise. Most things that were working on IVB should be available on HSW - and if they don't work it is probably a bug, so please let me know about it.

As usual, bikesheds, comments and suggestions are highly welcome :).

Eugeni

Eugeni Dodonov (10):
drm/i915: re-initialize DDI buffer translations after resume drm/i915: simplify FDI RX check for LPT
drm/i915: account for only one transcoder on LPT drm/i915: introduce lpt_enable_pch and cpt_enable_pch drm/i915: program FDI_RX TP and FDI delays drm/i915: support Haswell-style force waking drm/i915: add RPS configuration for Haswell drm/i915: introduce haswell_init_clock_gating drm/i915: prevent bogus intel_update_fbc notifications drm/i915: enable RC6 workaround on Haswell

Paulo Zanoni (1):
drm/i915: fix PIPE_WM_LINETIME definition

Shobhit Kumar (10):
drm/i915: Move DP structs to shared location drm/i915: Add support for DDI control DP outputs drm/i915: Add DP Helper functions for Haswell drm/i915: Haswell specific code for the DP Link Training drm/i915: Disable DDI Pipe Control on HSW while disabling pipe drm/i915: Hook DP init in ddi module
drm/i915: Add EDP Registers for Haswell
drm/i915: Timing initialization for eDP on HSW drm/i915: Modesetting for eDP on HSw
drm/i915: Hook eDP initialization on DDI A

 drivers/gpu/drm/i915/i915_drv.c      |  36 +++++-
 drivers/gpu/drm/i915/i915_drv.h      |   3 +
 drivers/gpu/drm/i915/i915_reg.h      |  74 ++++++++++-
 drivers/gpu/drm/i915/intel_ddi.c     | 235 +++++++++++++++++++++++++++++++++--
 drivers/gpu/drm/i915/intel_display.c | 194 ++++++++++++++++++++++-------
 drivers/gpu/drm/i915/intel_dp.c      | 171 +++++++++++++++++--------
 drivers/gpu/drm/i915/intel_drv.h     |  48 +++++++
 drivers/gpu/drm/i915/intel_pm.c      | 231 ++++++++++++++++++++++++++++++++--
 8 files changed, 878 insertions(+), 114 deletions(-)
Signature · 1 line
1.7.11.1
#1 lore raw

PATCH 1/21 drm/i915: Move DP structs to shared location

From: Shobhit Kumar <[email protected]>

Move the DP structure to shared location so that it can be used from within ddi module

---
 drivers/gpu/drm/i915/intel_dp.c  | 34 ----------------------------------
 drivers/gpu/drm/i915/intel_drv.h | 35 +++++++++++++++++++++++++++++++++++
 2 files changed, 35 insertions(+), 34 deletions(-)
drivers/gpu/drm/i915/intel_dp.c +0 -34
index 76a7080..65280a0a 100644
@@ -38,40 +38,6 @@
 #include "i915_drv.h"
 #include "drm_dp_helper.h"
 
-#define DP_RECEIVER_CAP_SIZE	0xf
-#define DP_LINK_STATUS_SIZE	6
-#define DP_LINK_CHECK_TIMEOUT	(10 * 1000)
-
-#define DP_LINK_CONFIGURATION_SIZE	9
-
-struct intel_dp {
-	struct intel_encoder base;
-	uint32_t output_reg;
-	uint32_t DP;
-	uint8_t  link_configuration[DP_LINK_CONFIGURATION_SIZE];
-	bool has_audio;
-	enum hdmi_force_audio force_audio;
-	uint32_t color_range;
-	int dpms_mode;
-	uint8_t link_bw;
-	uint8_t lane_count;
-	uint8_t dpcd[DP_RECEIVER_CAP_SIZE];
-	struct i2c_adapter adapter;
-	struct i2c_algo_dp_aux_data algo;
-	bool is_pch_edp;
-	uint8_t	train_set[4];
-	int panel_power_up_delay;
-	int panel_power_down_delay;
-	int panel_power_cycle_delay;
-	int backlight_on_delay;
-	int backlight_off_delay;
-	struct drm_display_mode *panel_fixed_mode;  /* for eDP */
-	struct delayed_work panel_vdd_work;
-	bool want_panel_vdd;
-	struct edid *edid; /* cached EDID for eDP */
-	int edid_mode_count;
-};
-
 /**
  * is_edp - is the given port attached to an eDP panel (either CPU or PCH)
  * @intel_dp: DP struct
drivers/gpu/drm/i915/intel_drv.h +35 -0
index cc1573b..4047b68 100644
@@ -31,6 +31,7 @@
 #include "drm_crtc.h"
 #include "drm_crtc_helper.h"
 #include "drm_fb_helper.h"
+#include "drm_dp_helper.h"
 
 #define _wait_for(COND, MS, W) ({ \
 	unsigned long timeout__ = jiffies + msecs_to_jiffies(MS);	\
@@ -305,6 +306,40 @@ struct intel_hdmi {
 			       struct drm_display_mode *adjusted_mode);
 };
 
+#define DP_RECEIVER_CAP_SIZE	0xf
+#define DP_LINK_STATUS_SIZE	6
+#define DP_LINK_CHECK_TIMEOUT	(10 * 1000)
+
+#define DP_LINK_CONFIGURATION_SIZE	9
+
+struct intel_dp {
+	struct intel_encoder base;
+	uint32_t output_reg;
+	uint32_t DP;
+	uint8_t  link_configuration[DP_LINK_CONFIGURATION_SIZE];
+	bool has_audio;
+	enum hdmi_force_audio force_audio;
+	uint32_t color_range;
+	int dpms_mode;
+	uint8_t link_bw;
+	uint8_t lane_count;
+	uint8_t dpcd[DP_RECEIVER_CAP_SIZE];
+	struct i2c_adapter adapter;
+	struct i2c_algo_dp_aux_data algo;
+	bool is_pch_edp;
+	uint8_t	train_set[4];
+	int panel_power_up_delay;
+	int panel_power_down_delay;
+	int panel_power_cycle_delay;
+	int backlight_on_delay;
+	int backlight_off_delay;
+	struct drm_display_mode *panel_fixed_mode;  /* for eDP */
+	struct delayed_work panel_vdd_work;
+	bool want_panel_vdd;
+	struct edid *edid; /* cached EDID for eDP */
+	int edid_mode_count;
+};
+
 static inline struct drm_crtc *
 intel_get_crtc_for_pipe(struct drm_device *dev, int pipe)
 {
Signature · 1 line
1.7.11.1
#2 lore raw

PATCH 2/21 drm/i915: Add support for DDI control DP outputs

In reply to Eugeni Dodonov

From: Shobhit Kumar <[email protected]>

These are driven by DDIs on Haswell architecture, so we need to keep track of which DDI is being used on each output.

---
 drivers/gpu/drm/i915/intel_dp.c  | 10 +++++++---
 drivers/gpu/drm/i915/intel_drv.h |  1 +
 2 files changed, 8 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/intel_dp.c +7 -3
index 65280a0a..e324c7a 100644
@@ -2458,12 +2458,16 @@ intel_dp_init(struct drm_device *dev, int output_reg)
 
 	connector->polled = DRM_CONNECTOR_POLL_HPD;
 
-	if (output_reg == DP_B || output_reg == PCH_DP_B)
+	if (output_reg == DP_B || output_reg == PCH_DP_B) {
 		intel_encoder->clone_mask = (1 << INTEL_DP_B_CLONE_BIT);
-	else if (output_reg == DP_C || output_reg == PCH_DP_C)
+		intel_dp->ddi_port = PORT_B;
+	} else if (output_reg == DP_C || output_reg == PCH_DP_C) {
 		intel_encoder->clone_mask = (1 << INTEL_DP_C_CLONE_BIT);
-	else if (output_reg == DP_D || output_reg == PCH_DP_D)
+		intel_dp->ddi_port = PORT_C;
+	} else if (output_reg == DP_D || output_reg == PCH_DP_D) {
 		intel_encoder->clone_mask = (1 << INTEL_DP_D_CLONE_BIT);
+		intel_dp->ddi_port = PORT_D;
+	}
 
 	if (is_edp(intel_dp)) {
 		intel_encoder->clone_mask = (1 << INTEL_EDP_CLONE_BIT);
drivers/gpu/drm/i915/intel_drv.h +1 -0
index 4047b68..7cffb12 100644
@@ -319,6 +319,7 @@ struct intel_dp {
 	uint8_t  link_configuration[DP_LINK_CONFIGURATION_SIZE];
 	bool has_audio;
 	enum hdmi_force_audio force_audio;
+	int ddi_port;
 	uint32_t color_range;
 	int dpms_mode;
 	uint8_t link_bw;
Signature · 1 line
1.7.11.1
#3 lore raw

PATCH 3/21 drm/i915: Add DP Helper functions for Haswell

In reply to Eugeni Dodonov

From: Shobhit Kumar <[email protected]>

Need to program new helpers for mode set and dpms as most of the stuff is done using DDI port. The current commit uses SPLL clock for SCC enabled panel and LCPLL for Non-SSC.

Also Haswell has LPT and DDIs are moved on CPU side and has DP_BUF_CTL and DP_TP_CTL so added a new TP (Transport) holder as well along with DP in intel_dp

Main stream attributes have to be set explicitely from HSW onwrads. So programmed the same.

Note that in DP mode the value of port width must match the one in the DDI_BUF_CTL for the DDI port attached to the pipe.

---
 drivers/gpu/drm/i915/i915_reg.h  |  26 +++++-
 drivers/gpu/drm/i915/intel_ddi.c | 167 +++++++++++++++++++++++++++++++++++++--
 drivers/gpu/drm/i915/intel_dp.c  |  18 ++++-
 drivers/gpu/drm/i915/intel_drv.h |   8 ++
 4 files changed, 210 insertions(+), 9 deletions(-)
drivers/gpu/drm/i915/i915_reg.h +25 -1
index 9dfc4c5..51398a8 100644
@@ -4289,6 +4289,7 @@
 #define  PIPE_DDI_BPC_6					(2<<20)
 #define  PIPE_DDI_BPC_12				(3<<20)
 #define  PIPE_DDI_BFI_ENABLE			(1<<4)
+#define  PIPE_DDI_PORT_WIDTH_MASK		(7<<1)
 #define  PIPE_DDI_PORT_WIDTH_X1			(0<<1)
 #define  PIPE_DDI_PORT_WIDTH_X2			(1<<1)
 #define  PIPE_DDI_PORT_WIDTH_X4			(3<<1)
@@ -4335,6 +4336,7 @@
 #define  DDI_BUF_EMP_800MV_3_5DB_HSW	(8<<24)   /* Sel8 */
 #define  DDI_BUF_EMP_MASK				(0xf<<24)
 #define  DDI_BUF_IS_IDLE				(1<<7)
+#define  DDI_PORT_WIDTH_MASK		(7<<1)
 #define  DDI_PORT_WIDTH_X1				(0<<1)
 #define  DDI_PORT_WIDTH_X2				(1<<1)
 #define  DDI_PORT_WIDTH_X4				(3<<1)
@@ -4427,9 +4429,31 @@
 #define LCPLL_CTL				0x130040
 #define  LCPLL_PLL_DISABLE		(1<<31)
 #define  LCPLL_PLL_LOCK			(1<<30)
-#define  LCPLL_CD_CLOCK_DISABLE	(1<<25)
+
+#define  LCPLL_PLL_REFERENCE_BLK	(2<<28)
+#define  LCPLL_PLL_REFERENCE_SSC	(3<<28)
+#define  LCPLL_PLL_REFERENCE_NON_SSC	(0<<28)
+#define	 LCPLL_PLL_REFERENCE_MASK	(0x3<<28)
+
+#define  LCPLL_CD_FREQ_SELECT_MASK      (0x3<<26)
+#define  LCPLL_CD_FREQ_450MHZ		(0x0<<26)
+#define  LCPLL_CD_FREQ_550MHZ		(0x1<<26)
+
+#define  LCPLL_CD_CLOCK_DISABLE		(1<<25)
+#define  LCPLL_ROOT_CD2X_CLOCK_DISABLE  (1<<24)
 #define  LCPLL_CD2X_CLOCK_DISABLE	(1<<23)
 
+#define HSW_MSA_CTL			0x60410
+#define  HSW_MSA_SYNC_CLK		(1<<0)
+#define  HSW_MSA_DYNAMIC_RANGE		(1<<3)
+#define  HSW_MSA_COLORIMETRY		(1<<4)
+#define  HSW_MSA_BPC_MASK		(7<<5)
+#define  HSW_MSA_BPC_6_BITS		(0<<5)
+#define  HSW_MSA_BPC_8_BITS		(1<<5)
+#define  HSW_MSA_BPC_10_BITS		(2<<5)
+#define  HSW_MSA_BPC_12_BITS		(3<<5)
+#define  HSW_MSA_BPC_16_BITS		(4<<5)
+
 /* Pipe WM_LINETIME - watermark line time */
 #define PIPE_WM_LINETIME_A		0x45270
 #define PIPE_WM_LINETIME_B		0x45274
drivers/gpu/drm/i915/intel_ddi.c +161 -6

181 lines not on this page · raw message

drivers/gpu/drm/i915/intel_dp.c +16 -2
index e324c7a..f4c6e98 100644
@@ -74,7 +74,7 @@ static bool is_cpu_edp(struct intel_dp *intel_dp)
 	return is_edp(intel_dp) && !is_pch_edp(intel_dp);
 }
 
-static struct intel_dp *enc_to_intel_dp(struct drm_encoder *encoder)
+struct intel_dp *enc_to_intel_dp(struct drm_encoder *encoder)
 {
 	return container_of(encoder, struct intel_dp, base.base);
 }
@@ -2339,6 +2339,14 @@ static const struct drm_encoder_helper_funcs intel_dp_helper_funcs = {
 	.commit = intel_dp_commit,
 };
 
+static const struct drm_encoder_helper_funcs intel_dp_helper_funcs_hsw = {
+	.dpms = intel_ddi_dpms,
+	.mode_fixup = intel_dp_mode_fixup,
+	.prepare = intel_dp_prepare,
+	.mode_set = intel_ddi_mode_set_dp,
+	.commit = intel_dp_commit,
+};
+
 static const struct drm_connector_funcs intel_dp_connector_funcs = {
 	.dpms = drm_helper_connector_dpms,
 	.detect = intel_dp_detect,
@@ -2482,7 +2490,13 @@ intel_dp_init(struct drm_device *dev, int output_reg)
 
 	drm_encoder_init(dev, &intel_encoder->base, &intel_dp_enc_funcs,
 			 DRM_MODE_ENCODER_TMDS);
-	drm_encoder_helper_add(&intel_encoder->base, &intel_dp_helper_funcs);
+
+	if (IS_HASWELL(dev))
+		drm_encoder_helper_add(&intel_encoder->base,
+						&intel_dp_helper_funcs_hsw);
+	else
+		drm_encoder_helper_add(&intel_encoder->base,
+						&intel_dp_helper_funcs);
 
 	intel_connector_attach_encoder(intel_connector, intel_encoder);
 	drm_sysfs_connector_add(connector);
drivers/gpu/drm/i915/intel_drv.h +8 -0
index 7cffb12..65b5b7c 100644
@@ -316,6 +316,7 @@ struct intel_dp {
 	struct intel_encoder base;
 	uint32_t output_reg;
 	uint32_t DP;
+	uint32_t TP;
 	uint8_t  link_configuration[DP_LINK_CONFIGURATION_SIZE];
 	bool has_audio;
 	enum hdmi_force_audio force_audio;
@@ -390,6 +391,8 @@ extern void intel_mark_busy(struct drm_device *dev,
 			    struct drm_i915_gem_object *obj);
 extern bool intel_lvds_init(struct drm_device *dev);
 extern void intel_dp_init(struct drm_device *dev, int dp_reg);
+extern struct intel_dp *enc_to_intel_dp(struct drm_encoder *encoder);
+
 void
 intel_dp_set_m_n(struct drm_crtc *crtc, struct drm_display_mode *mode,
 		 struct drm_display_mode *adjusted_mode);
@@ -534,5 +537,10 @@ extern void intel_ddi_dpms(struct drm_encoder *encoder, int mode);
 extern void intel_ddi_mode_set(struct drm_encoder *encoder,
 				struct drm_display_mode *mode,
 				struct drm_display_mode *adjusted_mode);
+void
+intel_ddi_mode_set_dp(struct drm_encoder *encoder,
+			struct drm_display_mode *mode,
+			struct drm_display_mode *adjusted_mode);
+
 
 #endif /* __INTEL_DRV_H__ */
Signature · 1 line
1.7.11.1
#4 lore raw

PATCH 4/21 drm/i915: Haswell specific code for the DP Link Training

In reply to Eugeni Dodonov

From: Shobhit Kumar <[email protected]>

In Haswell the DDIs have moved on to CPU. The AUX control is on LPT. Had to split the code in such a way so as to be able to do source training using DDI_BUF_CTL and DP_TP_CTL and sink side using AUX on LPT.

Also added new routine to get the Volatge and pre-emphashis values. This is now done for DDI_BUF_CTL which now has common bits indicating a combination of voltage and pre-emphasis values

---
 drivers/gpu/drm/i915/i915_reg.h  |   4 ++
 drivers/gpu/drm/i915/intel_ddi.c |  23 +++++++++
 drivers/gpu/drm/i915/intel_dp.c  | 100 ++++++++++++++++++++++++++++++++++++---
 drivers/gpu/drm/i915/intel_drv.h |   3 ++
 4 files changed, 123 insertions(+), 7 deletions(-)
drivers/gpu/drm/i915/i915_reg.h +4 -0
index 51398a8..5b0c5f6 100644
@@ -4308,7 +4308,10 @@
 #define  DP_TP_CTL_LINK_TRAIN_MASK		(7<<8)
 #define  DP_TP_CTL_LINK_TRAIN_PAT1		(0<<8)
 #define  DP_TP_CTL_LINK_TRAIN_PAT2		(1<<8)
+#define  DP_TP_CTL_LINK_TRAIN_IDLE		(2<<8)
 #define  DP_TP_CTL_LINK_TRAIN_NORMAL	(3<<8)
+#define  DP_TP_CTL_LINK_TRAIN_PAT3	(4<<8)
+#define	 DP_TP_CTL_SCRAMBLE_DISABLE	(1<<7)
 
 /* DisplayPort Transport Status */
 #define DP_TP_STATUS_A			0x64044
@@ -4317,6 +4320,7 @@
 					DP_TP_STATUS_A, \
 					DP_TP_STATUS_B)
 #define  DP_TP_STATUS_AUTOTRAIN_DONE	(1<<12)
+#define DP_TP_STATUS_IDLE_DONE		(1<<25)
 
 /* DDI Buffer Control */
 #define DDI_BUF_CTL_A				0x64000
drivers/gpu/drm/i915/intel_ddi.c +23 -0
index 91ed708..f08ce6c 100644
@@ -907,3 +907,26 @@ void intel_ddi_dpms(struct drm_encoder *encoder, int mode)
 				temp);
 	}
 }
+
+void intel_ddi_dp_set_link_train_src(struct intel_dp *intel_dp, u32 dp_reg_value)
+{
+	struct drm_device *dev = intel_dp->base.base.dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	int port = intel_dp->ddi_port;
+
+	/* Configure DP_TP_CTL with training pattern
+	 * TBD: Only SST now; Enable MST later
+	 * */
+
+	I915_WRITE(DP_TP_CTL(port),
+			intel_dp->TP | DP_TP_CTL_SCRAMBLE_DISABLE |
+			DP_TP_CTL_MODE_SST |
+			DP_TP_CTL_ENABLE);
+
+	/* Configure and enable DDI_BUF_CTL for DDI with next voltage */
+	I915_WRITE(DDI_BUF_CTL(port),
+			DDI_BUF_CTL_ENABLE |
+			dp_reg_value);
+
+	udelay(600);
+}
drivers/gpu/drm/i915/intel_dp.c +93 -7

173 lines not on this page · raw message

drivers/gpu/drm/i915/intel_drv.h +3 -0
index 65b5b7c..5ca133d 100644
@@ -543,4 +543,7 @@ intel_ddi_mode_set_dp(struct drm_encoder *encoder,
 			struct drm_display_mode *adjusted_mode);
 
 
+extern void intel_ddi_dp_set_link_train_src(struct intel_dp *intel_dp,
+				u32 dp_reg_value);
+
 #endif /* __INTEL_DRV_H__ */
Signature · 1 line
1.7.11.1
#5 lore raw

PATCH 5/21 drm/i915: Disable DDI Pipe Control on HSW while disabling pipe

In reply to Eugeni Dodonov

From: Shobhit Kumar <[email protected]>

In Haswell while disabling a pipe, we need to disable the DDI control as well along with the PIPECONF. Otherwise we will hit assertions during crtc disable

---
 drivers/gpu/drm/i915/intel_display.c | 7 +++++++
 1 file changed, 7 insertions(+)
drivers/gpu/drm/i915/intel_display.c +7 -0
index 3fbc802..28bee8a 100644
@@ -1794,6 +1794,13 @@ static void intel_disable_pipe(struct drm_i915_private *dev_priv,
 
 	I915_WRITE(reg, val & ~PIPECONF_ENABLE);
 	intel_wait_for_pipe_off(dev_priv->dev, pipe);
+
+	/* On HSW DDI Pipe control has to be disabled as well */
+	if (IS_HASWELL(dev_priv->dev)) {
+		val = I915_READ(DDI_FUNC_CTL(pipe));
+		val = val  & (~PIPE_DDI_FUNC_ENABLE);
+		I915_WRITE(DDI_FUNC_CTL(pipe), val);
+	}
 }
 
 /*
Signature · 1 line
1.7.11.1
#6 lore raw

PATCH 6/21 drm/i915: Hook DP init in ddi module

In reply to Eugeni Dodonov

From: Shobhit Kumar <[email protected]>

TBD: As of now just initializes DP connectors. Later we need to detect the actual output connected on the port(HDMI or DP) using AUX channel detection and appropriately call the initialization routines

v2: Init DP connector on DDIB and DDID, and HDMI on DDIC.

---
 drivers/gpu/drm/i915/intel_ddi.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_ddi.c +5 -1
index f08ce6c..effb263 100644
@@ -239,10 +239,14 @@ void intel_ddi_init(struct drm_device *dev, enum port port)
 		break;
 	/* Assume that the  ports B, C and D are working in HDMI mode for now */
 	case PORT_B:
+		intel_dp_init(dev, PCH_DP_B);
+		break;
 	case PORT_C:
-	case PORT_D:
 		intel_hdmi_init(dev, DDI_BUF_CTL(port));
 		break;
+	case PORT_D:
+		intel_dp_init(dev, PCH_DP_D);
+		break;
 	default:
 		DRM_DEBUG_DRIVER("No handlers defined for port %d, skipping DDI initialization\n",
 				port);
Signature · 1 line
1.7.11.1
#7 lore raw

PATCH 7/21 drm/i915: re-initialize DDI buffer translations after resume

In reply to Eugeni Dodonov

This is necessary for the modesetting to work correctly after a suspend-resume cycle. Without this, the pipes and clocks got the correct configuration, but the underlying DDI buffers configuration was lost.

---
 drivers/gpu/drm/i915/intel_display.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/intel_display.c +2 -2
index 28bee8a..76508a7 100644
@@ -7179,6 +7179,8 @@ static void ivb_pch_pwm_override(struct drm_device *dev)
 
 void intel_modeset_init_hw(struct drm_device *dev)
 {
+	intel_prepare_ddi(dev);
+
 	intel_init_clock_gating(dev);
 
 	mutex_lock(&dev->struct_mutex);
@@ -7208,8 +7210,6 @@ void intel_modeset_init(struct drm_device *dev)
 
 	intel_init_pm(dev);
 
-	intel_prepare_ddi(dev);
-
 	intel_init_display(dev);
 
 	if (IS_GEN2(dev)) {
Signature · 1 line
1.7.11.1
Reviewed-by on v1 7/21
#8 lore raw

Re: [PATCH 07/21] drm/i915: re-initialize DDI buffer translations after resume

2012/6/28 Eugeni Dodonov <[email protected]>:

Quoted 5 lines

This is necessary for the modesetting to work correctly after a suspend-resume cycle. Without this, the pipes and clocks got the correct configuration, but the underlying DDI buffers configuration was lost.

I couldn't make suspend/resume work on my machine yet, but running intel_prepare_ddi again after resume won't hurt us, so:

Quoted 33 lines
---
 drivers/gpu/drm/i915/intel_display.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/intel_display.c +2 -3
index 28bee8a..76508a7 100644
@@ -7179,6 +7179,8 @@ static void ivb_pch_pwm_override(struct drm_device *dev)
 
 void intel_modeset_init_hw(struct drm_device *dev)
 {
+       intel_prepare_ddi(dev);
+
        intel_init_clock_gating(dev);
 
        mutex_lock(&dev->struct_mutex);
@@ -7208,8 +7210,6 @@ void intel_modeset_init(struct drm_device *dev)
 
        intel_init_pm(dev);
 
-       intel_prepare_ddi(dev);
-
        intel_init_display(dev);
 
        if (IS_GEN2(dev)) {
--

1.7.11.1

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

Signature · 1 line
Paulo Zanoni
#9 lore raw

Re: [PATCH 07/21] drm/i915: re-initialize DDI buffer translations after resume

Quoted 11 lines · Paulo Zanoni

2012/6/28 Eugeni Dodonov <[email protected]>:

Quoted 5 lines

This is necessary for the modesetting to work correctly after a suspend-resume cycle. Without this, the pipes and clocks got the correct configuration, but the underlying DDI buffers configuration was lost.

I couldn't make suspend/resume work on my machine yet, but running intel_prepare_ddi again after resume won't hurt us, so:

Queued for -next, thanks for the patch. Quick aside: I think we have a similar issue with the hsw power well init code.
-Daniel

Quoted 43 lines
Quoted 34 lines
---
 drivers/gpu/drm/i915/intel_display.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/intel_display.c +2 -3
index 28bee8a..76508a7 100644
@@ -7179,6 +7179,8 @@ static void ivb_pch_pwm_override(struct drm_device *dev)
 
 void intel_modeset_init_hw(struct drm_device *dev)
 {
+       intel_prepare_ddi(dev);
+
        intel_init_clock_gating(dev);
 
        mutex_lock(&dev->struct_mutex);
@@ -7208,8 +7210,6 @@ void intel_modeset_init(struct drm_device *dev)
 
        intel_init_pm(dev);
 
-       intel_prepare_ddi(dev);
-
        intel_init_display(dev);
 
        if (IS_GEN2(dev)) {
--

1.7.11.1

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

Signature · 5 lines
Paulo Zanoni
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter
Mail: [email protected]
Mobile: +41 (0)79 365 57 48
#10 lore raw

Re: [PATCH 07/21] drm/i915: re-initialize DDI buffer translations after resume

Quoted 8 lines · Daniel Vetter
Quoted 7 lines · Paulo Zanoni

2012/6/28 Eugeni Dodonov <[email protected]>:

Quoted 5 lines

This is necessary for the modesetting to work correctly after a suspend-resume cycle. Without this, the pipes and clocks got the correct configuration, but the underlying DDI buffers configuration was lost.

Quoted 7 lines
Quoted 4 lines

I couldn't make suspend/resume work on my machine yet, but running intel_prepare_ddi again after resume won't hurt us, so:

Queued for -next, thanks for the patch. Quick aside: I think we have a similar issue with the hsw power well init code.

Yes, I'll send the power wells patch until the end of the week as well, sorry for taking that long!

Eugeni

#11 lore raw

PATCH 8/21 drm/i915: simplify FDI RX check for LPT

In reply to Eugeni Dodonov

On LPT onwards, there is only one FDI receiver, but any pipe can drive it. For now, we consider that only pipeA can work in CRT mode, so if any other pipe attempts to enable FDI receiver it is considered invalid behavior.

---
 drivers/gpu/drm/i915/intel_display.c | 36 ++++++++++++++++++++++++++----------
 1 file changed, 26 insertions(+), 10 deletions(-)
drivers/gpu/drm/i915/intel_display.c +26 -10
index 76508a7..e8fd6cc 100644
@@ -1128,14 +1128,23 @@ static void assert_fdi_rx(struct drm_i915_private *dev_priv,
 	u32 val;
 	bool cur_state;
 
-	if (IS_HASWELL(dev_priv->dev) && pipe > 0) {
-			DRM_ERROR("Attempting to enable FDI_RX on Haswell pipe > 0\n");
+	/* On LPT, there is only one FDI RX that can be used.  At the same
+	 * time, any pipe can connect to it, but only one at a time.  To
+	 * simplify the enabling, we consider that only pipeA can be used in
+	 * FDI configuration.
+	 */
+	if (HAS_PCH_LPT(dev_priv->dev)) {
+		if (pipe > 0) {
+			DRM_ERROR("Attempting to %s FDI_RX on Haswell pipe > 0\n",
+					(state==true) ? "enable" : "disable");
 			return;
-	} else {
-		reg = FDI_RX_CTL(pipe);
-		val = I915_READ(reg);
-		cur_state = !!(val & FDI_RX_ENABLE);
+		}
 	}
+
+	reg = FDI_RX_CTL(pipe);
+	val = I915_READ(reg);
+	cur_state = !!(val & FDI_RX_ENABLE);
+
 	WARN(cur_state != state,
 	     "FDI RX state assertion failure (expected %s, current %s)\n",
 	     state_string(state), state_string(cur_state));
@@ -1651,6 +1660,17 @@ static void intel_enable_transcoder(struct drm_i915_private *dev_priv,
 	/* PCH only available on ILK+ */
 	BUG_ON(dev_priv->info->gen < 5);
 
+	if (HAS_PCH_LPT(dev_priv->dev)) {
+		/* On LPT, there is only one transcoder, but any pipe can be used with it.
+		 * To simplify things, we consider that only pipeA can be used with CRT,
+		 * so we handle this check in the same way as with the FDI receiver.
+		 */
+		if (pipe > 0) {
+			DRM_ERROR("Attempting to enable transcoder on Lynx point with pipe > 0\n");
+			return;
+		}
+	}
+
 	/* Make sure PCH DPLL is enabled */
 	assert_pch_pll_enabled(dev_priv,
 			       to_intel_crtc(crtc)->pch_pll,
@@ -1660,10 +1680,6 @@ static void intel_enable_transcoder(struct drm_i915_private *dev_priv,
 	assert_fdi_tx_enabled(dev_priv, pipe);
 	assert_fdi_rx_enabled(dev_priv, pipe);
 
-	if (IS_HASWELL(dev_priv->dev) && pipe > 0) {
-		DRM_ERROR("Attempting to enable transcoder on Haswell with pipe > 0\n");
-		return;
-	}
 	reg = TRANSCONF(pipe);
 	val = I915_READ(reg);
 	pipeconf_val = I915_READ(PIPECONF(pipe));
Signature · 1 line
1.7.11.1
#12 lore raw

PATCH 9/21 drm/i915: account for only one transcoder on LPT

In reply to Eugeni Dodonov

Same as with FDI RX, Lynx point only has one transcoder, which can be driven by any pipe. So consider that only pipeA will be used with CRT for now.

---
 drivers/gpu/drm/i915/intel_display.c | 19 +++++++++++++++----
 1 file changed, 15 insertions(+), 4 deletions(-)
drivers/gpu/drm/i915/intel_display.c +15 -4
index e8fd6cc..d7b337b 100644
@@ -1660,11 +1660,12 @@ static void intel_enable_transcoder(struct drm_i915_private *dev_priv,
 	/* PCH only available on ILK+ */
 	BUG_ON(dev_priv->info->gen < 5);
 
+	/* On LPT, there is only one transcoder, but any pipe can be used with
+	 * it.  To simplify things, we consider that only pipeA can be used
+	 * with CRT, so we handle this check in the same way as with the FDI
+	 * receiver.
+	 */
 	if (HAS_PCH_LPT(dev_priv->dev)) {
-		/* On LPT, there is only one transcoder, but any pipe can be used with it.
-		 * To simplify things, we consider that only pipeA can be used with CRT,
-		 * so we handle this check in the same way as with the FDI receiver.
-		 */
 		if (pipe > 0) {
 			DRM_ERROR("Attempting to enable transcoder on Lynx point with pipe > 0\n");
 			return;
@@ -1714,6 +1715,16 @@ static void intel_disable_transcoder(struct drm_i915_private *dev_priv,
 	int reg;
 	u32 val;
 
+	/* On Lynx Point, there is only one transcoder, which we assume it is
+	 * mapped into pipe 0 for simplicity.
+	 */
+	if (HAS_PCH_LPT(dev_priv->dev)) {
+		if (pipe > 0) {
+			DRM_ERROR("Attempting to disable transcoder on Lynx Point with pipe > 0\n");
+			return;
+		}
+	}
+
 	/* FDI relies on the transcoder */
 	assert_fdi_tx_disabled(dev_priv, pipe);
 	assert_fdi_rx_disabled(dev_priv, pipe);
Signature · 1 line
1.7.11.1
#13 lore raw

PATCH 10/21 drm/i915: introduce lpt_enable_pch and cpt_enable_pch

In reply to Eugeni Dodonov

CPT/PPT and LPT have different functionality. So we introduce specific functions to handle each of them instead of using multiple if.. statements.

---
 drivers/gpu/drm/i915/intel_display.c | 70 ++++++++++++++++++++++++++++++++----
 1 file changed, 63 insertions(+), 7 deletions(-)
drivers/gpu/drm/i915/intel_display.c +63 -7
index d7b337b..9a695ab 100644
@@ -2968,6 +2968,51 @@ static void lpt_program_iclkip(struct drm_crtc *crtc)
 	I915_WRITE(PIXCLK_GATE, PIXCLK_GATE_UNGATE);
 }
 
+/* On Lynx Point, the PCH part is different, as it has one FDI RX, one
+ * transcoder, and the clock is driven by iCLKIP.
+ */
+static void lpt_pch_enable(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 pipe = intel_crtc->pipe;
+
+	/* On LPT, there is only one FDI RX and transcoder, which can be driven
+	 * by any pipe. To simplify, we consider that only pipeA can be used in
+	 * such mode.
+	 */
+	if (HAS_PCH_LPT(dev_priv->dev)) {
+		if (pipe > 0) {
+			DRM_ERROR("Attempting to enable PCH port with pipe > 0\n");
+			return;
+		}
+	}
+
+	assert_transcoder_disabled(dev_priv, pipe);
+
+	/* For PCH output, training FDI link */
+	dev_priv->display.fdi_link_train(crtc);
+
+	intel_enable_pch_pll(intel_crtc);
+
+	DRM_DEBUG_KMS("LPT detected: programming iCLKIP\n");
+	lpt_program_iclkip(crtc);
+
+	/* set transcoder timing, panel must allow it */
+	assert_panel_unlocked(dev_priv, pipe);
+	I915_WRITE(TRANS_HTOTAL(pipe), I915_READ(HTOTAL(pipe)));
+	I915_WRITE(TRANS_HBLANK(pipe), I915_READ(HBLANK(pipe)));
+	I915_WRITE(TRANS_HSYNC(pipe),  I915_READ(HSYNC(pipe)));
+
+	I915_WRITE(TRANS_VTOTAL(pipe), I915_READ(VTOTAL(pipe)));
+	I915_WRITE(TRANS_VBLANK(pipe), I915_READ(VBLANK(pipe)));
+	I915_WRITE(TRANS_VSYNC(pipe),  I915_READ(VSYNC(pipe)));
+	I915_WRITE(TRANS_VSYNCSHIFT(pipe),  I915_READ(VSYNCSHIFT(pipe)));
+
+	intel_enable_transcoder(dev_priv, pipe);
+}
+
 /*
  * Enable PCH resources required for PCH ports:
  *   - PCH PLLs
@@ -2976,7 +3021,7 @@ static void lpt_program_iclkip(struct drm_crtc *crtc)
  *   - DP transcoding bits
  *   - transcoder
  */
-static void ironlake_pch_enable(struct drm_crtc *crtc)
+static void cpt_pch_enable(struct drm_crtc *crtc)
 {
 	struct drm_device *dev = crtc->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -2991,10 +3036,7 @@ static void ironlake_pch_enable(struct drm_crtc *crtc)
 
 	intel_enable_pch_pll(intel_crtc);
 
-	if (HAS_PCH_LPT(dev)) {
-		DRM_DEBUG_KMS("LPT detected: programming iCLKIP\n");
-		lpt_program_iclkip(crtc);
-	} else if (HAS_PCH_CPT(dev)) {
+	if (HAS_PCH_CPT(dev)) {
 		u32 sel;
 
 		temp = I915_READ(PCH_DPLL_SEL);
@@ -3031,8 +3073,7 @@ static void ironlake_pch_enable(struct drm_crtc *crtc)
 	I915_WRITE(TRANS_VSYNC(pipe),  I915_READ(VSYNC(pipe)));
 	I915_WRITE(TRANS_VSYNCSHIFT(pipe),  I915_READ(VSYNCSHIFT(pipe)));
 
-	if (!IS_HASWELL(dev))
-		intel_fdi_normal_train(crtc);
+	intel_fdi_normal_train(crtc);
 
 	/* For PCH DP, enable TRANS_DP_CTL */
 	if (HAS_PCH_CPT(dev) &&
@@ -3075,6 +3116,21 @@ static void ironlake_pch_enable(struct drm_crtc *crtc)
 	intel_enable_transcoder(dev_priv, pipe);
 }
 
+/*
+ * When we need to use a PCH port, we can go through different paths on CPT/PPT
+ * and LPT chipsets.
+ */
+static void ironlake_pch_enable(struct drm_crtc *crtc)
+{
+	struct drm_device *dev = crtc->dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+
+	if (HAS_PCH_LPT(dev_priv->dev))
+		return lpt_pch_enable(crtc);
+	else
+		return cpt_pch_enable(crtc);
+}
+
 static void intel_put_pch_pll(struct intel_crtc *intel_crtc)
 {
 	struct intel_pch_pll *pll = intel_crtc->pch_pll;
Signature · 1 line
1.7.11.1
#14 lore raw

Re: [PATCH 10/21] drm/i915: introduce lpt_enable_pch and cpt_enable_pch

Hi

I certainly like this function split. Reviews inline:

2012/6/28 Eugeni Dodonov <[email protected]>:

Quoted 27 lines

CPT/PPT and LPT have different functionality. So we introduce specific functions to handle each of them instead of using multiple if.. statements.

---
 drivers/gpu/drm/i915/intel_display.c | 70 ++++++++++++++++++++++++++++++++----
 1 file changed, 63 insertions(+), 7 deletions(-)
drivers/gpu/drm/i915/intel_display.c +10 -0
index d7b337b..9a695ab 100644
@@ -2968,6 +2968,51 @@ static void lpt_program_iclkip(struct drm_crtc *crtc)
        I915_WRITE(PIXCLK_GATE, PIXCLK_GATE_UNGATE);
 }
 
+/* On Lynx Point, the PCH part is different, as it has one FDI RX, one
+ * transcoder, and the clock is driven by iCLKIP.
+ */
+static void lpt_pch_enable(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 pipe = intel_crtc->pipe;
+

drivers/gpu/drm/i915/intel_display.c static void lpt_program_iclkip(struct drm_crtc *crtc) · line 2990

Quoted 10 lines
drivers/gpu/drm/i915/intel_display.c +10 -0
+       /* On LPT, there is only one FDI RX and transcoder, which can be driven
+        * by any pipe. To simplify, we consider that only pipeA can be used in
+        * such mode.
+        */
+       if (HAS_PCH_LPT(dev_priv->dev)) {
+               if (pipe > 0) {
+                       DRM_ERROR("Attempting to enable PCH port with pipe > 0\n");
+                       return;
+               }
+       }

It looks like the check above could go on a separate commit. The first commit just moves code around, the second one adds the check.

I may have misunderstood something about this code which is not familiar to me, but instead of limiting everything to pipe 0, can't we just iterate through all pipes, read DDI_FUNC_CTL(pipe) and see if there's any other enabled pipe on FDI mode?

drivers/gpu/drm/i915/intel_display.c static void lpt_program_iclkip(struct drm_crtc *crtc) · line 2999

Quoted 9 lines
drivers/gpu/drm/i915/intel_display.c +9 -0
+
+       assert_transcoder_disabled(dev_priv, pipe);
+
+       /* For PCH output, training FDI link */
+       dev_priv->display.fdi_link_train(crtc);
+
+       intel_enable_pch_pll(intel_crtc);
+
+       DRM_DEBUG_KMS("LPT detected: programming iCLKIP\n");

This message could go inside lpt_program_iclkip :) We should also paint it green.

Quoted 20 lines
drivers/gpu/drm/i915/intel_display.c +16 -0
+       lpt_program_iclkip(crtc);
+
+       /* set transcoder timing, panel must allow it */
+       assert_panel_unlocked(dev_priv, pipe);
+       I915_WRITE(TRANS_HTOTAL(pipe), I915_READ(HTOTAL(pipe)));
+       I915_WRITE(TRANS_HBLANK(pipe), I915_READ(HBLANK(pipe)));
+       I915_WRITE(TRANS_HSYNC(pipe),  I915_READ(HSYNC(pipe)));
+
+       I915_WRITE(TRANS_VTOTAL(pipe), I915_READ(VTOTAL(pipe)));
+       I915_WRITE(TRANS_VBLANK(pipe), I915_READ(VBLANK(pipe)));
+       I915_WRITE(TRANS_VSYNC(pipe),  I915_READ(VSYNC(pipe)));
+       I915_WRITE(TRANS_VSYNCSHIFT(pipe),  I915_READ(VSYNCSHIFT(pipe)));
+
+       intel_enable_transcoder(dev_priv, pipe);
+}
+
 /*
  * Enable PCH resources required for PCH ports:
  *   - PCH PLLs

drivers/gpu/drm/i915/intel_display.c static void lpt_program_iclkip(struct drm_crtc *crtc) · line 3024

Quoted 5 lines
drivers/gpu/drm/i915/intel_display.c +1 -1
@@ -2976,7 +3021,7 @@ static void lpt_program_iclkip(struct drm_crtc *crtc)
  *   - DP transcoding bits
  *   - transcoder
  */
-static void ironlake_pch_enable(struct drm_crtc *crtc)
+static void cpt_pch_enable(struct drm_crtc *crtc)

Since this will run on both IBX and CPT, shouldn't we call it ibx_pch_enable? At least on intel_hdmi.c, the functions are named after the earliest generation that can run them.

Quoted 54 lines
drivers/gpu/drm/i915/intel_display.c +17 -7
 {
        struct drm_device *dev = crtc->dev;
        struct drm_i915_private *dev_priv = dev->dev_private;
@@ -2991,10 +3036,7 @@ static void ironlake_pch_enable(struct drm_crtc *crtc)
 
        intel_enable_pch_pll(intel_crtc);
 
-       if (HAS_PCH_LPT(dev)) {
-               DRM_DEBUG_KMS("LPT detected: programming iCLKIP\n");
-               lpt_program_iclkip(crtc);
-       } else if (HAS_PCH_CPT(dev)) {
+       if (HAS_PCH_CPT(dev)) {
                u32 sel;
 
                temp = I915_READ(PCH_DPLL_SEL);
@@ -3031,8 +3073,7 @@ static void ironlake_pch_enable(struct drm_crtc *crtc)
        I915_WRITE(TRANS_VSYNC(pipe),  I915_READ(VSYNC(pipe)));
        I915_WRITE(TRANS_VSYNCSHIFT(pipe),  I915_READ(VSYNCSHIFT(pipe)));
 
-       if (!IS_HASWELL(dev))
-               intel_fdi_normal_train(crtc);
+       intel_fdi_normal_train(crtc);
 
        /* For PCH DP, enable TRANS_DP_CTL */
        if (HAS_PCH_CPT(dev) &&
@@ -3075,6 +3116,21 @@ static void ironlake_pch_enable(struct drm_crtc *crtc)
        intel_enable_transcoder(dev_priv, pipe);
 }
 
+/*
+ * When we need to use a PCH port, we can go through different paths on CPT/PPT
+ * and LPT chipsets.
+ */
+static void ironlake_pch_enable(struct drm_crtc *crtc)
+{
+       struct drm_device *dev = crtc->dev;
+       struct drm_i915_private *dev_priv = dev->dev_private;
+
+       if (HAS_PCH_LPT(dev_priv->dev))
+               return lpt_pch_enable(crtc);
+       else
+               return cpt_pch_enable(crtc);
+}
+
 static void intel_put_pch_pll(struct intel_crtc *intel_crtc)
 {
        struct intel_pch_pll *pll = intel_crtc->pch_pll;
--

1.7.11.1

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

Signature · 1 line
Paulo Zanoni
#15 lore raw

Re: [PATCH 10/21] drm/i915: introduce lpt_enable_pch and cpt_enable_pch

Quoted 2 lines · Paulo Zanoni

It looks like the check above could go on a separate commit. The first commit just moves code around, the second one adds the check.

This patch attempts to reset cpt_pch_enable to what ironlake_pch_enable was before we added LPT support into it; and make lpt_pch_enable account for LPT checks and such.

But yes, I can split it into several commits.

Quoted 4 lines

I may have misunderstood something about this code which is not familiar to me, but instead of limiting everything to pipe 0, can't we just iterate through all pipes, read DDI_FUNC_CTL(pipe) and see if there's any other enabled pipe on FDI mode?

Yes, the idea in those asserts and checks was that as only 1 pipe can work in FDI mode, and as we only have 1 FDI RX, 1 PCH Transcoder and so on, we simple forced pipe1 to work in this mode, as this way most of older code which uses macros still works.

But there is certainly a room for improvements.

Quoted 6 lines

drivers/gpu/drm/i915/intel_display.c static void lpt_program_iclkip(struct drm_crtc *crtc) · line 3024

Quoted 2 lines
drivers/gpu/drm/i915/intel_display.c +1 -1
-static void ironlake_pch_enable(struct drm_crtc *crtc)
+static void cpt_pch_enable(struct drm_crtc *crtc)

Since this will run on both IBX and CPT, shouldn't we call it ibx_pch_enable? At least on intel_hdmi.c, the functions are named after the earliest generation that can run them.

Yes, I can rename it to ibx_pch_enable.

I'll do those changes and split this one into smaller commits and will resend later.

Thanks!
Eugeni

#16 lore raw

PATCH 11/21 drm/i915: program FDI_RX TP and FDI delays

In reply to Eugeni Dodonov

This is required for a stable FDI connection.

---
 drivers/gpu/drm/i915/i915_reg.h  |  3 +++
 drivers/gpu/drm/i915/intel_ddi.c | 10 ++++++++++
 2 files changed, 13 insertions(+)
drivers/gpu/drm/i915/i915_reg.h +3 -0
index 5b0c5f6..284965b 100644
@@ -3849,6 +3849,9 @@
 #define _FDI_RXA_TUSIZE2         0xf0038
 #define _FDI_RXB_TUSIZE1         0xf1030
 #define _FDI_RXB_TUSIZE2         0xf1038
+#define  FDI_RX_TP1_TO_TP2_48	(10<<20)
+#define  FDI_RX_TP1_TO_TP2_64	(11<<20)
+#define  FDI_RX_FDI_DELAY_90	(0x90<<0)
 #define FDI_RX_MISC(pipe) _PIPE(pipe, _FDI_RXA_MISC, _FDI_RXB_MISC)
 #define FDI_RX_TUSIZE1(pipe) _PIPE(pipe, _FDI_RXA_TUSIZE1, _FDI_RXB_TUSIZE1)
 #define FDI_RX_TUSIZE2(pipe) _PIPE(pipe, _FDI_RXA_TUSIZE2, _FDI_RXB_TUSIZE2)
drivers/gpu/drm/i915/intel_ddi.c +10 -0
index effb263..1c76d20 100644
@@ -170,6 +170,16 @@ void hsw_fdi_link_train(struct drm_crtc *crtc)
 
 		udelay(600);
 
+		/* We need to program FDI_RX_MISC with the default TP1 to TP2
+		 * values before enabling the receiver, and configure the delay
+		 * for the FDI timing generator to 90h.
+		 */
+		reg = FDI_RX_MISC(pipe);
+		temp = I915_READ(reg);
+		temp |= FDI_RX_TP1_TO_TP2_48 |
+			FDI_RX_FDI_DELAY_90;
+		I915_WRITE(reg, temp);
+
 		/* Enable CPU FDI Receiver with auto-training */
 		reg = FDI_RX_CTL(pipe);
 		I915_WRITE(reg,
Signature · 1 line
1.7.11.1
#17 lore raw

Re: [PATCH 11/21] drm/i915: program FDI_RX TP and FDI delays

Hi

2012/6/28 Eugeni Dodonov <[email protected]>:

Quoted 13 lines

This is required for a stable FDI connection.

---
 drivers/gpu/drm/i915/i915_reg.h  |  3 +++
 drivers/gpu/drm/i915/intel_ddi.c | 10 ++++++++++
 2 files changed, 13 insertions(+)
drivers/gpu/drm/i915/i915_reg.h +0 -0
index 5b0c5f6..284965b 100644

drivers/gpu/drm/i915/i915_reg.h line 3853

Quoted 5 lines
drivers/gpu/drm/i915/i915_reg.h +2 -0
@@ -3849,6 +3849,9 @@
 #define _FDI_RXA_TUSIZE2         0xf0038
 #define _FDI_RXB_TUSIZE1         0xf1030
 #define _FDI_RXB_TUSIZE2         0xf1038
+#define  FDI_RX_TP1_TO_TP2_48  (10<<20)
+#define  FDI_RX_TP1_TO_TP2_64  (11<<20)

Isn't this supposed to be (2 << 20) and (3 << 20)?

Quoted 10 lines
drivers/gpu/drm/i915/i915_reg.h +1 -0
+#define  FDI_RX_FDI_DELAY_90   (0x90<<0)
 #define FDI_RX_MISC(pipe) _PIPE(pipe, _FDI_RXA_MISC, _FDI_RXB_MISC)
 #define FDI_RX_TUSIZE1(pipe) _PIPE(pipe, _FDI_RXA_TUSIZE1, _FDI_RXB_TUSIZE1)
 #define FDI_RX_TUSIZE2(pipe) _PIPE(pipe, _FDI_RXA_TUSIZE2, _FDI_RXB_TUSIZE2)
drivers/gpu/drm/i915/intel_ddi.c +0 -0
index effb263..1c76d20 100644
@@ -170,6 +170,16 @@ void hsw_fdi_link_train(struct drm_crtc *crtc)
 

drivers/gpu/drm/i915/intel_ddi.c void hsw_fdi_link_train(struct drm_crtc *crtc) · line 180

Quoted 10 lines
drivers/gpu/drm/i915/intel_ddi.c +8 -0
                udelay(600);
 
+               /* We need to program FDI_RX_MISC with the default TP1 to TP2
+                * values before enabling the receiver, and configure the delay
+                * for the FDI timing generator to 90h.
+                */
+               reg = FDI_RX_MISC(pipe);
+               temp = I915_READ(reg);
+               temp |= FDI_RX_TP1_TO_TP2_48 |
+                       FDI_RX_FDI_DELAY_90;

Shouldn't we mask bits 27:26, 25:24, 21:20 and 12:0 first?

Quoted 12 lines
drivers/gpu/drm/i915/intel_ddi.c +2 -1
+               I915_WRITE(reg, temp);
+
                /* Enable CPU FDI Receiver with auto-training */
                reg = FDI_RX_CTL(pipe);
                I915_WRITE(reg,
--

1.7.11.1

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

Signature · 1 line
Paulo Zanoni
#18 lore raw

PATCH 10/31 drm/i915: program FDI_RX TP and FDI delays

This is required for a stable FDI connection.

v2: fix and simplify the FDI_RX_MISC bits as noticed by Paulo Zanoni.

---
 drivers/gpu/drm/i915/i915_reg.h  | 3 +++
 drivers/gpu/drm/i915/intel_ddi.c | 9 +++++++++
 2 files changed, 12 insertions(+)
drivers/gpu/drm/i915/i915_reg.h +3 -0
index 5b0c5f6..287d277 100644
@@ -3849,6 +3849,9 @@
 #define _FDI_RXA_TUSIZE2         0xf0038
 #define _FDI_RXB_TUSIZE1         0xf1030
 #define _FDI_RXB_TUSIZE2         0xf1038
+#define  FDI_RX_TP1_TO_TP2_48	(2<<20)
+#define  FDI_RX_TP1_TO_TP2_64	(3<<20)
+#define  FDI_RX_FDI_DELAY_90	(0x90<<0)
 #define FDI_RX_MISC(pipe) _PIPE(pipe, _FDI_RXA_MISC, _FDI_RXB_MISC)
 #define FDI_RX_TUSIZE1(pipe) _PIPE(pipe, _FDI_RXA_TUSIZE1, _FDI_RXB_TUSIZE1)
 #define FDI_RX_TUSIZE2(pipe) _PIPE(pipe, _FDI_RXA_TUSIZE2, _FDI_RXB_TUSIZE2)
drivers/gpu/drm/i915/intel_ddi.c +9 -0
index effb263..2d5acd2 100644
@@ -170,6 +170,15 @@ void hsw_fdi_link_train(struct drm_crtc *crtc)
 
 		udelay(600);
 
+		/* We need to program FDI_RX_MISC with the default TP1 to TP2
+		 * values before enabling the receiver, and configure the delay
+		 * for the FDI timing generator to 90h. Luckily, all the other
+		 * bits are supposed to be zeroed, so we can write those values
+		 * directly.
+		 */
+		I915_WRITE(FDI_RX_MISC(pipe), FDI_RX_TP1_TO_TP2_48 |
+				FDI_RX_FDI_DELAY_90);
+
 		/* Enable CPU FDI Receiver with auto-training */
 		reg = FDI_RX_CTL(pipe);
 		I915_WRITE(reg,
Signature · 1 line
1.7.11.1
Reviewed-by on v1 10/31
#19 lore raw

Re: [PATCH 10/31] drm/i915: program FDI_RX TP and FDI delays

2012/7/4 Eugeni Dodonov <[email protected]>:

Quoted 5 lines

This is required for a stable FDI connection.

v2: fix and simplify the FDI_RX_MISC bits as noticed by Paulo Zanoni.

CC Paulo Zanoni <[email protected]>
Quoted 47 lines
---
 drivers/gpu/drm/i915/i915_reg.h  | 3 +++
 drivers/gpu/drm/i915/intel_ddi.c | 9 +++++++++
 2 files changed, 12 insertions(+)
drivers/gpu/drm/i915/i915_reg.h +3 -0
index 5b0c5f6..287d277 100644
@@ -3849,6 +3849,9 @@
 #define _FDI_RXA_TUSIZE2         0xf0038
 #define _FDI_RXB_TUSIZE1         0xf1030
 #define _FDI_RXB_TUSIZE2         0xf1038
+#define  FDI_RX_TP1_TO_TP2_48  (2<<20)
+#define  FDI_RX_TP1_TO_TP2_64  (3<<20)
+#define  FDI_RX_FDI_DELAY_90   (0x90<<0)
 #define FDI_RX_MISC(pipe) _PIPE(pipe, _FDI_RXA_MISC, _FDI_RXB_MISC)
 #define FDI_RX_TUSIZE1(pipe) _PIPE(pipe, _FDI_RXA_TUSIZE1, _FDI_RXB_TUSIZE1)
 #define FDI_RX_TUSIZE2(pipe) _PIPE(pipe, _FDI_RXA_TUSIZE2, _FDI_RXB_TUSIZE2)
drivers/gpu/drm/i915/intel_ddi.c +9 -1
index effb263..2d5acd2 100644
@@ -170,6 +170,15 @@ void hsw_fdi_link_train(struct drm_crtc *crtc)
 
                udelay(600);
 
+               /* We need to program FDI_RX_MISC with the default TP1 to TP2
+                * values before enabling the receiver, and configure the delay
+                * for the FDI timing generator to 90h. Luckily, all the other
+                * bits are supposed to be zeroed, so we can write those values
+                * directly.
+                */
+               I915_WRITE(FDI_RX_MISC(pipe), FDI_RX_TP1_TO_TP2_48 |
+                               FDI_RX_FDI_DELAY_90);
+
                /* Enable CPU FDI Receiver with auto-training */
                reg = FDI_RX_CTL(pipe);
                I915_WRITE(reg,
--

1.7.11.1

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

Signature · 1 line
Paulo Zanoni
#20 lore raw

Re: [PATCH 10/31] drm/i915: program FDI_RX TP and FDI delays

Quoted 8 lines · Paulo Zanoni

2012/7/4 Eugeni Dodonov <[email protected]>:

Quoted 5 lines

This is required for a stable FDI connection.

v2: fix and simplify the FDI_RX_MISC bits as noticed by Paulo Zanoni.

CC Paulo Zanoni <[email protected]>

Patch queued for -next, thanks.
-Daniel

Signature · 3 lines
Daniel Vetter
Mail: [email protected]
Mobile: +41 (0)79 365 57 48
#21 lore raw

PATCH 12/21 drm/i915: support Haswell-style force waking

In reply to Eugeni Dodonov

On Haswell, there is a different register for reading force wake ACKs, and all the writes should go into the multi-threaded register, even for the legacy force wake.

Also, we have a theorical possibility for the force wake sequence to awaken the GT, but return while it hasn't finished bringing up the queue. So we properly check those bits as well, to ensure we won't end up in half-woken situation.

---
 drivers/gpu/drm/i915/i915_drv.c | 36 +++++++++++++++++++++++++++++++++++-
 drivers/gpu/drm/i915/i915_drv.h |  3 +++
 drivers/gpu/drm/i915/i915_reg.h |  1 +
 drivers/gpu/drm/i915/intel_pm.c |  4 +++-
 4 files changed, 42 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/i915_drv.c +35 -1
index 79be879..73fd38a 100644
@@ -536,6 +536,39 @@ int __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv)
 	return ret;
 }
 
+static inline void __hsw_wait_gt_awake(struct drm_i915_private *dev_priv)
+{
+	int count = 0;
+
+	while (count++ < 50) {
+		u32 tmp = I915_READ_NOTRACE(FORCEWAKE_ACK_HSW);
+		if ((tmp & 1) && !(tmp & _MASKED_BIT_ENABLE(~7)))
+			break;
+		udelay(10);
+	}
+}
+
+
+void hsw_gt_force_wake_get(struct drm_i915_private *dev_priv)
+{
+	__hsw_wait_gt_awake(dev_priv);
+
+	I915_WRITE_NOTRACE(FORCEWAKE_MT, _MASKED_BIT_ENABLE(1));
+	POSTING_READ(FORCEWAKE_MT);
+
+	__hsw_wait_gt_awake(dev_priv);
+}
+
+void hsw_gt_force_wake_put(struct drm_i915_private *dev_priv)
+{
+	I915_WRITE_NOTRACE(FORCEWAKE_MT, _MASKED_BIT_DISABLE(1));
+
+	__hsw_wait_gt_awake(dev_priv);
+
+	/* The below doubles as a POSTING_READ */
+	gen6_gt_check_fifodbg(dev_priv);
+}
+
 void vlv_force_wake_get(struct drm_i915_private *dev_priv)
 {
 	int count;
@@ -1161,7 +1194,8 @@ MODULE_LICENSE("GPL and additional rights");
 #define NEEDS_FORCE_WAKE(dev_priv, reg) \
 	((HAS_FORCE_WAKE((dev_priv)->dev)) && \
 	 ((reg) < 0x40000) &&            \
-	 ((reg) != FORCEWAKE))
+	 ((reg) != FORCEWAKE) &&         \
+	 ((reg) != FORCEWAKE_MT))
 
 static bool IS_DISPLAYREG(u32 reg)
 {
drivers/gpu/drm/i915/i915_drv.h +3 -0
index a0c15ab..e4916a0 100644
@@ -1525,6 +1525,9 @@ extern void __gen6_gt_force_wake_mt_put(struct drm_i915_private *dev_priv);
 extern void vlv_force_wake_get(struct drm_i915_private *dev_priv);
 extern void vlv_force_wake_put(struct drm_i915_private *dev_priv);
 
+extern void hsw_gt_force_wake_get(struct drm_i915_private *dev_priv);
+extern void hsw_gt_force_wake_put(struct drm_i915_private *dev_priv);
+
 /* overlay */
 #ifdef CONFIG_DEBUG_FS
 extern struct intel_overlay_error_state *intel_overlay_capture_error_state(struct drm_device *dev);
drivers/gpu/drm/i915/i915_reg.h +1 -0
index 284965b..2c4be2e 100644
@@ -4070,6 +4070,7 @@
 #define  FORCEWAKE				0xA18C
 #define  FORCEWAKE_VLV				0x1300b0
 #define  FORCEWAKE_ACK_VLV			0x1300b4
+#define  FORCEWAKE_ACK_HSW			0x130044
 #define  FORCEWAKE_ACK				0x130090
 #define  FORCEWAKE_MT				0xa188 /* multi-threaded */
 #define  FORCEWAKE_MT_ACK			0x130040
drivers/gpu/drm/i915/intel_pm.c +3 -1
index 99bc1f3..0334e42 100644
@@ -3769,7 +3769,7 @@ void intel_init_pm(struct drm_device *dev)
 		dev_priv->display.force_wake_put = __gen6_gt_force_wake_put;
 
 		/* IVB configs may use multi-threaded forcewake */
-		if (IS_IVYBRIDGE(dev) || IS_HASWELL(dev)) {
+		if (IS_IVYBRIDGE(dev)) {
 			u32	ecobus;
 
 			/* A small trick here - if the bios hasn't configured MT forcewake,
@@ -3842,6 +3842,8 @@ void intel_init_pm(struct drm_device *dev)
 			}
 			dev_priv->display.init_clock_gating = ivybridge_init_clock_gating;
 			dev_priv->display.sanitize_pm = gen6_sanitize_pm;
+			dev_priv->display.force_wake_get = hsw_gt_force_wake_get;
+			dev_priv->display.force_wake_put = hsw_gt_force_wake_put;
 		} else
 			dev_priv->display.update_wm = NULL;
 	} else if (IS_VALLEYVIEW(dev)) {
Signature · 1 line
1.7.11.1
#22 lore raw

Re: [PATCH 12/21] drm/i915: support Haswell-style force waking

Quoted 10 lines · Eugeni Dodonov

On Haswell, there is a different register for reading force wake ACKs, and all the writes should go into the multi-threaded register, even for the legacy force wake.

Also, we have a theorical possibility for the force wake sequence to awaken the GT, but return while it hasn't finished bringing up the queue. So we properly check those bits as well, to ensure we won't end up in half-woken situation.

This patch didn't survive review with the *shhh* intel top secret *ssshhh* power docs ;-) It looks like we're supposed to poll for the ack from the hw on the _same_ reg as we've used in the MT forcewake register until the bit is set (for get) or cleared (for put).

Also, calling the helper function hsw_wait_gt_ack makes more sense then. -Daniel

Quoted 116 lines
---
 drivers/gpu/drm/i915/i915_drv.c | 36 +++++++++++++++++++++++++++++++++++-
 drivers/gpu/drm/i915/i915_drv.h |  3 +++
 drivers/gpu/drm/i915/i915_reg.h |  1 +
 drivers/gpu/drm/i915/intel_pm.c |  4 +++-
 4 files changed, 42 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/i915_drv.c +35 -1
index 79be879..73fd38a 100644
@@ -536,6 +536,39 @@ int __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv)
 	return ret;
 }
 
+static inline void __hsw_wait_gt_awake(struct drm_i915_private *dev_priv)
+{
+	int count = 0;
+
+	while (count++ < 50) {
+		u32 tmp = I915_READ_NOTRACE(FORCEWAKE_ACK_HSW);
+		if ((tmp & 1) && !(tmp & _MASKED_BIT_ENABLE(~7)))
+			break;
+		udelay(10);
+	}
+}
+
+
+void hsw_gt_force_wake_get(struct drm_i915_private *dev_priv)
+{
+	__hsw_wait_gt_awake(dev_priv);
+
+	I915_WRITE_NOTRACE(FORCEWAKE_MT, _MASKED_BIT_ENABLE(1));
+	POSTING_READ(FORCEWAKE_MT);
+
+	__hsw_wait_gt_awake(dev_priv);
+}
+
+void hsw_gt_force_wake_put(struct drm_i915_private *dev_priv)
+{
+	I915_WRITE_NOTRACE(FORCEWAKE_MT, _MASKED_BIT_DISABLE(1));
+
+	__hsw_wait_gt_awake(dev_priv);
+
+	/* The below doubles as a POSTING_READ */
+	gen6_gt_check_fifodbg(dev_priv);
+}
+
 void vlv_force_wake_get(struct drm_i915_private *dev_priv)
 {
 	int count;
@@ -1161,7 +1194,8 @@ MODULE_LICENSE("GPL and additional rights");
 #define NEEDS_FORCE_WAKE(dev_priv, reg) \
 	((HAS_FORCE_WAKE((dev_priv)->dev)) && \
 	 ((reg) < 0x40000) &&            \
-	 ((reg) != FORCEWAKE))
+	 ((reg) != FORCEWAKE) &&         \
+	 ((reg) != FORCEWAKE_MT))
 
 static bool IS_DISPLAYREG(u32 reg)
 {
drivers/gpu/drm/i915/i915_drv.h +3 -0
index a0c15ab..e4916a0 100644
@@ -1525,6 +1525,9 @@ extern void __gen6_gt_force_wake_mt_put(struct drm_i915_private *dev_priv);
 extern void vlv_force_wake_get(struct drm_i915_private *dev_priv);
 extern void vlv_force_wake_put(struct drm_i915_private *dev_priv);
 
+extern void hsw_gt_force_wake_get(struct drm_i915_private *dev_priv);
+extern void hsw_gt_force_wake_put(struct drm_i915_private *dev_priv);
+
 /* overlay */
 #ifdef CONFIG_DEBUG_FS
 extern struct intel_overlay_error_state *intel_overlay_capture_error_state(struct drm_device *dev);
drivers/gpu/drm/i915/i915_reg.h +1 -0
index 284965b..2c4be2e 100644
@@ -4070,6 +4070,7 @@
 #define  FORCEWAKE				0xA18C
 #define  FORCEWAKE_VLV				0x1300b0
 #define  FORCEWAKE_ACK_VLV			0x1300b4
+#define  FORCEWAKE_ACK_HSW			0x130044
 #define  FORCEWAKE_ACK				0x130090
 #define  FORCEWAKE_MT				0xa188 /* multi-threaded */
 #define  FORCEWAKE_MT_ACK			0x130040
drivers/gpu/drm/i915/intel_pm.c +3 -1
index 99bc1f3..0334e42 100644
@@ -3769,7 +3769,7 @@ void intel_init_pm(struct drm_device *dev)
 		dev_priv->display.force_wake_put = __gen6_gt_force_wake_put;
 
 		/* IVB configs may use multi-threaded forcewake */
-		if (IS_IVYBRIDGE(dev) || IS_HASWELL(dev)) {
+		if (IS_IVYBRIDGE(dev)) {
 			u32	ecobus;
 
 			/* A small trick here - if the bios hasn't configured MT forcewake,
@@ -3842,6 +3842,8 @@ void intel_init_pm(struct drm_device *dev)
 			}
 			dev_priv->display.init_clock_gating = ivybridge_init_clock_gating;
 			dev_priv->display.sanitize_pm = gen6_sanitize_pm;
+			dev_priv->display.force_wake_get = hsw_gt_force_wake_get;
+			dev_priv->display.force_wake_put = hsw_gt_force_wake_put;
 		} else
 			dev_priv->display.update_wm = NULL;
 	} else if (IS_VALLEYVIEW(dev)) {
Signature · 6 lines
1.7.11.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter
Mail: [email protected]
Mobile: +41 (0)79 365 57 48
#23 lore raw

PATCH 12/21 drm/i915: support Haswell-style force waking

On Haswell, there is a different register for reading force wake ACKs.

v2: simplify forcewake dance and therefore the commit message as noticed by Daniel Vetter.

---
 drivers/gpu/drm/i915/i915_drv.c | 10 ++++++++--
 drivers/gpu/drm/i915/i915_reg.h |  1 +
 2 files changed, 9 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/i915_drv.c +8 -2
index 79be879..23ae450 100644
@@ -451,16 +451,22 @@ void __gen6_gt_force_wake_get(struct drm_i915_private *dev_priv)
 void __gen6_gt_force_wake_mt_get(struct drm_i915_private *dev_priv)
 {
 	int count;
+	u32 forcewake_ack;
+
+	if (IS_HASWELL(dev_priv->dev))
+		forcewake_ack = FORCEWAKE_ACK_HSW;
+	else
+		forcewake_ack = FORCEWAKE_MT_ACK;
 
 	count = 0;
-	while (count++ < 50 && (I915_READ_NOTRACE(FORCEWAKE_MT_ACK) & 1))
+	while (count++ < 50 && (I915_READ_NOTRACE(forcewake_ack) & 1))
 		udelay(10);
 
 	I915_WRITE_NOTRACE(FORCEWAKE_MT, _MASKED_BIT_ENABLE(1));
 	POSTING_READ(FORCEWAKE_MT);
 
 	count = 0;
-	while (count++ < 50 && (I915_READ_NOTRACE(FORCEWAKE_MT_ACK) & 1) == 0)
+	while (count++ < 50 && (I915_READ_NOTRACE(forcewake_ack) & 1) == 0)
 		udelay(10);
 }
 
drivers/gpu/drm/i915/i915_reg.h +1 -0
index 284965b..2c4be2e 100644
@@ -4070,6 +4070,7 @@
 #define  FORCEWAKE				0xA18C
 #define  FORCEWAKE_VLV				0x1300b0
 #define  FORCEWAKE_ACK_VLV			0x1300b4
+#define  FORCEWAKE_ACK_HSW			0x130044
 #define  FORCEWAKE_ACK				0x130090
 #define  FORCEWAKE_MT				0xa188 /* multi-threaded */
 #define  FORCEWAKE_MT_ACK			0x130040
Signature · 1 line
1.7.11.1
#24 lore raw

PATCH 13/21 drm/i915: add RPS configuration for Haswell

In reply to Eugeni Dodonov

Split Haswell-specific GT algorithms into its own function.

Note that Haswell only has RC6, so account for that as well.

---
 drivers/gpu/drm/i915/i915_reg.h |   1 +
 drivers/gpu/drm/i915/intel_pm.c | 160 ++++++++++++++++++++++++++++++++++++++--
 2 files changed, 155 insertions(+), 6 deletions(-)
drivers/gpu/drm/i915/i915_reg.h +1 -0
index 2c4be2e..0c53e4a 100644
@@ -4131,6 +4131,7 @@
 #define   GEN6_RP_UP_IDLE_MIN			(0x1<<3)
 #define   GEN6_RP_UP_BUSY_AVG			(0x2<<3)
 #define   GEN6_RP_UP_BUSY_CONT			(0x4<<3)
+#define   GEN7_RP_DOWN_IDLE_AVG			(0x2<<0)
 #define   GEN6_RP_DOWN_IDLE_CONT		(0x1<<0)
 #define GEN6_RP_UP_THRESHOLD			0xA02C
 #define GEN6_RP_DOWN_THRESHOLD			0xA030
drivers/gpu/drm/i915/intel_pm.c +154 -6

191 lines not on this page · raw message

Signature · 1 line
1.7.11.1
#25 lore raw

Re: [PATCH 13/21] drm/i915: add RPS configuration for Haswell

Quoted 39 lines · Eugeni Dodonov

Split Haswell-specific GT algorithms into its own function.

Note that Haswell only has RC6, so account for that as well.

---
 drivers/gpu/drm/i915/i915_reg.h |   1 +
 drivers/gpu/drm/i915/intel_pm.c | 160 ++++++++++++++++++++++++++++++++++++++--
 2 files changed, 155 insertions(+), 6 deletions(-)
drivers/gpu/drm/i915/i915_reg.h +1 -0
index 2c4be2e..0c53e4a 100644
@@ -4131,6 +4131,7 @@
 #define   GEN6_RP_UP_IDLE_MIN			(0x1<<3)
 #define   GEN6_RP_UP_BUSY_AVG			(0x2<<3)
 #define   GEN6_RP_UP_BUSY_CONT			(0x4<<3)
+#define   GEN7_RP_DOWN_IDLE_AVG			(0x2<<0)
 #define   GEN6_RP_DOWN_IDLE_CONT		(0x1<<0)
 #define GEN6_RP_UP_THRESHOLD			0xA02C
 #define GEN6_RP_DOWN_THRESHOLD			0xA030
drivers/gpu/drm/i915/intel_pm.c +9 -0
index 0334e42..0733f16 100644
@@ -2301,6 +2301,26 @@ void gen6_set_rps(struct drm_device *dev, u8 val)
 	dev_priv->cur_delay = val;
 }
 
+static void hsw_disable_rps(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+
+	I915_WRITE(GEN6_RC_CONTROL, 0);
+	I915_WRITE(GEN6_RPNSWREQ, 1 << 31);
+	I915_WRITE(GEN6_PMINTRMSK, 0xffffffff);
+	I915_WRITE(GEN6_PMIER, 0);
+	/* Complete PM interrupt masking here doesn't race with the rps work

drivers/gpu/drm/i915/intel_pm.c void gen6_set_rps(struct drm_device *dev, u8 val) · line 2322

Quoted 10 lines
drivers/gpu/drm/i915/intel_pm.c +10 -0
+	 * item again unmasking PM interrupts because that is using a different
+	 * register (PMIMR) to mask PM interrupts. The only risk is in leaving
+	 * stale bits in PMIIR and PMIMR which gen6_enable_rps will clean up. */
+
+	spin_lock_irq(&dev_priv->rps_lock);
+	dev_priv->pm_iir = 0;
+	spin_unlock_irq(&dev_priv->rps_lock);
+
+	I915_WRITE(GEN6_PMIIR, I915_READ(GEN6_PMIIR));
+}

The only difference I can see wrt gen6_disable_rps is clearing the RC_CONTROL reg. And that looks like a bugfix that gen6+ should get, too. Imo copy&pasting this function hence doesn't make sense.

Quoted 35 lines
drivers/gpu/drm/i915/intel_pm.c +18 -2
+
 static void gen6_disable_rps(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -2334,9 +2354,10 @@ int intel_enable_rc6(const struct drm_device *dev)
 	if (INTEL_INFO(dev)->gen == 5)
 		return 0;
 
-	/* Sorry Haswell, no RC6 for you for now. */
+	/* Haswell does not has RC6p nor RC6pp
+	 */
 	if (IS_HASWELL(dev))
-		return 0;
+		return INTEL_RC6_ENABLE;
 
 	/*
 	 * Disable rc6 on Sandybridge
@@ -2349,6 +2370,130 @@ int intel_enable_rc6(const struct drm_device *dev)
 	return (INTEL_RC6_ENABLE | INTEL_RC6p_ENABLE);
 }
 
+static void hsw_enable_rps(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_ring_buffer *ring;
+	u32 pcu_mbox, rc6_mask = 0;
+	int rc6_mode;
+	int i;
+
+	WARN_ON(!mutex_is_locked(&dev->struct_mutex));
+
+	/* Here begins a magic sequence of register writes to enable
+	 * auto-downclocking.
+	 *
+	 * Perhaps there might be some value in exposing these to

drivers/gpu/drm/i915/intel_pm.c int intel_enable_rc6(const struct drm_device *dev) · line 2396

Quoted 10 lines
drivers/gpu/drm/i915/intel_pm.c +10 -0
+	 * userspace...
+	 */
+	I915_WRITE(GEN6_RC_STATE, 0);
+
+	gen6_gt_force_wake_get(dev_priv);
+
+	/* In units of 100MHz */
+	dev_priv->max_delay = 18;
+	dev_priv->min_delay = 6;
+	dev_priv->cur_delay = 0;

gen6 enable_rps reads these magic values from regs. Shouldn't we do that for hsw, too?

Quoted 38 lines
drivers/gpu/drm/i915/intel_pm.c +38 -0
+
+	/* disable the counters and set deterministic thresholds */
+	I915_WRITE(GEN6_RC_CONTROL, 0);
+
+	I915_WRITE(GEN6_RC1_WAKE_RATE_LIMIT, 1000 << 16);
+	I915_WRITE(GEN6_RC6_WAKE_RATE_LIMIT, 40 << 16);
+	I915_WRITE(GEN6_RC_EVALUATION_INTERVAL, 125000);
+	I915_WRITE(GEN6_RC_IDLE_HYSTERSIS, 25);
+
+	for_each_ring(ring, dev_priv, i)
+		I915_WRITE(RING_MAX_IDLE(ring->mmio_base), 10);
+
+	I915_WRITE(GEN6_RC_SLEEP, 0);
+	I915_WRITE(GEN6_RC1e_THRESHOLD, 1000);
+	I915_WRITE(GEN6_RC6_THRESHOLD, 50000);
+
+	rc6_mode = intel_enable_rc6(dev_priv->dev);
+
+	if (rc6_mode & INTEL_RC6_ENABLE)
+		rc6_mask |= GEN6_RC_CTL_RC6_ENABLE;
+
+	DRM_INFO("Enabling RC6 states on Haswell: RC6 %s\n",
+			(rc6_mode & INTEL_RC6_ENABLE) ? "on" : "off");
+
+	I915_WRITE(GEN6_RC_CONTROL,
+		   rc6_mask |
+		   GEN6_RC_CTL_EI_MODE(1) |
+		   GEN6_RC_CTL_HW_ENABLE);
+
+	I915_WRITE(GEN6_RPNSWREQ,
+		   GEN6_FREQUENCY(10) |
+		   GEN6_OFFSET(0) |
+		   GEN6_AGGRESSIVE_TURBO);
+	I915_WRITE(GEN6_RC_VIDEO_FREQ,
+		   GEN6_FREQUENCY(12));
+
+	I915_WRITE(GEN6_RP_DOWN_TIMEOUT, 1000000);
+	I915_WRITE(GEN6_RP_INTERRUPT_LIMITS,

drivers/gpu/drm/i915/intel_pm.c int intel_enable_rc6(const struct drm_device *dev) · line 2444

Quoted 10 lines
drivers/gpu/drm/i915/intel_pm.c +10 -0
+		   dev_priv->max_delay << 24 |
+		   dev_priv->min_delay << 16);
+	I915_WRITE(GEN6_RP_UP_THRESHOLD, 59400);
+	I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 245000);
+	I915_WRITE(GEN6_RP_UP_EI, 66000);
+	I915_WRITE(GEN6_RP_DOWN_EI, 350000);
+	I915_WRITE(GEN6_RP_IDLE_HYSTERSIS, 10);
+	I915_WRITE(GEN6_RP_CONTROL,
+		   GEN6_RP_MEDIA_TURBO |
+		   GEN6_RP_MEDIA_HW_MODE |

I figure this should be changed like the gen6 rps code in 89ba829e ...

Quoted 41 lines
drivers/gpu/drm/i915/intel_pm.c +41 -0
+		   GEN6_RP_MEDIA_IS_GFX |
+		   GEN6_RP_ENABLE |
+		   GEN6_RP_UP_BUSY_AVG |
+		   GEN7_RP_DOWN_IDLE_AVG);
+
+	if (wait_for((I915_READ(GEN6_PCODE_MAILBOX) & GEN6_PCODE_READY) == 0,
+		     500))
+		DRM_ERROR("timeout waiting for pcode mailbox to become idle\n");
+
+	I915_WRITE(GEN6_PCODE_DATA, 0);
+	I915_WRITE(GEN6_PCODE_MAILBOX,
+		   GEN6_PCODE_READY |
+		   GEN6_PCODE_WRITE_MIN_FREQ_TABLE);
+	if (wait_for((I915_READ(GEN6_PCODE_MAILBOX) & GEN6_PCODE_READY) == 0,
+		     500))
+		DRM_ERROR("timeout waiting for pcode mailbox to finish\n");
+
+	/* Check for overclock support */
+	if (wait_for((I915_READ(GEN6_PCODE_MAILBOX) & GEN6_PCODE_READY) == 0,
+		     500))
+		DRM_ERROR("timeout waiting for pcode mailbox to become idle\n");
+	I915_WRITE(GEN6_PCODE_MAILBOX, GEN6_READ_OC_PARAMS);
+	pcu_mbox = I915_READ(GEN6_PCODE_DATA);
+	if (wait_for((I915_READ(GEN6_PCODE_MAILBOX) & GEN6_PCODE_READY) == 0,
+		     500))
+		DRM_ERROR("timeout waiting for pcode mailbox to finish\n");
+	if (pcu_mbox & (1<<31)) { /* OC supported */
+		dev_priv->max_delay = pcu_mbox & 0xff;
+		DRM_DEBUG_DRIVER("overclocking supported, adjusting frequency max to %dMHz\n", pcu_mbox * 50);
+	}
+
+	/* requires MSI enabled */
+	I915_WRITE(GEN6_PMIER,
+		   GEN6_PM_MBOX_EVENT |
+		   GEN6_PM_THERMAL_EVENT |
+		   GEN6_PM_RP_DOWN_TIMEOUT |
+		   GEN6_PM_RP_UP_THRESHOLD |
+		   GEN6_PM_RP_DOWN_THRESHOLD |
+		   GEN6_PM_RP_UP_EI_EXPIRED |
+		   GEN6_PM_RP_DOWN_EI_EXPIRED);
+

drivers/gpu/drm/i915/intel_pm.c int intel_enable_rc6(const struct drm_device *dev) · line 2495

Quoted 10 lines
drivers/gpu/drm/i915/intel_pm.c +10 -0
+	spin_lock_irq(&dev_priv->rps_lock);
+	WARN_ON(dev_priv->pm_iir != 0);
+	I915_WRITE(GEN6_PMIMR, 0);
+	spin_unlock_irq(&dev_priv->rps_lock);
+
+	/* enable all PM interrupts */
+	I915_WRITE(GEN6_PMINTRMSK, 0);
+
+	gen6_gt_force_wake_put(dev_priv);
+}

Ok, after reviewing this with the gen6 code and hsw power doc I see very few differences (besides the 2 things that look like mistakes):
- we don't set a few things related to rc6p&rc6pp
- we have slightly different timeout/threshold values.

When copy&pasting code we should balance the risk of introducing regressions with the risk to not apply bugfixes everywhere they're needed. Given that you've missed Jesse's fix while hacking on this and that this seems to work very much like rps on snb/ivb I'd vote for sprinkling a few ifs into gen6_enable_rps. Maybe set the timing values in local variables at the beginning of the function.
-Daniel

Quoted 36 lines
drivers/gpu/drm/i915/intel_pm.c +8 -4
+
 static void gen6_enable_rps(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -3227,7 +3372,9 @@ void intel_disable_gt_powersave(struct drm_device *dev)
 {
 	if (IS_IRONLAKE_M(dev))
 		ironlake_disable_drps(dev);
-	if (INTEL_INFO(dev)->gen >= 6 && !IS_VALLEYVIEW(dev))
+	else if (IS_HASWELL(dev))
+		hsw_disable_rps(dev);
+	else if (INTEL_INFO(dev)->gen >= 6 && !IS_VALLEYVIEW(dev))
 		gen6_disable_rps(dev);
 }
 
@@ -3237,9 +3384,10 @@ void intel_enable_gt_powersave(struct drm_device *dev)
 		ironlake_enable_drps(dev);
 		ironlake_enable_rc6(dev);
 		intel_init_emon(dev);
-	}
-
-	if ((IS_GEN6(dev) || IS_GEN7(dev)) && !IS_VALLEYVIEW(dev)) {
+	} else if (IS_HASWELL(dev)) {
+		hsw_enable_rps(dev);
+		gen6_update_ring_freq(dev);
+	} else if ((IS_GEN6(dev) || IS_GEN7(dev)) && !IS_VALLEYVIEW(dev)) {
 		gen6_enable_rps(dev);
 		gen6_update_ring_freq(dev);
 	}
Signature · 6 lines
1.7.11.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter
Mail: [email protected]
Mobile: +41 (0)79 365 57 48
#26 lore raw

Re: [PATCH 13/21] drm/i915: add RPS configuration for Haswell

Quoted 11 lines · Daniel Vetter

Ok, after reviewing this with the gen6 code and hsw power doc I see very few differences (besides the 2 things that look like mistakes):
- we don't set a few things related to rc6p&rc6pp
- we have slightly different timeout/threshold values.

When copy&pasting code we should balance the risk of introducing regressions with the risk to not apply bugfixes everywhere they're needed. Given that you've missed Jesse's fix while hacking on this and that this seems to work very much like rps on snb/ivb I'd vote for sprinkling a few ifs into gen6_enable_rps. Maybe set the timing values in local variables at the beginning of the function.

Sure, I'll do this and send the patch.

Eugeni

#27 lore raw

PATCH 14/21 drm/i915: Add EDP Registers for Haswell

In reply to Eugeni Dodonov

From: Shobhit Kumar <[email protected]>

PIPE EDP registers and timing registers are different for EDP on HSW

---
 drivers/gpu/drm/i915/i915_reg.h | 31 +++++++++++++++++++++++++++++++
 1 file changed, 31 insertions(+)
drivers/gpu/drm/i915/i915_reg.h +31 -0
index 0c53e4a..1e70fae 100644
@@ -1528,6 +1528,14 @@
 #define _BCLRPAT_B	0x61020
 #define _VSYNCSHIFT_B	0x61028
 
+/* Pipe EDP timing regs */
+#define HTOTAL_EDP	0x6F000
+#define HBLANK_EDP	0x6F004
+#define HSYNC_EDP	0x6F008
+#define VTOTAL_EDP	0x6F00C
+#define VBLANK_EDP	0x6F010
+#define VSYNC_EDP	0x6F014
+#define VSYNCSHIFT_EDP	0x6F028
 
 #define HTOTAL(pipe) _PIPE(pipe, _HTOTAL_A, _HTOTAL_B)
 #define HBLANK(pipe) _PIPE(pipe, _HBLANK_A, _HBLANK_B)
@@ -3243,6 +3251,19 @@
 #define _PIPEB_LINK_M2           0x61048
 #define _PIPEB_LINK_N2           0x6104c
 
+/* PIPE EDP timing regs */
+#define PIPEEDP_DATA_M1	0x6F030
+#define PIPEEDP_DATA_N1	0x6F034
+
+#define PIPDEDP_DATA_M2	0x6F038
+#define PIPEEDP_DATA_N2	0x6F03C
+
+#define PIPEEDP_LINK_M1	0x6F040
+#define PIPEEDP_LINK_N1	0x6F044
+
+#define PIPEEDP_LINK_M2	0x6F048
+#define PIPEEDP_LINK_N2	0x6F04C
+
 #define PIPE_DATA_M1(pipe) _PIPE(pipe, _PIPEA_DATA_M1, _PIPEB_DATA_M1)
 #define PIPE_DATA_N1(pipe) _PIPE(pipe, _PIPEA_DATA_N1, _PIPEB_DATA_N1)
 #define PIPE_DATA_M2(pipe) _PIPE(pipe, _PIPEA_DATA_M2, _PIPEB_DATA_M2)
@@ -4277,6 +4298,12 @@
 #define PIPE_DDI_FUNC_CTL_B			0x61400
 #define PIPE_DDI_FUNC_CTL_C			0x62400
 #define PIPE_DDI_FUNC_CTL_EDP		0x6F400
+#define PIPE_DDI_EDP_INPUT_SRC_MASK			(7<<12)
+#define PIPE_DDI_EDI_INPUT_SRC_A_ON			(0<<12)
+#define PIPE_DDI_EDI_INPUT_SRC_A_ON_OFF		(4<<12)
+#define PIPE_DDI_EDI_INPUT_SRC_B_ON_OFF		(5<<12)
+#define PIPE_DDI_EDI_INPUT_SRC_C_ON_OFF		(6<<12)
+
 #define DDI_FUNC_CTL(pipe) _PIPE(pipe, \
 					PIPE_DDI_FUNC_CTL_A, \
 					PIPE_DDI_FUNC_CTL_B)
@@ -4345,6 +4372,8 @@
 #define  DDI_BUF_EMP_800MV_3_5DB_HSW	(8<<24)   /* Sel8 */
 #define  DDI_BUF_EMP_MASK				(0xf<<24)
 #define  DDI_BUF_IS_IDLE				(1<<7)
+#define  DDI_PORT_LANE_CAP_DDIA_4		(1<<4)
+#define  DDI_PORT_LANE_CAP_DDIA_2		(0<<4)
 #define  DDI_PORT_WIDTH_MASK		(7<<1)
 #define  DDI_PORT_WIDTH_X1				(0<<1)
 #define  DDI_PORT_WIDTH_X2				(1<<1)
@@ -4463,6 +4492,8 @@
 #define  HSW_MSA_BPC_12_BITS		(3<<5)
 #define  HSW_MSA_BPC_16_BITS		(4<<5)
 
+#define HSW_MSA_EDP_CTL		0x6F410
+
 /* Pipe WM_LINETIME - watermark line time */
 #define PIPE_WM_LINETIME_A		0x45270
 #define PIPE_WM_LINETIME_B		0x45274
Signature · 1 line
1.7.11.1
#28 lore raw

PATCH 15/21 drm/i915: Timing initialization for eDP on HSW

In reply to Eugeni Dodonov

From: Shobhit Kumar <[email protected]>

v2: simplify horizontal and vertical timings calculations

---
 drivers/gpu/drm/i915/intel_display.c | 66 +++++++++++++++++++++++-------------
 1 file changed, 42 insertions(+), 24 deletions(-)
drivers/gpu/drm/i915/intel_display.c +42 -24
index 9a695ab..0438a10 100644
@@ -4669,6 +4669,7 @@ static int ironlake_crtc_mode_set(struct drm_crtc *crtc,
 	unsigned int pipe_bpp;
 	bool dither;
 	bool is_cpu_edp = false, is_pch_edp = false;
+	u32 htotal, hblank, hsync, vtotal, vblank, vsync;
 
 	list_for_each_entry(encoder, &mode_config->encoder_list, base.head) {
 		if (encoder->base.crtc != crtc)
@@ -4949,7 +4950,7 @@ static int ironlake_crtc_mode_set(struct drm_crtc *crtc,
 		pipeconf |= PIPECONF_DITHER_EN;
 		pipeconf |= PIPECONF_DITHER_TYPE_SP;
 	}
-	if (is_dp && !is_cpu_edp) {
+	if ((IS_HASWELL(dev) && is_cpu_edp) || (is_dp && !is_cpu_edp)) {
 		intel_dp_set_m_n(crtc, mode, adjusted_mode);
 	} else {
 		/* For non-DP output, clear any trans DP clock recovery setting.*/
@@ -4998,25 +4999,40 @@ static int ironlake_crtc_mode_set(struct drm_crtc *crtc,
 		I915_WRITE(VSYNCSHIFT(pipe), 0);
 	}
 
-	I915_WRITE(HTOTAL(pipe),
-		   (adjusted_mode->crtc_hdisplay - 1) |
-		   ((adjusted_mode->crtc_htotal - 1) << 16));
-	I915_WRITE(HBLANK(pipe),
-		   (adjusted_mode->crtc_hblank_start - 1) |
-		   ((adjusted_mode->crtc_hblank_end - 1) << 16));
-	I915_WRITE(HSYNC(pipe),
-		   (adjusted_mode->crtc_hsync_start - 1) |
-		   ((adjusted_mode->crtc_hsync_end - 1) << 16));
-
-	I915_WRITE(VTOTAL(pipe),
-		   (adjusted_mode->crtc_vdisplay - 1) |
-		   ((adjusted_mode->crtc_vtotal - 1) << 16));
-	I915_WRITE(VBLANK(pipe),
-		   (adjusted_mode->crtc_vblank_start - 1) |
-		   ((adjusted_mode->crtc_vblank_end - 1) << 16));
-	I915_WRITE(VSYNC(pipe),
-		   (adjusted_mode->crtc_vsync_start - 1) |
-		   ((adjusted_mode->crtc_vsync_end - 1) << 16));
+	if (is_cpu_edp && IS_HASWELL(dev)) {
+		htotal = HTOTAL_EDP;
+		hblank = HBLANK_EDP;
+		hsync = HSYNC_EDP;
+		vtotal = VTOTAL_EDP;
+		vblank = VBLANK_EDP;
+		vsync = VSYNC_EDP;
+	} else {
+		htotal = HTOTAL(pipe);
+		hblank = HBLANK(pipe);
+		hsync = HSYNC(pipe);
+		vtotal = VTOTAL(pipe);
+		vblank = VBLANK(pipe);
+		vsync = VSYNC(pipe);
+	}
+	I915_WRITE(htotal,
+		(adjusted_mode->crtc_hdisplay - 1) |
+		((adjusted_mode->crtc_htotal - 1) << 16));
+	I915_WRITE(hblank,
+		(adjusted_mode->crtc_hblank_start - 1) |
+		((adjusted_mode->crtc_hblank_end - 1) << 16));
+	I915_WRITE(hsync,
+		(adjusted_mode->crtc_hsync_start - 1) |
+		((adjusted_mode->crtc_hsync_end - 1) << 16));
+
+	I915_WRITE(vtotal,
+		(adjusted_mode->crtc_vdisplay - 1) |
+		((adjusted_mode->crtc_vtotal - 1) << 16));
+	I915_WRITE(vblank,
+		(adjusted_mode->crtc_vblank_start - 1) |
+		((adjusted_mode->crtc_vblank_end - 1) << 16));
+	I915_WRITE(vsync,
+		(adjusted_mode->crtc_vsync_start - 1) |
+		((adjusted_mode->crtc_vsync_end - 1) << 16));
 
 	/* pipesrc controls the size that is scaled from, which should
 	 * always be the user's requested size.
@@ -5024,10 +5040,12 @@ static int ironlake_crtc_mode_set(struct drm_crtc *crtc,
 	I915_WRITE(PIPESRC(pipe),
 		   ((mode->hdisplay - 1) << 16) | (mode->vdisplay - 1));
 
-	I915_WRITE(PIPE_DATA_M1(pipe), TU_SIZE(m_n.tu) | m_n.gmch_m);
-	I915_WRITE(PIPE_DATA_N1(pipe), m_n.gmch_n);
-	I915_WRITE(PIPE_LINK_M1(pipe), m_n.link_m);
-	I915_WRITE(PIPE_LINK_N1(pipe), m_n.link_n);
+	if (!is_cpu_edp) {
+		I915_WRITE(PIPE_DATA_M1(pipe), TU_SIZE(m_n.tu) | m_n.gmch_m);
+		I915_WRITE(PIPE_DATA_N1(pipe), m_n.gmch_n);
+		I915_WRITE(PIPE_LINK_M1(pipe), m_n.link_m);
+		I915_WRITE(PIPE_LINK_N1(pipe), m_n.link_n);
+	}
 
 	if (is_cpu_edp)
 		ironlake_set_pll_edp(crtc, adjusted_mode->clock);
Signature · 1 line
1.7.11.1
#29 lore raw

PATCH 16/21 drm/i915: Modesetting for eDP on HSw

In reply to Eugeni Dodonov

From: Shobhit Kumar <[email protected]>

The MSA register and PIPE EDP register are differnet than that of DP. Also link training flow though similar for DP had been corrected to follow DP path

---
 drivers/gpu/drm/i915/i915_reg.h  |  1 +
 drivers/gpu/drm/i915/intel_ddi.c | 70 ++++++++++++++++++++++++++++------------
 drivers/gpu/drm/i915/intel_dp.c  |  8 ++---
 drivers/gpu/drm/i915/intel_drv.h |  1 +
 4 files changed, 55 insertions(+), 25 deletions(-)
drivers/gpu/drm/i915/i915_reg.h +1 -0
index 1e70fae..1165d2e 100644
@@ -4316,6 +4316,7 @@
 #define  PIPE_DDI_MODE_SELECT_DP_SST	(2<<24)
 #define  PIPE_DDI_MODE_SELECT_DP_MST	(3<<24)
 #define  PIPE_DDI_MODE_SELECT_FDI		(4<<24)
+#define  PIPE_DDI_BPC_MASK		(0x7<<20)
 #define  PIPE_DDI_BPC_8					(0<<20)
 #define  PIPE_DDI_BPC_10				(1<<20)
 #define  PIPE_DDI_BPC_6					(2<<20)
drivers/gpu/drm/i915/intel_ddi.c +49 -21

129 lines not on this page · raw message

drivers/gpu/drm/i915/intel_dp.c +4 -4
index 38a1d3b..f086944 100644
@@ -45,7 +45,7 @@
  * If a CPU or PCH DP output is attached to an eDP panel, this function
  * will return true, and false otherwise.
  */
-static bool is_edp(struct intel_dp *intel_dp)
+bool is_edp(struct intel_dp *intel_dp)
 {
 	return intel_dp->base.type == INTEL_OUTPUT_EDP;
 }
@@ -363,7 +363,7 @@ intel_dp_aux_ch(struct intel_dp *intel_dp,
 	 * clock divider.
 	 */
 	if (is_cpu_edp(intel_dp)) {
-		if (IS_GEN6(dev) || IS_GEN7(dev))
+		if ((IS_GEN6(dev) || IS_GEN7(dev)) && !IS_HASWELL(dev))
 			aux_clock_divider = 200; /* SNB & IVB eDP input clock at 400Mhz */
 		else
 			aux_clock_divider = 225; /* eDP input clock at 450Mhz */
@@ -1753,7 +1753,7 @@ intel_dp_start_link_train(struct intel_dp *intel_dp)
 		uint32_t    signal_levels;
 
 
-		if (IS_GEN7(dev) && is_cpu_edp(intel_dp)) {
+		if (IS_GEN7(dev) && is_cpu_edp(intel_dp) && !IS_HASWELL(dev)) {
 			signal_levels = intel_gen7_edp_signal_levels(intel_dp->train_set[0]);
 			DP = (DP & ~EDP_LINK_TRAIN_VOL_EMP_MASK_IVB) | signal_levels;
 		} else if (IS_GEN6(dev) && is_cpu_edp(intel_dp)) {
@@ -1855,7 +1855,7 @@ intel_dp_complete_link_train(struct intel_dp *intel_dp)
 			break;
 		}
 
-		if (IS_GEN7(dev) && is_cpu_edp(intel_dp)) {
+		if (IS_GEN7(dev) && is_cpu_edp(intel_dp) && !IS_HASWELL(dev)) {
 			signal_levels = intel_gen7_edp_signal_levels(intel_dp->train_set[0]);
 			DP = (DP & ~EDP_LINK_TRAIN_VOL_EMP_MASK_IVB) | signal_levels;
 		} else if (IS_GEN6(dev) && is_cpu_edp(intel_dp)) {
drivers/gpu/drm/i915/intel_drv.h +1 -0
index 5ca133d..a741a02 100644
@@ -392,6 +392,7 @@ extern void intel_mark_busy(struct drm_device *dev,
 extern bool intel_lvds_init(struct drm_device *dev);
 extern void intel_dp_init(struct drm_device *dev, int dp_reg);
 extern struct intel_dp *enc_to_intel_dp(struct drm_encoder *encoder);
+extern bool is_edp(struct intel_dp *intel_dp);
 
 void
 intel_dp_set_m_n(struct drm_crtc *crtc, struct drm_display_mode *mode,
Signature · 1 line
1.7.11.1
#30 lore raw

PATCH 17/21 drm/i915: Hook eDP initialization on DDI A

In reply to Eugeni Dodonov

From: Shobhit Kumar <[email protected]>

---
 drivers/gpu/drm/i915/intel_ddi.c | 3 +--
 drivers/gpu/drm/i915/intel_dp.c  | 1 +
 2 files changed, 2 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/intel_ddi.c +1 -2
index 9d09f38..e8a5b80 100644
@@ -244,8 +244,7 @@ void intel_ddi_init(struct drm_device *dev, enum port port)
 
 	switch(port){
 	case PORT_A:
-		/* We don't handle eDP and DP yet */
-		DRM_DEBUG_DRIVER("Found digital output on DDI port A\n");
+		intel_dp_init(dev, DP_A);
 		break;
 	/* Assume that the  ports B, C and D are working in HDMI mode for now */
 	case PORT_B:
drivers/gpu/drm/i915/intel_dp.c +1 -0
index f086944..7b5da87 100644
@@ -2541,6 +2541,7 @@ intel_dp_init(struct drm_device *dev, int output_reg)
 	if (output_reg == DP_A || is_pch_edp(intel_dp)) {
 		type = DRM_MODE_CONNECTOR_eDP;
 		intel_encoder->type = INTEL_OUTPUT_EDP;
+		intel_dp->ddi_port = PORT_A;
 	} else {
 		type = DRM_MODE_CONNECTOR_DisplayPort;
 		intel_encoder->type = INTEL_OUTPUT_DISPLAYPORT;
Signature · 1 line
1.7.11.1
#31 lore raw

PATCH 18/21 drm/i915: introduce haswell_init_clock_gating

In reply to Eugeni Dodonov

This is based on Ivy Bridge clock gating for now, but is subject to changes in the future.

---
 drivers/gpu/drm/i915/intel_pm.c | 54 ++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 53 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_pm.c +53 -1
index 0733f16..a4372b3 100644
@@ -3551,6 +3551,58 @@ static void gen7_setup_fixed_func_scheduler(struct drm_i915_private *dev_priv)
 	I915_WRITE(GEN7_FF_THREAD_MODE, reg);
 }
 
+static void haswell_init_clock_gating(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	int pipe;
+	uint32_t dspclk_gate = VRHUNIT_CLOCK_GATE_DISABLE;
+
+	I915_WRITE(PCH_DSPCLK_GATE_D, dspclk_gate);
+
+	I915_WRITE(WM3_LP_ILK, 0);
+	I915_WRITE(WM2_LP_ILK, 0);
+	I915_WRITE(WM1_LP_ILK, 0);
+
+	/* According to the spec, bit 13 (RCZUNIT) must be set on IVB.
+	 * This implements the WaDisableRCZUnitClockGating workaround.
+	 */
+	I915_WRITE(GEN6_UCGCTL2, GEN6_RCZUNIT_CLOCK_GATE_DISABLE);
+
+	I915_WRITE(ILK_DSPCLK_GATE, IVB_VRHUNIT_CLK_GATE);
+
+	I915_WRITE(IVB_CHICKEN3,
+		   CHICKEN3_DGMG_REQ_OUT_FIX_DISABLE |
+		   CHICKEN3_DGMG_DONE_FIX_DISABLE);
+
+	/* Apply the WaDisableRHWOOptimizationForRenderHang workaround. */
+	I915_WRITE(GEN7_COMMON_SLICE_CHICKEN1,
+		   GEN7_CSC1_RHWO_OPT_DISABLE_IN_RCC);
+
+	/* WaApplyL3ControlAndL3ChickenMode requires those two on Ivy Bridge */
+	I915_WRITE(GEN7_L3CNTLREG1,
+			GEN7_WA_FOR_GEN7_L3_CONTROL);
+	I915_WRITE(GEN7_L3_CHICKEN_MODE_REGISTER,
+			GEN7_WA_L3_CHICKEN_MODE);
+
+	/* This is required by WaCatErrorRejectionIssue */
+	I915_WRITE(GEN7_SQ_CHICKEN_MBCUNIT_CONFIG,
+			I915_READ(GEN7_SQ_CHICKEN_MBCUNIT_CONFIG) |
+			GEN7_SQ_CHICKEN_MBCUNIT_SQINTMOB);
+
+	for_each_pipe(pipe) {
+		I915_WRITE(DSPCNTR(pipe),
+			   I915_READ(DSPCNTR(pipe)) |
+			   DISPPLANE_TRICKLE_FEED_DISABLE);
+		intel_flush_display_plane(dev_priv, pipe);
+	}
+
+	gen7_setup_fixed_func_scheduler(dev_priv);
+
+	/* WaDisable4x2SubspanOptimization */
+	I915_WRITE(CACHE_MODE_1,
+		   _MASKED_BIT_ENABLE(PIXEL_SUBSPAN_COLLECT_OPT_DISABLE));
+}
+
 static void ivybridge_init_clock_gating(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -3988,7 +4040,7 @@ void intel_init_pm(struct drm_device *dev)
 					      "Disable CxSR\n");
 				dev_priv->display.update_wm = NULL;
 			}
-			dev_priv->display.init_clock_gating = ivybridge_init_clock_gating;
+			dev_priv->display.init_clock_gating = haswell_init_clock_gating;
 			dev_priv->display.sanitize_pm = gen6_sanitize_pm;
 			dev_priv->display.force_wake_get = hsw_gt_force_wake_get;
 			dev_priv->display.force_wake_put = hsw_gt_force_wake_put;
Signature · 1 line
1.7.11.1
#32 lore raw

PATCH 19/21 drm/i915: prevent bogus intel_update_fbc notifications

In reply to Eugeni Dodonov

This pollutes dmesg output even if we do not have FBC for the device, so move the DRM_DEBUG_KMS statement lower.

---
 drivers/gpu/drm/i915/intel_pm.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/intel_pm.c +2 -2
index a4372b3..0ccdb96 100644
@@ -387,14 +387,14 @@ void intel_update_fbc(struct drm_device *dev)
 	struct drm_i915_gem_object *obj;
 	int enable_fbc;
 
-	DRM_DEBUG_KMS("\n");
-
 	if (!i915_powersave)
 		return;
 
 	if (!I915_HAS_FBC(dev))
 		return;
 
+	DRM_DEBUG_KMS("\n");
+
 	/*
 	 * If FBC is already on, we just have to verify that we can
 	 * keep it that way...
Signature · 1 line
1.7.11.1
#33 lore raw

Re: [PATCH 19/21] drm/i915: prevent bogus intel_update_fbc notifications

Quoted 4 lines · Eugeni Dodonov

This pollutes dmesg output even if we do not have FBC for the device, so move the DRM_DEBUG_KMS statement lower.

I suggest we just kill this right away ;-)
-Daniel

Quoted 32 lines
---
 drivers/gpu/drm/i915/intel_pm.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/intel_pm.c +2 -2
index a4372b3..0ccdb96 100644
@@ -387,14 +387,14 @@ void intel_update_fbc(struct drm_device *dev)
 	struct drm_i915_gem_object *obj;
 	int enable_fbc;
 
-	DRM_DEBUG_KMS("\n");
-
 	if (!i915_powersave)
 		return;
 
 	if (!I915_HAS_FBC(dev))
 		return;
 
+	DRM_DEBUG_KMS("\n");
+
 	/*
 	 * If FBC is already on, we just have to verify that we can
 	 * keep it that way...
Signature · 6 lines
1.7.11.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter
Mail: [email protected]
Mobile: +41 (0)79 365 57 48
#34 lore raw

PATCH 19/21 drm/i915: prevent bogus intel_update_fbc notifications

This pollutes dmesg output even if we do not have FBC for the device, so move the DRM_DEBUG_KMS statement lower.

v2: just kill the message as suggested by Daniel.

---
 drivers/gpu/drm/i915/intel_pm.c | 2 --
 1 file changed, 2 deletions(-)
drivers/gpu/drm/i915/intel_pm.c +0 -2
index 53645f1..42e847c 100644
@@ -387,8 +387,6 @@ void intel_update_fbc(struct drm_device *dev)
 	struct drm_i915_gem_object *obj;
 	int enable_fbc;
 
-	DRM_DEBUG_KMS("\n");
-
 	if (!i915_powersave)
 		return;
 
Signature · 1 line
1.7.11.1
Reviewed-by on v1 19/21
#35 lore raw

Re: [PATCH 19/21] drm/i915: prevent bogus intel_update_fbc notifications

2012/6/28 Eugeni Dodonov <[email protected]>:

Quoted 6 lines

This pollutes dmesg output even if we do not have FBC for the device, so move the DRM_DEBUG_KMS statement lower.

v2: just kill the message as suggested by Daniel.

The message first says we "move the statement lower", but then later it says we kill the message :)
With that fixed:

I'm happy with both v1 and v2.

Quoted 24 lines
---
 drivers/gpu/drm/i915/intel_pm.c | 2 --
 1 file changed, 2 deletions(-)
drivers/gpu/drm/i915/intel_pm.c +0 -3
index 53645f1..42e847c 100644
@@ -387,8 +387,6 @@ void intel_update_fbc(struct drm_device *dev)
        struct drm_i915_gem_object *obj;
        int enable_fbc;
 
-       DRM_DEBUG_KMS("\n");
-
        if (!i915_powersave)
                return;
 
--

1.7.11.1

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

Signature · 1 line
Paulo Zanoni
#36 lore raw

Re: [PATCH 19/21] drm/i915: prevent bogus intel_update_fbc notifications

Quoted 8 lines · Paulo Zanoni

2012/6/28 Eugeni Dodonov <[email protected]>:

Quoted 6 lines

This pollutes dmesg output even if we do not have FBC for the device, so move the DRM_DEBUG_KMS statement lower.

v2: just kill the message as suggested by Daniel.

Quoted 7 lines

The message first says we "move the statement lower", but then later it says we kill the message :)
With that fixed:

I'm happy with both v1 and v2.

I'll let Daniel decide the bike shed color on this one :). We can always say that 'move it down' was a slang term for 'killing' later :).

Eugeni

#37 lore raw

Re: [PATCH 19/21] drm/i915: prevent bogus intel_update_fbc notifications

Quoted 13 lines · Eugeni Dodonov
Quoted 12 lines · Paulo Zanoni

2012/6/28 Eugeni Dodonov <[email protected]>:

Quoted 6 lines

This pollutes dmesg output even if we do not have FBC for the device, so move the DRM_DEBUG_KMS statement lower.

v2: just kill the message as suggested by Daniel.

The message first says we "move the statement lower", but then later it says we kill the message :)
With that fixed:

Quoted 6 lines
Quoted 3 lines

I'm happy with both v1 and v2.

I'll let Daniel decide the bike shed color on this one :). We can always say that 'move it down' was a slang term for 'killing' later :).

I've already merged it, but forgot to send out the ack mail ;-) -Daniel

Signature · 3 lines
Daniel Vetter
Mail: [email protected]
Mobile: +41 (0)79 365 57 48
#38 lore raw

PATCH 20/21 drm/i915: fix PIPE_WM_LINETIME definition

In reply to Eugeni Dodonov

From: Paulo Zanoni <[email protected]>

Looks like a copy/paste error.

---
 drivers/gpu/drm/i915/i915_reg.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/i915/i915_reg.h +1 -1
index 1165d2e..78b6ba4 100644
@@ -4500,7 +4500,7 @@
 #define PIPE_WM_LINETIME_B		0x45274
 #define PIPE_WM_LINETIME(pipe) _PIPE(pipe, \
 					PIPE_WM_LINETIME_A, \
-					PIPE_WM_LINETIME_A)
+					PIPE_WM_LINETIME_B)
 #define   PIPE_WM_LINETIME_MASK		(0x1ff)
 #define   PIPE_WM_LINETIME_TIME(x)			((x))
 #define   PIPE_WM_LINETIME_IPS_LINETIME_MASK	(0x1ff<<16)
Signature · 1 line
1.7.11.1
#40 lore raw

PATCH 21/21 drm/i915: enable RC6 workaround on Haswell

In reply to Eugeni Dodonov

For Haswell, on some of the early hardware revisions, it is possible to run into issues when RC6 state is enabled and when pipes change state.

---
 drivers/gpu/drm/i915/i915_reg.h | 5 +++++
 drivers/gpu/drm/i915/intel_pm.c | 9 +++++++++
 2 files changed, 14 insertions(+)
drivers/gpu/drm/i915/i915_reg.h +5 -0
index 78b6ba4..fd55e91 100644
@@ -4512,4 +4512,9 @@
 #define  SFUSE_STRAP_DDIC_DETECTED	(1<<1)
 #define  SFUSE_STRAP_DDID_DETECTED	(1<<0)
 
+#define WM_DBG				0x45280
+#define  WM_DBG_DISALLOW_MULTIPLE_LP	(1<<0)
+#define  WM_DBG_DISALLOW_MAXFIFO	(1<<1)
+#define  WM_DBG_DISALLOW_SPRITE		(1<<2)
+
 #endif /* _I915_REG_H_ */
drivers/gpu/drm/i915/intel_pm.c +9 -0
index 0ccdb96..cc8628c 100644
@@ -3601,6 +3601,15 @@ static void haswell_init_clock_gating(struct drm_device *dev)
 	/* WaDisable4x2SubspanOptimization */
 	I915_WRITE(CACHE_MODE_1,
 		   _MASKED_BIT_ENABLE(PIXEL_SUBSPAN_COLLECT_OPT_DISABLE));
+
+	/* Work-around rc6 issue */
+	I915_WRITE(WM_DBG,
+			I915_READ(WM_DBG) |
+			WM_DBG_DISALLOW_MULTIPLE_LP |
+			WM_DBG_DISALLOW_SPRITE |
+			WM_DBG_DISALLOW_MAXFIFO
+		  );
+
 }
 
 static void ivybridge_init_clock_gating(struct drm_device *dev)
Signature · 1 line
1.7.11.1
#41 lore raw

Re: [PATCH 21/21] drm/i915: enable RC6 workaround on Haswell

Quoted 4 lines · Eugeni Dodonov

For Haswell, on some of the early hardware revisions, it is possible to run into issues when RC6 state is enabled and when pipes change state.

Given that this is a w/a that only applies for early silicon, please add a comment that this should get nuked once production stuff ships. -Daniel

Quoted 46 lines
---
 drivers/gpu/drm/i915/i915_reg.h | 5 +++++
 drivers/gpu/drm/i915/intel_pm.c | 9 +++++++++
 2 files changed, 14 insertions(+)
drivers/gpu/drm/i915/i915_reg.h +5 -0
index 78b6ba4..fd55e91 100644
@@ -4512,4 +4512,9 @@
 #define  SFUSE_STRAP_DDIC_DETECTED	(1<<1)
 #define  SFUSE_STRAP_DDID_DETECTED	(1<<0)
 
+#define WM_DBG				0x45280
+#define  WM_DBG_DISALLOW_MULTIPLE_LP	(1<<0)
+#define  WM_DBG_DISALLOW_MAXFIFO	(1<<1)
+#define  WM_DBG_DISALLOW_SPRITE		(1<<2)
+
 #endif /* _I915_REG_H_ */
drivers/gpu/drm/i915/intel_pm.c +9 -0
index 0ccdb96..cc8628c 100644
@@ -3601,6 +3601,15 @@ static void haswell_init_clock_gating(struct drm_device *dev)
 	/* WaDisable4x2SubspanOptimization */
 	I915_WRITE(CACHE_MODE_1,
 		   _MASKED_BIT_ENABLE(PIXEL_SUBSPAN_COLLECT_OPT_DISABLE));
+
+	/* Work-around rc6 issue */
+	I915_WRITE(WM_DBG,
+			I915_READ(WM_DBG) |
+			WM_DBG_DISALLOW_MULTIPLE_LP |
+			WM_DBG_DISALLOW_SPRITE |
+			WM_DBG_DISALLOW_MAXFIFO
+		  );
+
 }
 
 static void ivybridge_init_clock_gating(struct drm_device *dev)
Signature · 6 lines
1.7.11.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter
Mail: [email protected]
Mobile: +41 (0)79 365 57 48
#42 lore raw

PATCH 21/21 drm/i915: enable RC6 workaround on Haswell

For Haswell, on some of the early hardware revisions, it is possible to run into issues when RC6 state is enabled and when pipes change state.

v2: add comment saying that this is for early revisions only.

---
 drivers/gpu/drm/i915/i915_reg.h |  5 +++++
 drivers/gpu/drm/i915/intel_pm.c | 12 ++++++++++++
 2 files changed, 17 insertions(+)
drivers/gpu/drm/i915/i915_reg.h +5 -0
index 78b6ba4..fd55e91 100644
@@ -4512,4 +4512,9 @@
 #define  SFUSE_STRAP_DDIC_DETECTED	(1<<1)
 #define  SFUSE_STRAP_DDID_DETECTED	(1<<0)
 
+#define WM_DBG				0x45280
+#define  WM_DBG_DISALLOW_MULTIPLE_LP	(1<<0)
+#define  WM_DBG_DISALLOW_MAXFIFO	(1<<1)
+#define  WM_DBG_DISALLOW_SPRITE		(1<<2)
+
 #endif /* _I915_REG_H_ */
drivers/gpu/drm/i915/intel_pm.c +12 -0
index 0422e2d..40f2f53 100644
@@ -3601,6 +3601,18 @@ static void haswell_init_clock_gating(struct drm_device *dev)
 	/* WaDisable4x2SubspanOptimization */
 	I915_WRITE(CACHE_MODE_1,
 		   _MASKED_BIT_ENABLE(PIXEL_SUBSPAN_COLLECT_OPT_DISABLE));
+
+	/* Work-around rc6 issue */
+	/* XXX: This is a workaround for early silicon revisions and should be
+	 * removed later
+	 */
+	I915_WRITE(WM_DBG,
+			I915_READ(WM_DBG) |
+			WM_DBG_DISALLOW_MULTIPLE_LP |
+			WM_DBG_DISALLOW_SPRITE |
+			WM_DBG_DISALLOW_MAXFIFO
+		  );
+
 }
 
 static void ivybridge_init_clock_gating(struct drm_device *dev)
Signature · 1 line
1.7.11.1
#43 lore raw
In reply to Eugeni Dodonov

2012/6/28 Eugeni Dodonov <[email protected]>:>

Quoted 5 lines

Also, for DP and eDP, Paulo Zanoni and Shobhit Kumar are already working on a larger series of patches which should appear in the next few days as well.  So while the patches in this series should work on most configurations, considerable amount of improvements are still to come.

DP patches are 1->6 and eDP patches are 14->17. We could try to get patches 7->13 and 18->21 merged first if they don't receive objections. DP/eDP then can come on top of those. It seems there's no conflict in reordering the patches.

Signature · 1 line
Paulo Zanoni