[PATCH 0/7] Haswell RC6 updates

[PATCH 0/7] Haswell RC6 updates

[Eugeni Dodonov]

Hi,

After discussing with Daniel, I've rebased the RC6-specific items on top of
Chris Wilson's GT power management improvements, and also simplified the
forcewake/rps enable/disable code paths.

Please bikeshed :).


Chris Wilson (2):
  drm/i915: Group the GT routines together in both code and vtable
  drm/i915: Implement w/a for sporadic read failures on waking from rc6

Eugeni Dodonov (5):
  drm/i915: support Haswell-style force waking
  drm/i915: add RPS configuration for Haswell
  drm/i915: disable RC6 when disabling rps
  drm/i915: introduce haswell_init_clock_gating
  drm/i915: enable RC6 workaround on Haswell

 drivers/gpu/drm/i915/i915_dma.c      |   2 +-
 drivers/gpu/drm/i915/i915_drv.c      | 131 ++++++++++++++++++++++-------
 drivers/gpu/drm/i915/i915_drv.h      |  17 ++--
 drivers/gpu/drm/i915/i915_reg.h      |  11 +++
 drivers/gpu/drm/i915/intel_display.c |   3 -
 drivers/gpu/drm/i915/intel_pm.c      | 157 +++++++++++++++++++++++------------
 6 files changed, 227 insertions(+), 94 deletions(-)

-- 
1.7.11.1

[PATCH 1/7] drm/i915: Group the GT routines together in both code and vtable

[Eugeni Dodonov]

From: Chris Wilson <chris@chris-wilson.co.uk>

Tidy up the routines for interacting with the GT (in particular the
forcewake dance) which are scattered throughout the code in a single
structure.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
---
 drivers/gpu/drm/i915/i915_dma.c      |   2 +-
 drivers/gpu/drm/i915/i915_drv.c      | 100 +++++++++++++++++++++++++----------
 drivers/gpu/drm/i915/i915_drv.h      |  17 +++---
 drivers/gpu/drm/i915/intel_display.c |   3 --
 drivers/gpu/drm/i915/intel_pm.c      |  30 -----------
 5 files changed, 81 insertions(+), 71 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
index 2166519..4dc7697 100644
--- a/drivers/gpu/drm/i915/i915_dma.c
+++ b/drivers/gpu/drm/i915/i915_dma.c
@@ -1548,6 +1548,7 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
 	}
 
 	intel_irq_init(dev);
+	intel_gt_init(dev);
 
 	/* Try to make sure MCHBAR is enabled before poking at it */
 	intel_setup_mchbar(dev);
@@ -1580,7 +1581,6 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
 	if (!IS_I945G(dev) && !IS_I945GM(dev))
 		pci_enable_msi(dev->pdev);
 
-	spin_lock_init(&dev_priv->gt_lock);
 	spin_lock_init(&dev_priv->irq_lock);
 	spin_lock_init(&dev_priv->error_lock);
 	spin_lock_init(&dev_priv->rps_lock);
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 79be879..e7b989b 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -32,6 +32,7 @@
 #include "drm.h"
 #include "i915_drm.h"
 #include "i915_drv.h"
+#include "i915_trace.h"
 #include "intel_drv.h"
 
 #include <linux/console.h>
@@ -432,36 +433,38 @@ bool i915_semaphore_is_enabled(struct drm_device *dev)
 	return 1;
 }
 
-void __gen6_gt_force_wake_get(struct drm_i915_private *dev_priv)
+static void __gen6_gt_force_wake_get(struct drm_i915_private *dev_priv)
 {
-	int count;
+	unsigned long timeout;
 
-	count = 0;
-	while (count++ < 50 && (I915_READ_NOTRACE(FORCEWAKE_ACK) & 1))
-		udelay(10);
+	timeout = jiffies + msecs_to_jiffies(1);
+	while (I915_READ_NOTRACE(FORCEWAKE_ACK) & 1 &&
+	       time_before(jiffies, timeout))
+		;
 
 	I915_WRITE_NOTRACE(FORCEWAKE, 1);
-	POSTING_READ(FORCEWAKE);
 
-	count = 0;
-	while (count++ < 50 && (I915_READ_NOTRACE(FORCEWAKE_ACK) & 1) == 0)
-		udelay(10);
+	timeout = jiffies + msecs_to_jiffies(1);
+	while ((I915_READ_NOTRACE(FORCEWAKE_ACK) & 1) == 0 &&
+	       time_before(jiffies, timeout))
+		;
 }
 
-void __gen6_gt_force_wake_mt_get(struct drm_i915_private *dev_priv)
+static void __gen6_gt_force_wake_mt_get(struct drm_i915_private *dev_priv)
 {
-	int count;
+	unsigned long timeout;
 
-	count = 0;
-	while (count++ < 50 && (I915_READ_NOTRACE(FORCEWAKE_MT_ACK) & 1))
-		udelay(10);
+	timeout = jiffies + msecs_to_jiffies(1);
+	while (I915_READ_NOTRACE(FORCEWAKE_MT_ACK) & 1 &&
+	       time_before(jiffies, timeout))
+		;
 
 	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)
-		udelay(10);
+	timeout = jiffies + msecs_to_jiffies(1);
+	while ((I915_READ_NOTRACE(FORCEWAKE_MT_ACK) & 1) == 0 &&
+	       time_before(jiffies, timeout))
+		;
 }
 
 /*
@@ -476,7 +479,7 @@ void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv)
 
 	spin_lock_irqsave(&dev_priv->gt_lock, irqflags);
 	if (dev_priv->forcewake_count++ == 0)
-		dev_priv->display.force_wake_get(dev_priv);
+		dev_priv->gt.force_wake_get(dev_priv);
 	spin_unlock_irqrestore(&dev_priv->gt_lock, irqflags);
 }
 
@@ -489,14 +492,14 @@ static void gen6_gt_check_fifodbg(struct drm_i915_private *dev_priv)
 		I915_WRITE_NOTRACE(GTFIFODBG, GT_FIFO_CPU_ERROR_MASK);
 }
 
-void __gen6_gt_force_wake_put(struct drm_i915_private *dev_priv)
+static void __gen6_gt_force_wake_put(struct drm_i915_private *dev_priv)
 {
 	I915_WRITE_NOTRACE(FORCEWAKE, 0);
 	/* The below doubles as a POSTING_READ */
 	gen6_gt_check_fifodbg(dev_priv);
 }
 
-void __gen6_gt_force_wake_mt_put(struct drm_i915_private *dev_priv)
+static void __gen6_gt_force_wake_mt_put(struct drm_i915_private *dev_priv)
 {
 	I915_WRITE_NOTRACE(FORCEWAKE_MT, _MASKED_BIT_DISABLE(1));
 	/* The below doubles as a POSTING_READ */
@@ -512,7 +515,7 @@ void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv)
 
 	spin_lock_irqsave(&dev_priv->gt_lock, irqflags);
 	if (--dev_priv->forcewake_count == 0)
-		dev_priv->display.force_wake_put(dev_priv);
+		dev_priv->gt.force_wake_put(dev_priv);
 	spin_unlock_irqrestore(&dev_priv->gt_lock, irqflags);
 }
 
@@ -536,7 +539,7 @@ int __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv)
 	return ret;
 }
 
-void vlv_force_wake_get(struct drm_i915_private *dev_priv)
+static void vlv_force_wake_get(struct drm_i915_private *dev_priv)
 {
 	int count;
 
@@ -554,13 +557,54 @@ void vlv_force_wake_get(struct drm_i915_private *dev_priv)
 		udelay(10);
 }
 
-void vlv_force_wake_put(struct drm_i915_private *dev_priv)
+static void vlv_force_wake_put(struct drm_i915_private *dev_priv)
 {
 	I915_WRITE_NOTRACE(FORCEWAKE_VLV, 0xffff0000);
 	/* FIXME: confirm VLV behavior with Punit folks */
 	POSTING_READ(FORCEWAKE_VLV);
 }
 
+void intel_gt_init(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+
+	spin_lock_init(&dev_priv->gt_lock);
+
+	if (IS_VALLEYVIEW(dev)) {
+		dev_priv->gt.force_wake_get = vlv_force_wake_get;
+		dev_priv->gt.force_wake_put = vlv_force_wake_put;
+	} else if (INTEL_INFO(dev)->gen >= 6) {
+		dev_priv->gt.force_wake_get = __gen6_gt_force_wake_get;
+		dev_priv->gt.force_wake_put = __gen6_gt_force_wake_put;
+
+		/* IVB configs may use multi-threaded forcewake */
+		if (IS_IVYBRIDGE(dev) || IS_HASWELL(dev)) {
+			u32 ecobus;
+
+			/* A small trick here - if the bios hasn't configured
+			 * MT forcewake, and if the device is in RC6, then
+			 * force_wake_mt_get will not wake the device and the
+			 * ECOBUS read will return zero. Which will be
+			 * (correctly) interpreted by the test below as MT
+			 * forcewake being disabled.
+			 */
+			mutex_lock(&dev->struct_mutex);
+			__gen6_gt_force_wake_mt_get(dev_priv);
+			ecobus = I915_READ_NOTRACE(ECOBUS);
+			__gen6_gt_force_wake_mt_put(dev_priv);
+			mutex_unlock(&dev->struct_mutex);
+
+			if (ecobus & FORCEWAKE_MT_ENABLE) {
+				DRM_DEBUG_KMS("Using MT version of forcewake\n");
+				dev_priv->gt.force_wake_get =
+					__gen6_gt_force_wake_mt_get;
+				dev_priv->gt.force_wake_put =
+					__gen6_gt_force_wake_mt_put;
+			}
+		}
+	}
+}
+
 static int i915_drm_freeze(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -797,9 +841,9 @@ static int gen6_do_reset(struct drm_device *dev)
 
 	/* If reset with a user forcewake, try to restore, otherwise turn it off */
 	if (dev_priv->forcewake_count)
-		dev_priv->display.force_wake_get(dev_priv);
+		dev_priv->gt.force_wake_get(dev_priv);
 	else
-		dev_priv->display.force_wake_put(dev_priv);
+		dev_priv->gt.force_wake_put(dev_priv);
 
 	/* Restore fifo count */
 	dev_priv->gt_fifo_count = I915_READ_NOTRACE(GT_FIFO_FREE_ENTRIES);
@@ -1248,10 +1292,10 @@ u##x i915_read##x(struct drm_i915_private *dev_priv, u32 reg) { \
 		unsigned long irqflags; \
 		spin_lock_irqsave(&dev_priv->gt_lock, irqflags); \
 		if (dev_priv->forcewake_count == 0) \
-			dev_priv->display.force_wake_get(dev_priv); \
+			dev_priv->gt.force_wake_get(dev_priv); \
 		val = read##y(dev_priv->regs + reg); \
 		if (dev_priv->forcewake_count == 0) \
-			dev_priv->display.force_wake_put(dev_priv); \
+			dev_priv->gt.force_wake_put(dev_priv); \
 		spin_unlock_irqrestore(&dev_priv->gt_lock, irqflags); \
 	} else if (IS_VALLEYVIEW(dev_priv->dev) && IS_DISPLAYREG(reg)) { \
 		val = read##y(dev_priv->regs + reg + 0x180000);		\
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index a0c15ab..60f6974 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -262,8 +262,6 @@ struct drm_i915_display_funcs {
 			  struct drm_i915_gem_object *obj);
 	int (*update_plane)(struct drm_crtc *crtc, struct drm_framebuffer *fb,
 			    int x, int y);
-	void (*force_wake_get)(struct drm_i915_private *dev_priv);
-	void (*force_wake_put)(struct drm_i915_private *dev_priv);
 	/* clock updates for mode set */
 	/* cursor updates */
 	/* render clock increase/decrease */
@@ -271,6 +269,11 @@ struct drm_i915_display_funcs {
 	/* pll clock increase/decrease */
 };
 
+struct drm_i915_gt_funcs {
+	void (*force_wake_get)(struct drm_i915_private *dev_priv);
+	void (*force_wake_put)(struct drm_i915_private *dev_priv);
+};
+
 struct intel_device_info {
 	u8 gen;
 	u8 is_mobile:1;
@@ -362,6 +365,8 @@ typedef struct drm_i915_private {
 	int relative_constants_mode;
 
 	void __iomem *regs;
+
+	struct drm_i915_gt_funcs gt;
 	/** gt_fifo_count and the subsequent register write are synchronized
 	 * with dev->struct_mutex. */
 	unsigned gt_fifo_count;
@@ -1200,6 +1205,7 @@ void i915_hangcheck_elapsed(unsigned long data);
 void i915_handle_error(struct drm_device *dev, bool wedged);
 
 extern void intel_irq_init(struct drm_device *dev);
+extern void intel_gt_init(struct drm_device *dev);
 
 void i915_error_state_free(struct kref *error_ref);
 
@@ -1517,13 +1523,6 @@ extern int intel_trans_dp_port_sel(struct drm_crtc *crtc);
 extern int intel_enable_rc6(const struct drm_device *dev);
 
 extern bool i915_semaphore_is_enabled(struct drm_device *dev);
-extern void __gen6_gt_force_wake_get(struct drm_i915_private *dev_priv);
-extern void __gen6_gt_force_wake_mt_get(struct drm_i915_private *dev_priv);
-extern void __gen6_gt_force_wake_put(struct drm_i915_private *dev_priv);
-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);
 
 /* overlay */
 #ifdef CONFIG_DEBUG_FS
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index 0438a10..3ca91e2 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -7121,9 +7121,6 @@ static void intel_init_display(struct drm_device *dev)
 			dev_priv->display.write_eld = ironlake_write_eld;
 		} else
 			dev_priv->display.update_wm = NULL;
-	} else if (IS_VALLEYVIEW(dev)) {
-		dev_priv->display.force_wake_get = vlv_force_wake_get;
-		dev_priv->display.force_wake_put = vlv_force_wake_put;
 	} else if (IS_G4X(dev)) {
 		dev_priv->display.write_eld = g4x_write_eld;
 	}
diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
index a1d3da7..29720d2 100644
--- a/drivers/gpu/drm/i915/intel_pm.c
+++ b/drivers/gpu/drm/i915/intel_pm.c
@@ -3763,34 +3763,6 @@ void intel_init_pm(struct drm_device *dev)
 
 	/* For FIFO watermark updates */
 	if (HAS_PCH_SPLIT(dev)) {
-		dev_priv->display.force_wake_get = __gen6_gt_force_wake_get;
-		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)) {
-			u32	ecobus;
-
-			/* A small trick here - if the bios hasn't configured MT forcewake,
-			 * and if the device is in RC6, then force_wake_mt_get will not wake
-			 * the device and the ECOBUS read will return zero. Which will be
-			 * (correctly) interpreted by the test below as MT forcewake being
-			 * disabled.
-			 */
-			mutex_lock(&dev->struct_mutex);
-			__gen6_gt_force_wake_mt_get(dev_priv);
-			ecobus = I915_READ_NOTRACE(ECOBUS);
-			__gen6_gt_force_wake_mt_put(dev_priv);
-			mutex_unlock(&dev->struct_mutex);
-
-			if (ecobus & FORCEWAKE_MT_ENABLE) {
-				DRM_DEBUG_KMS("Using MT version of forcewake\n");
-				dev_priv->display.force_wake_get =
-					__gen6_gt_force_wake_mt_get;
-				dev_priv->display.force_wake_put =
-					__gen6_gt_force_wake_mt_put;
-			}
-		}
-
 		if (HAS_PCH_IBX(dev))
 			dev_priv->display.init_pch_clock_gating = ibx_init_clock_gating;
 		else if (HAS_PCH_CPT(dev))
@@ -3846,8 +3818,6 @@ void intel_init_pm(struct drm_device *dev)
 		dev_priv->display.update_wm = valleyview_update_wm;
 		dev_priv->display.init_clock_gating =
 			valleyview_init_clock_gating;
-		dev_priv->display.force_wake_get = vlv_force_wake_get;
-		dev_priv->display.force_wake_put = vlv_force_wake_put;
 	} else if (IS_PINEVIEW(dev)) {
 		if (!intel_get_cxsr_latency(IS_PINEVIEW_G(dev),
 					    dev_priv->is_ddr3,
-- 
1.7.11.1

[PATCH 2/7] drm/i915: Implement w/a for sporadic read failures on waking from rc6

[Eugeni Dodonov]

From: Chris Wilson <chris@chris-wilson.co.uk>

As a w/a to prevent reads sporadically returning 0, we need to wait for
the GT thread to return to TC0 before proceeding to read the registers.

v2: adapt for Haswell changes (Eugeni).

References: https://bugs.freedesktop.org/show_bug.cgi?id=50243
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
---
 drivers/gpu/drm/i915/i915_drv.c | 25 +++++++++++++++++++++++++
 drivers/gpu/drm/i915/i915_reg.h |  4 ++++
 2 files changed, 29 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index e7b989b..3ef2103 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -433,6 +433,25 @@ bool i915_semaphore_is_enabled(struct drm_device *dev)
 	return 1;
 }
 
+static void __gen6_gt_wait_for_thread_c0(struct drm_i915_private *dev_priv)
+{
+	unsigned long timeout;
+	u32 gt_thread_status_mask;
+
+	if (IS_HASWELL(dev_priv->dev))
+		gt_thread_status_mask = GEN6_GT_THREAD_STATUS_CORE_MASK_HSW;
+	else
+		gt_thread_status_mask = GEN6_GT_THREAD_STATUS_CORE_MASK;
+
+	/* w/a for a sporadic read returning 0 by waiting for the GT
+	 * thread to wake up.
+	 */
+	timeout = jiffies + msecs_to_jiffies(1);
+	while (I915_READ_NOTRACE(GEN6_GT_THREAD_STATUS_REG) & gt_thread_status_mask &&
+	       time_before(jiffies, timeout))
+		;
+}
+
 static void __gen6_gt_force_wake_get(struct drm_i915_private *dev_priv)
 {
 	unsigned long timeout;
@@ -448,6 +467,8 @@ static void __gen6_gt_force_wake_get(struct drm_i915_private *dev_priv)
 	while ((I915_READ_NOTRACE(FORCEWAKE_ACK) & 1) == 0 &&
 	       time_before(jiffies, timeout))
 		;
+
+	__gen6_gt_wait_for_thread_c0(dev_priv);
 }
 
 static void __gen6_gt_force_wake_mt_get(struct drm_i915_private *dev_priv)
@@ -465,6 +486,8 @@ static void __gen6_gt_force_wake_mt_get(struct drm_i915_private *dev_priv)
 	while ((I915_READ_NOTRACE(FORCEWAKE_MT_ACK) & 1) == 0 &&
 	       time_before(jiffies, timeout))
 		;
+
+	__gen6_gt_wait_for_thread_c0(dev_priv);
 }
 
 /*
@@ -555,6 +578,8 @@ static void vlv_force_wake_get(struct drm_i915_private *dev_priv)
 	count = 0;
 	while (count++ < 50 && (I915_READ_NOTRACE(FORCEWAKE_ACK_VLV) & 1) == 0)
 		udelay(10);
+
+	__gen6_gt_wait_for_thread_c0(dev_priv);
 }
 
 static void vlv_force_wake_put(struct drm_i915_private *dev_priv)
diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index d8f516b..20f7f0d 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -1458,6 +1458,10 @@
 #define DDRMPLL1		0X12c20
 #define PEG_BAND_GAP_DATA	0x14d68
 
+#define GEN6_GT_THREAD_STATUS_REG 0x13805c
+#define GEN6_GT_THREAD_STATUS_CORE_MASK 0x7
+#define GEN6_GT_THREAD_STATUS_CORE_MASK_HSW (0x7 | (0x07 << 16))
+
 #define GEN6_GT_PERF_STATUS	0x145948
 #define GEN6_RP_STATE_LIMITS	0x145994
 #define GEN6_RP_STATE_CAP	0x145998
-- 
1.7.11.1

[PATCH 3/7] drm/i915: support Haswell-style force waking

[Eugeni Dodonov]

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

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

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
---
 drivers/gpu/drm/i915/i915_drv.c | 10 ++++++++--
 drivers/gpu/drm/i915/i915_reg.h |  1 +
 2 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 3ef2103..cc8126e 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -474,16 +474,22 @@ static void __gen6_gt_force_wake_get(struct drm_i915_private *dev_priv)
 static void __gen6_gt_force_wake_mt_get(struct drm_i915_private *dev_priv)
 {
 	unsigned long timeout;
+	u32 forcewake_ack;
+
+	if (IS_HASWELL(dev_priv->dev))
+		forcewake_ack = FORCEWAKE_ACK_HSW;
+	else
+		forcewake_ack = FORCEWAKE_MT_ACK;
 
 	timeout = jiffies + msecs_to_jiffies(1);
-	while (I915_READ_NOTRACE(FORCEWAKE_MT_ACK) & 1 &&
+	while (I915_READ_NOTRACE(forcewake_ack) & 1 &&
 	       time_before(jiffies, timeout))
 		;
 
 	I915_WRITE_NOTRACE(FORCEWAKE_MT, _MASKED_BIT_ENABLE(1));
 
 	timeout = jiffies + msecs_to_jiffies(1);
-	while ((I915_READ_NOTRACE(FORCEWAKE_MT_ACK) & 1) == 0 &&
+	while ((I915_READ_NOTRACE(forcewake_ack) & 1) == 0 &&
 	       time_before(jiffies, timeout))
 		;
 
diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index 20f7f0d..f17de3d 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -4095,6 +4095,7 @@
 #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
-- 
1.7.11.1

[PATCH 4/7] drm/i915: add RPS configuration for Haswell

[Eugeni Dodonov]

Split Haswell-specific GT algorithms into its own function.

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

v2: reuse gen6 routines to simplify the code flow.

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
---
 drivers/gpu/drm/i915/i915_reg.h |  1 +
 drivers/gpu/drm/i915/intel_pm.c | 60 ++++++++++++++++++++++++++---------------
 2 files changed, 39 insertions(+), 22 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index f17de3d..9d5bf06 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -4156,6 +4156,7 @@
 #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
diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
index 29720d2..cf4dd16 100644
--- a/drivers/gpu/drm/i915/intel_pm.c
+++ b/drivers/gpu/drm/i915/intel_pm.c
@@ -2332,9 +2332,10 @@ 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
@@ -2396,26 +2397,34 @@ static void gen6_enable_rps(struct drm_device *dev)
 	for_each_ring(ring, dev_priv, i)
 		I915_WRITE(RING_MAX_IDLE(ring->mmio_base), 10);
 
+	/* Check if we are enabling RC6 */
+	rc6_mode = intel_enable_rc6(dev_priv->dev);
+
+	if (rc6_mode & INTEL_RC6_ENABLE)
+		rc6_mask |= GEN6_RC_CTL_RC6_ENABLE;
+
+	/* Write RC6 thresholds for RC1e and RC6 states */
 	I915_WRITE(GEN6_RC_SLEEP, 0);
 	I915_WRITE(GEN6_RC1e_THRESHOLD, 1000);
 	I915_WRITE(GEN6_RC6_THRESHOLD, 50000);
-	I915_WRITE(GEN6_RC6p_THRESHOLD, 100000);
-	I915_WRITE(GEN6_RC6pp_THRESHOLD, 64000); /* unused */
 
-	rc6_mode = intel_enable_rc6(dev_priv->dev);
-	if (rc6_mode & INTEL_RC6_ENABLE)
-		rc6_mask |= GEN6_RC_CTL_RC6_ENABLE;
+	/* We don't use those on Haswell */
+	if (!IS_HASWELL(dev)) {
+		I915_WRITE(GEN6_RC6p_THRESHOLD, 100000);
+		I915_WRITE(GEN6_RC6pp_THRESHOLD, 64000); /* unused */
+
+		if (rc6_mode & INTEL_RC6p_ENABLE)
+			rc6_mask |= GEN6_RC_CTL_RC6p_ENABLE;
 
-	if (rc6_mode & INTEL_RC6p_ENABLE)
-		rc6_mask |= GEN6_RC_CTL_RC6p_ENABLE;
+		if (rc6_mode & INTEL_RC6pp_ENABLE)
+			rc6_mask |= GEN6_RC_CTL_RC6pp_ENABLE;
 
-	if (rc6_mode & INTEL_RC6pp_ENABLE)
-		rc6_mask |= GEN6_RC_CTL_RC6pp_ENABLE;
+	}
 
 	DRM_INFO("Enabling RC6 states: RC6 %s, RC6p %s, RC6pp %s\n",
-			(rc6_mode & INTEL_RC6_ENABLE) ? "on" : "off",
-			(rc6_mode & INTEL_RC6p_ENABLE) ? "on" : "off",
-			(rc6_mode & INTEL_RC6pp_ENABLE) ? "on" : "off");
+			(rc6_mask & GEN6_RC_CTL_RC6_ENABLE) ? "on" : "off",
+			(rc6_mask & GEN6_RC_CTL_RC6p_ENABLE) ? "on" : "off",
+			(rc6_mask & GEN6_RC_CTL_RC6pp_ENABLE) ? "on" : "off");
 
 	I915_WRITE(GEN6_RC_CONTROL,
 		   rc6_mask |
@@ -2433,10 +2442,19 @@ static void gen6_enable_rps(struct drm_device *dev)
 	I915_WRITE(GEN6_RP_INTERRUPT_LIMITS,
 		   dev_priv->max_delay << 24 |
 		   dev_priv->min_delay << 16);
-	I915_WRITE(GEN6_RP_UP_THRESHOLD, 10000);
-	I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 1000000);
-	I915_WRITE(GEN6_RP_UP_EI, 100000);
-	I915_WRITE(GEN6_RP_DOWN_EI, 5000000);
+
+	if (IS_HASWELL(dev)) {
+		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);
+	} else {
+		I915_WRITE(GEN6_RP_UP_THRESHOLD, 10000);
+		I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 1000000);
+		I915_WRITE(GEN6_RP_UP_EI, 100000);
+		I915_WRITE(GEN6_RP_DOWN_EI, 5000000);
+	}
+
 	I915_WRITE(GEN6_RP_IDLE_HYSTERSIS, 10);
 	I915_WRITE(GEN6_RP_CONTROL,
 		   GEN6_RP_MEDIA_TURBO |
@@ -3225,7 +3243,7 @@ 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 (INTEL_INFO(dev)->gen >= 6 && !IS_VALLEYVIEW(dev))
 		gen6_disable_rps(dev);
 }
 
@@ -3235,9 +3253,7 @@ 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_GEN6(dev) || IS_GEN7(dev)) && !IS_VALLEYVIEW(dev)) {
 		gen6_enable_rps(dev);
 		gen6_update_ring_freq(dev);
 	}
-- 
1.7.11.1

[PATCH 5/7] drm/i915: disable RC6 when disabling rps

[Eugeni Dodonov]

We weren't disabling RC6 bits when bringing down RPS.

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
---
 drivers/gpu/drm/i915/intel_pm.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
index cf4dd16..64c02bc 100644
--- a/drivers/gpu/drm/i915/intel_pm.c
+++ b/drivers/gpu/drm/i915/intel_pm.c
@@ -2303,6 +2303,7 @@ static void gen6_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);
-- 
1.7.11.1

[PATCH 6/7] drm/i915: introduce haswell_init_clock_gating

[Eugeni Dodonov]

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

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
---
 drivers/gpu/drm/i915/intel_pm.c | 54 ++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 53 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
index 64c02bc..d7ea540 100644
--- a/drivers/gpu/drm/i915/intel_pm.c
+++ b/drivers/gpu/drm/i915/intel_pm.c
@@ -3418,6 +3418,58 @@ 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;
@@ -3827,7 +3879,7 @@ 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;
 		} else
 			dev_priv->display.update_wm = NULL;
-- 
1.7.11.1

[PATCH 7/7] drm/i915: enable RC6 workaround on Haswell

[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.

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

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
---
 drivers/gpu/drm/i915/i915_reg.h |  5 +++++
 drivers/gpu/drm/i915/intel_pm.c | 12 ++++++++++++
 2 files changed, 17 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index 9d5bf06..793a172 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -4516,4 +4516,9 @@
 #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_ */
diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
index d7ea540..8d2d0b9 100644
--- a/drivers/gpu/drm/i915/intel_pm.c
+++ b/drivers/gpu/drm/i915/intel_pm.c
@@ -3468,6 +3468,18 @@ 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)
-- 
1.7.11.1

[PATCH 4/7] drm/i915: add RPS configuration for Haswell

[Eugeni Dodonov]

Split Haswell-specific GT algorithms into its own function.

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

v2: reuse gen6 routines to simplify the code flow.
v2.1.: actually use proper RP controls for HSW.

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
---
 drivers/gpu/drm/i915/i915_reg.h |  1 +
 drivers/gpu/drm/i915/intel_pm.c | 73 ++++++++++++++++++++++++++++-------------
 2 files changed, 51 insertions(+), 23 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index f17de3d..9d5bf06 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -4156,6 +4156,7 @@
 #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
diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
index 29720d2..c6b5201 100644
--- a/drivers/gpu/drm/i915/intel_pm.c
+++ b/drivers/gpu/drm/i915/intel_pm.c
@@ -2332,9 +2332,10 @@ 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
@@ -2396,26 +2397,34 @@ static void gen6_enable_rps(struct drm_device *dev)
 	for_each_ring(ring, dev_priv, i)
 		I915_WRITE(RING_MAX_IDLE(ring->mmio_base), 10);
 
+	/* Check if we are enabling RC6 */
+	rc6_mode = intel_enable_rc6(dev_priv->dev);
+
+	if (rc6_mode & INTEL_RC6_ENABLE)
+		rc6_mask |= GEN6_RC_CTL_RC6_ENABLE;
+
+	/* Write RC6 thresholds for RC1e and RC6 states */
 	I915_WRITE(GEN6_RC_SLEEP, 0);
 	I915_WRITE(GEN6_RC1e_THRESHOLD, 1000);
 	I915_WRITE(GEN6_RC6_THRESHOLD, 50000);
-	I915_WRITE(GEN6_RC6p_THRESHOLD, 100000);
-	I915_WRITE(GEN6_RC6pp_THRESHOLD, 64000); /* unused */
 
-	rc6_mode = intel_enable_rc6(dev_priv->dev);
-	if (rc6_mode & INTEL_RC6_ENABLE)
-		rc6_mask |= GEN6_RC_CTL_RC6_ENABLE;
+	/* We don't use those on Haswell */
+	if (!IS_HASWELL(dev)) {
+		I915_WRITE(GEN6_RC6p_THRESHOLD, 100000);
+		I915_WRITE(GEN6_RC6pp_THRESHOLD, 64000); /* unused */
+
+		if (rc6_mode & INTEL_RC6p_ENABLE)
+			rc6_mask |= GEN6_RC_CTL_RC6p_ENABLE;
 
-	if (rc6_mode & INTEL_RC6p_ENABLE)
-		rc6_mask |= GEN6_RC_CTL_RC6p_ENABLE;
+		if (rc6_mode & INTEL_RC6pp_ENABLE)
+			rc6_mask |= GEN6_RC_CTL_RC6pp_ENABLE;
 
-	if (rc6_mode & INTEL_RC6pp_ENABLE)
-		rc6_mask |= GEN6_RC_CTL_RC6pp_ENABLE;
+	}
 
 	DRM_INFO("Enabling RC6 states: RC6 %s, RC6p %s, RC6pp %s\n",
-			(rc6_mode & INTEL_RC6_ENABLE) ? "on" : "off",
-			(rc6_mode & INTEL_RC6p_ENABLE) ? "on" : "off",
-			(rc6_mode & INTEL_RC6pp_ENABLE) ? "on" : "off");
+			(rc6_mask & GEN6_RC_CTL_RC6_ENABLE) ? "on" : "off",
+			(rc6_mask & GEN6_RC_CTL_RC6p_ENABLE) ? "on" : "off",
+			(rc6_mask & GEN6_RC_CTL_RC6pp_ENABLE) ? "on" : "off");
 
 	I915_WRITE(GEN6_RC_CONTROL,
 		   rc6_mask |
@@ -2433,12 +2442,32 @@ static void gen6_enable_rps(struct drm_device *dev)
 	I915_WRITE(GEN6_RP_INTERRUPT_LIMITS,
 		   dev_priv->max_delay << 24 |
 		   dev_priv->min_delay << 16);
-	I915_WRITE(GEN6_RP_UP_THRESHOLD, 10000);
-	I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 1000000);
-	I915_WRITE(GEN6_RP_UP_EI, 100000);
-	I915_WRITE(GEN6_RP_DOWN_EI, 5000000);
+
+	if (IS_HASWELL(dev)) {
+		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);
+	} else {
+		I915_WRITE(GEN6_RP_UP_THRESHOLD, 10000);
+		I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 1000000);
+		I915_WRITE(GEN6_RP_UP_EI, 100000);
+		I915_WRITE(GEN6_RP_DOWN_EI, 5000000);
+	}
+
 	I915_WRITE(GEN6_RP_IDLE_HYSTERSIS, 10);
-	I915_WRITE(GEN6_RP_CONTROL,
+
+	/* We have a bit different default configuration for HSW */
+	if (IS_HASWELL(dev))
+		I915_WRITE(GEN6_RP_CONTROL,
+		   GEN6_RP_MEDIA_TURBO |
+		   GEN6_RP_MEDIA_HW_MODE |
+		   GEN6_RP_MEDIA_IS_GFX |
+		   GEN6_RP_ENABLE |
+		   GEN6_RP_UP_BUSY_AVG |
+		   GEN7_RP_DOWN_IDLE_AVG);
+	else
+		I915_WRITE(GEN6_RP_CONTROL,
 		   GEN6_RP_MEDIA_TURBO |
 		   GEN6_RP_MEDIA_HW_NORMAL_MODE |
 		   GEN6_RP_MEDIA_IS_GFX |
@@ -3225,7 +3254,7 @@ 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 (INTEL_INFO(dev)->gen >= 6 && !IS_VALLEYVIEW(dev))
 		gen6_disable_rps(dev);
 }
 
@@ -3235,9 +3264,7 @@ 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_GEN6(dev) || IS_GEN7(dev)) && !IS_VALLEYVIEW(dev)) {
 		gen6_enable_rps(dev);
 		gen6_update_ring_freq(dev);
 	}
-- 
1.7.11.1

[PATCH 4/7] drm/i915: add RPS configuration for Haswell

[Eugeni Dodonov]

Split Haswell-specific GT algorithms into its own function.

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

v2: reuse gen6 routines to simplify the code flow.
v2.1.: actually use proper RP controls for HSW.

v2.2: experiments have shown that GEN6_RP_MEDIA_HW_NORMAL_MODE works
better indeed, so use it on Haswell as well.

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
---
 drivers/gpu/drm/i915/i915_reg.h |  1 +
 drivers/gpu/drm/i915/intel_pm.c | 73 ++++++++++++++++++++++++++++-------------
 2 files changed, 51 insertions(+), 23 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index f17de3d..9d5bf06 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -4156,6 +4156,7 @@
 #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
diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
index 29720d2..9e0dd5d 100644
--- a/drivers/gpu/drm/i915/intel_pm.c
+++ b/drivers/gpu/drm/i915/intel_pm.c
@@ -2332,9 +2332,10 @@ 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
@@ -2396,26 +2397,34 @@ static void gen6_enable_rps(struct drm_device *dev)
 	for_each_ring(ring, dev_priv, i)
 		I915_WRITE(RING_MAX_IDLE(ring->mmio_base), 10);
 
+	/* Check if we are enabling RC6 */
+	rc6_mode = intel_enable_rc6(dev_priv->dev);
+
+	if (rc6_mode & INTEL_RC6_ENABLE)
+		rc6_mask |= GEN6_RC_CTL_RC6_ENABLE;
+
+	/* Write RC6 thresholds for RC1e and RC6 states */
 	I915_WRITE(GEN6_RC_SLEEP, 0);
 	I915_WRITE(GEN6_RC1e_THRESHOLD, 1000);
 	I915_WRITE(GEN6_RC6_THRESHOLD, 50000);
-	I915_WRITE(GEN6_RC6p_THRESHOLD, 100000);
-	I915_WRITE(GEN6_RC6pp_THRESHOLD, 64000); /* unused */
 
-	rc6_mode = intel_enable_rc6(dev_priv->dev);
-	if (rc6_mode & INTEL_RC6_ENABLE)
-		rc6_mask |= GEN6_RC_CTL_RC6_ENABLE;
+	/* We don't use those on Haswell */
+	if (!IS_HASWELL(dev)) {
+		I915_WRITE(GEN6_RC6p_THRESHOLD, 100000);
+		I915_WRITE(GEN6_RC6pp_THRESHOLD, 64000); /* unused */
 
-	if (rc6_mode & INTEL_RC6p_ENABLE)
-		rc6_mask |= GEN6_RC_CTL_RC6p_ENABLE;
+		if (rc6_mode & INTEL_RC6p_ENABLE)
+			rc6_mask |= GEN6_RC_CTL_RC6p_ENABLE;
 
-	if (rc6_mode & INTEL_RC6pp_ENABLE)
-		rc6_mask |= GEN6_RC_CTL_RC6pp_ENABLE;
+		if (rc6_mode & INTEL_RC6pp_ENABLE)
+			rc6_mask |= GEN6_RC_CTL_RC6pp_ENABLE;
+
+	}
 
 	DRM_INFO("Enabling RC6 states: RC6 %s, RC6p %s, RC6pp %s\n",
-			(rc6_mode & INTEL_RC6_ENABLE) ? "on" : "off",
-			(rc6_mode & INTEL_RC6p_ENABLE) ? "on" : "off",
-			(rc6_mode & INTEL_RC6pp_ENABLE) ? "on" : "off");
+			(rc6_mask & GEN6_RC_CTL_RC6_ENABLE) ? "on" : "off",
+			(rc6_mask & GEN6_RC_CTL_RC6p_ENABLE) ? "on" : "off",
+			(rc6_mask & GEN6_RC_CTL_RC6pp_ENABLE) ? "on" : "off");
 
 	I915_WRITE(GEN6_RC_CONTROL,
 		   rc6_mask |
@@ -2433,12 +2442,32 @@ static void gen6_enable_rps(struct drm_device *dev)
 	I915_WRITE(GEN6_RP_INTERRUPT_LIMITS,
 		   dev_priv->max_delay << 24 |
 		   dev_priv->min_delay << 16);
-	I915_WRITE(GEN6_RP_UP_THRESHOLD, 10000);
-	I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 1000000);
-	I915_WRITE(GEN6_RP_UP_EI, 100000);
-	I915_WRITE(GEN6_RP_DOWN_EI, 5000000);
+
+	if (IS_HASWELL(dev)) {
+		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);
+	} else {
+		I915_WRITE(GEN6_RP_UP_THRESHOLD, 10000);
+		I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 1000000);
+		I915_WRITE(GEN6_RP_UP_EI, 100000);
+		I915_WRITE(GEN6_RP_DOWN_EI, 5000000);
+	}
+
 	I915_WRITE(GEN6_RP_IDLE_HYSTERSIS, 10);
-	I915_WRITE(GEN6_RP_CONTROL,
+
+	/* We have a bit different default configuration for HSW */
+	if (IS_HASWELL(dev))
+		I915_WRITE(GEN6_RP_CONTROL,
+		   GEN6_RP_MEDIA_TURBO |
+		   GEN6_RP_MEDIA_HW_NORMAL_MODE |
+		   GEN6_RP_MEDIA_IS_GFX |
+		   GEN6_RP_ENABLE |
+		   GEN6_RP_UP_BUSY_AVG |
+		   GEN7_RP_DOWN_IDLE_AVG);
+	else
+		I915_WRITE(GEN6_RP_CONTROL,
 		   GEN6_RP_MEDIA_TURBO |
 		   GEN6_RP_MEDIA_HW_NORMAL_MODE |
 		   GEN6_RP_MEDIA_IS_GFX |
@@ -3225,7 +3254,7 @@ 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 (INTEL_INFO(dev)->gen >= 6 && !IS_VALLEYVIEW(dev))
 		gen6_disable_rps(dev);
 }
 
@@ -3235,9 +3264,7 @@ 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_GEN6(dev) || IS_GEN7(dev)) && !IS_VALLEYVIEW(dev)) {
 		gen6_enable_rps(dev);
 		gen6_update_ring_freq(dev);
 	}
-- 
1.7.11.1

Re: [PATCH 4/7] drm/i915: add RPS configuration for Haswell

[Jesse Barnes]

On Fri, 29 Jun 2012 11:50:35 -0300
Eugeni Dodonov <eugeni.dodonov@intel.com> wrote:

> Split Haswell-specific GT algorithms into its own function.
> 
> Note that Haswell only has RC6, so account for that as well.
> 
> v2: reuse gen6 routines to simplify the code flow.
> v2.1.: actually use proper RP controls for HSW.
> 
> Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
> ---
>  drivers/gpu/drm/i915/i915_reg.h |  1 +
>  drivers/gpu/drm/i915/intel_pm.c | 73 ++++++++++++++++++++++++++++-------------
>  2 files changed, 51 insertions(+), 23 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> index f17de3d..9d5bf06 100644
> --- a/drivers/gpu/drm/i915/i915_reg.h
> +++ b/drivers/gpu/drm/i915/i915_reg.h
> @@ -4156,6 +4156,7 @@
>  #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
> diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
> index 29720d2..c6b5201 100644
> --- a/drivers/gpu/drm/i915/intel_pm.c
> +++ b/drivers/gpu/drm/i915/intel_pm.c
> @@ -2332,9 +2332,10 @@ 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;

This looks a little convoluted.  Should we call this routine on gen5?
If not, do we need these checks and return values?  If we got rid of
them we could just move the HSW check to enable_rps instead...

Or just make it return the bitmask directly and do rc6_mask =
calc_rc6_bits() or somesuch.  That would save us lines in enable_rps.

>  
>  	/*
>  	 * Disable rc6 on Sandybridge
> @@ -2396,26 +2397,34 @@ static void gen6_enable_rps(struct drm_device *dev)
>  	for_each_ring(ring, dev_priv, i)
>  		I915_WRITE(RING_MAX_IDLE(ring->mmio_base), 10);
>  
> +	/* Check if we are enabling RC6 */
> +	rc6_mode = intel_enable_rc6(dev_priv->dev);
> +
> +	if (rc6_mode & INTEL_RC6_ENABLE)
> +		rc6_mask |= GEN6_RC_CTL_RC6_ENABLE;
> +
> +	/* Write RC6 thresholds for RC1e and RC6 states */
>  	I915_WRITE(GEN6_RC_SLEEP, 0);
>  	I915_WRITE(GEN6_RC1e_THRESHOLD, 1000);
>  	I915_WRITE(GEN6_RC6_THRESHOLD, 50000);
> -	I915_WRITE(GEN6_RC6p_THRESHOLD, 100000);
> -	I915_WRITE(GEN6_RC6pp_THRESHOLD, 64000); /* unused */
>  
> -	rc6_mode = intel_enable_rc6(dev_priv->dev);
> -	if (rc6_mode & INTEL_RC6_ENABLE)
> -		rc6_mask |= GEN6_RC_CTL_RC6_ENABLE;
> +	/* We don't use those on Haswell */
> +	if (!IS_HASWELL(dev)) {
> +		I915_WRITE(GEN6_RC6p_THRESHOLD, 100000);
> +		I915_WRITE(GEN6_RC6pp_THRESHOLD, 64000); /* unused */
> +
> +		if (rc6_mode & INTEL_RC6p_ENABLE)
> +			rc6_mask |= GEN6_RC_CTL_RC6p_ENABLE;
>  
> -	if (rc6_mode & INTEL_RC6p_ENABLE)
> -		rc6_mask |= GEN6_RC_CTL_RC6p_ENABLE;
> +		if (rc6_mode & INTEL_RC6pp_ENABLE)
> +			rc6_mask |= GEN6_RC_CTL_RC6pp_ENABLE;
>  
> -	if (rc6_mode & INTEL_RC6pp_ENABLE)
> -		rc6_mask |= GEN6_RC_CTL_RC6pp_ENABLE;
> +	}
>  
>  	DRM_INFO("Enabling RC6 states: RC6 %s, RC6p %s, RC6pp %s\n",
> -			(rc6_mode & INTEL_RC6_ENABLE) ? "on" : "off",
> -			(rc6_mode & INTEL_RC6p_ENABLE) ? "on" : "off",
> -			(rc6_mode & INTEL_RC6pp_ENABLE) ? "on" : "off");
> +			(rc6_mask & GEN6_RC_CTL_RC6_ENABLE) ? "on" : "off",
> +			(rc6_mask & GEN6_RC_CTL_RC6p_ENABLE) ? "on" : "off",
> +			(rc6_mask & GEN6_RC_CTL_RC6pp_ENABLE) ? "on" : "off");
>  
>  	I915_WRITE(GEN6_RC_CONTROL,
>  		   rc6_mask |
> @@ -2433,12 +2442,32 @@ static void gen6_enable_rps(struct drm_device *dev)
>  	I915_WRITE(GEN6_RP_INTERRUPT_LIMITS,
>  		   dev_priv->max_delay << 24 |
>  		   dev_priv->min_delay << 16);
> -	I915_WRITE(GEN6_RP_UP_THRESHOLD, 10000);
> -	I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 1000000);
> -	I915_WRITE(GEN6_RP_UP_EI, 100000);
> -	I915_WRITE(GEN6_RP_DOWN_EI, 5000000);
> +
> +	if (IS_HASWELL(dev)) {
> +		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);
> +	} else {
> +		I915_WRITE(GEN6_RP_UP_THRESHOLD, 10000);
> +		I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 1000000);
> +		I915_WRITE(GEN6_RP_UP_EI, 100000);
> +		I915_WRITE(GEN6_RP_DOWN_EI, 5000000);
> +	}
> +
>  	I915_WRITE(GEN6_RP_IDLE_HYSTERSIS, 10);
> -	I915_WRITE(GEN6_RP_CONTROL,
> +
> +	/* We have a bit different default configuration for HSW */
> +	if (IS_HASWELL(dev))
> +		I915_WRITE(GEN6_RP_CONTROL,
> +		   GEN6_RP_MEDIA_TURBO |
> +		   GEN6_RP_MEDIA_HW_MODE |
> +		   GEN6_RP_MEDIA_IS_GFX |
> +		   GEN6_RP_ENABLE |
> +		   GEN6_RP_UP_BUSY_AVG |
> +		   GEN7_RP_DOWN_IDLE_AVG);
> +	else
> +		I915_WRITE(GEN6_RP_CONTROL,
>  		   GEN6_RP_MEDIA_TURBO |
>  		   GEN6_RP_MEDIA_HW_NORMAL_MODE |
>  		   GEN6_RP_MEDIA_IS_GFX |

The "else" here scares me a little since we may miss it with a new hw
generation.  Maybe a new 'gt' function pointer or two should be used to
write these?  That'll at least force us to implement it for new hw...

-- 
Jesse Barnes, Intel Open Source Technology Center