Skip to contentSkip to conversation

drm/i915: Rebalance runtime pm vs forcewake

PATCH In review Mika Kuoppala 9 participants 58 messages Dec 8, 2014 – Feb 23, 2015

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

PATCH 1/8 drm/i915: Rebalance runtime pm vs forcewake

From: Chris Wilson <[email protected]>

Calling intel_runtime_pm_put() is illegal from a soft-irq context, so revert the crude hack

commit aa0b3b5bb8768c1a6a6788869d9c7015eae7e80c

Author: Paulo Zanoni <[email protected]>
Date:   Tue Apr 1 14:55:07 2014 -0300

drm/i915: don't schedule force_wake_timer at gen6_read

and apply the single line corrective instead.

References: https://bugs.freedesktop.org/show_bug.cgi?id=80913

Cc Paulo Zanoni <[email protected]>
Cc Daniel Vetter <[email protected]>
---
 drivers/gpu/drm/i915/i915_drv.c     |  1 +
 drivers/gpu/drm/i915/intel_uncore.c | 18 ++++++------------
 2 files changed, 7 insertions(+), 12 deletions(-)
drivers/gpu/drm/i915/i915_drv.c +1 -0
index 71be3c9..706b122 100644
@@ -1402,6 +1402,7 @@ static int intel_runtime_suspend(struct device *device)
 	}
 
 	del_timer_sync(&dev_priv->gpu_error.hangcheck_timer);
+	intel_uncore_forcewake_reset(dev, false);
 	dev_priv->pm.suspended = true;
 
 	/*
drivers/gpu/drm/i915/intel_uncore.c +6 -12
index 46de8d7..38ac389 100644
@@ -449,8 +449,6 @@ static void gen6_force_wake_timer(unsigned long arg)
 	if (--dev_priv->uncore.forcewake_count == 0)
 		dev_priv->uncore.funcs.force_wake_put(dev_priv, FORCEWAKE_ALL);
 	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
-
-	intel_runtime_pm_put(dev_priv);
 }
 
 void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore)
@@ -586,7 +584,6 @@ void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
 void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
 {
 	unsigned long irqflags;
-	bool delayed = false;
 
 	if (!dev_priv->uncore.funcs.force_wake_put)
 		return;
@@ -603,21 +600,19 @@ void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
 		goto out;
 	}
 
-
 	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
 	WARN_ON(!dev_priv->uncore.forcewake_count);
 
 	if (--dev_priv->uncore.forcewake_count == 0) {
 		dev_priv->uncore.forcewake_count++;
-		delayed = true;
 		mod_timer_pinned(&dev_priv->uncore.force_wake_timer,
 				 jiffies + 1);
 	}
+
 	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
 
 out:
-	if (!delayed)
-		intel_runtime_pm_put(dev_priv);
+	intel_runtime_pm_put(dev_priv);
 }
 
 void assert_force_wake_inactive(struct drm_i915_private *dev_priv)
@@ -777,12 +772,11 @@ gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
 	    NEEDS_FORCE_WAKE((dev_priv), (reg))) { \
 		dev_priv->uncore.funcs.force_wake_get(dev_priv, \
 						      FORCEWAKE_ALL); \
-		val = __raw_i915_read##x(dev_priv, reg); \
-		dev_priv->uncore.funcs.force_wake_put(dev_priv, \
-						      FORCEWAKE_ALL); \
-	} else { \
-		val = __raw_i915_read##x(dev_priv, reg); \
+		dev_priv->uncore.forcewake_count++; \
+		mod_timer_pinned(&dev_priv->uncore.force_wake_timer, \
+				 jiffies + 1); \
 	} \
+	val = __raw_i915_read##x(dev_priv, reg); \
 	hsw_unclaimed_reg_debug(dev_priv, reg, true, false); \
 	REG_READ_FOOTER; \
 }
Signature · 6 lines
1.9.1

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

PATCH 2/8 drm/i915: Assert that runtime pm is active on user fw access

From: Chris Wilson <[email protected]>

On user forcewake access, assert that runtime pm reference is held. Fix and cleanup the callsites accordingly.

---
 drivers/gpu/drm/i915/i915_debugfs.c  |  2 +
 drivers/gpu/drm/i915/intel_display.c | 19 +--------
 drivers/gpu/drm/i915/intel_lrc.c     | 53 ++-----------------------
 drivers/gpu/drm/i915/intel_uncore.c  | 76 ++++++++++++------------------------
 4 files changed, 31 insertions(+), 119 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +2 -0
index d0e445e..e142629 100644
@@ -4250,6 +4250,7 @@ static int i915_forcewake_open(struct inode *inode, struct file *file)
 	if (INTEL_INFO(dev)->gen < 6)
 		return 0;
 
+	intel_runtime_pm_get(dev_priv);
 	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
 
 	return 0;
@@ -4264,6 +4265,7 @@ static int i915_forcewake_release(struct inode *inode, struct file *file)
 		return 0;
 
 	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_runtime_pm_put(dev_priv);
 
 	return 0;
 }
drivers/gpu/drm/i915/intel_display.c +2 -17
index d5153a4..86c2885 100644
@@ -7862,19 +7862,8 @@ static void hsw_restore_lcpll(struct drm_i915_private *dev_priv)
 	/*
 	 * Make sure we're not on PC8 state before disabling PC8, otherwise
 	 * we'll hang the machine. To prevent PC8 state, just enable force_wake.
-	 *
-	 * The other problem is that hsw_restore_lcpll() is called as part of
-	 * the runtime PM resume sequence, so we can't just call
-	 * gen6_gt_force_wake_get() because that function calls
-	 * intel_runtime_pm_get(), and we can't change the runtime PM refcount
-	 * while we are on the resume sequence. So to solve this problem we have
-	 * to call special forcewake code that doesn't touch runtime PM and
-	 * doesn't enable the forcewake delayed work.
 	 */
-	spin_lock_irq(&dev_priv->uncore.lock);
-	if (dev_priv->uncore.forcewake_count++ == 0)
-		dev_priv->uncore.funcs.force_wake_get(dev_priv, FORCEWAKE_ALL);
-	spin_unlock_irq(&dev_priv->uncore.lock);
+	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
 
 	if (val & LCPLL_POWER_DOWN_ALLOW) {
 		val &= ~LCPLL_POWER_DOWN_ALLOW;
@@ -7904,11 +7893,7 @@ static void hsw_restore_lcpll(struct drm_i915_private *dev_priv)
 			DRM_ERROR("Switching back to LCPLL failed\n");
 	}
 
-	/* See the big comment above. */
-	spin_lock_irq(&dev_priv->uncore.lock);
-	if (--dev_priv->uncore.forcewake_count == 0)
-		dev_priv->uncore.funcs.force_wake_put(dev_priv, FORCEWAKE_ALL);
-	spin_unlock_irq(&dev_priv->uncore.lock);
+	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
 }
 
 /*
drivers/gpu/drm/i915/intel_lrc.c +3 -50

73 lines not on this page · raw message

drivers/gpu/drm/i915/intel_uncore.c +24 -52

147 lines not on this page · raw message

Signature · 6 lines
1.9.1

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

Re: [PATCH 2/8] drm/i915: Assert that runtime pm is active on user fw access

Quoted 240 lines · Mika Kuoppala

From: Chris Wilson <[email protected]>

On user forcewake access, assert that runtime pm reference is held. Fix and cleanup the callsites accordingly.

---
  drivers/gpu/drm/i915/i915_debugfs.c  |  2 +
  drivers/gpu/drm/i915/intel_display.c | 19 +--------
  drivers/gpu/drm/i915/intel_lrc.c     | 53 ++-----------------------
  drivers/gpu/drm/i915/intel_uncore.c  | 76 ++++++++++++------------------------
  4 files changed, 31 insertions(+), 119 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +2 -0
index d0e445e..e142629 100644
@@ -4250,6 +4250,7 @@ static int i915_forcewake_open(struct inode *inode, struct file *file)
  	if (INTEL_INFO(dev)->gen < 6)
  		return 0;
 
+	intel_runtime_pm_get(dev_priv);
  	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
 
  	return 0;
@@ -4264,6 +4265,7 @@ static int i915_forcewake_release(struct inode *inode, struct file *file)
  		return 0;
 
  	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_runtime_pm_put(dev_priv);
 
  	return 0;
  }
drivers/gpu/drm/i915/intel_display.c +2 -17
index d5153a4..86c2885 100644
@@ -7862,19 +7862,8 @@ static void hsw_restore_lcpll(struct drm_i915_private *dev_priv)
  	/*
  	 * Make sure we're not on PC8 state before disabling PC8, otherwise
  	 * we'll hang the machine. To prevent PC8 state, just enable force_wake.
-	 *
-	 * The other problem is that hsw_restore_lcpll() is called as part of
-	 * the runtime PM resume sequence, so we can't just call
-	 * gen6_gt_force_wake_get() because that function calls
-	 * intel_runtime_pm_get(), and we can't change the runtime PM refcount
-	 * while we are on the resume sequence. So to solve this problem we have
-	 * to call special forcewake code that doesn't touch runtime PM and
-	 * doesn't enable the forcewake delayed work.
  	 */
-	spin_lock_irq(&dev_priv->uncore.lock);
-	if (dev_priv->uncore.forcewake_count++ == 0)
-		dev_priv->uncore.funcs.force_wake_get(dev_priv, FORCEWAKE_ALL);
-	spin_unlock_irq(&dev_priv->uncore.lock);
+	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
 
  	if (val & LCPLL_POWER_DOWN_ALLOW) {
  		val &= ~LCPLL_POWER_DOWN_ALLOW;
@@ -7904,11 +7893,7 @@ static void hsw_restore_lcpll(struct drm_i915_private *dev_priv)
  			DRM_ERROR("Switching back to LCPLL failed\n");
  	}
 
-	/* See the big comment above. */
-	spin_lock_irq(&dev_priv->uncore.lock);
-	if (--dev_priv->uncore.forcewake_count == 0)
-		dev_priv->uncore.funcs.force_wake_put(dev_priv, FORCEWAKE_ALL);
-	spin_unlock_irq(&dev_priv->uncore.lock);
+	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
  }
 
  /*
drivers/gpu/drm/i915/intel_lrc.c +3 -50
index a82020e..fcb5140 100644
@@ -284,7 +284,6 @@ static void execlists_elsp_write(struct intel_engine_cs *ring,
  	struct drm_i915_private *dev_priv = dev->dev_private;
  	uint64_t temp = 0;
  	uint32_t desc[4];
-	unsigned long flags;
 
  	/* XXX: You must always write both descriptors in the order below. */
  	if (ctx_obj1)
@@ -298,63 +297,17 @@ static void execlists_elsp_write(struct intel_engine_cs *ring,
  	desc[3] = (u32)(temp >> 32);
  	desc[2] = (u32)temp;
 
-	/* Set Force Wakeup bit to prevent GT from entering C6 while ELSP writes
-	 * are in progress.
-	 *
-	 * The other problem is that we can't just call gen6_gt_force_wake_get()
-	 * because that function calls intel_runtime_pm_get(), which might sleep.
-	 * Instead, we do the runtime_pm_get/put when creating/destroying requests.
-	 */
-	spin_lock_irqsave(&dev_priv->uncore.lock, flags);
-	if (IS_CHERRYVIEW(dev) || INTEL_INFO(dev)->gen >= 9) {
-		if (dev_priv->uncore.fw_rendercount++ == 0)
-			dev_priv->uncore.funcs.force_wake_get(dev_priv,
-							      FORCEWAKE_RENDER);
-		if (dev_priv->uncore.fw_mediacount++ == 0)
-			dev_priv->uncore.funcs.force_wake_get(dev_priv,
-							      FORCEWAKE_MEDIA);
-		if (INTEL_INFO(dev)->gen >= 9) {
-			if (dev_priv->uncore.fw_blittercount++ == 0)
-				dev_priv->uncore.funcs.force_wake_get(dev_priv,
-							FORCEWAKE_BLITTER);
-		}
-	} else {
-		if (dev_priv->uncore.forcewake_count++ == 0)
-			dev_priv->uncore.funcs.force_wake_get(dev_priv,
-							      FORCEWAKE_ALL);
-	}
-	spin_unlock_irqrestore(&dev_priv->uncore.lock, flags);
-
+	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
  	I915_WRITE(RING_ELSP(ring), desc[1]);
  	I915_WRITE(RING_ELSP(ring), desc[0]);
  	I915_WRITE(RING_ELSP(ring), desc[3]);
+
  	/* The context is automatically loaded after the following */
  	I915_WRITE(RING_ELSP(ring), desc[2]);
 
  	/* ELSP is a wo register, so use another nearby reg for posting instead */
  	POSTING_READ(RING_EXECLIST_STATUS(ring));
-
-	/* Release Force Wakeup (see the big comment above). */
-	spin_lock_irqsave(&dev_priv->uncore.lock, flags);
-	if (IS_CHERRYVIEW(dev) || INTEL_INFO(dev)->gen >= 9) {
-		if (--dev_priv->uncore.fw_rendercount == 0)
-			dev_priv->uncore.funcs.force_wake_put(dev_priv,
-							      FORCEWAKE_RENDER);
-		if (--dev_priv->uncore.fw_mediacount == 0)
-			dev_priv->uncore.funcs.force_wake_put(dev_priv,
-							      FORCEWAKE_MEDIA);
-		if (INTEL_INFO(dev)->gen >= 9) {
-			if (--dev_priv->uncore.fw_blittercount == 0)
-				dev_priv->uncore.funcs.force_wake_put(dev_priv,
-							FORCEWAKE_BLITTER);
-		}
-	} else {
-		if (--dev_priv->uncore.forcewake_count == 0)
-			dev_priv->uncore.funcs.force_wake_put(dev_priv,
-							      FORCEWAKE_ALL);
-	}
-
-	spin_unlock_irqrestore(&dev_priv->uncore.lock, flags);
+	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
  }
 
  static int execlists_update_context(struct drm_i915_gem_object *ctx_obj,
drivers/gpu/drm/i915/intel_uncore.c +2 -24
index 38ac389..a1ceb92 100644
@@ -24,6 +24,8 @@
  #include "i915_drv.h"
  #include "intel_drv.h"
 
+#include <linux/pm_runtime.h>
+
  #define FORCEWAKE_ACK_TIMEOUT_MS 2
 
  #define __raw_i915_read8(dev_priv__, reg__) readb((dev_priv__)->regs + (reg__))
@@ -247,10 +249,6 @@ static void __vlv_force_wake_put(struct drm_i915_private *dev_priv,
 
  static void vlv_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
  {
-	unsigned long irqflags;
-
-	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
-
  	if (fw_engine & FORCEWAKE_RENDER &&
  	    dev_priv->uncore.fw_rendercount++ != 0)
  		fw_engine &= ~FORCEWAKE_RENDER;
@@ -260,16 +258,10 @@ static void vlv_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
 
  	if (fw_engine)
  		dev_priv->uncore.funcs.force_wake_get(dev_priv, fw_engine);
-
-	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
  }
 
  static void vlv_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
  {
-	unsigned long irqflags;
-
-	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
-
  	if (fw_engine & FORCEWAKE_RENDER) {
  		WARN_ON(!dev_priv->uncore.fw_rendercount);
  		if (--dev_priv->uncore.fw_rendercount != 0)
@@ -284,8 +276,6 @@ static void vlv_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
 
  	if (fw_engine)
  		dev_priv->uncore.funcs.force_wake_put(dev_priv, fw_engine);
-
-	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
  }
 
  static void __gen9_gt_force_wake_mt_reset(struct drm_i915_private *dev_priv)
@@ -380,10 +370,6 @@ __gen9_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
  static void
  gen9_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
  {
-	unsigned long irqflags;
-
-	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
-
  	if (FORCEWAKE_RENDER & fw_engine) {
  		if (dev_priv->uncore.fw_rendercount++ == 0)
  			dev_priv->uncore.funcs.force_wake_get(dev_priv,
@@ -401,17 +387,11 @@ gen9_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
  			dev_priv->uncore.funcs.force_wake_get(dev_priv,
  							FORCEWAKE_BLITTER);
  	}
-
-	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
  }
 
  static void
  gen9_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
  {
-	unsigned long irqflags;
-
-	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
-
  	if (FORCEWAKE_RENDER & fw_engine) {
  		WARN_ON(dev_priv->uncore.fw_rendercount == 0);
  		if (--dev_priv->uncore.fw_rendercount == 0)
@@ -432,8 +412,6 @@ gen9_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
  			dev_priv->uncore.funcs.force_wake_put(dev_priv,
  							FORCEWAKE_BLITTER);
  	}
-
-	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
  }
 
  static void gen6_force_wake_timer(unsigned long arg)

drivers/gpu/drm/i915/intel_uncore.c void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine) · line 544

Quoted 3 lines
drivers/gpu/drm/i915/intel_uncore.c +0 -0
@@ -564,17 +542,20 @@ void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
 
  	intel_runtime_pm_get(dev_priv);
 

I think we need to remove "intel_runtime_pm_get" here? I see runtime_put is removed from "gen6_gt_force_wake_put"

Quoted 58 lines
drivers/gpu/drm/i915/intel_uncore.c +21 -25
-	/* Redirect to Gen9 specific routine */
-	if (IS_GEN9(dev_priv->dev))
-		return gen9_force_wake_get(dev_priv, fw_engine);
-
-	/* Redirect to VLV specific routine */
-	if (IS_VALLEYVIEW(dev_priv->dev))
-		return vlv_force_wake_get(dev_priv, fw_engine);
+	WARN_ON(!pm_runtime_active(&dev_priv->dev->pdev->dev));
 
  	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
-	if (dev_priv->uncore.forcewake_count++ == 0)
-		dev_priv->uncore.funcs.force_wake_get(dev_priv, FORCEWAKE_ALL);
+
+	if (IS_GEN9(dev_priv->dev)) {
+		gen9_force_wake_get(dev_priv, fw_engine);
+	} else if (IS_VALLEYVIEW(dev_priv->dev)) {
+		vlv_force_wake_get(dev_priv, fw_engine);
+	} else {
+		if (dev_priv->uncore.forcewake_count++ == 0)
+			dev_priv->uncore.funcs.force_wake_get(dev_priv,
+							      FORCEWAKE_ALL);
+	}
+
  	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
  }
 
@@ -588,31 +569,22 @@ void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
  	if (!dev_priv->uncore.funcs.force_wake_put)
  		return;
 
-	/* Redirect to Gen9 specific routine */
+	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
+
  	if (IS_GEN9(dev_priv->dev)) {
  		gen9_force_wake_put(dev_priv, fw_engine);
-		goto out;
-	}
-
-	/* Redirect to VLV specific routine */
-	if (IS_VALLEYVIEW(dev_priv->dev)) {
+	} else if (IS_VALLEYVIEW(dev_priv->dev)) {
  		vlv_force_wake_put(dev_priv, fw_engine);
-		goto out;
-	}
-
-	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
-	WARN_ON(!dev_priv->uncore.forcewake_count);
-
-	if (--dev_priv->uncore.forcewake_count == 0) {
-		dev_priv->uncore.forcewake_count++;
-		mod_timer_pinned(&dev_priv->uncore.force_wake_timer,
-				 jiffies + 1);
+	} else {
+		WARN_ON(!dev_priv->uncore.forcewake_count);
+		if (--dev_priv->uncore.forcewake_count == 0) {
+			dev_priv->uncore.forcewake_count++;
+			mod_timer_pinned(&dev_priv->uncore.force_wake_timer,
+					 jiffies + 1);

drivers/gpu/drm/i915/intel_uncore.c void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine) · line 590

Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +1 -3
+		}
  	}
 
  	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
-
-out:
-	intel_runtime_pm_put(dev_priv);
  }
 
  void assert_force_wake_inactive(struct drm_i915_private *dev_priv)

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

#3 lore raw

Re: [PATCH 2/8] drm/i915: Assert that runtime pm is active on user fw access

Quoted 6 lines · Deepak S

drivers/gpu/drm/i915/intel_uncore.c void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine) · line 542

Quoted 2 lines
drivers/gpu/drm/i915/intel_uncore.c +0 -0
@@ -564,17 +542,20 @@ void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
 	intel_runtime_pm_get(dev_priv);

I think we need to remove "intel_runtime_pm_get" here? I see runtime_put is removed from "gen6_gt_force_wake_put"

Quoted 8 lines

drivers/gpu/drm/i915/intel_uncore.c void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine) · line 545

Quoted 8 lines
drivers/gpu/drm/i915/intel_uncore.c +1 -7
-	/* Redirect to Gen9 specific routine */
-	if (IS_GEN9(dev_priv->dev))
-		return gen9_force_wake_get(dev_priv, fw_engine);
-
-	/* Redirect to VLV specific routine */
-	if (IS_VALLEYVIEW(dev_priv->dev))
-		return vlv_force_wake_get(dev_priv, fw_engine);
+	WARN_ON(!pm_runtime_active(&dev_priv->dev->pdev->dev));

Indeed, the purpose of this patch was to remove the rpm_get() and replace it with this WARN_ON (and so make sure that all call paths here already held the rpm reference). Something went wrong in the rebase I hope.
-Chris

Signature · 5 lines
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#4 lore raw

Re: [PATCH 2/8] drm/i915: Assert that runtime pm is active on user fw access

Quoted 6 lines · Chris Wilson
Quoted 5 lines · Deepak S

drivers/gpu/drm/i915/intel_uncore.c void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine) · line 542

Quoted 2 lines
drivers/gpu/drm/i915/intel_uncore.c +0 -0
@@ -564,17 +542,20 @@ void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
  	intel_runtime_pm_get(dev_priv);

I think we need to remove "intel_runtime_pm_get" here? I see runtime_put is removed from "gen6_gt_force_wake_put"

Quoted 13 lines
Quoted 8 lines

drivers/gpu/drm/i915/intel_uncore.c void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine) · line 545

Quoted 8 lines
drivers/gpu/drm/i915/intel_uncore.c +1 -7
-	/* Redirect to Gen9 specific routine */
-	if (IS_GEN9(dev_priv->dev))
-		return gen9_force_wake_get(dev_priv, fw_engine);
-
-	/* Redirect to VLV specific routine */
-	if (IS_VALLEYVIEW(dev_priv->dev))
-		return vlv_force_wake_get(dev_priv, fw_engine);
+	WARN_ON(!pm_runtime_active(&dev_priv->dev->pdev->dev));

Indeed, the purpose of this patch was to remove the rpm_get() and replace it with this WARN_ON (and so make sure that all call paths here already held the rpm reference). Something went wrong in the rebase I hope.
-Chris

Yeah, Looks like it got missed.

Hi Mika,
once we fix the issue, you can add my r-b, Other changes looks fine

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

#5 lore raw

PATCH 3/8 drm/i915: Skip uncore lock on earlier gens

In reply to Mika Kuoppala

From: Chris Wilson <[email protected]>

With gen < 6 we don't need to take uncore lock as we don't have anything to protect from concurrent access.

v2: rebase and account for gen9 changes

---
 drivers/gpu/drm/i915/intel_uncore.c | 158 +++++++++++++++++++++---------------
 1 file changed, 91 insertions(+), 67 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +91 -67
index a1ceb92..069fe7a 100644
@@ -707,38 +707,61 @@ hsw_unclaimed_reg_detect(struct drm_i915_private *dev_priv)
 	}
 }
 
-#define REG_READ_HEADER(x) \
-	unsigned long irqflags; \
+#define GEN2_READ_HEADER(x) \
 	u##x val = 0; \
-	assert_device_not_suspended(dev_priv); \
-	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags)
+	assert_device_not_suspended(dev_priv);
 
-#define REG_READ_FOOTER \
-	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); \
+#define GEN2_READ_FOOTER \
 	trace_i915_reg_rw(false, reg, val, sizeof(val), trace); \
 	return val
 
-#define __gen4_read(x) \
+#define __gen2_read(x) \
 static u##x \
-gen4_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
-	REG_READ_HEADER(x); \
+gen2_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
+	GEN2_READ_HEADER(x); \
 	val = __raw_i915_read##x(dev_priv, reg); \
-	REG_READ_FOOTER; \
+	GEN2_READ_FOOTER; \
 }
 
 #define __gen5_read(x) \
 static u##x \
 gen5_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
-	REG_READ_HEADER(x); \
+	GEN2_READ_HEADER(x); \
 	ilk_dummy_write(dev_priv); \
 	val = __raw_i915_read##x(dev_priv, reg); \
-	REG_READ_FOOTER; \
+	GEN2_READ_FOOTER; \
 }
 
+__gen5_read(8)
+__gen5_read(16)
+__gen5_read(32)
+__gen5_read(64)
+__gen2_read(8)
+__gen2_read(16)
+__gen2_read(32)
+__gen2_read(64)
+
+#undef __gen5_read
+#undef __gen2_read
+
+#undef GEN2_READ_FOOTER
+#undef GEN2_READ_HEADER
+
+#define GEN6_READ_HEADER(x) \
+	unsigned long irqflags; \
+	u##x val = 0; \
+	assert_device_not_suspended(dev_priv); \
+	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags)
+
+#define GEN6_READ_FOOTER \
+	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); \
+	trace_i915_reg_rw(false, reg, val, sizeof(val), trace); \
+	return val
+
 #define __gen6_read(x) \
 static u##x \
 gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
-	REG_READ_HEADER(x); \
+	GEN6_READ_HEADER(x); \
 	hsw_unclaimed_reg_debug(dev_priv, reg, true, true); \
 	if (dev_priv->uncore.forcewake_count == 0 && \
 	    NEEDS_FORCE_WAKE((dev_priv), (reg))) { \
@@ -750,14 +773,14 @@ gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
 	} \
 	val = __raw_i915_read##x(dev_priv, reg); \
 	hsw_unclaimed_reg_debug(dev_priv, reg, true, false); \
-	REG_READ_FOOTER; \
+	GEN6_READ_FOOTER; \
 }
 
 #define __vlv_read(x) \
 static u##x \
 vlv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
 	unsigned fwengine = 0; \
-	REG_READ_HEADER(x); \
+	GEN6_READ_HEADER(x); \
 	if (FORCEWAKE_VLV_RENDER_RANGE_OFFSET(reg)) { \
 		if (dev_priv->uncore.fw_rendercount == 0) \
 			fwengine = FORCEWAKE_RENDER; \
@@ -770,14 +793,14 @@ vlv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
 	val = __raw_i915_read##x(dev_priv, reg); \
 	if (fwengine) \
 		dev_priv->uncore.funcs.force_wake_put(dev_priv, fwengine); \
-	REG_READ_FOOTER; \
+	GEN6_READ_FOOTER; \
 }
 
 #define __chv_read(x) \
 static u##x \
 chv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
 	unsigned fwengine = 0; \
-	REG_READ_HEADER(x); \
+	GEN6_READ_HEADER(x); \
 	if (FORCEWAKE_CHV_RENDER_RANGE_OFFSET(reg)) { \
 		if (dev_priv->uncore.fw_rendercount == 0) \
 			fwengine = FORCEWAKE_RENDER; \
@@ -795,7 +818,7 @@ chv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
 	val = __raw_i915_read##x(dev_priv, reg); \
 	if (fwengine) \
 		dev_priv->uncore.funcs.force_wake_put(dev_priv, fwengine); \
-	REG_READ_FOOTER; \
+	GEN6_READ_FOOTER; \
 }
 
 #define SKL_NEEDS_FORCE_WAKE(dev_priv, reg)	\
@@ -804,7 +827,7 @@ chv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
 #define __gen9_read(x) \
 static u##x \
 gen9_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
-	REG_READ_HEADER(x); \
+	GEN6_READ_HEADER(x); \
 	if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg))) { \
 		val = __raw_i915_read##x(dev_priv, reg); \
 	} else { \
@@ -830,7 +853,7 @@ gen9_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
 		if (fwengine) \
 			dev_priv->uncore.funcs.force_wake_put(dev_priv, fwengine); \
 	} \
-	REG_READ_FOOTER; \
+	GEN6_READ_FOOTER; \
 }
 
 __gen9_read(8)
@@ -849,55 +872,66 @@ __gen6_read(8)
 __gen6_read(16)
 __gen6_read(32)
 __gen6_read(64)
-__gen5_read(8)
-__gen5_read(16)
-__gen5_read(32)
-__gen5_read(64)
-__gen4_read(8)
-__gen4_read(16)
-__gen4_read(32)
-__gen4_read(64)
 
 #undef __gen9_read
 #undef __chv_read
 #undef __vlv_read
 #undef __gen6_read
-#undef __gen5_read
-#undef __gen4_read
-#undef REG_READ_FOOTER
-#undef REG_READ_HEADER
+#undef GEN6_READ_FOOTER
+#undef GEN6_READ_HEADER
 
-#define REG_WRITE_HEADER \
-	unsigned long irqflags; \
+#define GEN2_WRITE_HEADER \
 	trace_i915_reg_rw(true, reg, val, sizeof(val), trace); \
 	assert_device_not_suspended(dev_priv); \
-	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags)
 
-#define REG_WRITE_FOOTER \
-	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags)
+#define GEN2_WRITE_FOOTER
 
-#define __gen4_write(x) \
+#define __gen2_write(x) \
 static void \
-gen4_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) { \
-	REG_WRITE_HEADER; \
+gen2_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) { \
+	GEN2_WRITE_HEADER; \
 	__raw_i915_write##x(dev_priv, reg, val); \
-	REG_WRITE_FOOTER; \
+	GEN2_WRITE_FOOTER; \
 }
 
 #define __gen5_write(x) \
 static void \
 gen5_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) { \
-	REG_WRITE_HEADER; \
+	GEN2_WRITE_HEADER; \
 	ilk_dummy_write(dev_priv); \
 	__raw_i915_write##x(dev_priv, reg, val); \
-	REG_WRITE_FOOTER; \
+	GEN2_WRITE_FOOTER; \
 }
 
+__gen5_write(8)
+__gen5_write(16)
+__gen5_write(32)
+__gen5_write(64)
+__gen2_write(8)
+__gen2_write(16)
+__gen2_write(32)
+__gen2_write(64)
+
+#undef __gen5_write
+#undef __gen2_write
+
+#undef GEN2_WRITE_FOOTER
+#undef GEN2_WRITE_HEADER
+
+#define GEN6_WRITE_HEADER \
+	unsigned long irqflags; \
+	trace_i915_reg_rw(true, reg, val, sizeof(val), trace); \
+	assert_device_not_suspended(dev_priv); \
+	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags)
+
+#define GEN6_WRITE_FOOTER \
+	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags)
+
 #define __gen6_write(x) \
 static void \
 gen6_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) { \
 	u32 __fifo_ret = 0; \
-	REG_WRITE_HEADER; \
+	GEN6_WRITE_HEADER; \
 	if (NEEDS_FORCE_WAKE((dev_priv), (reg))) { \
 		__fifo_ret = __gen6_gt_wait_for_fifo(dev_priv); \
 	} \
@@ -905,14 +939,14 @@ gen6_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace
 	if (unlikely(__fifo_ret)) { \
 		gen6_gt_check_fifodbg(dev_priv); \
 	} \
-	REG_WRITE_FOOTER; \
+	GEN6_WRITE_FOOTER; \
 }
 
 #define __hsw_write(x) \
 static void \
 hsw_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) { \
 	u32 __fifo_ret = 0; \
-	REG_WRITE_HEADER; \
+	GEN6_WRITE_HEADER; \
 	if (NEEDS_FORCE_WAKE((dev_priv), (reg))) { \
 		__fifo_ret = __gen6_gt_wait_for_fifo(dev_priv); \
 	} \
@@ -923,7 +957,7 @@ hsw_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace)
 	} \
 	hsw_unclaimed_reg_debug(dev_priv, reg, false, false); \
 	hsw_unclaimed_reg_detect(dev_priv); \
-	REG_WRITE_FOOTER; \
+	GEN6_WRITE_FOOTER; \
 }
 
 static const u32 gen8_shadowed_regs[] = {
@@ -950,7 +984,7 @@ static bool is_gen8_shadowed(struct drm_i915_private *dev_priv, u32 reg)
 #define __gen8_write(x) \
 static void \
 gen8_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) { \
-	REG_WRITE_HEADER; \
+	GEN6_WRITE_HEADER; \
 	hsw_unclaimed_reg_debug(dev_priv, reg, false, true); \
 	if (reg < 0x40000 && !is_gen8_shadowed(dev_priv, reg)) { \
 		if (dev_priv->uncore.forcewake_count == 0) \
@@ -965,7 +999,7 @@ gen8_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace
 	} \
 	hsw_unclaimed_reg_debug(dev_priv, reg, false, false); \
 	hsw_unclaimed_reg_detect(dev_priv); \
-	REG_WRITE_FOOTER; \
+	GEN6_WRITE_FOOTER; \
 }
 
 #define __chv_write(x) \
@@ -973,7 +1007,7 @@ static void \
 chv_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) { \
 	unsigned fwengine = 0; \
 	bool shadowed = is_gen8_shadowed(dev_priv, reg); \
-	REG_WRITE_HEADER; \
+	GEN6_WRITE_HEADER; \
 	if (!shadowed) { \
 		if (FORCEWAKE_CHV_RENDER_RANGE_OFFSET(reg)) { \
 			if (dev_priv->uncore.fw_rendercount == 0) \
@@ -993,7 +1027,7 @@ chv_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace)
 	__raw_i915_write##x(dev_priv, reg, val); \
 	if (fwengine) \
 		dev_priv->uncore.funcs.force_wake_put(dev_priv, fwengine); \
-	REG_WRITE_FOOTER; \
+	GEN6_WRITE_FOOTER; \
 }
 
 static const u32 gen9_shadowed_regs[] = {
@@ -1023,7 +1057,7 @@ static bool is_gen9_shadowed(struct drm_i915_private *dev_priv, u32 reg)
 static void \
 gen9_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, \
 		bool trace) { \
-	REG_WRITE_HEADER; \
+	GEN6_WRITE_HEADER; \
 	if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg)) || \
 			is_gen9_shadowed(dev_priv, reg)) { \
 		__raw_i915_write##x(dev_priv, reg, val); \
@@ -1052,7 +1086,7 @@ gen9_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, \
 			dev_priv->uncore.funcs.force_wake_put(dev_priv, \
 					fwengine); \
 	} \
-	REG_WRITE_FOOTER; \
+	GEN6_WRITE_FOOTER; \
 }
 
 __gen9_write(8)
@@ -1075,24 +1109,14 @@ __gen6_write(8)
 __gen6_write(16)
 __gen6_write(32)
 __gen6_write(64)
-__gen5_write(8)
-__gen5_write(16)
-__gen5_write(32)
-__gen5_write(64)
-__gen4_write(8)
-__gen4_write(16)
-__gen4_write(32)
-__gen4_write(64)
 
 #undef __gen9_write
 #undef __chv_write
 #undef __gen8_write
 #undef __hsw_write
 #undef __gen6_write
-#undef __gen5_write
-#undef __gen4_write
-#undef REG_WRITE_FOOTER
-#undef REG_WRITE_HEADER
+#undef GEN6_WRITE_FOOTER
+#undef GEN6_WRITE_HEADER
 
 #define ASSIGN_WRITE_MMIO_VFUNCS(x) \
 do { \
@@ -1205,8 +1229,8 @@ void intel_uncore_init(struct drm_device *dev)
 	case 4:
 	case 3:
 	case 2:
-		ASSIGN_WRITE_MMIO_VFUNCS(gen4);
-		ASSIGN_READ_MMIO_VFUNCS(gen4);
+		ASSIGN_WRITE_MMIO_VFUNCS(gen2);
+		ASSIGN_READ_MMIO_VFUNCS(gen2);
 		break;
 	}
 
Signature · 6 lines
1.9.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Reviewed-by on v1 3/8
#6 lore raw

Re: [PATCH 3/8] drm/i915: Skip uncore lock on earlier gens

Quoted 352 lines · Mika Kuoppala

From: Chris Wilson <[email protected]>

With gen < 6 we don't need to take uncore lock as we don't have anything to protect from concurrent access.

v2: rebase and account for gen9 changes

---
  drivers/gpu/drm/i915/intel_uncore.c | 158 +++++++++++++++++++++---------------
  1 file changed, 91 insertions(+), 67 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +89 -65
index a1ceb92..069fe7a 100644
@@ -707,38 +707,61 @@ hsw_unclaimed_reg_detect(struct drm_i915_private *dev_priv)
  	}
  }
 
-#define REG_READ_HEADER(x) \
-	unsigned long irqflags; \
+#define GEN2_READ_HEADER(x) \
  	u##x val = 0; \
-	assert_device_not_suspended(dev_priv); \
-	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags)
+	assert_device_not_suspended(dev_priv);
 
-#define REG_READ_FOOTER \
-	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); \
+#define GEN2_READ_FOOTER \
  	trace_i915_reg_rw(false, reg, val, sizeof(val), trace); \
  	return val
 
-#define __gen4_read(x) \
+#define __gen2_read(x) \
  static u##x \
-gen4_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
-	REG_READ_HEADER(x); \
+gen2_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
+	GEN2_READ_HEADER(x); \
  	val = __raw_i915_read##x(dev_priv, reg); \
-	REG_READ_FOOTER; \
+	GEN2_READ_FOOTER; \
  }
 
  #define __gen5_read(x) \
  static u##x \
  gen5_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
-	REG_READ_HEADER(x); \
+	GEN2_READ_HEADER(x); \
  	ilk_dummy_write(dev_priv); \
  	val = __raw_i915_read##x(dev_priv, reg); \
-	REG_READ_FOOTER; \
+	GEN2_READ_FOOTER; \
  }
 
+__gen5_read(8)
+__gen5_read(16)
+__gen5_read(32)
+__gen5_read(64)
+__gen2_read(8)
+__gen2_read(16)
+__gen2_read(32)
+__gen2_read(64)
+
+#undef __gen5_read
+#undef __gen2_read
+
+#undef GEN2_READ_FOOTER
+#undef GEN2_READ_HEADER
+
+#define GEN6_READ_HEADER(x) \
+	unsigned long irqflags; \
+	u##x val = 0; \
+	assert_device_not_suspended(dev_priv); \
+	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags)
+
+#define GEN6_READ_FOOTER \
+	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags); \
+	trace_i915_reg_rw(false, reg, val, sizeof(val), trace); \
+	return val
+
  #define __gen6_read(x) \
  static u##x \
  gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
-	REG_READ_HEADER(x); \
+	GEN6_READ_HEADER(x); \
  	hsw_unclaimed_reg_debug(dev_priv, reg, true, true); \
  	if (dev_priv->uncore.forcewake_count == 0 && \
  	    NEEDS_FORCE_WAKE((dev_priv), (reg))) { \
@@ -750,14 +773,14 @@ gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
  	} \
  	val = __raw_i915_read##x(dev_priv, reg); \
  	hsw_unclaimed_reg_debug(dev_priv, reg, true, false); \
-	REG_READ_FOOTER; \
+	GEN6_READ_FOOTER; \
  }
 
  #define __vlv_read(x) \
  static u##x \
  vlv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
  	unsigned fwengine = 0; \
-	REG_READ_HEADER(x); \
+	GEN6_READ_HEADER(x); \
  	if (FORCEWAKE_VLV_RENDER_RANGE_OFFSET(reg)) { \
  		if (dev_priv->uncore.fw_rendercount == 0) \
  			fwengine = FORCEWAKE_RENDER; \
@@ -770,14 +793,14 @@ vlv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
  	val = __raw_i915_read##x(dev_priv, reg); \
  	if (fwengine) \
  		dev_priv->uncore.funcs.force_wake_put(dev_priv, fwengine); \
-	REG_READ_FOOTER; \
+	GEN6_READ_FOOTER; \
  }
 
  #define __chv_read(x) \
  static u##x \
  chv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
  	unsigned fwengine = 0; \
-	REG_READ_HEADER(x); \
+	GEN6_READ_HEADER(x); \
  	if (FORCEWAKE_CHV_RENDER_RANGE_OFFSET(reg)) { \
  		if (dev_priv->uncore.fw_rendercount == 0) \
  			fwengine = FORCEWAKE_RENDER; \
@@ -795,7 +818,7 @@ chv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
  	val = __raw_i915_read##x(dev_priv, reg); \
  	if (fwengine) \
  		dev_priv->uncore.funcs.force_wake_put(dev_priv, fwengine); \
-	REG_READ_FOOTER; \
+	GEN6_READ_FOOTER; \
  }
 
  #define SKL_NEEDS_FORCE_WAKE(dev_priv, reg)	\
@@ -804,7 +827,7 @@ chv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
  #define __gen9_read(x) \
  static u##x \
  gen9_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
-	REG_READ_HEADER(x); \
+	GEN6_READ_HEADER(x); \
  	if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg))) { \
  		val = __raw_i915_read##x(dev_priv, reg); \
  	} else { \
@@ -830,7 +853,7 @@ gen9_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
  		if (fwengine) \
  			dev_priv->uncore.funcs.force_wake_put(dev_priv, fwengine); \
  	} \
-	REG_READ_FOOTER; \
+	GEN6_READ_FOOTER; \
  }
 
  __gen9_read(8)
@@ -849,55 +872,66 @@ __gen6_read(8)
  __gen6_read(16)
  __gen6_read(32)
  __gen6_read(64)
-__gen5_read(8)
-__gen5_read(16)
-__gen5_read(32)
-__gen5_read(64)
-__gen4_read(8)
-__gen4_read(16)
-__gen4_read(32)
-__gen4_read(64)
 
  #undef __gen9_read
  #undef __chv_read
  #undef __vlv_read
  #undef __gen6_read
-#undef __gen5_read
-#undef __gen4_read
-#undef REG_READ_FOOTER
-#undef REG_READ_HEADER
+#undef GEN6_READ_FOOTER
+#undef GEN6_READ_HEADER
 
-#define REG_WRITE_HEADER \
-	unsigned long irqflags; \
+#define GEN2_WRITE_HEADER \
  	trace_i915_reg_rw(true, reg, val, sizeof(val), trace); \
  	assert_device_not_suspended(dev_priv); \
-	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags)
 
-#define REG_WRITE_FOOTER \
-	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags)
+#define GEN2_WRITE_FOOTER
 
-#define __gen4_write(x) \
+#define __gen2_write(x) \
  static void \
-gen4_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) { \
-	REG_WRITE_HEADER; \
+gen2_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) { \
+	GEN2_WRITE_HEADER; \
  	__raw_i915_write##x(dev_priv, reg, val); \
-	REG_WRITE_FOOTER; \
+	GEN2_WRITE_FOOTER; \
  }
 
  #define __gen5_write(x) \
  static void \
  gen5_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) { \
-	REG_WRITE_HEADER; \
+	GEN2_WRITE_HEADER; \
  	ilk_dummy_write(dev_priv); \
  	__raw_i915_write##x(dev_priv, reg, val); \
-	REG_WRITE_FOOTER; \
+	GEN2_WRITE_FOOTER; \
  }
 
+__gen5_write(8)
+__gen5_write(16)
+__gen5_write(32)
+__gen5_write(64)
+__gen2_write(8)
+__gen2_write(16)
+__gen2_write(32)
+__gen2_write(64)
+
+#undef __gen5_write
+#undef __gen2_write
+
+#undef GEN2_WRITE_FOOTER
+#undef GEN2_WRITE_HEADER
+
+#define GEN6_WRITE_HEADER \
+	unsigned long irqflags; \
+	trace_i915_reg_rw(true, reg, val, sizeof(val), trace); \
+	assert_device_not_suspended(dev_priv); \
+	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags)
+
+#define GEN6_WRITE_FOOTER \
+	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags)
+
  #define __gen6_write(x) \
  static void \
  gen6_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) { \
  	u32 __fifo_ret = 0; \
-	REG_WRITE_HEADER; \
+	GEN6_WRITE_HEADER; \
  	if (NEEDS_FORCE_WAKE((dev_priv), (reg))) { \
  		__fifo_ret = __gen6_gt_wait_for_fifo(dev_priv); \
  	} \
@@ -905,14 +939,14 @@ gen6_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace
  	if (unlikely(__fifo_ret)) { \
  		gen6_gt_check_fifodbg(dev_priv); \
  	} \
-	REG_WRITE_FOOTER; \
+	GEN6_WRITE_FOOTER; \
  }
 
  #define __hsw_write(x) \
  static void \
  hsw_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) { \
  	u32 __fifo_ret = 0; \
-	REG_WRITE_HEADER; \
+	GEN6_WRITE_HEADER; \
  	if (NEEDS_FORCE_WAKE((dev_priv), (reg))) { \
  		__fifo_ret = __gen6_gt_wait_for_fifo(dev_priv); \
  	} \
@@ -923,7 +957,7 @@ hsw_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace)
  	} \
  	hsw_unclaimed_reg_debug(dev_priv, reg, false, false); \
  	hsw_unclaimed_reg_detect(dev_priv); \
-	REG_WRITE_FOOTER; \
+	GEN6_WRITE_FOOTER; \
  }
 
  static const u32 gen8_shadowed_regs[] = {
@@ -950,7 +984,7 @@ static bool is_gen8_shadowed(struct drm_i915_private *dev_priv, u32 reg)
  #define __gen8_write(x) \
  static void \
  gen8_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) { \
-	REG_WRITE_HEADER; \
+	GEN6_WRITE_HEADER; \
  	hsw_unclaimed_reg_debug(dev_priv, reg, false, true); \
  	if (reg < 0x40000 && !is_gen8_shadowed(dev_priv, reg)) { \
  		if (dev_priv->uncore.forcewake_count == 0) \
@@ -965,7 +999,7 @@ gen8_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace
  	} \
  	hsw_unclaimed_reg_debug(dev_priv, reg, false, false); \
  	hsw_unclaimed_reg_detect(dev_priv); \
-	REG_WRITE_FOOTER; \
+	GEN6_WRITE_FOOTER; \
  }
 
  #define __chv_write(x) \
@@ -973,7 +1007,7 @@ static void \
  chv_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) { \
  	unsigned fwengine = 0; \
  	bool shadowed = is_gen8_shadowed(dev_priv, reg); \
-	REG_WRITE_HEADER; \
+	GEN6_WRITE_HEADER; \
  	if (!shadowed) { \
  		if (FORCEWAKE_CHV_RENDER_RANGE_OFFSET(reg)) { \
  			if (dev_priv->uncore.fw_rendercount == 0) \
@@ -993,7 +1027,7 @@ chv_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace)
  	__raw_i915_write##x(dev_priv, reg, val); \
  	if (fwengine) \
  		dev_priv->uncore.funcs.force_wake_put(dev_priv, fwengine); \
-	REG_WRITE_FOOTER; \
+	GEN6_WRITE_FOOTER; \
  }
 
  static const u32 gen9_shadowed_regs[] = {
@@ -1023,7 +1057,7 @@ static bool is_gen9_shadowed(struct drm_i915_private *dev_priv, u32 reg)
  static void \
  gen9_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, \
  		bool trace) { \
-	REG_WRITE_HEADER; \
+	GEN6_WRITE_HEADER; \
  	if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg)) || \
  			is_gen9_shadowed(dev_priv, reg)) { \
  		__raw_i915_write##x(dev_priv, reg, val); \
@@ -1052,7 +1086,7 @@ gen9_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, \
  			dev_priv->uncore.funcs.force_wake_put(dev_priv, \
  					fwengine); \
  	} \
-	REG_WRITE_FOOTER; \
+	GEN6_WRITE_FOOTER; \
  }
 
  __gen9_write(8)
@@ -1075,24 +1109,14 @@ __gen6_write(8)
  __gen6_write(16)
  __gen6_write(32)
  __gen6_write(64)
-__gen5_write(8)
-__gen5_write(16)
-__gen5_write(32)
-__gen5_write(64)
-__gen4_write(8)
-__gen4_write(16)
-__gen4_write(32)
-__gen4_write(64)
 
  #undef __gen9_write
  #undef __chv_write
  #undef __gen8_write
  #undef __hsw_write
  #undef __gen6_write
-#undef __gen5_write
-#undef __gen4_write
-#undef REG_WRITE_FOOTER
-#undef REG_WRITE_HEADER
+#undef GEN6_WRITE_FOOTER
+#undef GEN6_WRITE_HEADER
 
  #define ASSIGN_WRITE_MMIO_VFUNCS(x) \
  do { \

drivers/gpu/drm/i915/intel_uncore.c void intel_uncore_init(struct drm_device *dev) · line 1236

Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +2 -2
@@ -1205,8 +1229,8 @@ void intel_uncore_init(struct drm_device *dev)
  	case 4:
  	case 3:
  	case 2:
-		ASSIGN_WRITE_MMIO_VFUNCS(gen4);
-		ASSIGN_READ_MMIO_VFUNCS(gen4);
+		ASSIGN_WRITE_MMIO_VFUNCS(gen2);
+		ASSIGN_READ_MMIO_VFUNCS(gen2);
  		break;
  	}
 

changes looks fine to me

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

#7 lore raw

PATCH 4/8 drm/i915: Reduce duplicated forcewake logic

In reply to Mika Kuoppala

From: Chris Wilson <[email protected]>

Introduce a structure to track the individual forcewake domains and use that to eliminate duplicate logic.

v2: - Rebase on latest dinq (Mika)
- for_each_fw_domain macro (Mika)
- Handle reset atomically, keeping the timer running (Mika)
- for_each_fw_domain parameter ordering (Chris)
- defer timer on new register access (Mika)

---
 drivers/gpu/drm/i915/i915_debugfs.c |  65 +++---
 drivers/gpu/drm/i915/i915_drv.h     |  54 +++--
 drivers/gpu/drm/i915/intel_uncore.c | 410 +++++++++++++-----------------------
 3 files changed, 208 insertions(+), 321 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +26 -39
index e142629..5cc838b 100644
@@ -1235,14 +1235,36 @@ static int ironlake_drpc_info(struct seq_file *m)
 	return 0;
 }
 
-static int vlv_drpc_info(struct seq_file *m)
+static int i915_gen6_forcewake_count_info(struct seq_file *m, void *data)
 {
+	struct drm_info_node *node = m->private;
+	struct drm_device *dev = node->minor->dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_uncore_forcewake_domain *fw_domain;
+	const char *domain_names[] = {
+		"render",
+		"blitter",
+		"media",
+	};
+	int i;
+
+	spin_lock_irq(&dev_priv->uncore.lock);
+	for_each_fw_domain(fw_domain, dev_priv, i) {
+		seq_printf(m, "%s.wake_count = %u\n",
+			   domain_names[i],
+			   fw_domain->wake_count);
+	}
+	spin_unlock_irq(&dev_priv->uncore.lock);
 
+	return 0;
+}
+
+static int vlv_drpc_info(struct seq_file *m)
+{
 	struct drm_info_node *node = m->private;
 	struct drm_device *dev = node->minor->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	u32 rpmodectl1, rcctl1, pw_status;
-	unsigned fw_rendercount = 0, fw_mediacount = 0;
 
 	intel_runtime_pm_get(dev_priv);
 
@@ -1274,22 +1296,11 @@ static int vlv_drpc_info(struct seq_file *m)
 	seq_printf(m, "Media RC6 residency since boot: %u\n",
 		   I915_READ(VLV_GT_MEDIA_RC6));
 
-	spin_lock_irq(&dev_priv->uncore.lock);
-	fw_rendercount = dev_priv->uncore.fw_rendercount;
-	fw_mediacount = dev_priv->uncore.fw_mediacount;
-	spin_unlock_irq(&dev_priv->uncore.lock);
-
-	seq_printf(m, "Forcewake Render Count = %u\n", fw_rendercount);
-	seq_printf(m, "Forcewake Media Count = %u\n", fw_mediacount);
-
-
-	return 0;
+	return i915_gen6_forcewake_count_info(m, NULL);
 }
 
-
 static int gen6_drpc_info(struct seq_file *m)
 {
-
 	struct drm_info_node *node = m->private;
 	struct drm_device *dev = node->minor->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -1303,7 +1314,7 @@ static int gen6_drpc_info(struct seq_file *m)
 	intel_runtime_pm_get(dev_priv);
 
 	spin_lock_irq(&dev_priv->uncore.lock);
-	forcewake_count = dev_priv->uncore.forcewake_count;
+	forcewake_count = dev_priv->uncore.fw_domain[FW_DOMAIN_ID_RENDER].wake_count;
 	spin_unlock_irq(&dev_priv->uncore.lock);
 
 	if (forcewake_count) {
@@ -1931,30 +1942,6 @@ static int i915_execlists(struct seq_file *m, void *data)
 	return 0;
 }
 
-static int i915_gen6_forcewake_count_info(struct seq_file *m, void *data)
-{
-	struct drm_info_node *node = m->private;
-	struct drm_device *dev = node->minor->dev;
-	struct drm_i915_private *dev_priv = dev->dev_private;
-	unsigned forcewake_count = 0, fw_rendercount = 0, fw_mediacount = 0;
-
-	spin_lock_irq(&dev_priv->uncore.lock);
-	if (IS_VALLEYVIEW(dev)) {
-		fw_rendercount = dev_priv->uncore.fw_rendercount;
-		fw_mediacount = dev_priv->uncore.fw_mediacount;
-	} else
-		forcewake_count = dev_priv->uncore.forcewake_count;
-	spin_unlock_irq(&dev_priv->uncore.lock);
-
-	if (IS_VALLEYVIEW(dev)) {
-		seq_printf(m, "fw_rendercount = %u\n", fw_rendercount);
-		seq_printf(m, "fw_mediacount = %u\n", fw_mediacount);
-	} else
-		seq_printf(m, "forcewake count = %u\n", forcewake_count);
-
-	return 0;
-}
-
 static const char *swizzle_string(unsigned swizzle)
 {
 	switch (swizzle) {
drivers/gpu/drm/i915/i915_drv.h +37 -17

78 lines not on this page · raw message

drivers/gpu/drm/i915/intel_uncore.c +145 -265

606 lines not on this page · raw message

Signature · 6 lines
1.9.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Acked-by on v1 4/8
#8 lore raw

Re: [PATCH 4/8] drm/i915: Reduce duplicated forcewake logic

Quoted 397 lines · Mika Kuoppala

From: Chris Wilson <[email protected]>

Introduce a structure to track the individual forcewake domains and use that to eliminate duplicate logic.

v2: - Rebase on latest dinq (Mika)
- for_each_fw_domain macro (Mika)
- Handle reset atomically, keeping the timer running (Mika)
- for_each_fw_domain parameter ordering (Chris)
- defer timer on new register access (Mika)

---
  drivers/gpu/drm/i915/i915_debugfs.c |  65 +++---
  drivers/gpu/drm/i915/i915_drv.h     |  54 +++--
  drivers/gpu/drm/i915/intel_uncore.c | 410 +++++++++++++-----------------------
  3 files changed, 208 insertions(+), 321 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +26 -39
index e142629..5cc838b 100644
@@ -1235,14 +1235,36 @@ static int ironlake_drpc_info(struct seq_file *m)
  	return 0;
  }
 
-static int vlv_drpc_info(struct seq_file *m)
+static int i915_gen6_forcewake_count_info(struct seq_file *m, void *data)
  {
+	struct drm_info_node *node = m->private;
+	struct drm_device *dev = node->minor->dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_uncore_forcewake_domain *fw_domain;
+	const char *domain_names[] = {
+		"render",
+		"blitter",
+		"media",
+	};
+	int i;
+
+	spin_lock_irq(&dev_priv->uncore.lock);
+	for_each_fw_domain(fw_domain, dev_priv, i) {
+		seq_printf(m, "%s.wake_count = %u\n",
+			   domain_names[i],
+			   fw_domain->wake_count);
+	}
+	spin_unlock_irq(&dev_priv->uncore.lock);
 
+	return 0;
+}
+
+static int vlv_drpc_info(struct seq_file *m)
+{
  	struct drm_info_node *node = m->private;
  	struct drm_device *dev = node->minor->dev;
  	struct drm_i915_private *dev_priv = dev->dev_private;
  	u32 rpmodectl1, rcctl1, pw_status;
-	unsigned fw_rendercount = 0, fw_mediacount = 0;
 
  	intel_runtime_pm_get(dev_priv);
 
@@ -1274,22 +1296,11 @@ static int vlv_drpc_info(struct seq_file *m)
  	seq_printf(m, "Media RC6 residency since boot: %u\n",
  		   I915_READ(VLV_GT_MEDIA_RC6));
 
-	spin_lock_irq(&dev_priv->uncore.lock);
-	fw_rendercount = dev_priv->uncore.fw_rendercount;
-	fw_mediacount = dev_priv->uncore.fw_mediacount;
-	spin_unlock_irq(&dev_priv->uncore.lock);
-
-	seq_printf(m, "Forcewake Render Count = %u\n", fw_rendercount);
-	seq_printf(m, "Forcewake Media Count = %u\n", fw_mediacount);
-
-
-	return 0;
+	return i915_gen6_forcewake_count_info(m, NULL);
  }
 
-
  static int gen6_drpc_info(struct seq_file *m)
  {
-
  	struct drm_info_node *node = m->private;
  	struct drm_device *dev = node->minor->dev;
  	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -1303,7 +1314,7 @@ static int gen6_drpc_info(struct seq_file *m)
  	intel_runtime_pm_get(dev_priv);
 
  	spin_lock_irq(&dev_priv->uncore.lock);
-	forcewake_count = dev_priv->uncore.forcewake_count;
+	forcewake_count = dev_priv->uncore.fw_domain[FW_DOMAIN_ID_RENDER].wake_count;
  	spin_unlock_irq(&dev_priv->uncore.lock);
 
  	if (forcewake_count) {
@@ -1931,30 +1942,6 @@ static int i915_execlists(struct seq_file *m, void *data)
  	return 0;
  }
 
-static int i915_gen6_forcewake_count_info(struct seq_file *m, void *data)
-{
-	struct drm_info_node *node = m->private;
-	struct drm_device *dev = node->minor->dev;
-	struct drm_i915_private *dev_priv = dev->dev_private;
-	unsigned forcewake_count = 0, fw_rendercount = 0, fw_mediacount = 0;
-
-	spin_lock_irq(&dev_priv->uncore.lock);
-	if (IS_VALLEYVIEW(dev)) {
-		fw_rendercount = dev_priv->uncore.fw_rendercount;
-		fw_mediacount = dev_priv->uncore.fw_mediacount;
-	} else
-		forcewake_count = dev_priv->uncore.forcewake_count;
-	spin_unlock_irq(&dev_priv->uncore.lock);
-
-	if (IS_VALLEYVIEW(dev)) {
-		seq_printf(m, "fw_rendercount = %u\n", fw_rendercount);
-		seq_printf(m, "fw_mediacount = %u\n", fw_mediacount);
-	} else
-		seq_printf(m, "forcewake count = %u\n", forcewake_count);
-
-	return 0;
-}
-
  static const char *swizzle_string(unsigned swizzle)
  {
  	switch (swizzle) {
drivers/gpu/drm/i915/i915_drv.h +37 -17
index 95dfa2d..410558a 100644
@@ -554,20 +554,45 @@ struct intel_uncore_funcs {
  				uint64_t val, bool trace);
  };
 
+enum {
+	FW_DOMAIN_ID_RENDER = 0,
+	FW_DOMAIN_ID_BLITTER,
+	FW_DOMAIN_ID_MEDIA,
+
+	FW_DOMAIN_ID_COUNT
+};
+
  struct intel_uncore {
  	spinlock_t lock; /** lock is also taken in irq contexts. */
 
  	struct intel_uncore_funcs funcs;
 
  	unsigned fifo_count;
-	unsigned forcewake_count;
-
-	unsigned fw_rendercount;
-	unsigned fw_mediacount;
-	unsigned fw_blittercount;
-
-	struct timer_list force_wake_timer;
-};
+	unsigned fw_domains;
+
+	struct intel_uncore_forcewake_domain {
+		struct drm_i915_private *i915;
+		int id;
+		unsigned wake_count;
+		struct timer_list timer;
+	} fw_domain[FW_DOMAIN_ID_COUNT];
+#define FORCEWAKE_RENDER	(1 << FW_DOMAIN_ID_RENDER)
+#define FORCEWAKE_BLITTER	(1 << FW_DOMAIN_ID_BLITTER)
+#define FORCEWAKE_MEDIA		(1 << FW_DOMAIN_ID_MEDIA)
+#define FORCEWAKE_ALL		(FORCEWAKE_RENDER | \
+				 FORCEWAKE_BLITTER | \
+				 FORCEWAKE_MEDIA)
+};
+
+/* Iterate over initialised fw domains */
+#define for_each_fw_domain_mask(domain__, mask__, dev_priv__, i__) \
+	for ((i__) = 0, (domain__) = &(dev_priv__)->uncore.fw_domain[0]; \
+	     (i__) < FW_DOMAIN_ID_COUNT; \
+	     (i__)++, (domain__) = &(dev_priv__)->uncore.fw_domain[i__]) \
+		if (((mask__) & (dev_priv__)->uncore.fw_domains) & (1 << (i__)))
+
+#define for_each_fw_domain(domain__, dev_priv__, i__) \
+	for_each_fw_domain_mask(domain__, FORCEWAKE_ALL, dev_priv__, i__)
 
  #define DEV_INFO_FOR_EACH_FLAG(func, sep) \
  	func(is_mobile) sep \
@@ -2998,8 +3023,10 @@ extern void intel_display_print_error_state(struct drm_i915_error_state_buf *e,
   * must be set to prevent GT core from power down and stale values being
   * returned.
   */
-void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine);
-void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine);
+void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv,
+			    unsigned fw_domains);
+void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv,
+			    unsigned fw_domains);
  void assert_force_wake_inactive(struct drm_i915_private *dev_priv);
 
  int sandybridge_pcode_read(struct drm_i915_private *dev_priv, u32 mbox, u32 *val);
@@ -3031,13 +3058,6 @@ void vlv_flisdsi_write(struct drm_i915_private *dev_priv, u32 reg, u32 val);
  int vlv_gpu_freq(struct drm_i915_private *dev_priv, int val);
  int vlv_freq_opcode(struct drm_i915_private *dev_priv, int val);
 
-#define FORCEWAKE_RENDER	(1 << 0)
-#define FORCEWAKE_MEDIA		(1 << 1)
-#define FORCEWAKE_BLITTER	(1 << 2)
-#define FORCEWAKE_ALL		(FORCEWAKE_RENDER | FORCEWAKE_MEDIA | \
-					FORCEWAKE_BLITTER)
-
-
  #define I915_READ8(reg)		dev_priv->uncore.funcs.mmio_readb(dev_priv, (reg), true)
  #define I915_WRITE8(reg, val)	dev_priv->uncore.funcs.mmio_writeb(dev_priv, (reg), (val), true)
 
drivers/gpu/drm/i915/intel_uncore.c +18 -96
index 069fe7a..85e46b0 100644
@@ -67,7 +67,7 @@ static void __gen6_gt_force_wake_reset(struct drm_i915_private *dev_priv)
  }
 
  static void __gen6_gt_force_wake_get(struct drm_i915_private *dev_priv,
-							int fw_engine)
+				     int fw_engine)
  {
  	if (wait_for_atomic((__raw_i915_read32(dev_priv, FORCEWAKE_ACK) & 1) == 0,
  			    FORCEWAKE_ACK_TIMEOUT_MS))
@@ -93,7 +93,7 @@ static void __gen7_gt_force_wake_mt_reset(struct drm_i915_private *dev_priv)
  }
 
  static void __gen7_gt_force_wake_mt_get(struct drm_i915_private *dev_priv,
-							int fw_engine)
+					int fw_engine)
  {
  	u32 forcewake_ack;
 
@@ -129,7 +129,7 @@ static void gen6_gt_check_fifodbg(struct drm_i915_private *dev_priv)
  }
 
  static void __gen6_gt_force_wake_put(struct drm_i915_private *dev_priv,
-							int fw_engine)
+				     int fw_engine)
  {
  	__raw_i915_write32(dev_priv, FORCEWAKE, 0);
  	/* something from same cacheline, but !FORCEWAKE */
@@ -138,7 +138,7 @@ static void __gen6_gt_force_wake_put(struct drm_i915_private *dev_priv,
  }
 
  static void __gen7_gt_force_wake_mt_put(struct drm_i915_private *dev_priv,
-							int fw_engine)
+					int fw_engine)
  {
  	__raw_i915_write32(dev_priv, FORCEWAKE_MT,
  			   _MASKED_BIT_DISABLE(FORCEWAKE_KERNEL));
@@ -187,7 +187,7 @@ static void vlv_force_wake_reset(struct drm_i915_private *dev_priv)
  }
 
  static void __vlv_force_wake_get(struct drm_i915_private *dev_priv,
-						int fw_engine)
+				 int fw_engine)
  {
  	/* Check for Render Engine */
  	if (FORCEWAKE_RENDER & fw_engine) {
@@ -227,9 +227,8 @@ static void __vlv_force_wake_get(struct drm_i915_private *dev_priv,
  }
 
  static void __vlv_force_wake_put(struct drm_i915_private *dev_priv,
-					int fw_engine)
+				 int fw_engine)
  {
-
  	/* Check for Render Engine */
  	if (FORCEWAKE_RENDER & fw_engine)
  		__raw_i915_write32(dev_priv, FORCEWAKE_VLV,
@@ -247,37 +246,6 @@ static void __vlv_force_wake_put(struct drm_i915_private *dev_priv,
  		gen6_gt_check_fifodbg(dev_priv);
  }
 
-static void vlv_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
-{
-	if (fw_engine & FORCEWAKE_RENDER &&
-	    dev_priv->uncore.fw_rendercount++ != 0)
-		fw_engine &= ~FORCEWAKE_RENDER;
-	if (fw_engine & FORCEWAKE_MEDIA &&
-	    dev_priv->uncore.fw_mediacount++ != 0)
-		fw_engine &= ~FORCEWAKE_MEDIA;
-
-	if (fw_engine)
-		dev_priv->uncore.funcs.force_wake_get(dev_priv, fw_engine);
-}
-
-static void vlv_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
-{
-	if (fw_engine & FORCEWAKE_RENDER) {
-		WARN_ON(!dev_priv->uncore.fw_rendercount);
-		if (--dev_priv->uncore.fw_rendercount != 0)
-			fw_engine &= ~FORCEWAKE_RENDER;
-	}
-
-	if (fw_engine & FORCEWAKE_MEDIA) {
-		WARN_ON(!dev_priv->uncore.fw_mediacount);
-		if (--dev_priv->uncore.fw_mediacount != 0)
-			fw_engine &= ~FORCEWAKE_MEDIA;
-	}
-
-	if (fw_engine)
-		dev_priv->uncore.funcs.force_wake_put(dev_priv, fw_engine);
-}
-
  static void __gen9_gt_force_wake_mt_reset(struct drm_i915_private *dev_priv)
  {
  	__raw_i915_write32(dev_priv, FORCEWAKE_RENDER_GEN9,
@@ -367,80 +335,40 @@ __gen9_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
  				_MASKED_BIT_DISABLE(FORCEWAKE_KERNEL));
  }
 
-static void
-gen9_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
-{
-	if (FORCEWAKE_RENDER & fw_engine) {
-		if (dev_priv->uncore.fw_rendercount++ == 0)
-			dev_priv->uncore.funcs.force_wake_get(dev_priv,
-							FORCEWAKE_RENDER);
-	}
-
-	if (FORCEWAKE_MEDIA & fw_engine) {
-		if (dev_priv->uncore.fw_mediacount++ == 0)
-			dev_priv->uncore.funcs.force_wake_get(dev_priv,
-							FORCEWAKE_MEDIA);
-	}
-
-	if (FORCEWAKE_BLITTER & fw_engine) {
-		if (dev_priv->uncore.fw_blittercount++ == 0)
-			dev_priv->uncore.funcs.force_wake_get(dev_priv,
-							FORCEWAKE_BLITTER);
-	}
-}
-
-static void
-gen9_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
-{
-	if (FORCEWAKE_RENDER & fw_engine) {
-		WARN_ON(dev_priv->uncore.fw_rendercount == 0);
-		if (--dev_priv->uncore.fw_rendercount == 0)
-			dev_priv->uncore.funcs.force_wake_put(dev_priv,
-							FORCEWAKE_RENDER);
-	}
-
-	if (FORCEWAKE_MEDIA & fw_engine) {
-		WARN_ON(dev_priv->uncore.fw_mediacount == 0);
-		if (--dev_priv->uncore.fw_mediacount == 0)
-			dev_priv->uncore.funcs.force_wake_put(dev_priv,
-							FORCEWAKE_MEDIA);
-	}
-
-	if (FORCEWAKE_BLITTER & fw_engine) {
-		WARN_ON(dev_priv->uncore.fw_blittercount == 0);
-		if (--dev_priv->uncore.fw_blittercount == 0)
-			dev_priv->uncore.funcs.force_wake_put(dev_priv,
-							FORCEWAKE_BLITTER);
-	}
-}
-
  static void gen6_force_wake_timer(unsigned long arg)
  {
-	struct drm_i915_private *dev_priv = (void *)arg;
+	struct intel_uncore_forcewake_domain *domain = (void *)arg;
  	unsigned long irqflags;
 
-	assert_device_not_suspended(dev_priv);
+	assert_device_not_suspended(domain->i915);
 
-	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
-	WARN_ON(!dev_priv->uncore.forcewake_count);
+	spin_lock_irqsave(&domain->i915->uncore.lock, irqflags);
+	WARN_ON(!domain->wake_count);
 
-	if (--dev_priv->uncore.forcewake_count == 0)
-		dev_priv->uncore.funcs.force_wake_put(dev_priv, FORCEWAKE_ALL);
-	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
+	if (--domain->wake_count == 0)
+		domain->i915->uncore.funcs.force_wake_put(domain->i915,
+							  1 << domain->id);
+	spin_unlock_irqrestore(&domain->i915->uncore.lock, irqflags);
  }
 
  void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore)
  {
  	struct drm_i915_private *dev_priv = dev->dev_private;
-	unsigned long irqflags;
-
-	if (del_timer_sync(&dev_priv->uncore.force_wake_timer))
-		gen6_force_wake_timer((unsigned long)dev_priv);
+	unsigned long irqflags, fw = 0;
+	struct intel_uncore_forcewake_domain *domain;
+	int id;
 
+	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
  	/* Hold uncore.lock across reset to prevent any register access
  	 * with forcewake not set correctly

drivers/gpu/drm/i915/intel_uncore.c __gen9_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine) · line 372

Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +7 -1
  	 */
-	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
+
+	for_each_fw_domain(domain, dev_priv, id)
+		if (domain->wake_count)
+			fw |= 1 << id;
+
+	if (fw)
+		dev_priv->uncore.funcs.force_wake_put(dev_priv, fw);
 

Hi Mika,

Why do put?

Is there a possibility of getting get & put count mismatch? for example,
1. We Read the register
-> forcewake_get & start domain_timer. (count = 1)
2. Now we call "intel_uncore_forcewake_reset" which does a put. (count = 0)
3. now the domain_timer from step #1 comes (count -1)? right? or i am missing something here?

Others Looks fine to me.

Quoted 38 lines
drivers/gpu/drm/i915/intel_uncore.c +2 -23
  	if (IS_VALLEYVIEW(dev))
  		vlv_force_wake_reset(dev_priv);
@@ -454,28 +382,6 @@ void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore)
  		__gen9_gt_force_wake_mt_reset(dev_priv);
 
  	if (restore) { /* If reset with a user forcewake, try to restore */
-		unsigned fw = 0;
-
-		if (IS_VALLEYVIEW(dev)) {
-			if (dev_priv->uncore.fw_rendercount)
-				fw |= FORCEWAKE_RENDER;
-
-			if (dev_priv->uncore.fw_mediacount)
-				fw |= FORCEWAKE_MEDIA;
-		} else if (IS_GEN9(dev)) {
-			if (dev_priv->uncore.fw_rendercount)
-				fw |= FORCEWAKE_RENDER;
-
-			if (dev_priv->uncore.fw_mediacount)
-				fw |= FORCEWAKE_MEDIA;
-
-			if (dev_priv->uncore.fw_blittercount)
-				fw |= FORCEWAKE_BLITTER;
-		} else {
-			if (dev_priv->uncore.forcewake_count)
-				fw = FORCEWAKE_ALL;
-		}
-
  		if (fw)
  			dev_priv->uncore.funcs.force_wake_get(dev_priv, fw);
 
@@ -533,28 +439,28 @@ void intel_uncore_sanitize(struct drm_device *dev)
   * be called at the beginning of the sequence followed by a call to
   * gen6_gt_force_wake_put() at the end of the sequence.
   */
-void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
+void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv,
+			    unsigned fw_domains)

drivers/gpu/drm/i915/intel_uncore.c void intel_uncore_sanitize(struct drm_device *dev) · line 544

Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +2 -2
  {
  	unsigned long irqflags;
+	struct intel_uncore_forcewake_domain *domain;
+	int id;
 
  	if (!dev_priv->uncore.funcs.force_wake_get)
  		return;
 
-	intel_runtime_pm_get(dev_priv);
-

you need to change this in Patch #2

Quoted 376 lines
drivers/gpu/drm/i915/intel_uncore.c +109 -143
  	WARN_ON(!pm_runtime_active(&dev_priv->dev->pdev->dev));
 
+	fw_domains &= dev_priv->uncore.fw_domains;
+
  	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
 
-	if (IS_GEN9(dev_priv->dev)) {
-		gen9_force_wake_get(dev_priv, fw_engine);
-	} else if (IS_VALLEYVIEW(dev_priv->dev)) {
-		vlv_force_wake_get(dev_priv, fw_engine);
-	} else {
-		if (dev_priv->uncore.forcewake_count++ == 0)
-			dev_priv->uncore.funcs.force_wake_get(dev_priv,
-							      FORCEWAKE_ALL);
-	}
+	for_each_fw_domain_mask(domain, fw_domains, dev_priv, id)
+		if (domain->wake_count++)
+			fw_domains &= ~(1 << id);
+
+	if (fw_domains)
+		dev_priv->uncore.funcs.force_wake_get(dev_priv, fw_domains);
 
  	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
  }
@@ -562,26 +468,27 @@ void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
  /*
   * see gen6_gt_force_wake_get()
   */
-void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
+void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv,
+			    unsigned fw_domains)
  {
  	unsigned long irqflags;
+	struct intel_uncore_forcewake_domain *domain;
+	int id;
 
  	if (!dev_priv->uncore.funcs.force_wake_put)
  		return;
 
+	fw_domains &= dev_priv->uncore.fw_domains;
+
  	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
 
-	if (IS_GEN9(dev_priv->dev)) {
-		gen9_force_wake_put(dev_priv, fw_engine);
-	} else if (IS_VALLEYVIEW(dev_priv->dev)) {
-		vlv_force_wake_put(dev_priv, fw_engine);
-	} else {
-		WARN_ON(!dev_priv->uncore.forcewake_count);
-		if (--dev_priv->uncore.forcewake_count == 0) {
-			dev_priv->uncore.forcewake_count++;
-			mod_timer_pinned(&dev_priv->uncore.force_wake_timer,
-					 jiffies + 1);
-		}
+	for_each_fw_domain_mask(domain, fw_domains, dev_priv, id) {
+		WARN_ON(!domain->wake_count);
+		if (--domain->wake_count)
+			continue;
+
+		domain->wake_count++;
+		mod_timer_pinned(&domain->timer, jiffies + 1);
  	}
 
  	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
@@ -589,10 +496,14 @@ void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
 
  void assert_force_wake_inactive(struct drm_i915_private *dev_priv)
  {
+	struct intel_uncore_forcewake_domain *domain;
+	int i;
+
  	if (!dev_priv->uncore.funcs.force_wake_get)
  		return;
 
-	WARN_ON(dev_priv->uncore.forcewake_count > 0);
+	for_each_fw_domain(domain, dev_priv, i)
+		WARN_ON(domain->wake_count);
  }
 
  /* We give fast paths for the really cool registers */
@@ -758,19 +669,36 @@ __gen2_read(64)
  	trace_i915_reg_rw(false, reg, val, sizeof(val), trace); \
  	return val
 
+static inline void __force_wake_get(struct drm_i915_private *dev_priv,
+				    unsigned fw_domains)
+{
+	struct intel_uncore_forcewake_domain *domain;
+	int i;
+
+	if (WARN_ON(!fw_domains))
+		return;
+
+	/* Ideally GCC would be constant-fold and eliminate this loop */
+	for_each_fw_domain_mask(domain, fw_domains, dev_priv, i) {
+		if (domain->wake_count)
+			fw_domains &= ~(1 << i);
+		else
+			domain->wake_count++;
+
+		mod_timer_pinned(&domain->timer, jiffies + 1);
+	}
+
+	if (fw_domains)
+		dev_priv->uncore.funcs.force_wake_get(dev_priv, fw_domains);
+}
+
  #define __gen6_read(x) \
  static u##x \
  gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
  	GEN6_READ_HEADER(x); \
  	hsw_unclaimed_reg_debug(dev_priv, reg, true, true); \
-	if (dev_priv->uncore.forcewake_count == 0 && \
-	    NEEDS_FORCE_WAKE((dev_priv), (reg))) { \
-		dev_priv->uncore.funcs.force_wake_get(dev_priv, \
-						      FORCEWAKE_ALL); \
-		dev_priv->uncore.forcewake_count++; \
-		mod_timer_pinned(&dev_priv->uncore.force_wake_timer, \
-				 jiffies + 1); \
-	} \
+	if (NEEDS_FORCE_WAKE((dev_priv), (reg))) \
+		__force_wake_get(dev_priv, FORCEWAKE_RENDER); \
  	val = __raw_i915_read##x(dev_priv, reg); \
  	hsw_unclaimed_reg_debug(dev_priv, reg, true, false); \
  	GEN6_READ_FOOTER; \
@@ -779,45 +707,27 @@ gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
  #define __vlv_read(x) \
  static u##x \
  vlv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
-	unsigned fwengine = 0; \
  	GEN6_READ_HEADER(x); \
-	if (FORCEWAKE_VLV_RENDER_RANGE_OFFSET(reg)) { \
-		if (dev_priv->uncore.fw_rendercount == 0) \
-			fwengine = FORCEWAKE_RENDER; \
-	} else if (FORCEWAKE_VLV_MEDIA_RANGE_OFFSET(reg)) { \
-		if (dev_priv->uncore.fw_mediacount == 0) \
-			fwengine = FORCEWAKE_MEDIA; \
-	}  \
-	if (fwengine) \
-		dev_priv->uncore.funcs.force_wake_get(dev_priv, fwengine); \
+	if (FORCEWAKE_VLV_RENDER_RANGE_OFFSET(reg)) \
+		__force_wake_get(dev_priv, FORCEWAKE_RENDER); \
+	else if (FORCEWAKE_VLV_MEDIA_RANGE_OFFSET(reg)) \
+		__force_wake_get(dev_priv, FORCEWAKE_MEDIA); \
  	val = __raw_i915_read##x(dev_priv, reg); \
-	if (fwengine) \
-		dev_priv->uncore.funcs.force_wake_put(dev_priv, fwengine); \
  	GEN6_READ_FOOTER; \
  }
 
  #define __chv_read(x) \
  static u##x \
  chv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
-	unsigned fwengine = 0; \
  	GEN6_READ_HEADER(x); \
-	if (FORCEWAKE_CHV_RENDER_RANGE_OFFSET(reg)) { \
-		if (dev_priv->uncore.fw_rendercount == 0) \
-			fwengine = FORCEWAKE_RENDER; \
-	} else if (FORCEWAKE_CHV_MEDIA_RANGE_OFFSET(reg)) { \
-		if (dev_priv->uncore.fw_mediacount == 0) \
-			fwengine = FORCEWAKE_MEDIA; \
-	} else if (FORCEWAKE_CHV_COMMON_RANGE_OFFSET(reg)) { \
-		if (dev_priv->uncore.fw_rendercount == 0) \
-			fwengine |= FORCEWAKE_RENDER; \
-		if (dev_priv->uncore.fw_mediacount == 0) \
-			fwengine |= FORCEWAKE_MEDIA; \
-	} \
-	if (fwengine) \
-		dev_priv->uncore.funcs.force_wake_get(dev_priv, fwengine); \
+	if (FORCEWAKE_CHV_RENDER_RANGE_OFFSET(reg)) \
+		__force_wake_get(dev_priv, FORCEWAKE_RENDER); \
+	else if (FORCEWAKE_CHV_MEDIA_RANGE_OFFSET(reg)) \
+		__force_wake_get(dev_priv, FORCEWAKE_MEDIA); \
+	else if (FORCEWAKE_CHV_COMMON_RANGE_OFFSET(reg)) \
+		__force_wake_get(dev_priv, \
+				 FORCEWAKE_RENDER | FORCEWAKE_MEDIA); \
  	val = __raw_i915_read##x(dev_priv, reg); \
-	if (fwengine) \
-		dev_priv->uncore.funcs.force_wake_put(dev_priv, fwengine); \
  	GEN6_READ_FOOTER; \
  }
 
@@ -827,32 +737,21 @@ chv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
  #define __gen9_read(x) \
  static u##x \
  gen9_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
+	unsigned fw_engine; \
  	GEN6_READ_HEADER(x); \
-	if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg))) { \
-		val = __raw_i915_read##x(dev_priv, reg); \
-	} else { \
-		unsigned fwengine = 0; \
-		if (FORCEWAKE_GEN9_RENDER_RANGE_OFFSET(reg)) { \
-			if (dev_priv->uncore.fw_rendercount == 0) \
-				fwengine = FORCEWAKE_RENDER; \
-		} else if (FORCEWAKE_GEN9_MEDIA_RANGE_OFFSET(reg)) { \
-			if (dev_priv->uncore.fw_mediacount == 0) \
-				fwengine = FORCEWAKE_MEDIA; \
-		} else if (FORCEWAKE_GEN9_COMMON_RANGE_OFFSET(reg)) { \
-			if (dev_priv->uncore.fw_rendercount == 0) \
-				fwengine |= FORCEWAKE_RENDER; \
-			if (dev_priv->uncore.fw_mediacount == 0) \
-				fwengine |= FORCEWAKE_MEDIA; \
-		} else { \
-			if (dev_priv->uncore.fw_blittercount == 0) \
-				fwengine = FORCEWAKE_BLITTER; \
-		} \
-		if (fwengine) \
-			dev_priv->uncore.funcs.force_wake_get(dev_priv, fwengine); \
-		val = __raw_i915_read##x(dev_priv, reg); \
-		if (fwengine) \
-			dev_priv->uncore.funcs.force_wake_put(dev_priv, fwengine); \
-	} \
+	if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg)))	\
+		fw_engine = 0; \
+	else if (FORCEWAKE_GEN9_RENDER_RANGE_OFFSET(reg))	\
+		fw_engine = FORCEWAKE_RENDER; \
+	else if (FORCEWAKE_GEN9_MEDIA_RANGE_OFFSET(reg)) \
+		fw_engine = FORCEWAKE_MEDIA; \
+	else if (FORCEWAKE_GEN9_COMMON_RANGE_OFFSET(reg)) \
+		fw_engine = FORCEWAKE_RENDER | FORCEWAKE_MEDIA; \
+	else \
+		fw_engine = FORCEWAKE_BLITTER; \
+	if (fw_engine) \
+		__force_wake_get(dev_priv, fw_engine); \
+	val = __raw_i915_read##x(dev_priv, reg); \
  	GEN6_READ_FOOTER; \
  }
 
@@ -986,17 +885,9 @@ static void \
  gen8_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) { \
  	GEN6_WRITE_HEADER; \
  	hsw_unclaimed_reg_debug(dev_priv, reg, false, true); \
-	if (reg < 0x40000 && !is_gen8_shadowed(dev_priv, reg)) { \
-		if (dev_priv->uncore.forcewake_count == 0) \
-			dev_priv->uncore.funcs.force_wake_get(dev_priv,	\
-							      FORCEWAKE_ALL); \
-		__raw_i915_write##x(dev_priv, reg, val); \
-		if (dev_priv->uncore.forcewake_count == 0) \
-			dev_priv->uncore.funcs.force_wake_put(dev_priv, \
-							      FORCEWAKE_ALL); \
-	} else { \
-		__raw_i915_write##x(dev_priv, reg, val); \
-	} \
+	if (reg < 0x40000 && !is_gen8_shadowed(dev_priv, reg)) \
+		__force_wake_get(dev_priv, FORCEWAKE_RENDER); \
+	__raw_i915_write##x(dev_priv, reg, val); \
  	hsw_unclaimed_reg_debug(dev_priv, reg, false, false); \
  	hsw_unclaimed_reg_detect(dev_priv); \
  	GEN6_WRITE_FOOTER; \
@@ -1005,28 +896,17 @@ gen8_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace
  #define __chv_write(x) \
  static void \
  chv_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) { \
-	unsigned fwengine = 0; \
  	bool shadowed = is_gen8_shadowed(dev_priv, reg); \
  	GEN6_WRITE_HEADER; \
  	if (!shadowed) { \
-		if (FORCEWAKE_CHV_RENDER_RANGE_OFFSET(reg)) { \
-			if (dev_priv->uncore.fw_rendercount == 0) \
-				fwengine = FORCEWAKE_RENDER; \
-		} else if (FORCEWAKE_CHV_MEDIA_RANGE_OFFSET(reg)) { \
-			if (dev_priv->uncore.fw_mediacount == 0) \
-				fwengine = FORCEWAKE_MEDIA; \
-		} else if (FORCEWAKE_CHV_COMMON_RANGE_OFFSET(reg)) { \
-			if (dev_priv->uncore.fw_rendercount == 0) \
-				fwengine |= FORCEWAKE_RENDER; \
-			if (dev_priv->uncore.fw_mediacount == 0) \
-				fwengine |= FORCEWAKE_MEDIA; \
-		} \
+		if (FORCEWAKE_CHV_RENDER_RANGE_OFFSET(reg)) \
+			__force_wake_get(dev_priv, FORCEWAKE_RENDER); \
+		else if (FORCEWAKE_CHV_MEDIA_RANGE_OFFSET(reg)) \
+			__force_wake_get(dev_priv, FORCEWAKE_MEDIA); \
+		else if (FORCEWAKE_CHV_COMMON_RANGE_OFFSET(reg)) \
+			__force_wake_get(dev_priv, FORCEWAKE_RENDER | FORCEWAKE_MEDIA); \
  	} \
-	if (fwengine) \
-		dev_priv->uncore.funcs.force_wake_get(dev_priv, fwengine); \
  	__raw_i915_write##x(dev_priv, reg, val); \
-	if (fwengine) \
-		dev_priv->uncore.funcs.force_wake_put(dev_priv, fwengine); \
  	GEN6_WRITE_FOOTER; \
  }
 
@@ -1057,35 +937,22 @@ static bool is_gen9_shadowed(struct drm_i915_private *dev_priv, u32 reg)
  static void \
  gen9_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, \
  		bool trace) { \
+	unsigned fw_engine; \
  	GEN6_WRITE_HEADER; \
-	if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg)) || \
-			is_gen9_shadowed(dev_priv, reg)) { \
-		__raw_i915_write##x(dev_priv, reg, val); \
-	} else { \
-		unsigned fwengine = 0; \
-		if (FORCEWAKE_GEN9_RENDER_RANGE_OFFSET(reg)) { \
-			if (dev_priv->uncore.fw_rendercount == 0) \
-				fwengine = FORCEWAKE_RENDER; \
-		} else if (FORCEWAKE_GEN9_MEDIA_RANGE_OFFSET(reg)) { \
-			if (dev_priv->uncore.fw_mediacount == 0) \
-				fwengine = FORCEWAKE_MEDIA; \
-		} else if (FORCEWAKE_GEN9_COMMON_RANGE_OFFSET(reg)) { \
-			if (dev_priv->uncore.fw_rendercount == 0) \
-				fwengine |= FORCEWAKE_RENDER; \
-			if (dev_priv->uncore.fw_mediacount == 0) \
-				fwengine |= FORCEWAKE_MEDIA; \
-		} else { \
-			if (dev_priv->uncore.fw_blittercount == 0) \
-				fwengine = FORCEWAKE_BLITTER; \
-		} \
-		if (fwengine) \
-			dev_priv->uncore.funcs.force_wake_get(dev_priv, \
-					fwengine); \
-		__raw_i915_write##x(dev_priv, reg, val); \
-		if (fwengine) \
-			dev_priv->uncore.funcs.force_wake_put(dev_priv, \
-					fwengine); \
-	} \
+	if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg)) ||	\
+	    is_gen9_shadowed(dev_priv, reg)) \
+		fw_engine = 0; \
+	else if (FORCEWAKE_GEN9_RENDER_RANGE_OFFSET(reg)) \
+		fw_engine = FORCEWAKE_RENDER; \
+	else if (FORCEWAKE_GEN9_MEDIA_RANGE_OFFSET(reg)) \
+		fw_engine = FORCEWAKE_MEDIA; \
+	else if (FORCEWAKE_GEN9_COMMON_RANGE_OFFSET(reg)) \
+		fw_engine = FORCEWAKE_RENDER | FORCEWAKE_MEDIA; \
+	else \
+		fw_engine = FORCEWAKE_BLITTER; \
+	if (fw_engine) \
+		__force_wake_get(dev_priv, fw_engine); \
+	__raw_i915_write##x(dev_priv, reg, val); \
  	GEN6_WRITE_FOOTER; \
  }
 
@@ -1137,21 +1004,24 @@ do { \
  void intel_uncore_init(struct drm_device *dev)
  {
  	struct drm_i915_private *dev_priv = dev->dev_private;
-
-	setup_timer(&dev_priv->uncore.force_wake_timer,
-		    gen6_force_wake_timer, (unsigned long)dev_priv);
+	struct intel_uncore_forcewake_domain *domain;
+	int i;
 
  	__intel_uncore_early_sanitize(dev, false);
 
  	if (IS_GEN9(dev)) {
  		dev_priv->uncore.funcs.force_wake_get = __gen9_force_wake_get;
  		dev_priv->uncore.funcs.force_wake_put = __gen9_force_wake_put;
+		dev_priv->uncore.fw_domains = FORCEWAKE_RENDER |
+			FORCEWAKE_BLITTER | FORCEWAKE_MEDIA;
  	} else if (IS_VALLEYVIEW(dev)) {
  		dev_priv->uncore.funcs.force_wake_get = __vlv_force_wake_get;
  		dev_priv->uncore.funcs.force_wake_put = __vlv_force_wake_put;
+		dev_priv->uncore.fw_domains = FORCEWAKE_RENDER | FORCEWAKE_MEDIA;
  	} else if (IS_HASWELL(dev) || IS_BROADWELL(dev)) {
  		dev_priv->uncore.funcs.force_wake_get = __gen7_gt_force_wake_mt_get;
  		dev_priv->uncore.funcs.force_wake_put = __gen7_gt_force_wake_mt_put;
+		dev_priv->uncore.fw_domains = FORCEWAKE_RENDER;
  	} else if (IS_IVYBRIDGE(dev)) {
  		u32 ecobus;
 
@@ -1183,11 +1053,21 @@ void intel_uncore_init(struct drm_device *dev)
  			dev_priv->uncore.funcs.force_wake_put =
  				__gen6_gt_force_wake_put;
  		}
+		dev_priv->uncore.fw_domains = FORCEWAKE_RENDER;
  	} else if (IS_GEN6(dev)) {
  		dev_priv->uncore.funcs.force_wake_get =
  			__gen6_gt_force_wake_get;
  		dev_priv->uncore.funcs.force_wake_put =
  			__gen6_gt_force_wake_put;
+		dev_priv->uncore.fw_domains = FORCEWAKE_RENDER;
+	}

drivers/gpu/drm/i915/intel_uncore.c void intel_uncore_init(struct drm_device *dev) · line 1073

Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +7 -0
+
+	for_each_fw_domain(domain, dev_priv, i) {
+		domain->i915 = dev_priv;
+		domain->id = i;
+
+		setup_timer(&domain->timer, gen6_force_wake_timer,
+			    (unsigned long)domain);
  	}
 
  	switch (INTEL_INFO(dev)->gen) {

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

#9 lore raw

Re: [PATCH 4/8] drm/i915: Reduce duplicated forcewake logic

Quoted 411 lines · Deepak S
Quoted 397 lines · Mika Kuoppala

From: Chris Wilson <[email protected]>

Introduce a structure to track the individual forcewake domains and use that to eliminate duplicate logic.

v2: - Rebase on latest dinq (Mika)
- for_each_fw_domain macro (Mika)
- Handle reset atomically, keeping the timer running (Mika)
- for_each_fw_domain parameter ordering (Chris)
- defer timer on new register access (Mika)

---
  drivers/gpu/drm/i915/i915_debugfs.c |  65 +++---
  drivers/gpu/drm/i915/i915_drv.h     |  54 +++--
  drivers/gpu/drm/i915/intel_uncore.c | 410 +++++++++++++-----------------------
  3 files changed, 208 insertions(+), 321 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +26 -39
index e142629..5cc838b 100644
@@ -1235,14 +1235,36 @@ static int ironlake_drpc_info(struct seq_file *m)
  	return 0;
  }
 
-static int vlv_drpc_info(struct seq_file *m)
+static int i915_gen6_forcewake_count_info(struct seq_file *m, void *data)
  {
+	struct drm_info_node *node = m->private;
+	struct drm_device *dev = node->minor->dev;
+	struct drm_i915_private *dev_priv = dev->dev_private;
+	struct intel_uncore_forcewake_domain *fw_domain;
+	const char *domain_names[] = {
+		"render",
+		"blitter",
+		"media",
+	};
+	int i;
+
+	spin_lock_irq(&dev_priv->uncore.lock);
+	for_each_fw_domain(fw_domain, dev_priv, i) {
+		seq_printf(m, "%s.wake_count = %u\n",
+			   domain_names[i],
+			   fw_domain->wake_count);
+	}
+	spin_unlock_irq(&dev_priv->uncore.lock);
 
+	return 0;
+}
+
+static int vlv_drpc_info(struct seq_file *m)
+{
  	struct drm_info_node *node = m->private;
  	struct drm_device *dev = node->minor->dev;
  	struct drm_i915_private *dev_priv = dev->dev_private;
  	u32 rpmodectl1, rcctl1, pw_status;
-	unsigned fw_rendercount = 0, fw_mediacount = 0;
 
  	intel_runtime_pm_get(dev_priv);
 
@@ -1274,22 +1296,11 @@ static int vlv_drpc_info(struct seq_file *m)
  	seq_printf(m, "Media RC6 residency since boot: %u\n",
  		   I915_READ(VLV_GT_MEDIA_RC6));
 
-	spin_lock_irq(&dev_priv->uncore.lock);
-	fw_rendercount = dev_priv->uncore.fw_rendercount;
-	fw_mediacount = dev_priv->uncore.fw_mediacount;
-	spin_unlock_irq(&dev_priv->uncore.lock);
-
-	seq_printf(m, "Forcewake Render Count = %u\n", fw_rendercount);
-	seq_printf(m, "Forcewake Media Count = %u\n", fw_mediacount);
-
-
-	return 0;
+	return i915_gen6_forcewake_count_info(m, NULL);
  }
 
-
  static int gen6_drpc_info(struct seq_file *m)
  {
-
  	struct drm_info_node *node = m->private;
  	struct drm_device *dev = node->minor->dev;
  	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -1303,7 +1314,7 @@ static int gen6_drpc_info(struct seq_file *m)
  	intel_runtime_pm_get(dev_priv);
 
  	spin_lock_irq(&dev_priv->uncore.lock);
-	forcewake_count = dev_priv->uncore.forcewake_count;
+	forcewake_count = dev_priv->uncore.fw_domain[FW_DOMAIN_ID_RENDER].wake_count;
  	spin_unlock_irq(&dev_priv->uncore.lock);
 
  	if (forcewake_count) {
@@ -1931,30 +1942,6 @@ static int i915_execlists(struct seq_file *m, void *data)
  	return 0;
  }
 
-static int i915_gen6_forcewake_count_info(struct seq_file *m, void *data)
-{
-	struct drm_info_node *node = m->private;
-	struct drm_device *dev = node->minor->dev;
-	struct drm_i915_private *dev_priv = dev->dev_private;
-	unsigned forcewake_count = 0, fw_rendercount = 0, fw_mediacount = 0;
-
-	spin_lock_irq(&dev_priv->uncore.lock);
-	if (IS_VALLEYVIEW(dev)) {
-		fw_rendercount = dev_priv->uncore.fw_rendercount;
-		fw_mediacount = dev_priv->uncore.fw_mediacount;
-	} else
-		forcewake_count = dev_priv->uncore.forcewake_count;
-	spin_unlock_irq(&dev_priv->uncore.lock);
-
-	if (IS_VALLEYVIEW(dev)) {
-		seq_printf(m, "fw_rendercount = %u\n", fw_rendercount);
-		seq_printf(m, "fw_mediacount = %u\n", fw_mediacount);
-	} else
-		seq_printf(m, "forcewake count = %u\n", forcewake_count);
-
-	return 0;
-}
-
  static const char *swizzle_string(unsigned swizzle)
  {
  	switch (swizzle) {
drivers/gpu/drm/i915/i915_drv.h +37 -17
index 95dfa2d..410558a 100644
@@ -554,20 +554,45 @@ struct intel_uncore_funcs {
  				uint64_t val, bool trace);
  };
 
+enum {
+	FW_DOMAIN_ID_RENDER = 0,
+	FW_DOMAIN_ID_BLITTER,
+	FW_DOMAIN_ID_MEDIA,
+
+	FW_DOMAIN_ID_COUNT
+};
+
  struct intel_uncore {
  	spinlock_t lock; /** lock is also taken in irq contexts. */
 
  	struct intel_uncore_funcs funcs;
 
  	unsigned fifo_count;
-	unsigned forcewake_count;
-
-	unsigned fw_rendercount;
-	unsigned fw_mediacount;
-	unsigned fw_blittercount;
-
-	struct timer_list force_wake_timer;
-};
+	unsigned fw_domains;
+
+	struct intel_uncore_forcewake_domain {
+		struct drm_i915_private *i915;
+		int id;
+		unsigned wake_count;
+		struct timer_list timer;
+	} fw_domain[FW_DOMAIN_ID_COUNT];
+#define FORCEWAKE_RENDER	(1 << FW_DOMAIN_ID_RENDER)
+#define FORCEWAKE_BLITTER	(1 << FW_DOMAIN_ID_BLITTER)
+#define FORCEWAKE_MEDIA		(1 << FW_DOMAIN_ID_MEDIA)
+#define FORCEWAKE_ALL		(FORCEWAKE_RENDER | \
+				 FORCEWAKE_BLITTER | \
+				 FORCEWAKE_MEDIA)
+};
+
+/* Iterate over initialised fw domains */
+#define for_each_fw_domain_mask(domain__, mask__, dev_priv__, i__) \
+	for ((i__) = 0, (domain__) = &(dev_priv__)->uncore.fw_domain[0]; \
+	     (i__) < FW_DOMAIN_ID_COUNT; \
+	     (i__)++, (domain__) = &(dev_priv__)->uncore.fw_domain[i__]) \
+		if (((mask__) & (dev_priv__)->uncore.fw_domains) & (1 << (i__)))
+
+#define for_each_fw_domain(domain__, dev_priv__, i__) \
+	for_each_fw_domain_mask(domain__, FORCEWAKE_ALL, dev_priv__, i__)
 
  #define DEV_INFO_FOR_EACH_FLAG(func, sep) \
  	func(is_mobile) sep \
@@ -2998,8 +3023,10 @@ extern void intel_display_print_error_state(struct drm_i915_error_state_buf *e,
   * must be set to prevent GT core from power down and stale values being
   * returned.
   */
-void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine);
-void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine);
+void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv,
+			    unsigned fw_domains);
+void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv,
+			    unsigned fw_domains);
  void assert_force_wake_inactive(struct drm_i915_private *dev_priv);
 
  int sandybridge_pcode_read(struct drm_i915_private *dev_priv, u32 mbox, u32 *val);
@@ -3031,13 +3058,6 @@ void vlv_flisdsi_write(struct drm_i915_private *dev_priv, u32 reg, u32 val);
  int vlv_gpu_freq(struct drm_i915_private *dev_priv, int val);
  int vlv_freq_opcode(struct drm_i915_private *dev_priv, int val);
 
-#define FORCEWAKE_RENDER	(1 << 0)
-#define FORCEWAKE_MEDIA		(1 << 1)
-#define FORCEWAKE_BLITTER	(1 << 2)
-#define FORCEWAKE_ALL		(FORCEWAKE_RENDER | FORCEWAKE_MEDIA | \
-					FORCEWAKE_BLITTER)
-
-
  #define I915_READ8(reg)		dev_priv->uncore.funcs.mmio_readb(dev_priv, (reg), true)
  #define I915_WRITE8(reg, val)	dev_priv->uncore.funcs.mmio_writeb(dev_priv, (reg), (val), true)
 
drivers/gpu/drm/i915/intel_uncore.c +18 -96
index 069fe7a..85e46b0 100644
@@ -67,7 +67,7 @@ static void __gen6_gt_force_wake_reset(struct drm_i915_private *dev_priv)
  }
 
  static void __gen6_gt_force_wake_get(struct drm_i915_private *dev_priv,
-							int fw_engine)
+				     int fw_engine)
  {
  	if (wait_for_atomic((__raw_i915_read32(dev_priv, FORCEWAKE_ACK) & 1) == 0,
  			    FORCEWAKE_ACK_TIMEOUT_MS))
@@ -93,7 +93,7 @@ static void __gen7_gt_force_wake_mt_reset(struct drm_i915_private *dev_priv)
  }
 
  static void __gen7_gt_force_wake_mt_get(struct drm_i915_private *dev_priv,
-							int fw_engine)
+					int fw_engine)
  {
  	u32 forcewake_ack;
 
@@ -129,7 +129,7 @@ static void gen6_gt_check_fifodbg(struct drm_i915_private *dev_priv)
  }
 
  static void __gen6_gt_force_wake_put(struct drm_i915_private *dev_priv,
-							int fw_engine)
+				     int fw_engine)
  {
  	__raw_i915_write32(dev_priv, FORCEWAKE, 0);
  	/* something from same cacheline, but !FORCEWAKE */
@@ -138,7 +138,7 @@ static void __gen6_gt_force_wake_put(struct drm_i915_private *dev_priv,
  }
 
  static void __gen7_gt_force_wake_mt_put(struct drm_i915_private *dev_priv,
-							int fw_engine)
+					int fw_engine)
  {
  	__raw_i915_write32(dev_priv, FORCEWAKE_MT,
  			   _MASKED_BIT_DISABLE(FORCEWAKE_KERNEL));
@@ -187,7 +187,7 @@ static void vlv_force_wake_reset(struct drm_i915_private *dev_priv)
  }
 
  static void __vlv_force_wake_get(struct drm_i915_private *dev_priv,
-						int fw_engine)
+				 int fw_engine)
  {
  	/* Check for Render Engine */
  	if (FORCEWAKE_RENDER & fw_engine) {
@@ -227,9 +227,8 @@ static void __vlv_force_wake_get(struct drm_i915_private *dev_priv,
  }
 
  static void __vlv_force_wake_put(struct drm_i915_private *dev_priv,
-					int fw_engine)
+				 int fw_engine)
  {
-
  	/* Check for Render Engine */
  	if (FORCEWAKE_RENDER & fw_engine)
  		__raw_i915_write32(dev_priv, FORCEWAKE_VLV,
@@ -247,37 +246,6 @@ static void __vlv_force_wake_put(struct drm_i915_private *dev_priv,
  		gen6_gt_check_fifodbg(dev_priv);
  }
 
-static void vlv_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
-{
-	if (fw_engine & FORCEWAKE_RENDER &&
-	    dev_priv->uncore.fw_rendercount++ != 0)
-		fw_engine &= ~FORCEWAKE_RENDER;
-	if (fw_engine & FORCEWAKE_MEDIA &&
-	    dev_priv->uncore.fw_mediacount++ != 0)
-		fw_engine &= ~FORCEWAKE_MEDIA;
-
-	if (fw_engine)
-		dev_priv->uncore.funcs.force_wake_get(dev_priv, fw_engine);
-}
-
-static void vlv_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
-{
-	if (fw_engine & FORCEWAKE_RENDER) {
-		WARN_ON(!dev_priv->uncore.fw_rendercount);
-		if (--dev_priv->uncore.fw_rendercount != 0)
-			fw_engine &= ~FORCEWAKE_RENDER;
-	}
-
-	if (fw_engine & FORCEWAKE_MEDIA) {
-		WARN_ON(!dev_priv->uncore.fw_mediacount);
-		if (--dev_priv->uncore.fw_mediacount != 0)
-			fw_engine &= ~FORCEWAKE_MEDIA;
-	}
-
-	if (fw_engine)
-		dev_priv->uncore.funcs.force_wake_put(dev_priv, fw_engine);
-}
-
  static void __gen9_gt_force_wake_mt_reset(struct drm_i915_private *dev_priv)
  {
  	__raw_i915_write32(dev_priv, FORCEWAKE_RENDER_GEN9,
@@ -367,80 +335,40 @@ __gen9_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
  				_MASKED_BIT_DISABLE(FORCEWAKE_KERNEL));
  }
 
-static void
-gen9_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
-{
-	if (FORCEWAKE_RENDER & fw_engine) {
-		if (dev_priv->uncore.fw_rendercount++ == 0)
-			dev_priv->uncore.funcs.force_wake_get(dev_priv,
-							FORCEWAKE_RENDER);
-	}
-
-	if (FORCEWAKE_MEDIA & fw_engine) {
-		if (dev_priv->uncore.fw_mediacount++ == 0)
-			dev_priv->uncore.funcs.force_wake_get(dev_priv,
-							FORCEWAKE_MEDIA);
-	}
-
-	if (FORCEWAKE_BLITTER & fw_engine) {
-		if (dev_priv->uncore.fw_blittercount++ == 0)
-			dev_priv->uncore.funcs.force_wake_get(dev_priv,
-							FORCEWAKE_BLITTER);
-	}
-}
-
-static void
-gen9_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
-{
-	if (FORCEWAKE_RENDER & fw_engine) {
-		WARN_ON(dev_priv->uncore.fw_rendercount == 0);
-		if (--dev_priv->uncore.fw_rendercount == 0)
-			dev_priv->uncore.funcs.force_wake_put(dev_priv,
-							FORCEWAKE_RENDER);
-	}
-
-	if (FORCEWAKE_MEDIA & fw_engine) {
-		WARN_ON(dev_priv->uncore.fw_mediacount == 0);
-		if (--dev_priv->uncore.fw_mediacount == 0)
-			dev_priv->uncore.funcs.force_wake_put(dev_priv,
-							FORCEWAKE_MEDIA);
-	}
-
-	if (FORCEWAKE_BLITTER & fw_engine) {
-		WARN_ON(dev_priv->uncore.fw_blittercount == 0);
-		if (--dev_priv->uncore.fw_blittercount == 0)
-			dev_priv->uncore.funcs.force_wake_put(dev_priv,
-							FORCEWAKE_BLITTER);
-	}
-}
-
  static void gen6_force_wake_timer(unsigned long arg)
  {
-	struct drm_i915_private *dev_priv = (void *)arg;
+	struct intel_uncore_forcewake_domain *domain = (void *)arg;
  	unsigned long irqflags;
 
-	assert_device_not_suspended(dev_priv);
+	assert_device_not_suspended(domain->i915);
 
-	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
-	WARN_ON(!dev_priv->uncore.forcewake_count);
+	spin_lock_irqsave(&domain->i915->uncore.lock, irqflags);
+	WARN_ON(!domain->wake_count);
 
-	if (--dev_priv->uncore.forcewake_count == 0)
-		dev_priv->uncore.funcs.force_wake_put(dev_priv, FORCEWAKE_ALL);
-	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
+	if (--domain->wake_count == 0)
+		domain->i915->uncore.funcs.force_wake_put(domain->i915,
+							  1 << domain->id);
+	spin_unlock_irqrestore(&domain->i915->uncore.lock, irqflags);
  }
 
  void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore)
  {
  	struct drm_i915_private *dev_priv = dev->dev_private;
-	unsigned long irqflags;
-
-	if (del_timer_sync(&dev_priv->uncore.force_wake_timer))
-		gen6_force_wake_timer((unsigned long)dev_priv);
+	unsigned long irqflags, fw = 0;
+	struct intel_uncore_forcewake_domain *domain;
+	int id;
 
+	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
  	/* Hold uncore.lock across reset to prevent any register access
  	 * with forcewake not set correctly

drivers/gpu/drm/i915/intel_uncore.c __gen9_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine) · line 372

Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +7 -1
  	 */
-	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
+
+	for_each_fw_domain(domain, dev_priv, id)
+		if (domain->wake_count)
+			fw |= 1 << id;
+
+	if (fw)
+		dev_priv->uncore.funcs.force_wake_put(dev_priv, fw);
 

Hi Mika,

Quoted 8 lines

Why do put?

Is there a possibility of getting get & put count mismatch? for example,
1. We Read the register

    -> forcewake_get & start domain_timer. (count = 1)

2. Now we call "intel_uncore_forcewake_reset" which does a put. (count = 0)
3. now the domain_timer from step #1 comes (count -1)? right? or i am missing something here?

1. We read the register, wake_count = 1 and domain timer is armed.
2. In reset we see which domains are awake and store them in fw
3. We do a put for all domains with the hw callback,
this doesn't touch the wake_count. I wanted this so that we first try with the soft approach by doing the disable and all the related fifodbg dance here like this was a normal put.
4. Regardless if the above did the trick, we then we take the gloves off and use more blunt approach by resetting the register.
5. We restore all the forcewakes

But while I was reading this through, there is one other problem which is that if we don't restore, we mess the counts and the timers will be still running.

I posted a new series with a fix for this.

-Mika

Quoted 441 lines

Others Looks fine to me.

Quoted 38 lines
drivers/gpu/drm/i915/intel_uncore.c +2 -23
  	if (IS_VALLEYVIEW(dev))
  		vlv_force_wake_reset(dev_priv);
@@ -454,28 +382,6 @@ void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore)
  		__gen9_gt_force_wake_mt_reset(dev_priv);
 
  	if (restore) { /* If reset with a user forcewake, try to restore */
-		unsigned fw = 0;
-
-		if (IS_VALLEYVIEW(dev)) {
-			if (dev_priv->uncore.fw_rendercount)
-				fw |= FORCEWAKE_RENDER;
-
-			if (dev_priv->uncore.fw_mediacount)
-				fw |= FORCEWAKE_MEDIA;
-		} else if (IS_GEN9(dev)) {
-			if (dev_priv->uncore.fw_rendercount)
-				fw |= FORCEWAKE_RENDER;
-
-			if (dev_priv->uncore.fw_mediacount)
-				fw |= FORCEWAKE_MEDIA;
-
-			if (dev_priv->uncore.fw_blittercount)
-				fw |= FORCEWAKE_BLITTER;
-		} else {
-			if (dev_priv->uncore.forcewake_count)
-				fw = FORCEWAKE_ALL;
-		}
-
  		if (fw)
  			dev_priv->uncore.funcs.force_wake_get(dev_priv, fw);
 
@@ -533,28 +439,28 @@ void intel_uncore_sanitize(struct drm_device *dev)
   * be called at the beginning of the sequence followed by a call to
   * gen6_gt_force_wake_put() at the end of the sequence.
   */
-void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
+void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv,
+			    unsigned fw_domains)

drivers/gpu/drm/i915/intel_uncore.c void intel_uncore_sanitize(struct drm_device *dev) · line 544

Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +2 -2
  {
  	unsigned long irqflags;
+	struct intel_uncore_forcewake_domain *domain;
+	int id;
 
  	if (!dev_priv->uncore.funcs.force_wake_get)
  		return;
 
-	intel_runtime_pm_get(dev_priv);
-

you need to change this in Patch #2

Quoted 385 lines
drivers/gpu/drm/i915/intel_uncore.c +116 -143
  	WARN_ON(!pm_runtime_active(&dev_priv->dev->pdev->dev));
 
+	fw_domains &= dev_priv->uncore.fw_domains;
+
  	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
 
-	if (IS_GEN9(dev_priv->dev)) {
-		gen9_force_wake_get(dev_priv, fw_engine);
-	} else if (IS_VALLEYVIEW(dev_priv->dev)) {
-		vlv_force_wake_get(dev_priv, fw_engine);
-	} else {
-		if (dev_priv->uncore.forcewake_count++ == 0)
-			dev_priv->uncore.funcs.force_wake_get(dev_priv,
-							      FORCEWAKE_ALL);
-	}
+	for_each_fw_domain_mask(domain, fw_domains, dev_priv, id)
+		if (domain->wake_count++)
+			fw_domains &= ~(1 << id);
+
+	if (fw_domains)
+		dev_priv->uncore.funcs.force_wake_get(dev_priv, fw_domains);
 
  	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
  }
@@ -562,26 +468,27 @@ void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
  /*
   * see gen6_gt_force_wake_get()
   */
-void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
+void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv,
+			    unsigned fw_domains)
  {
  	unsigned long irqflags;
+	struct intel_uncore_forcewake_domain *domain;
+	int id;
 
  	if (!dev_priv->uncore.funcs.force_wake_put)
  		return;
 
+	fw_domains &= dev_priv->uncore.fw_domains;
+
  	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
 
-	if (IS_GEN9(dev_priv->dev)) {
-		gen9_force_wake_put(dev_priv, fw_engine);
-	} else if (IS_VALLEYVIEW(dev_priv->dev)) {
-		vlv_force_wake_put(dev_priv, fw_engine);
-	} else {
-		WARN_ON(!dev_priv->uncore.forcewake_count);
-		if (--dev_priv->uncore.forcewake_count == 0) {
-			dev_priv->uncore.forcewake_count++;
-			mod_timer_pinned(&dev_priv->uncore.force_wake_timer,
-					 jiffies + 1);
-		}
+	for_each_fw_domain_mask(domain, fw_domains, dev_priv, id) {
+		WARN_ON(!domain->wake_count);
+		if (--domain->wake_count)
+			continue;
+
+		domain->wake_count++;
+		mod_timer_pinned(&domain->timer, jiffies + 1);
  	}
 
  	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
@@ -589,10 +496,14 @@ void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
 
  void assert_force_wake_inactive(struct drm_i915_private *dev_priv)
  {
+	struct intel_uncore_forcewake_domain *domain;
+	int i;
+
  	if (!dev_priv->uncore.funcs.force_wake_get)
  		return;
 
-	WARN_ON(dev_priv->uncore.forcewake_count > 0);
+	for_each_fw_domain(domain, dev_priv, i)
+		WARN_ON(domain->wake_count);
  }
 
  /* We give fast paths for the really cool registers */
@@ -758,19 +669,36 @@ __gen2_read(64)
  	trace_i915_reg_rw(false, reg, val, sizeof(val), trace); \
  	return val
 
+static inline void __force_wake_get(struct drm_i915_private *dev_priv,
+				    unsigned fw_domains)
+{
+	struct intel_uncore_forcewake_domain *domain;
+	int i;
+
+	if (WARN_ON(!fw_domains))
+		return;
+
+	/* Ideally GCC would be constant-fold and eliminate this loop */
+	for_each_fw_domain_mask(domain, fw_domains, dev_priv, i) {
+		if (domain->wake_count)
+			fw_domains &= ~(1 << i);
+		else
+			domain->wake_count++;
+
+		mod_timer_pinned(&domain->timer, jiffies + 1);
+	}
+
+	if (fw_domains)
+		dev_priv->uncore.funcs.force_wake_get(dev_priv, fw_domains);
+}
+
  #define __gen6_read(x) \
  static u##x \
  gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
  	GEN6_READ_HEADER(x); \
  	hsw_unclaimed_reg_debug(dev_priv, reg, true, true); \
-	if (dev_priv->uncore.forcewake_count == 0 && \
-	    NEEDS_FORCE_WAKE((dev_priv), (reg))) { \
-		dev_priv->uncore.funcs.force_wake_get(dev_priv, \
-						      FORCEWAKE_ALL); \
-		dev_priv->uncore.forcewake_count++; \
-		mod_timer_pinned(&dev_priv->uncore.force_wake_timer, \
-				 jiffies + 1); \
-	} \
+	if (NEEDS_FORCE_WAKE((dev_priv), (reg))) \
+		__force_wake_get(dev_priv, FORCEWAKE_RENDER); \
  	val = __raw_i915_read##x(dev_priv, reg); \
  	hsw_unclaimed_reg_debug(dev_priv, reg, true, false); \
  	GEN6_READ_FOOTER; \
@@ -779,45 +707,27 @@ gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
  #define __vlv_read(x) \
  static u##x \
  vlv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
-	unsigned fwengine = 0; \
  	GEN6_READ_HEADER(x); \
-	if (FORCEWAKE_VLV_RENDER_RANGE_OFFSET(reg)) { \
-		if (dev_priv->uncore.fw_rendercount == 0) \
-			fwengine = FORCEWAKE_RENDER; \
-	} else if (FORCEWAKE_VLV_MEDIA_RANGE_OFFSET(reg)) { \
-		if (dev_priv->uncore.fw_mediacount == 0) \
-			fwengine = FORCEWAKE_MEDIA; \
-	}  \
-	if (fwengine) \
-		dev_priv->uncore.funcs.force_wake_get(dev_priv, fwengine); \
+	if (FORCEWAKE_VLV_RENDER_RANGE_OFFSET(reg)) \
+		__force_wake_get(dev_priv, FORCEWAKE_RENDER); \
+	else if (FORCEWAKE_VLV_MEDIA_RANGE_OFFSET(reg)) \
+		__force_wake_get(dev_priv, FORCEWAKE_MEDIA); \
  	val = __raw_i915_read##x(dev_priv, reg); \
-	if (fwengine) \
-		dev_priv->uncore.funcs.force_wake_put(dev_priv, fwengine); \
  	GEN6_READ_FOOTER; \
  }
 
  #define __chv_read(x) \
  static u##x \
  chv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
-	unsigned fwengine = 0; \
  	GEN6_READ_HEADER(x); \
-	if (FORCEWAKE_CHV_RENDER_RANGE_OFFSET(reg)) { \
-		if (dev_priv->uncore.fw_rendercount == 0) \
-			fwengine = FORCEWAKE_RENDER; \
-	} else if (FORCEWAKE_CHV_MEDIA_RANGE_OFFSET(reg)) { \
-		if (dev_priv->uncore.fw_mediacount == 0) \
-			fwengine = FORCEWAKE_MEDIA; \
-	} else if (FORCEWAKE_CHV_COMMON_RANGE_OFFSET(reg)) { \
-		if (dev_priv->uncore.fw_rendercount == 0) \
-			fwengine |= FORCEWAKE_RENDER; \
-		if (dev_priv->uncore.fw_mediacount == 0) \
-			fwengine |= FORCEWAKE_MEDIA; \
-	} \
-	if (fwengine) \
-		dev_priv->uncore.funcs.force_wake_get(dev_priv, fwengine); \
+	if (FORCEWAKE_CHV_RENDER_RANGE_OFFSET(reg)) \
+		__force_wake_get(dev_priv, FORCEWAKE_RENDER); \
+	else if (FORCEWAKE_CHV_MEDIA_RANGE_OFFSET(reg)) \
+		__force_wake_get(dev_priv, FORCEWAKE_MEDIA); \
+	else if (FORCEWAKE_CHV_COMMON_RANGE_OFFSET(reg)) \
+		__force_wake_get(dev_priv, \
+				 FORCEWAKE_RENDER | FORCEWAKE_MEDIA); \
  	val = __raw_i915_read##x(dev_priv, reg); \
-	if (fwengine) \
-		dev_priv->uncore.funcs.force_wake_put(dev_priv, fwengine); \
  	GEN6_READ_FOOTER; \
  }
 
@@ -827,32 +737,21 @@ chv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
  #define __gen9_read(x) \
  static u##x \
  gen9_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
+	unsigned fw_engine; \
  	GEN6_READ_HEADER(x); \
-	if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg))) { \
-		val = __raw_i915_read##x(dev_priv, reg); \
-	} else { \
-		unsigned fwengine = 0; \
-		if (FORCEWAKE_GEN9_RENDER_RANGE_OFFSET(reg)) { \
-			if (dev_priv->uncore.fw_rendercount == 0) \
-				fwengine = FORCEWAKE_RENDER; \
-		} else if (FORCEWAKE_GEN9_MEDIA_RANGE_OFFSET(reg)) { \
-			if (dev_priv->uncore.fw_mediacount == 0) \
-				fwengine = FORCEWAKE_MEDIA; \
-		} else if (FORCEWAKE_GEN9_COMMON_RANGE_OFFSET(reg)) { \
-			if (dev_priv->uncore.fw_rendercount == 0) \
-				fwengine |= FORCEWAKE_RENDER; \
-			if (dev_priv->uncore.fw_mediacount == 0) \
-				fwengine |= FORCEWAKE_MEDIA; \
-		} else { \
-			if (dev_priv->uncore.fw_blittercount == 0) \
-				fwengine = FORCEWAKE_BLITTER; \
-		} \
-		if (fwengine) \
-			dev_priv->uncore.funcs.force_wake_get(dev_priv, fwengine); \
-		val = __raw_i915_read##x(dev_priv, reg); \
-		if (fwengine) \
-			dev_priv->uncore.funcs.force_wake_put(dev_priv, fwengine); \
-	} \
+	if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg)))	\
+		fw_engine = 0; \
+	else if (FORCEWAKE_GEN9_RENDER_RANGE_OFFSET(reg))	\
+		fw_engine = FORCEWAKE_RENDER; \
+	else if (FORCEWAKE_GEN9_MEDIA_RANGE_OFFSET(reg)) \
+		fw_engine = FORCEWAKE_MEDIA; \
+	else if (FORCEWAKE_GEN9_COMMON_RANGE_OFFSET(reg)) \
+		fw_engine = FORCEWAKE_RENDER | FORCEWAKE_MEDIA; \
+	else \
+		fw_engine = FORCEWAKE_BLITTER; \
+	if (fw_engine) \
+		__force_wake_get(dev_priv, fw_engine); \
+	val = __raw_i915_read##x(dev_priv, reg); \
  	GEN6_READ_FOOTER; \
  }
 
@@ -986,17 +885,9 @@ static void \
  gen8_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) { \
  	GEN6_WRITE_HEADER; \
  	hsw_unclaimed_reg_debug(dev_priv, reg, false, true); \
-	if (reg < 0x40000 && !is_gen8_shadowed(dev_priv, reg)) { \
-		if (dev_priv->uncore.forcewake_count == 0) \
-			dev_priv->uncore.funcs.force_wake_get(dev_priv,	\
-							      FORCEWAKE_ALL); \
-		__raw_i915_write##x(dev_priv, reg, val); \
-		if (dev_priv->uncore.forcewake_count == 0) \
-			dev_priv->uncore.funcs.force_wake_put(dev_priv, \
-							      FORCEWAKE_ALL); \
-	} else { \
-		__raw_i915_write##x(dev_priv, reg, val); \
-	} \
+	if (reg < 0x40000 && !is_gen8_shadowed(dev_priv, reg)) \
+		__force_wake_get(dev_priv, FORCEWAKE_RENDER); \
+	__raw_i915_write##x(dev_priv, reg, val); \
  	hsw_unclaimed_reg_debug(dev_priv, reg, false, false); \
  	hsw_unclaimed_reg_detect(dev_priv); \
  	GEN6_WRITE_FOOTER; \
@@ -1005,28 +896,17 @@ gen8_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace
  #define __chv_write(x) \
  static void \
  chv_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, bool trace) { \
-	unsigned fwengine = 0; \
  	bool shadowed = is_gen8_shadowed(dev_priv, reg); \
  	GEN6_WRITE_HEADER; \
  	if (!shadowed) { \
-		if (FORCEWAKE_CHV_RENDER_RANGE_OFFSET(reg)) { \
-			if (dev_priv->uncore.fw_rendercount == 0) \
-				fwengine = FORCEWAKE_RENDER; \
-		} else if (FORCEWAKE_CHV_MEDIA_RANGE_OFFSET(reg)) { \
-			if (dev_priv->uncore.fw_mediacount == 0) \
-				fwengine = FORCEWAKE_MEDIA; \
-		} else if (FORCEWAKE_CHV_COMMON_RANGE_OFFSET(reg)) { \
-			if (dev_priv->uncore.fw_rendercount == 0) \
-				fwengine |= FORCEWAKE_RENDER; \
-			if (dev_priv->uncore.fw_mediacount == 0) \
-				fwengine |= FORCEWAKE_MEDIA; \
-		} \
+		if (FORCEWAKE_CHV_RENDER_RANGE_OFFSET(reg)) \
+			__force_wake_get(dev_priv, FORCEWAKE_RENDER); \
+		else if (FORCEWAKE_CHV_MEDIA_RANGE_OFFSET(reg)) \
+			__force_wake_get(dev_priv, FORCEWAKE_MEDIA); \
+		else if (FORCEWAKE_CHV_COMMON_RANGE_OFFSET(reg)) \
+			__force_wake_get(dev_priv, FORCEWAKE_RENDER | FORCEWAKE_MEDIA); \
  	} \
-	if (fwengine) \
-		dev_priv->uncore.funcs.force_wake_get(dev_priv, fwengine); \
  	__raw_i915_write##x(dev_priv, reg, val); \
-	if (fwengine) \
-		dev_priv->uncore.funcs.force_wake_put(dev_priv, fwengine); \
  	GEN6_WRITE_FOOTER; \
  }
 
@@ -1057,35 +937,22 @@ static bool is_gen9_shadowed(struct drm_i915_private *dev_priv, u32 reg)
  static void \
  gen9_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, \
  		bool trace) { \
+	unsigned fw_engine; \
  	GEN6_WRITE_HEADER; \
-	if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg)) || \
-			is_gen9_shadowed(dev_priv, reg)) { \
-		__raw_i915_write##x(dev_priv, reg, val); \
-	} else { \
-		unsigned fwengine = 0; \
-		if (FORCEWAKE_GEN9_RENDER_RANGE_OFFSET(reg)) { \
-			if (dev_priv->uncore.fw_rendercount == 0) \
-				fwengine = FORCEWAKE_RENDER; \
-		} else if (FORCEWAKE_GEN9_MEDIA_RANGE_OFFSET(reg)) { \
-			if (dev_priv->uncore.fw_mediacount == 0) \
-				fwengine = FORCEWAKE_MEDIA; \
-		} else if (FORCEWAKE_GEN9_COMMON_RANGE_OFFSET(reg)) { \
-			if (dev_priv->uncore.fw_rendercount == 0) \
-				fwengine |= FORCEWAKE_RENDER; \
-			if (dev_priv->uncore.fw_mediacount == 0) \
-				fwengine |= FORCEWAKE_MEDIA; \
-		} else { \
-			if (dev_priv->uncore.fw_blittercount == 0) \
-				fwengine = FORCEWAKE_BLITTER; \
-		} \
-		if (fwengine) \
-			dev_priv->uncore.funcs.force_wake_get(dev_priv, \
-					fwengine); \
-		__raw_i915_write##x(dev_priv, reg, val); \
-		if (fwengine) \
-			dev_priv->uncore.funcs.force_wake_put(dev_priv, \
-					fwengine); \
-	} \
+	if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg)) ||	\
+	    is_gen9_shadowed(dev_priv, reg)) \
+		fw_engine = 0; \
+	else if (FORCEWAKE_GEN9_RENDER_RANGE_OFFSET(reg)) \
+		fw_engine = FORCEWAKE_RENDER; \
+	else if (FORCEWAKE_GEN9_MEDIA_RANGE_OFFSET(reg)) \
+		fw_engine = FORCEWAKE_MEDIA; \
+	else if (FORCEWAKE_GEN9_COMMON_RANGE_OFFSET(reg)) \
+		fw_engine = FORCEWAKE_RENDER | FORCEWAKE_MEDIA; \
+	else \
+		fw_engine = FORCEWAKE_BLITTER; \
+	if (fw_engine) \
+		__force_wake_get(dev_priv, fw_engine); \
+	__raw_i915_write##x(dev_priv, reg, val); \
  	GEN6_WRITE_FOOTER; \
  }
 
@@ -1137,21 +1004,24 @@ do { \
  void intel_uncore_init(struct drm_device *dev)
  {
  	struct drm_i915_private *dev_priv = dev->dev_private;
-
-	setup_timer(&dev_priv->uncore.force_wake_timer,
-		    gen6_force_wake_timer, (unsigned long)dev_priv);
+	struct intel_uncore_forcewake_domain *domain;
+	int i;
 
  	__intel_uncore_early_sanitize(dev, false);
 
  	if (IS_GEN9(dev)) {
  		dev_priv->uncore.funcs.force_wake_get = __gen9_force_wake_get;
  		dev_priv->uncore.funcs.force_wake_put = __gen9_force_wake_put;
+		dev_priv->uncore.fw_domains = FORCEWAKE_RENDER |
+			FORCEWAKE_BLITTER | FORCEWAKE_MEDIA;
  	} else if (IS_VALLEYVIEW(dev)) {
  		dev_priv->uncore.funcs.force_wake_get = __vlv_force_wake_get;
  		dev_priv->uncore.funcs.force_wake_put = __vlv_force_wake_put;
+		dev_priv->uncore.fw_domains = FORCEWAKE_RENDER | FORCEWAKE_MEDIA;
  	} else if (IS_HASWELL(dev) || IS_BROADWELL(dev)) {
  		dev_priv->uncore.funcs.force_wake_get = __gen7_gt_force_wake_mt_get;
  		dev_priv->uncore.funcs.force_wake_put = __gen7_gt_force_wake_mt_put;
+		dev_priv->uncore.fw_domains = FORCEWAKE_RENDER;
  	} else if (IS_IVYBRIDGE(dev)) {
  		u32 ecobus;
 
@@ -1183,11 +1053,21 @@ void intel_uncore_init(struct drm_device *dev)
  			dev_priv->uncore.funcs.force_wake_put =
  				__gen6_gt_force_wake_put;
  		}
+		dev_priv->uncore.fw_domains = FORCEWAKE_RENDER;
  	} else if (IS_GEN6(dev)) {
  		dev_priv->uncore.funcs.force_wake_get =
  			__gen6_gt_force_wake_get;
  		dev_priv->uncore.funcs.force_wake_put =
  			__gen6_gt_force_wake_put;
+		dev_priv->uncore.fw_domains = FORCEWAKE_RENDER;
+	}
+
+	for_each_fw_domain(domain, dev_priv, i) {
+		domain->i915 = dev_priv;
+		domain->id = i;
+
+		setup_timer(&domain->timer, gen6_force_wake_timer,
+			    (unsigned long)domain);
  	}
 
Quoted 7 lines

drivers/gpu/drm/i915/intel_uncore.c void intel_uncore_init(struct drm_device *dev) · line 1073

Quoted 1 line
drivers/gpu/drm/i915/intel_uncore.c +0 -0
  	switch (INTEL_INFO(dev)->gen) {

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

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

#10 lore raw

PATCH 5/8 drm/i915: Consolidate forcewake code

In reply to Mika Kuoppala

As we now have forcewake domains, take advantage of it by putting the differences in gen fw handling in data rather than in code.

In past we have opencoded this quite extensively as the fw handling is in the fast path. There has also been a lot of cargo-culted copy'n'pasting from older gens to newer ones.

Now when the releasing of the forcewake is done by deferred timer, it gives chance to consolidate more. Due to the frequency of actual hw access being significantly less.

Take advantage of this and generalize the fw handling code as much as possible. But we still aim to keep the forcewake sequence particularities for each gen intact. So the access pattern to fw engines should remain the same.

v2: - s/old_ack/clear_ack (Chris)
- s/post_read/posting_read (Chris)
- less polite commit msg (Chris)

---
 drivers/gpu/drm/i915/i915_debugfs.c |   7 +-
 drivers/gpu/drm/i915/i915_drv.h     |   7 +
 drivers/gpu/drm/i915/intel_uncore.c | 469 ++++++++++++++++--------------------
 3 files changed, 212 insertions(+), 271 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +1 -6
index 5cc838b..93390c9 100644
@@ -1241,17 +1241,12 @@ static int i915_gen6_forcewake_count_info(struct seq_file *m, void *data)
 	struct drm_device *dev = node->minor->dev;
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_uncore_forcewake_domain *fw_domain;
-	const char *domain_names[] = {
-		"render",
-		"blitter",
-		"media",
-	};
 	int i;
 
 	spin_lock_irq(&dev_priv->uncore.lock);
 	for_each_fw_domain(fw_domain, dev_priv, i) {
 		seq_printf(m, "%s.wake_count = %u\n",
-			   domain_names[i],
+			   intel_uncore_forcewake_domain_to_str(i),
 			   fw_domain->wake_count);
 	}
 	spin_unlock_irq(&dev_priv->uncore.lock);
drivers/gpu/drm/i915/i915_drv.h +7 -0
index 410558a..4263084 100644
@@ -575,6 +575,12 @@ struct intel_uncore {
 		int id;
 		unsigned wake_count;
 		struct timer_list timer;
+		u32 reg_set;
+		u32 val_set;
+		u32 val_clear;
+		u32 reg_ack;
+		u32 reg_post;
+		u32 val_reset;
 	} fw_domain[FW_DOMAIN_ID_COUNT];
 #define FORCEWAKE_RENDER	(1 << FW_DOMAIN_ID_RENDER)
 #define FORCEWAKE_BLITTER	(1 << FW_DOMAIN_ID_BLITTER)
@@ -2443,6 +2449,7 @@ extern void intel_uncore_init(struct drm_device *dev);
 extern void intel_uncore_check_errors(struct drm_device *dev);
 extern void intel_uncore_fini(struct drm_device *dev);
 extern void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore);
+const char *intel_uncore_forcewake_domain_to_str(const int domain_id);
 
 void
 i915_enable_pipestat(struct drm_i915_private *dev_priv, enum pipe pipe,
drivers/gpu/drm/i915/intel_uncore.c +204 -265

608 lines not on this page · raw message

Signature · 6 lines
1.9.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Reviewed-by on v1 5/8
#11 lore raw

Re: [PATCH 5/8] drm/i915: Consolidate forcewake code

Quoted 689 lines · Mika Kuoppala

As we now have forcewake domains, take advantage of it by putting the differences in gen fw handling in data rather than in code.

In past we have opencoded this quite extensively as the fw handling is in the fast path. There has also been a lot of cargo-culted copy'n'pasting from older gens to newer ones.

Now when the releasing of the forcewake is done by deferred timer, it gives chance to consolidate more. Due to the frequency of actual hw access being significantly less.

Take advantage of this and generalize the fw handling code as much as possible. But we still aim to keep the forcewake sequence particularities for each gen intact. So the access pattern to fw engines should remain the same.

v2: - s/old_ack/clear_ack (Chris)
- s/post_read/posting_read (Chris)
- less polite commit msg (Chris)

---
  drivers/gpu/drm/i915/i915_debugfs.c |   7 +-
  drivers/gpu/drm/i915/i915_drv.h     |   7 +
  drivers/gpu/drm/i915/intel_uncore.c | 469 ++++++++++++++++--------------------
  3 files changed, 212 insertions(+), 271 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +1 -6
index 5cc838b..93390c9 100644
@@ -1241,17 +1241,12 @@ static int i915_gen6_forcewake_count_info(struct seq_file *m, void *data)
  	struct drm_device *dev = node->minor->dev;
  	struct drm_i915_private *dev_priv = dev->dev_private;
  	struct intel_uncore_forcewake_domain *fw_domain;
-	const char *domain_names[] = {
-		"render",
-		"blitter",
-		"media",
-	};
  	int i;
 
  	spin_lock_irq(&dev_priv->uncore.lock);
  	for_each_fw_domain(fw_domain, dev_priv, i) {
  		seq_printf(m, "%s.wake_count = %u\n",
-			   domain_names[i],
+			   intel_uncore_forcewake_domain_to_str(i),
  			   fw_domain->wake_count);
  	}
  	spin_unlock_irq(&dev_priv->uncore.lock);
drivers/gpu/drm/i915/i915_drv.h +7 -0
index 410558a..4263084 100644
@@ -575,6 +575,12 @@ struct intel_uncore {
  		int id;
  		unsigned wake_count;
  		struct timer_list timer;
+		u32 reg_set;
+		u32 val_set;
+		u32 val_clear;
+		u32 reg_ack;
+		u32 reg_post;
+		u32 val_reset;
  	} fw_domain[FW_DOMAIN_ID_COUNT];
  #define FORCEWAKE_RENDER	(1 << FW_DOMAIN_ID_RENDER)
  #define FORCEWAKE_BLITTER	(1 << FW_DOMAIN_ID_BLITTER)
@@ -2443,6 +2449,7 @@ extern void intel_uncore_init(struct drm_device *dev);
  extern void intel_uncore_check_errors(struct drm_device *dev);
  extern void intel_uncore_fini(struct drm_device *dev);
  extern void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore);
+const char *intel_uncore_forcewake_domain_to_str(const int domain_id);
 
  void
  i915_enable_pipestat(struct drm_i915_private *dev_priv, enum pipe pipe,
drivers/gpu/drm/i915/intel_uncore.c +201 -261
index 85e46b0..e883790 100644
@@ -42,6 +42,26 @@
 
  #define __raw_posting_read(dev_priv__, reg__) (void)__raw_i915_read32(dev_priv__, reg__)
 
+static const char * const forcewake_domain_names[] = {
+	"render",
+	"blitter",
+	"media",
+};
+
+const char *
+intel_uncore_forcewake_domain_to_str(const int id)
+{
+	BUILD_BUG_ON((sizeof(forcewake_domain_names)/sizeof(const char *)) !=
+		     FW_DOMAIN_ID_COUNT);
+
+	if (id >= 0 && id < FW_DOMAIN_ID_COUNT)
+		return forcewake_domain_names[id];
+
+	WARN_ON(id);
+
+	return "unknown";
+}
+
  static void
  assert_device_not_suspended(struct drm_i915_private *dev_priv)
  {
@@ -49,73 +69,123 @@ assert_device_not_suspended(struct drm_i915_private *dev_priv)
  		  "Device suspended\n");
  }
 
-static void __gen6_gt_wait_for_thread_c0(struct drm_i915_private *dev_priv)
+static inline void
+fw_domain_reset(const struct intel_uncore_forcewake_domain *d)
  {
-	/* w/a for a sporadic read returning 0 by waiting for the GT
-	 * thread to wake up.
-	 */
-	if (wait_for_atomic_us((__raw_i915_read32(dev_priv, GEN6_GT_THREAD_STATUS_REG) &
-				GEN6_GT_THREAD_STATUS_CORE_MASK) == 0, 500))
-		DRM_ERROR("GT thread status wait timed out\n");
+	__raw_i915_write32(d->i915, d->reg_set, d->val_reset);
  }
 
-static void __gen6_gt_force_wake_reset(struct drm_i915_private *dev_priv)
+static inline void
+fw_domain_arm_timer(struct intel_uncore_forcewake_domain *d)
  {
-	__raw_i915_write32(dev_priv, FORCEWAKE, 0);
-	/* something from same cacheline, but !FORCEWAKE */
-	__raw_posting_read(dev_priv, ECOBUS);
+	mod_timer_pinned(&d->timer, jiffies + 1);
  }
 
-static void __gen6_gt_force_wake_get(struct drm_i915_private *dev_priv,
-				     int fw_engine)
+static inline void
+fw_domain_wait_ack_clear(const struct intel_uncore_forcewake_domain *d)
  {
-	if (wait_for_atomic((__raw_i915_read32(dev_priv, FORCEWAKE_ACK) & 1) == 0,
+	if (wait_for_atomic((__raw_i915_read32(d->i915, d->reg_ack) &
+			     FORCEWAKE_KERNEL) == 0,
  			    FORCEWAKE_ACK_TIMEOUT_MS))
-		DRM_ERROR("Timed out waiting for forcewake old ack to clear.\n");
+		DRM_ERROR("%s: timed out waiting for forcewake ack to clear.\n",
+			  intel_uncore_forcewake_domain_to_str(d->id));
+}
 
-	__raw_i915_write32(dev_priv, FORCEWAKE, 1);
-	/* something from same cacheline, but !FORCEWAKE */
-	__raw_posting_read(dev_priv, ECOBUS);
+static inline void
+fw_domain_get(const struct intel_uncore_forcewake_domain *d)
+{
+	__raw_i915_write32(d->i915, d->reg_set, d->val_set);
+}
 
-	if (wait_for_atomic((__raw_i915_read32(dev_priv, FORCEWAKE_ACK) & 1),
+static inline void
+fw_domain_wait_ack(const struct intel_uncore_forcewake_domain *d)
+{
+	if (wait_for_atomic((__raw_i915_read32(d->i915, d->reg_ack) &
+			     FORCEWAKE_KERNEL),
  			    FORCEWAKE_ACK_TIMEOUT_MS))
-		DRM_ERROR("Timed out waiting for forcewake to ack request.\n");
+		DRM_ERROR("%s: timed out waiting for forcewake ack request.\n",
+			  intel_uncore_forcewake_domain_to_str(d->id));
+}
 
-	/* WaRsForcewakeWaitTC0:snb */
-	__gen6_gt_wait_for_thread_c0(dev_priv);
+static inline void
+fw_domain_put(const struct intel_uncore_forcewake_domain *d)
+{
+	__raw_i915_write32(d->i915, d->reg_set, d->val_clear);
  }
 
-static void __gen7_gt_force_wake_mt_reset(struct drm_i915_private *dev_priv)
+static inline void
+fw_domain_posting_read(const struct intel_uncore_forcewake_domain *d)
  {
-	__raw_i915_write32(dev_priv, FORCEWAKE_MT, _MASKED_BIT_DISABLE(0xffff));
-	/* something from same cacheline, but !FORCEWAKE_MT */
-	__raw_posting_read(dev_priv, ECOBUS);
+	/* something from same cacheline, but not from the set register */
+	if (d->reg_post)
+		__raw_posting_read(d->i915, d->reg_post);
  }
 
-static void __gen7_gt_force_wake_mt_get(struct drm_i915_private *dev_priv,
-					int fw_engine)
+static void
+fw_domains_get(struct drm_i915_private *dev_priv, int fw_domains)
  {
-	u32 forcewake_ack;
+	struct intel_uncore_forcewake_domain *d;
+	int id;
 
-	if (IS_HASWELL(dev_priv->dev) || IS_BROADWELL(dev_priv->dev))
-		forcewake_ack = FORCEWAKE_ACK_HSW;
-	else
-		forcewake_ack = FORCEWAKE_MT_ACK;
+	for_each_fw_domain_mask(d, fw_domains, dev_priv, id) {
+		fw_domain_wait_ack_clear(d);
+		fw_domain_get(d);
+		fw_domain_posting_read(d);
+		fw_domain_wait_ack(d);
+	}
+}
 
-	if (wait_for_atomic((__raw_i915_read32(dev_priv, forcewake_ack) & FORCEWAKE_KERNEL) == 0,
-			    FORCEWAKE_ACK_TIMEOUT_MS))
-		DRM_ERROR("Timed out waiting for forcewake old ack to clear.\n");
+static void
+fw_domains_put(struct drm_i915_private *dev_priv, int fw_domains)
+{
+	struct intel_uncore_forcewake_domain *d;
+	int id;
 
-	__raw_i915_write32(dev_priv, FORCEWAKE_MT,
-			   _MASKED_BIT_ENABLE(FORCEWAKE_KERNEL));
-	/* something from same cacheline, but !FORCEWAKE_MT */
-	__raw_posting_read(dev_priv, ECOBUS);
+	for_each_fw_domain_mask(d, fw_domains, dev_priv, id)
+		fw_domain_put(d);
+}
 
-	if (wait_for_atomic((__raw_i915_read32(dev_priv, forcewake_ack) & FORCEWAKE_KERNEL),
-			    FORCEWAKE_ACK_TIMEOUT_MS))
-		DRM_ERROR("Timed out waiting for forcewake to ack request.\n");
+static void
+fw_domains_posting_read(struct drm_i915_private *dev_priv)
+{
+	struct intel_uncore_forcewake_domain *d;
+	int id;
+
+	/* No need to do for all, just do for first found */
+	for_each_fw_domain(d, dev_priv, id) {
+		fw_domain_posting_read(d);
+		break;
+	}
+}
+
+static void
+fw_domains_reset(struct drm_i915_private *dev_priv, const unsigned fw_domains)
+{
+	struct intel_uncore_forcewake_domain *d;
+	int id;
+
+	for_each_fw_domain_mask(d, fw_domains, dev_priv, id)
+		fw_domain_reset(d);
+
+	fw_domains_posting_read(dev_priv);
+}
+
+static void __gen6_gt_wait_for_thread_c0(struct drm_i915_private *dev_priv)
+{
+	/* w/a for a sporadic read returning 0 by waiting for the GT
+	 * thread to wake up.
+	 */
+	if (wait_for_atomic_us((__raw_i915_read32(dev_priv, GEN6_GT_THREAD_STATUS_REG) &
+				GEN6_GT_THREAD_STATUS_CORE_MASK) == 0, 500))
+		DRM_ERROR("GT thread status wait timed out\n");
+}
 
-	/* WaRsForcewakeWaitTC0:ivb,hsw */
+static void fw_domains_get_with_thread_status(struct drm_i915_private *dev_priv,
+					      int fw_domains)
+{
+	fw_domains_get(dev_priv, fw_domains);
+
+	/* WaRsForcewakeWaitTC0:snb,ivb,hsw,bdw,vlv */
  	__gen6_gt_wait_for_thread_c0(dev_priv);
  }
 
@@ -128,27 +198,13 @@ static void gen6_gt_check_fifodbg(struct drm_i915_private *dev_priv)
  		__raw_i915_write32(dev_priv, GTFIFODBG, gtfifodbg);
  }
 
-static void __gen6_gt_force_wake_put(struct drm_i915_private *dev_priv,
-				     int fw_engine)
+static void fw_domains_put_with_fifo(struct drm_i915_private *dev_priv,
+				     int fw_domains)
  {
-	__raw_i915_write32(dev_priv, FORCEWAKE, 0);
-	/* something from same cacheline, but !FORCEWAKE */
-	__raw_posting_read(dev_priv, ECOBUS);
+	fw_domains_put(dev_priv, fw_domains);
  	gen6_gt_check_fifodbg(dev_priv);
  }
 
-static void __gen7_gt_force_wake_mt_put(struct drm_i915_private *dev_priv,
-					int fw_engine)
-{
-	__raw_i915_write32(dev_priv, FORCEWAKE_MT,
-			   _MASKED_BIT_DISABLE(FORCEWAKE_KERNEL));
-	/* something from same cacheline, but !FORCEWAKE_MT */
-	__raw_posting_read(dev_priv, ECOBUS);
-
-	if (IS_GEN7(dev_priv->dev))
-		gen6_gt_check_fifodbg(dev_priv);
-}
-
  static int __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv)
  {
  	int ret = 0;
@@ -176,165 +232,16 @@ static int __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv)
  	return ret;
  }
 
-static void vlv_force_wake_reset(struct drm_i915_private *dev_priv)
-{
-	__raw_i915_write32(dev_priv, FORCEWAKE_VLV,
-			   _MASKED_BIT_DISABLE(0xffff));
-	__raw_i915_write32(dev_priv, FORCEWAKE_MEDIA_VLV,
-			   _MASKED_BIT_DISABLE(0xffff));
-	/* something from same cacheline, but !FORCEWAKE_VLV */
-	__raw_posting_read(dev_priv, FORCEWAKE_ACK_VLV);
-}
-
-static void __vlv_force_wake_get(struct drm_i915_private *dev_priv,
-				 int fw_engine)
-{
-	/* Check for Render Engine */
-	if (FORCEWAKE_RENDER & fw_engine) {
-		if (wait_for_atomic((__raw_i915_read32(dev_priv,
-						FORCEWAKE_ACK_VLV) &
-						FORCEWAKE_KERNEL) == 0,
-					FORCEWAKE_ACK_TIMEOUT_MS))
-			DRM_ERROR("Timed out: Render forcewake old ack to clear.\n");
-
-		__raw_i915_write32(dev_priv, FORCEWAKE_VLV,
-				   _MASKED_BIT_ENABLE(FORCEWAKE_KERNEL));
-
-		if (wait_for_atomic((__raw_i915_read32(dev_priv,
-						FORCEWAKE_ACK_VLV) &
-						FORCEWAKE_KERNEL),
-					FORCEWAKE_ACK_TIMEOUT_MS))
-			DRM_ERROR("Timed out: waiting for Render to ack.\n");
-	}
-
-	/* Check for Media Engine */
-	if (FORCEWAKE_MEDIA & fw_engine) {
-		if (wait_for_atomic((__raw_i915_read32(dev_priv,
-						FORCEWAKE_ACK_MEDIA_VLV) &
-						FORCEWAKE_KERNEL) == 0,
-					FORCEWAKE_ACK_TIMEOUT_MS))
-			DRM_ERROR("Timed out: Media forcewake old ack to clear.\n");
-
-		__raw_i915_write32(dev_priv, FORCEWAKE_MEDIA_VLV,
-				   _MASKED_BIT_ENABLE(FORCEWAKE_KERNEL));
-
-		if (wait_for_atomic((__raw_i915_read32(dev_priv,
-						FORCEWAKE_ACK_MEDIA_VLV) &
-						FORCEWAKE_KERNEL),
-					FORCEWAKE_ACK_TIMEOUT_MS))
-			DRM_ERROR("Timed out: waiting for media to ack.\n");
-	}
-}
-
  static void __vlv_force_wake_put(struct drm_i915_private *dev_priv,
  				 int fw_engine)
  {
-	/* Check for Render Engine */
-	if (FORCEWAKE_RENDER & fw_engine)
-		__raw_i915_write32(dev_priv, FORCEWAKE_VLV,
-					_MASKED_BIT_DISABLE(FORCEWAKE_KERNEL));
+	fw_domains_put(dev_priv, fw_engine);
+	fw_domains_posting_read(dev_priv);
 
-
-	/* Check for Media Engine */
-	if (FORCEWAKE_MEDIA & fw_engine)
-		__raw_i915_write32(dev_priv, FORCEWAKE_MEDIA_VLV,
-				_MASKED_BIT_DISABLE(FORCEWAKE_KERNEL));
-
-	/* something from same cacheline, but !FORCEWAKE_VLV */
-	__raw_posting_read(dev_priv, FORCEWAKE_ACK_VLV);
  	if (!IS_CHERRYVIEW(dev_priv->dev))
  		gen6_gt_check_fifodbg(dev_priv);
  }
 
-static void __gen9_gt_force_wake_mt_reset(struct drm_i915_private *dev_priv)
-{
-	__raw_i915_write32(dev_priv, FORCEWAKE_RENDER_GEN9,
-			_MASKED_BIT_DISABLE(0xffff));
-
-	__raw_i915_write32(dev_priv, FORCEWAKE_MEDIA_GEN9,
-			_MASKED_BIT_DISABLE(0xffff));
-
-	__raw_i915_write32(dev_priv, FORCEWAKE_BLITTER_GEN9,
-			_MASKED_BIT_DISABLE(0xffff));
-}
-
-static void
-__gen9_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
-{
-	/* Check for Render Engine */
-	if (FORCEWAKE_RENDER & fw_engine) {
-		if (wait_for_atomic((__raw_i915_read32(dev_priv,
-						FORCEWAKE_ACK_RENDER_GEN9) &
-						FORCEWAKE_KERNEL) == 0,
-					FORCEWAKE_ACK_TIMEOUT_MS))
-			DRM_ERROR("Timed out: Render forcewake old ack to clear.\n");
-
-		__raw_i915_write32(dev_priv, FORCEWAKE_RENDER_GEN9,
-				   _MASKED_BIT_ENABLE(FORCEWAKE_KERNEL));
-
-		if (wait_for_atomic((__raw_i915_read32(dev_priv,
-						FORCEWAKE_ACK_RENDER_GEN9) &
-						FORCEWAKE_KERNEL),
-					FORCEWAKE_ACK_TIMEOUT_MS))
-			DRM_ERROR("Timed out: waiting for Render to ack.\n");
-	}
-
-	/* Check for Media Engine */
-	if (FORCEWAKE_MEDIA & fw_engine) {
-		if (wait_for_atomic((__raw_i915_read32(dev_priv,
-						FORCEWAKE_ACK_MEDIA_GEN9) &
-						FORCEWAKE_KERNEL) == 0,
-					FORCEWAKE_ACK_TIMEOUT_MS))
-			DRM_ERROR("Timed out: Media forcewake old ack to clear.\n");
-
-		__raw_i915_write32(dev_priv, FORCEWAKE_MEDIA_GEN9,
-				   _MASKED_BIT_ENABLE(FORCEWAKE_KERNEL));
-
-		if (wait_for_atomic((__raw_i915_read32(dev_priv,
-						FORCEWAKE_ACK_MEDIA_GEN9) &
-						FORCEWAKE_KERNEL),
-					FORCEWAKE_ACK_TIMEOUT_MS))
-			DRM_ERROR("Timed out: waiting for Media to ack.\n");
-	}
-
-	/* Check for Blitter Engine */
-	if (FORCEWAKE_BLITTER & fw_engine) {
-		if (wait_for_atomic((__raw_i915_read32(dev_priv,
-						FORCEWAKE_ACK_BLITTER_GEN9) &
-						FORCEWAKE_KERNEL) == 0,
-					FORCEWAKE_ACK_TIMEOUT_MS))
-			DRM_ERROR("Timed out: Blitter forcewake old ack to clear.\n");
-
-		__raw_i915_write32(dev_priv, FORCEWAKE_BLITTER_GEN9,
-				   _MASKED_BIT_ENABLE(FORCEWAKE_KERNEL));
-
-		if (wait_for_atomic((__raw_i915_read32(dev_priv,
-						FORCEWAKE_ACK_BLITTER_GEN9) &
-						FORCEWAKE_KERNEL),
-					FORCEWAKE_ACK_TIMEOUT_MS))
-			DRM_ERROR("Timed out: waiting for Blitter to ack.\n");
-	}
-}
-
-static void
-__gen9_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
-{
-	/* Check for Render Engine */
-	if (FORCEWAKE_RENDER & fw_engine)
-		__raw_i915_write32(dev_priv, FORCEWAKE_RENDER_GEN9,
-				_MASKED_BIT_DISABLE(FORCEWAKE_KERNEL));
-
-	/* Check for Media Engine */
-	if (FORCEWAKE_MEDIA & fw_engine)
-		__raw_i915_write32(dev_priv, FORCEWAKE_MEDIA_GEN9,
-				_MASKED_BIT_DISABLE(FORCEWAKE_KERNEL));
-
-	/* Check for Blitter Engine */
-	if (FORCEWAKE_BLITTER & fw_engine)
-		__raw_i915_write32(dev_priv, FORCEWAKE_BLITTER_GEN9,
-				_MASKED_BIT_DISABLE(FORCEWAKE_KERNEL));
-}
-
  static void gen6_force_wake_timer(unsigned long arg)
  {
  	struct intel_uncore_forcewake_domain *domain = (void *)arg;
@@ -358,28 +265,20 @@ void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore)
  	struct intel_uncore_forcewake_domain *domain;
  	int id;
 
-	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
  	/* Hold uncore.lock across reset to prevent any register access
  	 * with forcewake not set correctly
  	 */
+	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
 
  	for_each_fw_domain(domain, dev_priv, id)
  		if (domain->wake_count)
  			fw |= 1 << id;
 
+	/* Don't stop the timer, just turn domains off for reset */
  	if (fw)
  		dev_priv->uncore.funcs.force_wake_put(dev_priv, fw);
 
-	if (IS_VALLEYVIEW(dev))
-		vlv_force_wake_reset(dev_priv);
-	else if (IS_GEN6(dev) || IS_GEN7(dev))
-		__gen6_gt_force_wake_reset(dev_priv);
-
-	if (IS_IVYBRIDGE(dev) || IS_HASWELL(dev) || IS_BROADWELL(dev))
-		__gen7_gt_force_wake_mt_reset(dev_priv);
-
-	if (IS_GEN9(dev))
-		__gen9_gt_force_wake_mt_reset(dev_priv);
+	fw_domains_reset(dev_priv, FORCEWAKE_ALL);
 
  	if (restore) { /* If reset with a user forcewake, try to restore */
  		if (fw)
@@ -488,7 +387,7 @@ void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv,
  			continue;
 
  		domain->wake_count++;
-		mod_timer_pinned(&domain->timer, jiffies + 1);
+		fw_domain_arm_timer(domain);
  	}
 
  	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
@@ -497,12 +396,12 @@ void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv,
  void assert_force_wake_inactive(struct drm_i915_private *dev_priv)
  {
  	struct intel_uncore_forcewake_domain *domain;
-	int i;
+	int id;
 
  	if (!dev_priv->uncore.funcs.force_wake_get)
  		return;
 
-	for_each_fw_domain(domain, dev_priv, i)
+	for_each_fw_domain(domain, dev_priv, id)
  		WARN_ON(domain->wake_count);
  }
 
@@ -673,19 +572,19 @@ static inline void __force_wake_get(struct drm_i915_private *dev_priv,
  				    unsigned fw_domains)
  {
  	struct intel_uncore_forcewake_domain *domain;
-	int i;
+	int id;
 
  	if (WARN_ON(!fw_domains))
  		return;
 
  	/* Ideally GCC would be constant-fold and eliminate this loop */
-	for_each_fw_domain_mask(domain, fw_domains, dev_priv, i) {
+	for_each_fw_domain_mask(domain, fw_domains, dev_priv, id) {
  		if (domain->wake_count)
-			fw_domains &= ~(1 << i);
+			fw_domains &= ~(1 << id);
  		else
  			domain->wake_count++;
 
-		mod_timer_pinned(&domain->timer, jiffies + 1);
+		fw_domain_arm_timer(domain);
  	}
 
  	if (fw_domains)
@@ -1001,27 +900,75 @@ do { \
  	dev_priv->uncore.funcs.mmio_readq = x##_read64; \
  } while (0)
 
+
+static void fw_domain_init(struct drm_i915_private *dev_priv,
+			   u32 domain_id, u32 reg_set, u32 reg_ack)
+{
+	struct intel_uncore_forcewake_domain *d;
+
+	if (WARN_ON(domain_id >= FW_DOMAIN_ID_COUNT))
+		return;
+
+	d = &dev_priv->uncore.fw_domain[domain_id];
+
+	d->reg_set = reg_set;
+	d->reg_ack = reg_ack;
+
+	if (IS_GEN6(dev_priv)) {
+		d->val_reset = 0;
+		d->val_set = FORCEWAKE_KERNEL;
+		d->val_clear = 0;
+	} else {
+		d->val_reset = _MASKED_BIT_DISABLE(0xffff);
+		d->val_set = _MASKED_BIT_ENABLE(FORCEWAKE_KERNEL);
+		d->val_clear = _MASKED_BIT_DISABLE(FORCEWAKE_KERNEL);
+	}
+
+	if (IS_VALLEYVIEW(dev_priv))
+		d->reg_post = FORCEWAKE_ACK_VLV;
+	else if (IS_GEN6(dev_priv) || IS_GEN7(dev_priv))
+		d->reg_post = ECOBUS;
+	else
+		d->reg_post = 0;
+
+	d->i915 = dev_priv;
+	d->id = domain_id;
+
+	setup_timer(&d->timer, gen6_force_wake_timer, (unsigned long)d);
+
+	dev_priv->uncore.fw_domains |= (1 << domain_id);
+}
+
  void intel_uncore_init(struct drm_device *dev)
  {
  	struct drm_i915_private *dev_priv = dev->dev_private;
-	struct intel_uncore_forcewake_domain *domain;
-	int i;
 
  	__intel_uncore_early_sanitize(dev, false);
 
  	if (IS_GEN9(dev)) {
-		dev_priv->uncore.funcs.force_wake_get = __gen9_force_wake_get;
-		dev_priv->uncore.funcs.force_wake_put = __gen9_force_wake_put;
-		dev_priv->uncore.fw_domains = FORCEWAKE_RENDER |
-			FORCEWAKE_BLITTER | FORCEWAKE_MEDIA;
+		dev_priv->uncore.funcs.force_wake_get = fw_domains_get;
+		dev_priv->uncore.funcs.force_wake_put = fw_domains_put;
+		fw_domain_init(dev_priv, FW_DOMAIN_ID_RENDER,
+			       FORCEWAKE_RENDER_GEN9,
+			       FORCEWAKE_ACK_RENDER_GEN9);
+		fw_domain_init(dev_priv, FW_DOMAIN_ID_BLITTER,
+			       FORCEWAKE_BLITTER_GEN9,
+			       FORCEWAKE_ACK_BLITTER_GEN9);
+		fw_domain_init(dev_priv, FW_DOMAIN_ID_MEDIA,
+			       FORCEWAKE_MEDIA_GEN9, FORCEWAKE_ACK_MEDIA_GEN9);
  	} else if (IS_VALLEYVIEW(dev)) {
-		dev_priv->uncore.funcs.force_wake_get = __vlv_force_wake_get;
+		dev_priv->uncore.funcs.force_wake_get = fw_domains_get;
  		dev_priv->uncore.funcs.force_wake_put = __vlv_force_wake_put;
-		dev_priv->uncore.fw_domains = FORCEWAKE_RENDER | FORCEWAKE_MEDIA;
+		fw_domain_init(dev_priv, FW_DOMAIN_ID_RENDER,
+			       FORCEWAKE_VLV, FORCEWAKE_ACK_VLV);
+		fw_domain_init(dev_priv, FW_DOMAIN_ID_MEDIA,
+			       FORCEWAKE_MEDIA_VLV, FORCEWAKE_ACK_MEDIA_VLV);
  	} else if (IS_HASWELL(dev) || IS_BROADWELL(dev)) {
-		dev_priv->uncore.funcs.force_wake_get = __gen7_gt_force_wake_mt_get;
-		dev_priv->uncore.funcs.force_wake_put = __gen7_gt_force_wake_mt_put;
-		dev_priv->uncore.fw_domains = FORCEWAKE_RENDER;
+		dev_priv->uncore.funcs.force_wake_get =
+			fw_domains_get_with_thread_status;
+		dev_priv->uncore.funcs.force_wake_put = fw_domains_put;
+		fw_domain_init(dev_priv, FW_DOMAIN_ID_RENDER,
+			       FORCEWAKE_MT, FORCEWAKE_ACK_HSW);
  	} else if (IS_IVYBRIDGE(dev)) {
  		u32 ecobus;
 
@@ -1034,40 +981,32 @@ void intel_uncore_init(struct drm_device *dev)
  		 * (correctly) interpreted by the test below as MT
  		 * forcewake being disabled.
  		 */
+		dev_priv->uncore.funcs.force_wake_get =
+			fw_domains_get_with_thread_status;
+		dev_priv->uncore.funcs.force_wake_put =
+			fw_domains_put_with_fifo;
+
+		fw_domain_init(dev_priv, FW_DOMAIN_ID_RENDER,
+			       FORCEWAKE_MT, FORCEWAKE_MT_ACK);
  		mutex_lock(&dev->struct_mutex);
-		__gen7_gt_force_wake_mt_get(dev_priv, FORCEWAKE_ALL);
+		fw_domains_get_with_thread_status(dev_priv, FORCEWAKE_ALL);
  		ecobus = __raw_i915_read32(dev_priv, ECOBUS);
-		__gen7_gt_force_wake_mt_put(dev_priv, FORCEWAKE_ALL);
+		fw_domains_put_with_fifo(dev_priv, FORCEWAKE_ALL);
  		mutex_unlock(&dev->struct_mutex);
 
-		if (ecobus & FORCEWAKE_MT_ENABLE) {
-			dev_priv->uncore.funcs.force_wake_get =
-				__gen7_gt_force_wake_mt_get;
-			dev_priv->uncore.funcs.force_wake_put =
-				__gen7_gt_force_wake_mt_put;
-		} else {
+		if (!(ecobus & FORCEWAKE_MT_ENABLE)) {
  			DRM_INFO("No MT forcewake available on Ivybridge, this can result in issues\n");
  			DRM_INFO("when using vblank-synced partial screen updates.\n");
-			dev_priv->uncore.funcs.force_wake_get =
-				__gen6_gt_force_wake_get;
-			dev_priv->uncore.funcs.force_wake_put =
-				__gen6_gt_force_wake_put;
+			fw_domain_init(dev_priv, FW_DOMAIN_ID_RENDER,
+				       FORCEWAKE, FORCEWAKE_ACK);
  		}
-		dev_priv->uncore.fw_domains = FORCEWAKE_RENDER;
  	} else if (IS_GEN6(dev)) {
  		dev_priv->uncore.funcs.force_wake_get =
-			__gen6_gt_force_wake_get;
+			fw_domains_get_with_thread_status;
  		dev_priv->uncore.funcs.force_wake_put =
-			__gen6_gt_force_wake_put;
-		dev_priv->uncore.fw_domains = FORCEWAKE_RENDER;
-	}
-
-	for_each_fw_domain(domain, dev_priv, i) {
-		domain->i915 = dev_priv;

drivers/gpu/drm/i915/intel_uncore.c void intel_uncore_init(struct drm_device *dev) · line 1012

Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +3 -4
-		domain->id = i;
-
-		setup_timer(&domain->timer, gen6_force_wake_timer,
-			    (unsigned long)domain);
+			fw_domains_put_with_fifo;
+		fw_domain_init(dev_priv, FW_DOMAIN_ID_RENDER,
+			       FORCEWAKE, FORCEWAKE_ACK);
  	}
 
  	switch (INTEL_INFO(dev)->gen) {

Looks fine

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

#12 lore raw

PATCH 6/8 drm/i915: Make vlv and chv forcewake put generic.

In reply to Mika Kuoppala

These two were using a fw dance logic where posting read was done after both domain bit were set. When in other gens, the posting read is done immediately after setting the forcewake bit for each domain.

Now bring these in line with other gens.

---
 drivers/gpu/drm/i915/intel_uncore.c | 16 +++++-----------
 1 file changed, 5 insertions(+), 11 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +5 -11
index e883790..8021bec 100644
@@ -232,16 +232,6 @@ static int __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv)
 	return ret;
 }
 
-static void __vlv_force_wake_put(struct drm_i915_private *dev_priv,
-				 int fw_engine)
-{
-	fw_domains_put(dev_priv, fw_engine);
-	fw_domains_posting_read(dev_priv);
-
-	if (!IS_CHERRYVIEW(dev_priv->dev))
-		gen6_gt_check_fifodbg(dev_priv);
-}
-
 static void gen6_force_wake_timer(unsigned long arg)
 {
 	struct intel_uncore_forcewake_domain *domain = (void *)arg;
@@ -958,7 +948,11 @@ void intel_uncore_init(struct drm_device *dev)
 			       FORCEWAKE_MEDIA_GEN9, FORCEWAKE_ACK_MEDIA_GEN9);
 	} else if (IS_VALLEYVIEW(dev)) {
 		dev_priv->uncore.funcs.force_wake_get = fw_domains_get;
-		dev_priv->uncore.funcs.force_wake_put = __vlv_force_wake_put;
+		if (!IS_CHERRYVIEW(dev))
+			dev_priv->uncore.funcs.force_wake_put =
+				fw_domains_put_with_fifo;
+		else
+			dev_priv->uncore.funcs.force_wake_put = fw_domains_put;
 		fw_domain_init(dev_priv, FW_DOMAIN_ID_RENDER,
 			       FORCEWAKE_VLV, FORCEWAKE_ACK_VLV);
 		fw_domain_init(dev_priv, FW_DOMAIN_ID_MEDIA,
Signature · 6 lines
1.9.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Reviewed-by on v1 6/8
#13 lore raw

Re: [PATCH 6/8] drm/i915: Make vlv and chv forcewake put generic.

Quoted 37 lines · Mika Kuoppala

These two were using a fw dance logic where posting read was done after both domain bit were set. When in other gens, the posting read is done immediately after setting the forcewake bit for each domain.

Now bring these in line with other gens.

---
  drivers/gpu/drm/i915/intel_uncore.c | 16 +++++-----------
  1 file changed, 5 insertions(+), 11 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +0 -10
index e883790..8021bec 100644
@@ -232,16 +232,6 @@ static int __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv)
  	return ret;
  }
 
-static void __vlv_force_wake_put(struct drm_i915_private *dev_priv,
-				 int fw_engine)
-{
-	fw_domains_put(dev_priv, fw_engine);
-	fw_domains_posting_read(dev_priv);
-
-	if (!IS_CHERRYVIEW(dev_priv->dev))
-		gen6_gt_check_fifodbg(dev_priv);
-}
-
  static void gen6_force_wake_timer(unsigned long arg)
  {
  	struct intel_uncore_forcewake_domain *domain = (void *)arg;
@@ -958,7 +948,11 @@ void intel_uncore_init(struct drm_device *dev)
  			       FORCEWAKE_MEDIA_GEN9, FORCEWAKE_ACK_MEDIA_GEN9);
  	} else if (IS_VALLEYVIEW(dev)) {

drivers/gpu/drm/i915/intel_uncore.c void intel_uncore_init(struct drm_device *dev) · line 958

Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +5 -1
  		dev_priv->uncore.funcs.force_wake_get = fw_domains_get;
-		dev_priv->uncore.funcs.force_wake_put = __vlv_force_wake_put;
+		if (!IS_CHERRYVIEW(dev))
+			dev_priv->uncore.funcs.force_wake_put =
+				fw_domains_put_with_fifo;
+		else
+			dev_priv->uncore.funcs.force_wake_put = fw_domains_put;
  		fw_domain_init(dev_priv, FW_DOMAIN_ID_RENDER,
  			       FORCEWAKE_VLV, FORCEWAKE_ACK_VLV);
  		fw_domain_init(dev_priv, FW_DOMAIN_ID_MEDIA,

Looks fine

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

#14 lore raw

PATCH 7/8 drm/i915: Rename the forcewake get/put functions

In reply to Mika Kuoppala

We have multiple forcewake domains now on recent gens. Change the function naming to reflect this.

v2: More verbose names (Chris)

---
 drivers/gpu/drm/i915/i915_debugfs.c     |  8 ++++----
 drivers/gpu/drm/i915/i915_drv.c         |  2 +-
 drivers/gpu/drm/i915/i915_drv.h         | 15 +++++----------
 drivers/gpu/drm/i915/intel_display.c    |  4 ++--
 drivers/gpu/drm/i915/intel_lrc.c        |  4 ++--
 drivers/gpu/drm/i915/intel_pm.c         | 28 ++++++++++++++--------------
 drivers/gpu/drm/i915/intel_ringbuffer.c |  4 ++--
 drivers/gpu/drm/i915/intel_uncore.c     | 18 +++++++++---------
 8 files changed, 39 insertions(+), 44 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +4 -4
index 93390c9..ecc4b42 100644
@@ -1052,7 +1052,7 @@ static int i915_frequency_info(struct seq_file *m, void *unused)
 		if (ret)
 			goto out;
 
-		gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+		intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
 		reqf = I915_READ(GEN6_RPNSWREQ);
 		reqf &= ~GEN6_TURBO_DISABLE;
@@ -1079,7 +1079,7 @@ static int i915_frequency_info(struct seq_file *m, void *unused)
 			cagf = (rpstat & GEN6_CAGF_MASK) >> GEN6_CAGF_SHIFT;
 		cagf *= GT_FREQUENCY_MULTIPLIER;
 
-		gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+		intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
 		mutex_unlock(&dev->struct_mutex);
 
 		if (IS_GEN6(dev) || IS_GEN7(dev)) {
@@ -4233,7 +4233,7 @@ static int i915_forcewake_open(struct inode *inode, struct file *file)
 		return 0;
 
 	intel_runtime_pm_get(dev_priv);
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
 	return 0;
 }
@@ -4246,7 +4246,7 @@ static int i915_forcewake_release(struct inode *inode, struct file *file)
 	if (INTEL_INFO(dev)->gen < 6)
 		return 0;
 
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
 	intel_runtime_pm_put(dev_priv);
 
 	return 0;
drivers/gpu/drm/i915/i915_drv.c +1 -1
index 706b122..011caa2 100644
@@ -1362,7 +1362,7 @@ static int intel_runtime_suspend(struct device *device)
 	if (WARN_ON_ONCE(!HAS_RUNTIME_PM(dev)))
 		return -ENODEV;
 
-	assert_force_wake_inactive(dev_priv);
+	assert_forcewakes_inactive(dev_priv);
 
 	DRM_DEBUG_KMS("Suspending device\n");
 
drivers/gpu/drm/i915/i915_drv.h +5 -10
index 4263084..a2a8536 100644
@@ -2450,6 +2450,11 @@ extern void intel_uncore_check_errors(struct drm_device *dev);
 extern void intel_uncore_fini(struct drm_device *dev);
 extern void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore);
 const char *intel_uncore_forcewake_domain_to_str(const int domain_id);
+void intel_uncore_forcewake_get(struct drm_i915_private *dev_priv,
+				unsigned fw_domains);
+void intel_uncore_forcewake_put(struct drm_i915_private *dev_priv,
+				unsigned fw_domains);
+void assert_forcewakes_inactive(struct drm_i915_private *dev_priv);
 
 void
 i915_enable_pipestat(struct drm_i915_private *dev_priv, enum pipe pipe,
@@ -3026,16 +3031,6 @@ extern void intel_display_print_error_state(struct drm_i915_error_state_buf *e,
 					    struct drm_device *dev,
 					    struct intel_display_error_state *error);
 
-/* On SNB platform, before reading ring registers forcewake bit
- * must be set to prevent GT core from power down and stale values being
- * returned.
- */
-void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv,
-			    unsigned fw_domains);
-void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv,
-			    unsigned fw_domains);
-void assert_force_wake_inactive(struct drm_i915_private *dev_priv);
-
 int sandybridge_pcode_read(struct drm_i915_private *dev_priv, u32 mbox, u32 *val);
 int sandybridge_pcode_write(struct drm_i915_private *dev_priv, u32 mbox, u32 val);
 
drivers/gpu/drm/i915/intel_display.c +2 -2
index 86c2885..d2d60f0 100644
@@ -7863,7 +7863,7 @@ static void hsw_restore_lcpll(struct drm_i915_private *dev_priv)
 	 * Make sure we're not on PC8 state before disabling PC8, otherwise
 	 * we'll hang the machine. To prevent PC8 state, just enable force_wake.
 	 */
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
 	if (val & LCPLL_POWER_DOWN_ALLOW) {
 		val &= ~LCPLL_POWER_DOWN_ALLOW;
@@ -7893,7 +7893,7 @@ static void hsw_restore_lcpll(struct drm_i915_private *dev_priv)
 			DRM_ERROR("Switching back to LCPLL failed\n");
 	}
 
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
 }
 
 /*
drivers/gpu/drm/i915/intel_lrc.c +2 -2
index fcb5140..be9d00b 100644
@@ -297,7 +297,7 @@ static void execlists_elsp_write(struct intel_engine_cs *ring,
 	desc[3] = (u32)(temp >> 32);
 	desc[2] = (u32)temp;
 
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 	I915_WRITE(RING_ELSP(ring), desc[1]);
 	I915_WRITE(RING_ELSP(ring), desc[0]);
 	I915_WRITE(RING_ELSP(ring), desc[3]);
@@ -307,7 +307,7 @@ static void execlists_elsp_write(struct intel_engine_cs *ring,
 
 	/* ELSP is a wo register, so use another nearby reg for posting instead */
 	POSTING_READ(RING_EXECLIST_STATUS(ring));
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
 }
 
 static int execlists_update_context(struct drm_i915_gem_object *ctx_obj,
drivers/gpu/drm/i915/intel_pm.c +14 -14
index 78911e2..64e06a3 100644
@@ -228,7 +228,7 @@ static void sandybridge_blit_fbc_update(struct drm_device *dev)
 
 	/* Blitter is part of Media powerwell on VLV. No impact of
 	 * his param in other platforms for now */
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_MEDIA);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_MEDIA);
 
 	blt_ecoskpd = I915_READ(GEN6_BLITTER_ECOSKPD);
 	blt_ecoskpd |= GEN6_BLITTER_FBC_NOTIFY <<
@@ -241,7 +241,7 @@ static void sandybridge_blit_fbc_update(struct drm_device *dev)
 	I915_WRITE(GEN6_BLITTER_ECOSKPD, blt_ecoskpd);
 	POSTING_READ(GEN6_BLITTER_ECOSKPD);
 
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_MEDIA);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_MEDIA);
 }
 
 static void ironlake_enable_fbc(struct drm_crtc *crtc)
@@ -4541,11 +4541,11 @@ static void valleyview_disable_rps(struct drm_device *dev)
 
 	/* we're doing forcewake before Disabling RC6,
 	 * This what the BIOS expects when going into suspend */
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
 	I915_WRITE(GEN6_RC_CONTROL, 0);
 
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
 }
 
 static void intel_print_rc6_info(struct drm_device *dev, u32 mode)
@@ -4663,7 +4663,7 @@ static void gen9_enable_rps(struct drm_device *dev)
 
 	/* 1b: Get forcewake during program sequence. Although the driver
 	 * hasn't enabled a state yet where we need forcewake, BIOS may have.*/
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
 	/* 2a: Disable RC states. */
 	I915_WRITE(GEN6_RC_CONTROL, 0);
@@ -4686,7 +4686,7 @@ static void gen9_enable_rps(struct drm_device *dev)
 				   GEN6_RC_CTL_EI_MODE(1) |
 				   rc6_mask);
 
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
 
 }
 
@@ -4702,7 +4702,7 @@ static void gen8_enable_rps(struct drm_device *dev)
 
 	/* 1c & 1d: Get forcewake during program sequence. Although the driver
 	 * hasn't enabled a state yet where we need forcewake, BIOS may have.*/
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
 	/* 2a: Disable RC states. */
 	I915_WRITE(GEN6_RC_CONTROL, 0);
@@ -4769,7 +4769,7 @@ static void gen8_enable_rps(struct drm_device *dev)
 	dev_priv->rps.power = HIGH_POWER; /* force a reset */
 	gen6_set_rps(dev_priv->dev, dev_priv->rps.min_freq_softlimit);
 
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
 }
 
 static void gen6_enable_rps(struct drm_device *dev)
@@ -4797,7 +4797,7 @@ static void gen6_enable_rps(struct drm_device *dev)
 		I915_WRITE(GTFIFODBG, gtfifodbg);
 	}
 
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
 	/* Initialize rps frequencies */
 	gen6_init_rps_frequencies(dev);
@@ -4877,7 +4877,7 @@ static void gen6_enable_rps(struct drm_device *dev)
 			DRM_ERROR("Couldn't fix incorrect rc6 voltage\n");
 	}
 
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
 }
 
 static void __gen6_update_ring_freq(struct drm_device *dev)
@@ -5296,7 +5296,7 @@ static void cherryview_enable_rps(struct drm_device *dev)
 
 	/* 1a & 1b: Get forcewake during program sequence. Although the driver
 	 * hasn't enabled a state yet where we need forcewake, BIOS may have.*/
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
 	/* 2a: Program RC6 thresholds.*/
 	I915_WRITE(GEN6_RC6_WAKE_RATE_LIMIT, 40 << 16);
@@ -5364,7 +5364,7 @@ static void cherryview_enable_rps(struct drm_device *dev)
 
 	valleyview_set_rps(dev_priv->dev, dev_priv->rps.efficient_freq);
 
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
 }
 
 static void valleyview_enable_rps(struct drm_device *dev)
@@ -5385,7 +5385,7 @@ static void valleyview_enable_rps(struct drm_device *dev)
 	}
 
 	/* If VLV, Forcewake all wells, else re-direct to regular path */
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
 	I915_WRITE(GEN6_RP_UP_THRESHOLD, 59400);
 	I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 245000);
@@ -5445,7 +5445,7 @@ static void valleyview_enable_rps(struct drm_device *dev)
 
 	valleyview_set_rps(dev_priv->dev, dev_priv->rps.efficient_freq);
 
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
 }
 
 void ironlake_teardown_rc6(struct drm_device *dev)
drivers/gpu/drm/i915/intel_ringbuffer.c +2 -2
index 3887f1a..210728f 100644
@@ -536,7 +536,7 @@ static int init_ring_common(struct intel_engine_cs *ring)
 	struct drm_i915_gem_object *obj = ringbuf->obj;
 	int ret = 0;
 
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
 	if (!stop_ring(ring)) {
 		/* G45 ring initialization often fails to reset head to zero */
@@ -608,7 +608,7 @@ static int init_ring_common(struct intel_engine_cs *ring)
 	memset(&ring->hangcheck, 0, sizeof(ring->hangcheck));
 
 out:
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
 
 	return ret;
 }
drivers/gpu/drm/i915/intel_uncore.c +9 -9
index 8021bec..509b9c9 100644
@@ -232,7 +232,7 @@ static int __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv)
 	return ret;
 }
 
-static void gen6_force_wake_timer(unsigned long arg)
+static void intel_uncore_fw_release_timer(unsigned long arg)
 {
 	struct intel_uncore_forcewake_domain *domain = (void *)arg;
 	unsigned long irqflags;
@@ -326,10 +326,10 @@ void intel_uncore_sanitize(struct drm_device *dev)
  * Generally this is called implicitly by the register read function. However,
  * if some sequence requires the GT to not power down then this function should
  * be called at the beginning of the sequence followed by a call to
- * gen6_gt_force_wake_put() at the end of the sequence.
+ * intel_uncore_forcewake_put() at the end of the sequence.
  */
-void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv,
-			    unsigned fw_domains)
+void intel_uncore_forcewake_get(struct drm_i915_private *dev_priv,
+				unsigned fw_domains)
 {
 	unsigned long irqflags;
 	struct intel_uncore_forcewake_domain *domain;
@@ -355,10 +355,10 @@ void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv,
 }
 
 /*
- * see gen6_gt_force_wake_get()
+ * see intel_uncore_forcewake_get()
  */
-void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv,
-			    unsigned fw_domains)
+void intel_uncore_forcewake_put(struct drm_i915_private *dev_priv,
+				unsigned fw_domains)
 {
 	unsigned long irqflags;
 	struct intel_uncore_forcewake_domain *domain;
@@ -383,7 +383,7 @@ void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv,
 	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
 }
 
-void assert_force_wake_inactive(struct drm_i915_private *dev_priv)
+void assert_forcewakes_inactive(struct drm_i915_private *dev_priv)
 {
 	struct intel_uncore_forcewake_domain *domain;
 	int id;
@@ -924,7 +924,7 @@ static void fw_domain_init(struct drm_i915_private *dev_priv,
 	d->i915 = dev_priv;
 	d->id = domain_id;
 
-	setup_timer(&d->timer, gen6_force_wake_timer, (unsigned long)d);
+	setup_timer(&d->timer, intel_uncore_fw_release_timer, (unsigned long)d);
 
 	dev_priv->uncore.fw_domains |= (1 << domain_id);
 }
Signature · 6 lines
1.9.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Reviewed-by on v1 7/8
#15 lore raw

Re: [PATCH 7/8] drm/i915: Rename the forcewake get/put functions

Quoted 346 lines · Mika Kuoppala

We have multiple forcewake domains now on recent gens. Change the function naming to reflect this.

v2: More verbose names (Chris)

---
  drivers/gpu/drm/i915/i915_debugfs.c     |  8 ++++----
  drivers/gpu/drm/i915/i915_drv.c         |  2 +-
  drivers/gpu/drm/i915/i915_drv.h         | 15 +++++----------
  drivers/gpu/drm/i915/intel_display.c    |  4 ++--
  drivers/gpu/drm/i915/intel_lrc.c        |  4 ++--
  drivers/gpu/drm/i915/intel_pm.c         | 28 ++++++++++++++--------------
  drivers/gpu/drm/i915/intel_ringbuffer.c |  4 ++--
  drivers/gpu/drm/i915/intel_uncore.c     | 18 +++++++++---------
  8 files changed, 39 insertions(+), 44 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +4 -4
index 93390c9..ecc4b42 100644
@@ -1052,7 +1052,7 @@ static int i915_frequency_info(struct seq_file *m, void *unused)
  		if (ret)
  			goto out;
 
-		gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+		intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
  		reqf = I915_READ(GEN6_RPNSWREQ);
  		reqf &= ~GEN6_TURBO_DISABLE;
@@ -1079,7 +1079,7 @@ static int i915_frequency_info(struct seq_file *m, void *unused)
  			cagf = (rpstat & GEN6_CAGF_MASK) >> GEN6_CAGF_SHIFT;
  		cagf *= GT_FREQUENCY_MULTIPLIER;
 
-		gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+		intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
  		mutex_unlock(&dev->struct_mutex);
 
  		if (IS_GEN6(dev) || IS_GEN7(dev)) {
@@ -4233,7 +4233,7 @@ static int i915_forcewake_open(struct inode *inode, struct file *file)
  		return 0;
 
  	intel_runtime_pm_get(dev_priv);
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
  	return 0;
  }
@@ -4246,7 +4246,7 @@ static int i915_forcewake_release(struct inode *inode, struct file *file)
  	if (INTEL_INFO(dev)->gen < 6)
  		return 0;
 
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
  	intel_runtime_pm_put(dev_priv);
 
  	return 0;
drivers/gpu/drm/i915/i915_drv.c +1 -1
index 706b122..011caa2 100644
@@ -1362,7 +1362,7 @@ static int intel_runtime_suspend(struct device *device)
  	if (WARN_ON_ONCE(!HAS_RUNTIME_PM(dev)))
  		return -ENODEV;
 
-	assert_force_wake_inactive(dev_priv);
+	assert_forcewakes_inactive(dev_priv);
 
  	DRM_DEBUG_KMS("Suspending device\n");
 
drivers/gpu/drm/i915/i915_drv.h +5 -10
index 4263084..a2a8536 100644
@@ -2450,6 +2450,11 @@ extern void intel_uncore_check_errors(struct drm_device *dev);
  extern void intel_uncore_fini(struct drm_device *dev);
  extern void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore);
  const char *intel_uncore_forcewake_domain_to_str(const int domain_id);
+void intel_uncore_forcewake_get(struct drm_i915_private *dev_priv,
+				unsigned fw_domains);
+void intel_uncore_forcewake_put(struct drm_i915_private *dev_priv,
+				unsigned fw_domains);
+void assert_forcewakes_inactive(struct drm_i915_private *dev_priv);
 
  void
  i915_enable_pipestat(struct drm_i915_private *dev_priv, enum pipe pipe,
@@ -3026,16 +3031,6 @@ extern void intel_display_print_error_state(struct drm_i915_error_state_buf *e,
  					    struct drm_device *dev,
  					    struct intel_display_error_state *error);
 
-/* On SNB platform, before reading ring registers forcewake bit
- * must be set to prevent GT core from power down and stale values being
- * returned.
- */
-void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv,
-			    unsigned fw_domains);
-void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv,
-			    unsigned fw_domains);
-void assert_force_wake_inactive(struct drm_i915_private *dev_priv);
-
  int sandybridge_pcode_read(struct drm_i915_private *dev_priv, u32 mbox, u32 *val);
  int sandybridge_pcode_write(struct drm_i915_private *dev_priv, u32 mbox, u32 val);
 
drivers/gpu/drm/i915/intel_display.c +2 -2
index 86c2885..d2d60f0 100644
@@ -7863,7 +7863,7 @@ static void hsw_restore_lcpll(struct drm_i915_private *dev_priv)
  	 * Make sure we're not on PC8 state before disabling PC8, otherwise
  	 * we'll hang the machine. To prevent PC8 state, just enable force_wake.
  	 */
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
  	if (val & LCPLL_POWER_DOWN_ALLOW) {
  		val &= ~LCPLL_POWER_DOWN_ALLOW;
@@ -7893,7 +7893,7 @@ static void hsw_restore_lcpll(struct drm_i915_private *dev_priv)
  			DRM_ERROR("Switching back to LCPLL failed\n");
  	}
 
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
  }
 
  /*
drivers/gpu/drm/i915/intel_lrc.c +2 -2
index fcb5140..be9d00b 100644
@@ -297,7 +297,7 @@ static void execlists_elsp_write(struct intel_engine_cs *ring,
  	desc[3] = (u32)(temp >> 32);
  	desc[2] = (u32)temp;
 
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
  	I915_WRITE(RING_ELSP(ring), desc[1]);
  	I915_WRITE(RING_ELSP(ring), desc[0]);
  	I915_WRITE(RING_ELSP(ring), desc[3]);
@@ -307,7 +307,7 @@ static void execlists_elsp_write(struct intel_engine_cs *ring,
 
  	/* ELSP is a wo register, so use another nearby reg for posting instead */
  	POSTING_READ(RING_EXECLIST_STATUS(ring));
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
  }
 
  static int execlists_update_context(struct drm_i915_gem_object *ctx_obj,
drivers/gpu/drm/i915/intel_pm.c +14 -14
index 78911e2..64e06a3 100644
@@ -228,7 +228,7 @@ static void sandybridge_blit_fbc_update(struct drm_device *dev)
 
  	/* Blitter is part of Media powerwell on VLV. No impact of
  	 * his param in other platforms for now */
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_MEDIA);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_MEDIA);
 
  	blt_ecoskpd = I915_READ(GEN6_BLITTER_ECOSKPD);
  	blt_ecoskpd |= GEN6_BLITTER_FBC_NOTIFY <<
@@ -241,7 +241,7 @@ static void sandybridge_blit_fbc_update(struct drm_device *dev)
  	I915_WRITE(GEN6_BLITTER_ECOSKPD, blt_ecoskpd);
  	POSTING_READ(GEN6_BLITTER_ECOSKPD);
 
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_MEDIA);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_MEDIA);
  }
 
  static void ironlake_enable_fbc(struct drm_crtc *crtc)
@@ -4541,11 +4541,11 @@ static void valleyview_disable_rps(struct drm_device *dev)
 
  	/* we're doing forcewake before Disabling RC6,
  	 * This what the BIOS expects when going into suspend */
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
  	I915_WRITE(GEN6_RC_CONTROL, 0);
 
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
  }
 
  static void intel_print_rc6_info(struct drm_device *dev, u32 mode)
@@ -4663,7 +4663,7 @@ static void gen9_enable_rps(struct drm_device *dev)
 
  	/* 1b: Get forcewake during program sequence. Although the driver
  	 * hasn't enabled a state yet where we need forcewake, BIOS may have.*/
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
  	/* 2a: Disable RC states. */
  	I915_WRITE(GEN6_RC_CONTROL, 0);
@@ -4686,7 +4686,7 @@ static void gen9_enable_rps(struct drm_device *dev)
  				   GEN6_RC_CTL_EI_MODE(1) |
  				   rc6_mask);
 
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
 
  }
 
@@ -4702,7 +4702,7 @@ static void gen8_enable_rps(struct drm_device *dev)
 
  	/* 1c & 1d: Get forcewake during program sequence. Although the driver
  	 * hasn't enabled a state yet where we need forcewake, BIOS may have.*/
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
  	/* 2a: Disable RC states. */
  	I915_WRITE(GEN6_RC_CONTROL, 0);
@@ -4769,7 +4769,7 @@ static void gen8_enable_rps(struct drm_device *dev)
  	dev_priv->rps.power = HIGH_POWER; /* force a reset */
  	gen6_set_rps(dev_priv->dev, dev_priv->rps.min_freq_softlimit);
 
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
  }
 
  static void gen6_enable_rps(struct drm_device *dev)
@@ -4797,7 +4797,7 @@ static void gen6_enable_rps(struct drm_device *dev)
  		I915_WRITE(GTFIFODBG, gtfifodbg);
  	}
 
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
  	/* Initialize rps frequencies */
  	gen6_init_rps_frequencies(dev);
@@ -4877,7 +4877,7 @@ static void gen6_enable_rps(struct drm_device *dev)
  			DRM_ERROR("Couldn't fix incorrect rc6 voltage\n");
  	}
 
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
  }
 
  static void __gen6_update_ring_freq(struct drm_device *dev)
@@ -5296,7 +5296,7 @@ static void cherryview_enable_rps(struct drm_device *dev)
 
  	/* 1a & 1b: Get forcewake during program sequence. Although the driver
  	 * hasn't enabled a state yet where we need forcewake, BIOS may have.*/
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
  	/* 2a: Program RC6 thresholds.*/
  	I915_WRITE(GEN6_RC6_WAKE_RATE_LIMIT, 40 << 16);
@@ -5364,7 +5364,7 @@ static void cherryview_enable_rps(struct drm_device *dev)
 
  	valleyview_set_rps(dev_priv->dev, dev_priv->rps.efficient_freq);
 
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
  }
 
  static void valleyview_enable_rps(struct drm_device *dev)
@@ -5385,7 +5385,7 @@ static void valleyview_enable_rps(struct drm_device *dev)
  	}
 
  	/* If VLV, Forcewake all wells, else re-direct to regular path */
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
  	I915_WRITE(GEN6_RP_UP_THRESHOLD, 59400);
  	I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 245000);
@@ -5445,7 +5445,7 @@ static void valleyview_enable_rps(struct drm_device *dev)
 
  	valleyview_set_rps(dev_priv->dev, dev_priv->rps.efficient_freq);
 
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
  }
 
  void ironlake_teardown_rc6(struct drm_device *dev)
drivers/gpu/drm/i915/intel_ringbuffer.c +2 -2
index 3887f1a..210728f 100644
@@ -536,7 +536,7 @@ static int init_ring_common(struct intel_engine_cs *ring)
  	struct drm_i915_gem_object *obj = ringbuf->obj;
  	int ret = 0;
 
-	gen6_gt_force_wake_get(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
 
  	if (!stop_ring(ring)) {
  		/* G45 ring initialization often fails to reset head to zero */
@@ -608,7 +608,7 @@ static int init_ring_common(struct intel_engine_cs *ring)
  	memset(&ring->hangcheck, 0, sizeof(ring->hangcheck));
 
  out:
-	gen6_gt_force_wake_put(dev_priv, FORCEWAKE_ALL);
+	intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
 
  	return ret;
  }
drivers/gpu/drm/i915/intel_uncore.c +8 -8
index 8021bec..509b9c9 100644
@@ -232,7 +232,7 @@ static int __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv)
  	return ret;
  }
 
-static void gen6_force_wake_timer(unsigned long arg)
+static void intel_uncore_fw_release_timer(unsigned long arg)
  {
  	struct intel_uncore_forcewake_domain *domain = (void *)arg;
  	unsigned long irqflags;
@@ -326,10 +326,10 @@ void intel_uncore_sanitize(struct drm_device *dev)
   * Generally this is called implicitly by the register read function. However,
   * if some sequence requires the GT to not power down then this function should
   * be called at the beginning of the sequence followed by a call to
- * gen6_gt_force_wake_put() at the end of the sequence.
+ * intel_uncore_forcewake_put() at the end of the sequence.
   */
-void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv,
-			    unsigned fw_domains)
+void intel_uncore_forcewake_get(struct drm_i915_private *dev_priv,
+				unsigned fw_domains)
  {
  	unsigned long irqflags;
  	struct intel_uncore_forcewake_domain *domain;
@@ -355,10 +355,10 @@ void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv,
  }
 
  /*
- * see gen6_gt_force_wake_get()
+ * see intel_uncore_forcewake_get()
   */
-void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv,
-			    unsigned fw_domains)
+void intel_uncore_forcewake_put(struct drm_i915_private *dev_priv,
+				unsigned fw_domains)
  {
  	unsigned long irqflags;
  	struct intel_uncore_forcewake_domain *domain;
@@ -383,7 +383,7 @@ void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv,
  	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
  }
 
-void assert_force_wake_inactive(struct drm_i915_private *dev_priv)
+void assert_forcewakes_inactive(struct drm_i915_private *dev_priv)
  {
  	struct intel_uncore_forcewake_domain *domain;
  	int id;

drivers/gpu/drm/i915/intel_uncore.c static void fw_domain_init(struct drm_i915_private *dev_priv, · line 930

Quoted 8 lines
drivers/gpu/drm/i915/intel_uncore.c +1 -1
@@ -924,7 +924,7 @@ static void fw_domain_init(struct drm_i915_private *dev_priv,
  	d->i915 = dev_priv;
  	d->id = domain_id;
 
-	setup_timer(&d->timer, gen6_force_wake_timer, (unsigned long)d);
+	setup_timer(&d->timer, intel_uncore_fw_release_timer, (unsigned long)d);
 
  	dev_priv->uncore.fw_domains |= (1 << domain_id);
  }

Looks fine

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

#16 lore raw

PATCH 8/8 drm/i915: Follow the forcewake domains type on hw accessors

In reply to Mika Kuoppala

Forcewake domain code uses unsigned int as a type for 'domains mask'. Bring the hw accessors inline with this.

---
 drivers/gpu/drm/i915/i915_drv.h     | 4 ++--
 drivers/gpu/drm/i915/intel_uncore.c | 8 ++++----
 2 files changed, 6 insertions(+), 6 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +2 -2
index a2a8536..917614e 100644
@@ -535,9 +535,9 @@ struct drm_i915_display_funcs {
 
 struct intel_uncore_funcs {
 	void (*force_wake_get)(struct drm_i915_private *dev_priv,
-							int fw_engine);
+							unsigned fw_domains);
 	void (*force_wake_put)(struct drm_i915_private *dev_priv,
-							int fw_engine);
+							unsigned fw_domains);
 
 	uint8_t  (*mmio_readb)(struct drm_i915_private *dev_priv, off_t offset, bool trace);
 	uint16_t (*mmio_readw)(struct drm_i915_private *dev_priv, off_t offset, bool trace);
drivers/gpu/drm/i915/intel_uncore.c +4 -4
index 509b9c9..be02aab 100644
@@ -122,7 +122,7 @@ fw_domain_posting_read(const struct intel_uncore_forcewake_domain *d)
 }
 
 static void
-fw_domains_get(struct drm_i915_private *dev_priv, int fw_domains)
+fw_domains_get(struct drm_i915_private *dev_priv, unsigned fw_domains)
 {
 	struct intel_uncore_forcewake_domain *d;
 	int id;
@@ -136,7 +136,7 @@ fw_domains_get(struct drm_i915_private *dev_priv, int fw_domains)
 }
 
 static void
-fw_domains_put(struct drm_i915_private *dev_priv, int fw_domains)
+fw_domains_put(struct drm_i915_private *dev_priv, unsigned fw_domains)
 {
 	struct intel_uncore_forcewake_domain *d;
 	int id;
@@ -181,7 +181,7 @@ static void __gen6_gt_wait_for_thread_c0(struct drm_i915_private *dev_priv)
 }
 
 static void fw_domains_get_with_thread_status(struct drm_i915_private *dev_priv,
-					      int fw_domains)
+					      unsigned fw_domains)
 {
 	fw_domains_get(dev_priv, fw_domains);
 
@@ -199,7 +199,7 @@ static void gen6_gt_check_fifodbg(struct drm_i915_private *dev_priv)
 }
 
 static void fw_domains_put_with_fifo(struct drm_i915_private *dev_priv,
-				     int fw_domains)
+				     unsigned fw_domains)
 {
 	fw_domains_put(dev_priv, fw_domains);
 	gen6_gt_check_fifodbg(dev_priv);
Signature · 6 lines
1.9.1

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

PATCH 8/8 drm/i915: Enum forcewake domains and domain identifiers

Make the domains and domain identifiers enums. To emphasize the difference in order to avoid mistakes.

---
 drivers/gpu/drm/i915/i915_drv.h     | 41 +++++++++++++++++----------------
 drivers/gpu/drm/i915/intel_uncore.c | 45 +++++++++++++++++++------------------
 2 files changed, 45 insertions(+), 41 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +22 -19
index 0d47397..5c6c372 100644
@@ -533,11 +533,28 @@ struct drm_i915_display_funcs {
 	void (*enable_backlight)(struct intel_connector *connector);
 };
 
+enum fw_domain_id {
+	FW_DOMAIN_ID_RENDER = 0,
+	FW_DOMAIN_ID_BLITTER,
+	FW_DOMAIN_ID_MEDIA,
+
+	FW_DOMAIN_ID_COUNT
+};
+
+enum fw_domains {
+	FORCEWAKE_RENDER = (1 << FW_DOMAIN_ID_RENDER),
+	FORCEWAKE_BLITTER = (1 << FW_DOMAIN_ID_BLITTER),
+	FORCEWAKE_MEDIA	= (1 << FW_DOMAIN_ID_MEDIA),
+	FORCEWAKE_ALL = (FORCEWAKE_RENDER |
+			 FORCEWAKE_BLITTER |
+			 FORCEWAKE_MEDIA)
+};
+
 struct intel_uncore_funcs {
 	void (*force_wake_get)(struct drm_i915_private *dev_priv,
-							int fw_engine);
+							enum fw_domains domains);
 	void (*force_wake_put)(struct drm_i915_private *dev_priv,
-							int fw_engine);
+							enum fw_domains domains);
 
 	uint8_t  (*mmio_readb)(struct drm_i915_private *dev_priv, off_t offset, bool trace);
 	uint16_t (*mmio_readw)(struct drm_i915_private *dev_priv, off_t offset, bool trace);
@@ -554,25 +571,17 @@ struct intel_uncore_funcs {
 				uint64_t val, bool trace);
 };
 
-enum {
-	FW_DOMAIN_ID_RENDER = 0,
-	FW_DOMAIN_ID_BLITTER,
-	FW_DOMAIN_ID_MEDIA,
-
-	FW_DOMAIN_ID_COUNT
-};
-
 struct intel_uncore {
 	spinlock_t lock; /** lock is also taken in irq contexts. */
 
 	struct intel_uncore_funcs funcs;
 
 	unsigned fifo_count;
-	unsigned fw_domains;
+	enum fw_domains fw_domains;
 
 	struct intel_uncore_forcewake_domain {
 		struct drm_i915_private *i915;
-		int id;
+		enum fw_domain_id id;
 		unsigned wake_count;
 		struct timer_list timer;
 		u32 reg_set;
@@ -582,12 +591,6 @@ struct intel_uncore {
 		u32 reg_post;
 		u32 val_reset;
 	} fw_domain[FW_DOMAIN_ID_COUNT];
-#define FORCEWAKE_RENDER	(1 << FW_DOMAIN_ID_RENDER)
-#define FORCEWAKE_BLITTER	(1 << FW_DOMAIN_ID_BLITTER)
-#define FORCEWAKE_MEDIA		(1 << FW_DOMAIN_ID_MEDIA)
-#define FORCEWAKE_ALL		(FORCEWAKE_RENDER | \
-				 FORCEWAKE_BLITTER | \
-				 FORCEWAKE_MEDIA)
 };
 
 /* Iterate over initialised fw domains */
@@ -2449,7 +2452,7 @@ extern void intel_uncore_init(struct drm_device *dev);
 extern void intel_uncore_check_errors(struct drm_device *dev);
 extern void intel_uncore_fini(struct drm_device *dev);
 extern void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore);
-const char *intel_uncore_forcewake_domain_to_str(const int domain_id);
+const char *intel_uncore_forcewake_domain_to_str(const enum fw_domain_id id);
 void intel_uncore_forcewake_get(struct drm_i915_private *dev_priv,
 				unsigned fw_domains);
 void intel_uncore_forcewake_put(struct drm_i915_private *dev_priv,
drivers/gpu/drm/i915/intel_uncore.c +23 -22

150 lines not on this page · raw message

Signature · 6 lines
1.9.1

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

Re: [PATCH 8/8] drm/i915: Enum forcewake domains and domain identifiers

Quoted 15 lines · Mika Kuoppala

Make the domains and domain identifiers enums. To emphasize the difference in order to avoid mistakes.

---
 drivers/gpu/drm/i915/i915_drv.h     | 41 +++++++++++++++++----------------
 drivers/gpu/drm/i915/intel_uncore.c | 45 +++++++++++++++++++------------------
 2 files changed, 45 insertions(+), 41 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +0 -0
index 0d47397..5c6c372 100644

drivers/gpu/drm/i915/i915_drv.h struct drm_i915_display_funcs { · line 536

Quoted 4 lines
drivers/gpu/drm/i915/i915_drv.h +1 -0
@@ -533,11 +533,28 @@ struct drm_i915_display_funcs {
 	void (*enable_backlight)(struct intel_connector *connector);
 };
 
+enum fw_domain_id {

Does anyone else think of "firmware" first when seeing "fw", and "forcewake" only after a moment?

Jani.

Quoted 252 lines
drivers/gpu/drm/i915/i915_drv.h +21 -19
+	FW_DOMAIN_ID_RENDER = 0,
+	FW_DOMAIN_ID_BLITTER,
+	FW_DOMAIN_ID_MEDIA,
+
+	FW_DOMAIN_ID_COUNT
+};
+
+enum fw_domains {
+	FORCEWAKE_RENDER = (1 << FW_DOMAIN_ID_RENDER),
+	FORCEWAKE_BLITTER = (1 << FW_DOMAIN_ID_BLITTER),
+	FORCEWAKE_MEDIA	= (1 << FW_DOMAIN_ID_MEDIA),
+	FORCEWAKE_ALL = (FORCEWAKE_RENDER |
+			 FORCEWAKE_BLITTER |
+			 FORCEWAKE_MEDIA)
+};
+
 struct intel_uncore_funcs {
 	void (*force_wake_get)(struct drm_i915_private *dev_priv,
-							int fw_engine);
+							enum fw_domains domains);
 	void (*force_wake_put)(struct drm_i915_private *dev_priv,
-							int fw_engine);
+							enum fw_domains domains);
 
 	uint8_t  (*mmio_readb)(struct drm_i915_private *dev_priv, off_t offset, bool trace);
 	uint16_t (*mmio_readw)(struct drm_i915_private *dev_priv, off_t offset, bool trace);
@@ -554,25 +571,17 @@ struct intel_uncore_funcs {
 				uint64_t val, bool trace);
 };
 
-enum {
-	FW_DOMAIN_ID_RENDER = 0,
-	FW_DOMAIN_ID_BLITTER,
-	FW_DOMAIN_ID_MEDIA,
-
-	FW_DOMAIN_ID_COUNT
-};
-
 struct intel_uncore {
 	spinlock_t lock; /** lock is also taken in irq contexts. */
 
 	struct intel_uncore_funcs funcs;
 
 	unsigned fifo_count;
-	unsigned fw_domains;
+	enum fw_domains fw_domains;
 
 	struct intel_uncore_forcewake_domain {
 		struct drm_i915_private *i915;
-		int id;
+		enum fw_domain_id id;
 		unsigned wake_count;
 		struct timer_list timer;
 		u32 reg_set;
@@ -582,12 +591,6 @@ struct intel_uncore {
 		u32 reg_post;
 		u32 val_reset;
 	} fw_domain[FW_DOMAIN_ID_COUNT];
-#define FORCEWAKE_RENDER	(1 << FW_DOMAIN_ID_RENDER)
-#define FORCEWAKE_BLITTER	(1 << FW_DOMAIN_ID_BLITTER)
-#define FORCEWAKE_MEDIA		(1 << FW_DOMAIN_ID_MEDIA)
-#define FORCEWAKE_ALL		(FORCEWAKE_RENDER | \
-				 FORCEWAKE_BLITTER | \
-				 FORCEWAKE_MEDIA)
 };
 
 /* Iterate over initialised fw domains */
@@ -2449,7 +2452,7 @@ extern void intel_uncore_init(struct drm_device *dev);
 extern void intel_uncore_check_errors(struct drm_device *dev);
 extern void intel_uncore_fini(struct drm_device *dev);
 extern void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore);
-const char *intel_uncore_forcewake_domain_to_str(const int domain_id);
+const char *intel_uncore_forcewake_domain_to_str(const enum fw_domain_id id);
 void intel_uncore_forcewake_get(struct drm_i915_private *dev_priv,
 				unsigned fw_domains);
 void intel_uncore_forcewake_put(struct drm_i915_private *dev_priv,
drivers/gpu/drm/i915/intel_uncore.c +23 -22
index 509b9c9..e802486 100644
@@ -49,7 +49,7 @@ static const char * const forcewake_domain_names[] = {
 };
 
 const char *
-intel_uncore_forcewake_domain_to_str(const int id)
+intel_uncore_forcewake_domain_to_str(const enum fw_domain_id id)
 {
 	BUILD_BUG_ON((sizeof(forcewake_domain_names)/sizeof(const char *)) !=
 		     FW_DOMAIN_ID_COUNT);
@@ -122,10 +122,10 @@ fw_domain_posting_read(const struct intel_uncore_forcewake_domain *d)
 }
 
 static void
-fw_domains_get(struct drm_i915_private *dev_priv, int fw_domains)
+fw_domains_get(struct drm_i915_private *dev_priv, enum fw_domains fw_domains)
 {
 	struct intel_uncore_forcewake_domain *d;
-	int id;
+	enum fw_domain_id id;
 
 	for_each_fw_domain_mask(d, fw_domains, dev_priv, id) {
 		fw_domain_wait_ack_clear(d);
@@ -136,10 +136,10 @@ fw_domains_get(struct drm_i915_private *dev_priv, int fw_domains)
 }
 
 static void
-fw_domains_put(struct drm_i915_private *dev_priv, int fw_domains)
+fw_domains_put(struct drm_i915_private *dev_priv, enum fw_domains fw_domains)
 {
 	struct intel_uncore_forcewake_domain *d;
-	int id;
+	enum fw_domain_id id;
 
 	for_each_fw_domain_mask(d, fw_domains, dev_priv, id)
 		fw_domain_put(d);
@@ -149,7 +149,7 @@ static void
 fw_domains_posting_read(struct drm_i915_private *dev_priv)
 {
 	struct intel_uncore_forcewake_domain *d;
-	int id;
+	enum fw_domain_id id;
 
 	/* No need to do for all, just do for first found */
 	for_each_fw_domain(d, dev_priv, id) {
@@ -159,10 +159,10 @@ fw_domains_posting_read(struct drm_i915_private *dev_priv)
 }
 
 static void
-fw_domains_reset(struct drm_i915_private *dev_priv, const unsigned fw_domains)
+fw_domains_reset(struct drm_i915_private *dev_priv, enum fw_domains fw_domains)
 {
 	struct intel_uncore_forcewake_domain *d;
-	int id;
+	enum fw_domain_id id;
 
 	for_each_fw_domain_mask(d, fw_domains, dev_priv, id)
 		fw_domain_reset(d);
@@ -181,7 +181,7 @@ static void __gen6_gt_wait_for_thread_c0(struct drm_i915_private *dev_priv)
 }
 
 static void fw_domains_get_with_thread_status(struct drm_i915_private *dev_priv,
-					      int fw_domains)
+					      enum fw_domains fw_domains)
 {
 	fw_domains_get(dev_priv, fw_domains);
 
@@ -199,7 +199,7 @@ static void gen6_gt_check_fifodbg(struct drm_i915_private *dev_priv)
 }
 
 static void fw_domains_put_with_fifo(struct drm_i915_private *dev_priv,
-				     int fw_domains)
+				     enum fw_domains fw_domains)
 {
 	fw_domains_put(dev_priv, fw_domains);
 	gen6_gt_check_fifodbg(dev_priv);
@@ -251,9 +251,10 @@ static void intel_uncore_fw_release_timer(unsigned long arg)
 void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
-	unsigned long irqflags, fw = 0;
+	unsigned long irqflags;
 	struct intel_uncore_forcewake_domain *domain;
-	int id;
+	enum fw_domain_id id;
+	enum fw_domains fw = 0;
 
 	/* Hold uncore.lock across reset to prevent any register access
 	 * with forcewake not set correctly
@@ -329,11 +330,11 @@ void intel_uncore_sanitize(struct drm_device *dev)
  * intel_uncore_forcewake_put() at the end of the sequence.
  */
 void intel_uncore_forcewake_get(struct drm_i915_private *dev_priv,
-				unsigned fw_domains)
+				enum fw_domains fw_domains)
 {
 	unsigned long irqflags;
 	struct intel_uncore_forcewake_domain *domain;
-	int id;
+	enum fw_domain_id id;
 
 	if (!dev_priv->uncore.funcs.force_wake_get)
 		return;
@@ -358,11 +359,11 @@ void intel_uncore_forcewake_get(struct drm_i915_private *dev_priv,
  * see intel_uncore_forcewake_get()
  */
 void intel_uncore_forcewake_put(struct drm_i915_private *dev_priv,
-				unsigned fw_domains)
+				enum fw_domains fw_domains)
 {
 	unsigned long irqflags;
 	struct intel_uncore_forcewake_domain *domain;
-	int id;
+	enum fw_domain_id id;
 
 	if (!dev_priv->uncore.funcs.force_wake_put)
 		return;
@@ -386,7 +387,7 @@ void intel_uncore_forcewake_put(struct drm_i915_private *dev_priv,
 void assert_forcewakes_inactive(struct drm_i915_private *dev_priv)
 {
 	struct intel_uncore_forcewake_domain *domain;
-	int id;
+	enum fw_domain_id id;
 
 	if (!dev_priv->uncore.funcs.force_wake_get)
 		return;
@@ -559,10 +560,10 @@ __gen2_read(64)
 	return val
 
 static inline void __force_wake_get(struct drm_i915_private *dev_priv,
-				    unsigned fw_domains)
+				    enum fw_domains fw_domains)
 {
 	struct intel_uncore_forcewake_domain *domain;
-	int id;
+	enum fw_domain_id id;
 
 	if (WARN_ON(!fw_domains))
 		return;
@@ -626,7 +627,7 @@ chv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
 #define __gen9_read(x) \
 static u##x \
 gen9_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
-	unsigned fw_engine; \
+	enum fw_domains fw_engine; \
 	GEN6_READ_HEADER(x); \
 	if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg)))	\
 		fw_engine = 0; \
@@ -826,7 +827,7 @@ static bool is_gen9_shadowed(struct drm_i915_private *dev_priv, u32 reg)
 static void \
 gen9_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, \
 		bool trace) { \
-	unsigned fw_engine; \
+	enum fw_domains fw_engine; \
 	GEN6_WRITE_HEADER; \
 	if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg)) ||	\
 	    is_gen9_shadowed(dev_priv, reg)) \
@@ -892,7 +893,7 @@ do { \
 
 
 static void fw_domain_init(struct drm_i915_private *dev_priv,
-			   u32 domain_id, u32 reg_set, u32 reg_ack)
+			   enum fw_domain_id domain_id, u32 reg_set, u32 reg_ack)
 {
 	struct intel_uncore_forcewake_domain *d;
 
Signature · 6 lines
1.9.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 5 lines
Jani Nikula, Intel Open Source Technology Center
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#19 lore raw

Re: [PATCH 8/8] drm/i915: Enum forcewake domains and domain identifiers

Quoted 19 lines · Jani Nikula
Quoted 18 lines · Mika Kuoppala

Make the domains and domain identifiers enums. To emphasize the difference in order to avoid mistakes.

---
 drivers/gpu/drm/i915/i915_drv.h     | 41 +++++++++++++++++----------------
 drivers/gpu/drm/i915/intel_uncore.c | 45 +++++++++++++++++++------------------
 2 files changed, 45 insertions(+), 41 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +0 -0
index 0d47397..5c6c372 100644
@@ -533,11 +533,28 @@ struct drm_i915_display_funcs {
 	void (*enable_backlight)(struct intel_connector *connector);
 };
 
Quoted 4 lines

drivers/gpu/drm/i915/i915_drv.h struct drm_i915_display_funcs { · line 536

Quoted 1 line
drivers/gpu/drm/i915/i915_drv.h +1 -0
+enum fw_domain_id {

Does anyone else think of "firmware" first when seeing "fw", and "forcewake" only after a moment?

Hm yeah.
-Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#20 lore raw

PATCH 8/8 drm/i915: Enum forcewake domains and domain identifiers

In reply to Jani Nikula

Make the domains and domain identifiers enums. To emphasize the difference in order to avoid mistakes.

v2: s/fw_domain/forcewake_domain (Jani)

---
 drivers/gpu/drm/i915/i915_drv.h     | 45 +++++++++++++++++++-----------------
 drivers/gpu/drm/i915/intel_uncore.c | 46 +++++++++++++++++++------------------
 2 files changed, 48 insertions(+), 43 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +24 -21
index 0d47397..8591a5c 100644
@@ -533,11 +533,28 @@ struct drm_i915_display_funcs {
 	void (*enable_backlight)(struct intel_connector *connector);
 };
 
+enum forcewake_domain_id {
+	FW_DOMAIN_ID_RENDER = 0,
+	FW_DOMAIN_ID_BLITTER,
+	FW_DOMAIN_ID_MEDIA,
+
+	FW_DOMAIN_ID_COUNT
+};
+
+enum forcewake_domains {
+	FORCEWAKE_RENDER = (1 << FW_DOMAIN_ID_RENDER),
+	FORCEWAKE_BLITTER = (1 << FW_DOMAIN_ID_BLITTER),
+	FORCEWAKE_MEDIA	= (1 << FW_DOMAIN_ID_MEDIA),
+	FORCEWAKE_ALL = (FORCEWAKE_RENDER |
+			 FORCEWAKE_BLITTER |
+			 FORCEWAKE_MEDIA)
+};
+
 struct intel_uncore_funcs {
 	void (*force_wake_get)(struct drm_i915_private *dev_priv,
-							int fw_engine);
+							enum forcewake_domains domains);
 	void (*force_wake_put)(struct drm_i915_private *dev_priv,
-							int fw_engine);
+							enum forcewake_domains domains);
 
 	uint8_t  (*mmio_readb)(struct drm_i915_private *dev_priv, off_t offset, bool trace);
 	uint16_t (*mmio_readw)(struct drm_i915_private *dev_priv, off_t offset, bool trace);
@@ -554,25 +571,17 @@ struct intel_uncore_funcs {
 				uint64_t val, bool trace);
 };
 
-enum {
-	FW_DOMAIN_ID_RENDER = 0,
-	FW_DOMAIN_ID_BLITTER,
-	FW_DOMAIN_ID_MEDIA,
-
-	FW_DOMAIN_ID_COUNT
-};
-
 struct intel_uncore {
 	spinlock_t lock; /** lock is also taken in irq contexts. */
 
 	struct intel_uncore_funcs funcs;
 
 	unsigned fifo_count;
-	unsigned fw_domains;
+	enum forcewake_domains fw_domains;
 
 	struct intel_uncore_forcewake_domain {
 		struct drm_i915_private *i915;
-		int id;
+		enum forcewake_domain_id id;
 		unsigned wake_count;
 		struct timer_list timer;
 		u32 reg_set;
@@ -582,12 +591,6 @@ struct intel_uncore {
 		u32 reg_post;
 		u32 val_reset;
 	} fw_domain[FW_DOMAIN_ID_COUNT];
-#define FORCEWAKE_RENDER	(1 << FW_DOMAIN_ID_RENDER)
-#define FORCEWAKE_BLITTER	(1 << FW_DOMAIN_ID_BLITTER)
-#define FORCEWAKE_MEDIA		(1 << FW_DOMAIN_ID_MEDIA)
-#define FORCEWAKE_ALL		(FORCEWAKE_RENDER | \
-				 FORCEWAKE_BLITTER | \
-				 FORCEWAKE_MEDIA)
 };
 
 /* Iterate over initialised fw domains */
@@ -2449,11 +2452,11 @@ extern void intel_uncore_init(struct drm_device *dev);
 extern void intel_uncore_check_errors(struct drm_device *dev);
 extern void intel_uncore_fini(struct drm_device *dev);
 extern void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore);
-const char *intel_uncore_forcewake_domain_to_str(const int domain_id);
+const char *intel_uncore_forcewake_domain_to_str(const enum forcewake_domain_id id);
 void intel_uncore_forcewake_get(struct drm_i915_private *dev_priv,
-				unsigned fw_domains);
+				enum forcewake_domains domains);
 void intel_uncore_forcewake_put(struct drm_i915_private *dev_priv,
-				unsigned fw_domains);
+				enum forcewake_domains domains);
 void assert_forcewakes_inactive(struct drm_i915_private *dev_priv);
 
 void
drivers/gpu/drm/i915/intel_uncore.c +24 -22

151 lines not on this page · raw message

Signature · 6 lines
1.9.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Reviewed-by on v1 8/8
#21 lore raw

Re: [PATCH 8/8] drm/i915: Enum forcewake domains and domain identifiers

In reply to Mika Kuoppala
Quoted 256 lines · Mika Kuoppala

Make the domains and domain identifiers enums. To emphasize the difference in order to avoid mistakes.

---
  drivers/gpu/drm/i915/i915_drv.h     | 41 +++++++++++++++++----------------
  drivers/gpu/drm/i915/intel_uncore.c | 45 +++++++++++++++++++------------------
  2 files changed, 45 insertions(+), 41 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +22 -19
index 0d47397..5c6c372 100644
@@ -533,11 +533,28 @@ struct drm_i915_display_funcs {
  	void (*enable_backlight)(struct intel_connector *connector);
  };
 
+enum fw_domain_id {
+	FW_DOMAIN_ID_RENDER = 0,
+	FW_DOMAIN_ID_BLITTER,
+	FW_DOMAIN_ID_MEDIA,
+
+	FW_DOMAIN_ID_COUNT
+};
+
+enum fw_domains {
+	FORCEWAKE_RENDER = (1 << FW_DOMAIN_ID_RENDER),
+	FORCEWAKE_BLITTER = (1 << FW_DOMAIN_ID_BLITTER),
+	FORCEWAKE_MEDIA	= (1 << FW_DOMAIN_ID_MEDIA),
+	FORCEWAKE_ALL = (FORCEWAKE_RENDER |
+			 FORCEWAKE_BLITTER |
+			 FORCEWAKE_MEDIA)
+};
+
  struct intel_uncore_funcs {
  	void (*force_wake_get)(struct drm_i915_private *dev_priv,
-							int fw_engine);
+							enum fw_domains domains);
  	void (*force_wake_put)(struct drm_i915_private *dev_priv,
-							int fw_engine);
+							enum fw_domains domains);
 
  	uint8_t  (*mmio_readb)(struct drm_i915_private *dev_priv, off_t offset, bool trace);
  	uint16_t (*mmio_readw)(struct drm_i915_private *dev_priv, off_t offset, bool trace);
@@ -554,25 +571,17 @@ struct intel_uncore_funcs {
  				uint64_t val, bool trace);
  };
 
-enum {
-	FW_DOMAIN_ID_RENDER = 0,
-	FW_DOMAIN_ID_BLITTER,
-	FW_DOMAIN_ID_MEDIA,
-
-	FW_DOMAIN_ID_COUNT
-};
-
  struct intel_uncore {
  	spinlock_t lock; /** lock is also taken in irq contexts. */
 
  	struct intel_uncore_funcs funcs;
 
  	unsigned fifo_count;
-	unsigned fw_domains;
+	enum fw_domains fw_domains;
 
  	struct intel_uncore_forcewake_domain {
  		struct drm_i915_private *i915;
-		int id;
+		enum fw_domain_id id;
  		unsigned wake_count;
  		struct timer_list timer;
  		u32 reg_set;
@@ -582,12 +591,6 @@ struct intel_uncore {
  		u32 reg_post;
  		u32 val_reset;
  	} fw_domain[FW_DOMAIN_ID_COUNT];
-#define FORCEWAKE_RENDER	(1 << FW_DOMAIN_ID_RENDER)
-#define FORCEWAKE_BLITTER	(1 << FW_DOMAIN_ID_BLITTER)
-#define FORCEWAKE_MEDIA		(1 << FW_DOMAIN_ID_MEDIA)
-#define FORCEWAKE_ALL		(FORCEWAKE_RENDER | \
-				 FORCEWAKE_BLITTER | \
-				 FORCEWAKE_MEDIA)
  };
 
  /* Iterate over initialised fw domains */
@@ -2449,7 +2452,7 @@ extern void intel_uncore_init(struct drm_device *dev);
  extern void intel_uncore_check_errors(struct drm_device *dev);
  extern void intel_uncore_fini(struct drm_device *dev);
  extern void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore);
-const char *intel_uncore_forcewake_domain_to_str(const int domain_id);
+const char *intel_uncore_forcewake_domain_to_str(const enum fw_domain_id id);
  void intel_uncore_forcewake_get(struct drm_i915_private *dev_priv,
  				unsigned fw_domains);
  void intel_uncore_forcewake_put(struct drm_i915_private *dev_priv,
drivers/gpu/drm/i915/intel_uncore.c +22 -21
index 509b9c9..e802486 100644
@@ -49,7 +49,7 @@ static const char * const forcewake_domain_names[] = {
  };
 
  const char *
-intel_uncore_forcewake_domain_to_str(const int id)
+intel_uncore_forcewake_domain_to_str(const enum fw_domain_id id)
  {
  	BUILD_BUG_ON((sizeof(forcewake_domain_names)/sizeof(const char *)) !=
  		     FW_DOMAIN_ID_COUNT);
@@ -122,10 +122,10 @@ fw_domain_posting_read(const struct intel_uncore_forcewake_domain *d)
  }
 
  static void
-fw_domains_get(struct drm_i915_private *dev_priv, int fw_domains)
+fw_domains_get(struct drm_i915_private *dev_priv, enum fw_domains fw_domains)
  {
  	struct intel_uncore_forcewake_domain *d;
-	int id;
+	enum fw_domain_id id;
 
  	for_each_fw_domain_mask(d, fw_domains, dev_priv, id) {
  		fw_domain_wait_ack_clear(d);
@@ -136,10 +136,10 @@ fw_domains_get(struct drm_i915_private *dev_priv, int fw_domains)
  }
 
  static void
-fw_domains_put(struct drm_i915_private *dev_priv, int fw_domains)
+fw_domains_put(struct drm_i915_private *dev_priv, enum fw_domains fw_domains)
  {
  	struct intel_uncore_forcewake_domain *d;
-	int id;
+	enum fw_domain_id id;
 
  	for_each_fw_domain_mask(d, fw_domains, dev_priv, id)
  		fw_domain_put(d);
@@ -149,7 +149,7 @@ static void
  fw_domains_posting_read(struct drm_i915_private *dev_priv)
  {
  	struct intel_uncore_forcewake_domain *d;
-	int id;
+	enum fw_domain_id id;
 
  	/* No need to do for all, just do for first found */
  	for_each_fw_domain(d, dev_priv, id) {
@@ -159,10 +159,10 @@ fw_domains_posting_read(struct drm_i915_private *dev_priv)
  }
 
  static void
-fw_domains_reset(struct drm_i915_private *dev_priv, const unsigned fw_domains)
+fw_domains_reset(struct drm_i915_private *dev_priv, enum fw_domains fw_domains)
  {
  	struct intel_uncore_forcewake_domain *d;
-	int id;
+	enum fw_domain_id id;
 
  	for_each_fw_domain_mask(d, fw_domains, dev_priv, id)
  		fw_domain_reset(d);
@@ -181,7 +181,7 @@ static void __gen6_gt_wait_for_thread_c0(struct drm_i915_private *dev_priv)
  }
 
  static void fw_domains_get_with_thread_status(struct drm_i915_private *dev_priv,
-					      int fw_domains)
+					      enum fw_domains fw_domains)
  {
  	fw_domains_get(dev_priv, fw_domains);
 
@@ -199,7 +199,7 @@ static void gen6_gt_check_fifodbg(struct drm_i915_private *dev_priv)
  }
 
  static void fw_domains_put_with_fifo(struct drm_i915_private *dev_priv,
-				     int fw_domains)
+				     enum fw_domains fw_domains)
  {
  	fw_domains_put(dev_priv, fw_domains);
  	gen6_gt_check_fifodbg(dev_priv);
@@ -251,9 +251,10 @@ static void intel_uncore_fw_release_timer(unsigned long arg)
  void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore)
  {
  	struct drm_i915_private *dev_priv = dev->dev_private;
-	unsigned long irqflags, fw = 0;
+	unsigned long irqflags;
  	struct intel_uncore_forcewake_domain *domain;
-	int id;
+	enum fw_domain_id id;
+	enum fw_domains fw = 0;
 
  	/* Hold uncore.lock across reset to prevent any register access
  	 * with forcewake not set correctly
@@ -329,11 +330,11 @@ void intel_uncore_sanitize(struct drm_device *dev)
   * intel_uncore_forcewake_put() at the end of the sequence.
   */
  void intel_uncore_forcewake_get(struct drm_i915_private *dev_priv,
-				unsigned fw_domains)
+				enum fw_domains fw_domains)
  {
  	unsigned long irqflags;
  	struct intel_uncore_forcewake_domain *domain;
-	int id;
+	enum fw_domain_id id;
 
  	if (!dev_priv->uncore.funcs.force_wake_get)
  		return;
@@ -358,11 +359,11 @@ void intel_uncore_forcewake_get(struct drm_i915_private *dev_priv,
   * see intel_uncore_forcewake_get()
   */
  void intel_uncore_forcewake_put(struct drm_i915_private *dev_priv,
-				unsigned fw_domains)
+				enum fw_domains fw_domains)
  {
  	unsigned long irqflags;
  	struct intel_uncore_forcewake_domain *domain;
-	int id;
+	enum fw_domain_id id;
 
  	if (!dev_priv->uncore.funcs.force_wake_put)
  		return;
@@ -386,7 +387,7 @@ void intel_uncore_forcewake_put(struct drm_i915_private *dev_priv,
  void assert_forcewakes_inactive(struct drm_i915_private *dev_priv)
  {
  	struct intel_uncore_forcewake_domain *domain;
-	int id;
+	enum fw_domain_id id;
 
  	if (!dev_priv->uncore.funcs.force_wake_get)
  		return;
@@ -559,10 +560,10 @@ __gen2_read(64)
  	return val
 
  static inline void __force_wake_get(struct drm_i915_private *dev_priv,
-				    unsigned fw_domains)
+				    enum fw_domains fw_domains)
  {
  	struct intel_uncore_forcewake_domain *domain;
-	int id;
+	enum fw_domain_id id;
 
  	if (WARN_ON(!fw_domains))
  		return;
@@ -626,7 +627,7 @@ chv_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
  #define __gen9_read(x) \
  static u##x \
  gen9_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
-	unsigned fw_engine; \
+	enum fw_domains fw_engine; \
  	GEN6_READ_HEADER(x); \
  	if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg)))	\
  		fw_engine = 0; \
@@ -826,7 +827,7 @@ static bool is_gen9_shadowed(struct drm_i915_private *dev_priv, u32 reg)
  static void \
  gen9_write##x(struct drm_i915_private *dev_priv, off_t reg, u##x val, \
  		bool trace) { \
-	unsigned fw_engine; \
+	enum fw_domains fw_engine; \
  	GEN6_WRITE_HEADER; \
  	if (!SKL_NEEDS_FORCE_WAKE((dev_priv), (reg)) ||	\
  	    is_gen9_shadowed(dev_priv, reg)) \

drivers/gpu/drm/i915/intel_uncore.c do { \ · line 899

Quoted 8 lines
drivers/gpu/drm/i915/intel_uncore.c +1 -1
@@ -892,7 +893,7 @@ do { \
 
 
  static void fw_domain_init(struct drm_i915_private *dev_priv,
-			   u32 domain_id, u32 reg_set, u32 reg_ack)
+			   enum fw_domain_id domain_id, u32 reg_set, u32 reg_ack)
  {
  	struct intel_uncore_forcewake_domain *d;
 

Looks fine

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

Tested-by on v1 8/8
#22 lore raw

Re: [PATCH 8/8] drm/i915: Follow the forcewake domains type on hw accessors

In reply to Mika Kuoppala

-------------------------------------Summary-------------------------------------
Platform Delta drm-intel-nightly Series Applied
PNV 364/364 364/364
ILK +1-3 364/366 362/366

SNB                 -19              448/450              429/450
IVB                                  497/498              497/498
BYT                                  289/289              289/289
HSW                 -83              563/564              480/564
BDW                 -4              417/417              413/417
-------------------------------------Detailed-------------------------------------
Platform  Test                                drm-intel-nightly          Series Applied
 ILK  igt_kms_flip_blocking-absolute-wf_vblank-interruptible      DMESG_WARN(1, M26)PASS(3, M26)      DMESG_WARN(1, M26)
*ILK  igt_kms_flip_busy-flip-interruptible      PASS(4, M26)      DMESG_WARN(1, M26)
*ILK  igt_kms_flip_flip-vs-panning      PASS(3, M26)      DMESG_WARN(1, M26)
 ILK  igt_kms_flip_wf_vblank-ts-check      DMESG_WARN(1, M26)PASS(10, M26M37)      PASS(1, M26)
*SNB  igt_pm_rpm_cursor      PASS(2, M35)      DMESG_WARN(1, M22)
*SNB  igt_pm_rpm_cursor-dpms      PASS(2, M35)      DMESG_WARN(1, M22)
*SNB  igt_pm_rpm_debugfs-forcewake-user      PASS(1, M35)      DMESG_WARN(1, M22)
*SNB  igt_pm_rpm_dpms-mode-unset-non-lpsp      PASS(1, M35)      DMESG_WARN(1, M22)
*SNB  igt_pm_rpm_dpms-non-lpsp      PASS(1, M35)      DMESG_WARN(1, M22)
*SNB  igt_pm_rpm_drm-resources-equal      PASS(1, M35)      DMESG_WARN(1, M22)
*SNB  igt_pm_rpm_fences      PASS(1, M35)      DMESG_WARN(1, M22)
*SNB  igt_pm_rpm_fences-dpms      PASS(1, M35)      DMESG_WARN(1, M22)
*SNB  igt_pm_rpm_gem-execbuf      PASS(1, M35)      DMESG_WARN(1, M22)
*SNB  igt_pm_rpm_gem-idle      PASS(1, M35)      DMESG_WARN(1, M22)
*SNB  igt_pm_rpm_gem-mmap-cpu      PASS(1, M35)      DMESG_WARN(1, M22)
*SNB  igt_pm_rpm_gem-mmap-gtt      PASS(1, M35)      DMESG_WARN(1, M22)
*SNB  igt_pm_rpm_gem-pread      PASS(1, M35)      DMESG_WARN(1, M22)
*SNB  igt_pm_rpm_i2c      PASS(1, M35)      DMESG_WARN(1, M22)
*SNB  igt_pm_rpm_modeset-non-lpsp      PASS(1, M35)      DMESG_WARN(1, M22)
*SNB  igt_pm_rpm_modeset-non-lpsp-stress-no-wait      PASS(1, M35)      DMESG_WARN(1, M22)
*SNB  igt_pm_rpm_pci-d3-state      PASS(1, M35)      DMESG_WARN(1, M22)
*SNB  igt_pm_rpm_reg-read-ioctl      PASS(1, M35)      DMESG_WARN(1, M22)
*SNB  igt_pm_rpm_rte      PASS(1, M35)      DMESG_WARN(1, M22)
*HSW  igt_gem_concurrent_blit_cpu-bcs-overwrite-source-forked      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_vblank-vs-dpms-rpm      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_pipe_crc_basic_read-crc-pipe-A      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_pipe_crc_basic_read-crc-pipe-A-frame-sequence      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_pipe_crc_basic_read-crc-pipe-B      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_pipe_crc_basic_read-crc-pipe-B-frame-sequence      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_pipe_crc_basic_read-crc-pipe-C      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_pipe_crc_basic_read-crc-pipe-C-frame-sequence      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-panning-bottom-right-pipe-A-plane-1      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-panning-bottom-right-pipe-A-plane-2      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-panning-bottom-right-pipe-B-plane-1      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-panning-bottom-right-pipe-B-plane-2      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-panning-bottom-right-pipe-C-plane-1      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-panning-bottom-right-pipe-C-plane-2      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-panning-top-left-pipe-A-plane-1      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-panning-top-left-pipe-A-plane-2      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-panning-top-left-pipe-B-plane-1      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-panning-top-left-pipe-B-plane-2      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-panning-top-left-pipe-C-plane-1      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-panning-top-left-pipe-C-plane-2      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-position-covered-pipe-A-plane-1      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-position-covered-pipe-A-plane-2      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-position-covered-pipe-B-plane-1      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-position-covered-pipe-C-plane-1      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-position-covered-pipe-C-plane-2      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-position-hole-pipe-A-plane-1      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-position-hole-pipe-A-plane-2      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-position-hole-pipe-B-plane-1      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-position-hole-pipe-B-plane-2      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-position-hole-pipe-C-plane-1      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_plane_plane-position-hole-pipe-C-plane-2      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_lpsp_non-edp      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_lpsp_screens-disabled      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_cursor      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_cursor-dpms      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_debugfs-forcewake-user      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_debugfs-read      PASS(3, M40M20)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_dpms-mode-unset-non-lpsp      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_dpms-non-lpsp      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_drm-resources-equal      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_fences      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_fences-dpms      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_gem-execbuf      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_gem-idle      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_gem-mmap-cpu      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_gem-mmap-gtt      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_gem-pread      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_i2c      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_modeset-non-lpsp      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_modeset-non-lpsp-stress-no-wait      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_pci-d3-state      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_reg-read-ioctl      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_rte      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_pm_rpm_sysfs-read      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_gem_partial_pwrite_pread_reads-snoop      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_absolute-wf_vblank      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_absolute-wf_vblank-interruptible      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_blocking-absolute-wf_vblank      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_blocking-absolute-wf_vblank-interruptible      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_blocking-wf_vblank      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_busy-flip      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_busy-flip-interruptible      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_dpms-off-confusion      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_dpms-off-confusion-interruptible      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_flip-vs-absolute-wf_vblank      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_flip-vs-absolute-wf_vblank-interruptible      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_flip-vs-blocking-wf-vblank      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_flip-vs-panning      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_flip-vs-panning-interruptible      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_flip-vs-rmfb      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_flip-vs-rmfb-interruptible      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_flip-vs-wf_vblank      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_flip-vs-wf_vblank-interruptible      PASS(2, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_plain-flip      PASS(1, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_plain-flip-fb-recreate      PASS(1, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_plain-flip-fb-recreate-interruptible      PASS(1, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_plain-flip-interruptible      PASS(1, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_plain-flip-ts-check      PASS(1, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_plain-flip-ts-check-interruptible      PASS(1, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_wf_vblank      PASS(1, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_wf_vblank-interruptible      PASS(1, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_wf_vblank-ts-check      PASS(1, M40)      DMESG_WARN(1, M40)
*HSW  igt_kms_flip_wf_vblank-ts-check-interruptible      PASS(1, M40)      DMESG_WARN(1, M40)
*BDW  igt_gem_persistent_relocs_forked-interruptible      PASS(2, M30)      DMESG_WARN(1, M30)
*BDW  igt_gem_persistent_relocs_forked-interruptible-faulting-reloc-thrash-inactive      PASS(2, M30)      DMESG_WARN(1, M30)
 BDW  igt_gem_render_linear_blits      TIMEOUT(1, M30)PASS(2, M30M28)      TIMEOUT(1, M30)
 BDW  igt_gem_render_tiled_blits      TIMEOUT(1, M30)PASS(2, M30M28)      TIMEOUT(1, M30)

Note: You need to pay more attention to line start with '*'
_______________________________________________
Intel-gfx mailing list [email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Reviewed-by on v1 8/8
#23 lore raw

Re: [PATCH 8/8] drm/i915: Follow the forcewake domains type on hw accessors

In reply to Mika Kuoppala
Quoted 58 lines · Mika Kuoppala

Forcewake domain code uses unsigned int as a type for 'domains mask'. Bring the hw accessors inline with this.

---
  drivers/gpu/drm/i915/i915_drv.h     | 4 ++--
  drivers/gpu/drm/i915/intel_uncore.c | 8 ++++----
  2 files changed, 6 insertions(+), 6 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +2 -2
index a2a8536..917614e 100644
@@ -535,9 +535,9 @@ struct drm_i915_display_funcs {
 
  struct intel_uncore_funcs {
  	void (*force_wake_get)(struct drm_i915_private *dev_priv,
-							int fw_engine);
+							unsigned fw_domains);
  	void (*force_wake_put)(struct drm_i915_private *dev_priv,
-							int fw_engine);
+							unsigned fw_domains);
 
  	uint8_t  (*mmio_readb)(struct drm_i915_private *dev_priv, off_t offset, bool trace);
  	uint16_t (*mmio_readw)(struct drm_i915_private *dev_priv, off_t offset, bool trace);
drivers/gpu/drm/i915/intel_uncore.c +3 -3
index 509b9c9..be02aab 100644
@@ -122,7 +122,7 @@ fw_domain_posting_read(const struct intel_uncore_forcewake_domain *d)
  }
 
  static void
-fw_domains_get(struct drm_i915_private *dev_priv, int fw_domains)
+fw_domains_get(struct drm_i915_private *dev_priv, unsigned fw_domains)
  {
  	struct intel_uncore_forcewake_domain *d;
  	int id;
@@ -136,7 +136,7 @@ fw_domains_get(struct drm_i915_private *dev_priv, int fw_domains)
  }
 
  static void
-fw_domains_put(struct drm_i915_private *dev_priv, int fw_domains)
+fw_domains_put(struct drm_i915_private *dev_priv, unsigned fw_domains)
  {
  	struct intel_uncore_forcewake_domain *d;
  	int id;
@@ -181,7 +181,7 @@ static void __gen6_gt_wait_for_thread_c0(struct drm_i915_private *dev_priv)
  }
 
  static void fw_domains_get_with_thread_status(struct drm_i915_private *dev_priv,
-					      int fw_domains)
+					      unsigned fw_domains)
  {
  	fw_domains_get(dev_priv, fw_domains);
 

drivers/gpu/drm/i915/intel_uncore.c static void gen6_gt_check_fifodbg(struct drm_i915_private *dev_priv) · line 205

Quoted 8 lines
drivers/gpu/drm/i915/intel_uncore.c +1 -1
@@ -199,7 +199,7 @@ static void gen6_gt_check_fifodbg(struct drm_i915_private *dev_priv)
  }
 
  static void fw_domains_put_with_fifo(struct drm_i915_private *dev_priv,
-				     int fw_domains)
+				     unsigned fw_domains)
  {
  	fw_domains_put(dev_priv, fw_domains);
  	gen6_gt_check_fifodbg(dev_priv);

Looks fine

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

#24 lore raw

PATCH 1/1 drm/i915: FIFO space query code refactor

In reply to Mika Kuoppala

When querying the GTFIFOCTL register to check the FIFO space, the read value must be masked. The operation is repeated explicitly in several places. This change refactors the read-and-mask code into a function call.

v2*: rebased on top of Mika's forcewake changes

Change-Id Id1a9f3785cb20b82d4caa330c37b31e4e384a3ef
---
 drivers/gpu/drm/i915/intel_uncore.c |   19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +12 -7
index 5e2b11f..fc3bcf5 100644
@@ -205,6 +205,13 @@ static void fw_domains_put_with_fifo(struct drm_i915_private *dev_priv,
 	gen6_gt_check_fifodbg(dev_priv);
 }
 
+static inline u32 fifo_free_entries(struct drm_i915_private *dev_priv)
+{
+	u32 count = __raw_i915_read32(dev_priv, GTFIFOCTL);
+
+	return count & GT_FIFO_FREE_ENTRIES_MASK;
+}
+
 static int __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv)
 {
 	int ret = 0;
@@ -212,16 +219,15 @@ static int __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv)
 	/* On VLV, FIFO will be shared by both SW and HW.
 	 * So, we need to read the FREE_ENTRIES everytime */
 	if (IS_VALLEYVIEW(dev_priv->dev))
-		dev_priv->uncore.fifo_count =
-			__raw_i915_read32(dev_priv, GTFIFOCTL) &
-						GT_FIFO_FREE_ENTRIES_MASK;
+		dev_priv->uncore.fifo_count = fifo_free_entries(dev_priv);
 
 	if (dev_priv->uncore.fifo_count < GT_FIFO_NUM_RESERVED_ENTRIES) {
 		int loop = 500;
-		u32 fifo = __raw_i915_read32(dev_priv, GTFIFOCTL) & GT_FIFO_FREE_ENTRIES_MASK;
+		u32 fifo = fifo_free_entries(dev_priv);
+
 		while (fifo <= GT_FIFO_NUM_RESERVED_ENTRIES && loop--) {
 			udelay(10);
-			fifo = __raw_i915_read32(dev_priv, GTFIFOCTL) & GT_FIFO_FREE_ENTRIES_MASK;
+			fifo = fifo_free_entries(dev_priv);
 		}
 		if (WARN_ON(loop < 0 && fifo <= GT_FIFO_NUM_RESERVED_ENTRIES))
 			++ret;
@@ -277,8 +283,7 @@ void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore)
 
 		if (IS_GEN6(dev) || IS_GEN7(dev))
 			dev_priv->uncore.fifo_count =
-				__raw_i915_read32(dev_priv, GTFIFOCTL) &
-				GT_FIFO_FREE_ENTRIES_MASK;
+				fifo_free_entries(dev_priv);
 	}
 
 	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Reviewed-by on v1 1/8
#25 lore raw
In reply to Mika Kuoppala
Quoted 81 lines · Mika Kuoppala

From: Chris Wilson <[email protected]>

Calling intel_runtime_pm_put() is illegal from a soft-irq context, so revert the crude hack

commit aa0b3b5bb8768c1a6a6788869d9c7015eae7e80c

Author: Paulo Zanoni <[email protected]>
Date:   Tue Apr 1 14:55:07 2014 -0300

drm/i915: don't schedule force_wake_timer at gen6_read

and apply the single line corrective instead.

References: https://bugs.freedesktop.org/show_bug.cgi?id=80913

Cc Paulo Zanoni <[email protected]>
Cc Daniel Vetter <[email protected]>
---
  drivers/gpu/drm/i915/i915_drv.c     |  1 +
  drivers/gpu/drm/i915/intel_uncore.c | 18 ++++++------------
  2 files changed, 7 insertions(+), 12 deletions(-)
drivers/gpu/drm/i915/i915_drv.c +1 -0
index 71be3c9..706b122 100644
@@ -1402,6 +1402,7 @@ static int intel_runtime_suspend(struct device *device)
  	}
 
  	del_timer_sync(&dev_priv->gpu_error.hangcheck_timer);
+	intel_uncore_forcewake_reset(dev, false);
  	dev_priv->pm.suspended = true;
 
  	/*
drivers/gpu/drm/i915/intel_uncore.c +2 -7
index 46de8d7..38ac389 100644
@@ -449,8 +449,6 @@ static void gen6_force_wake_timer(unsigned long arg)
  	if (--dev_priv->uncore.forcewake_count == 0)
  		dev_priv->uncore.funcs.force_wake_put(dev_priv, FORCEWAKE_ALL);
  	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
-
-	intel_runtime_pm_put(dev_priv);
  }
 
  void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore)
@@ -586,7 +584,6 @@ void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
  void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
  {
  	unsigned long irqflags;
-	bool delayed = false;
 
  	if (!dev_priv->uncore.funcs.force_wake_put)
  		return;
@@ -603,21 +600,19 @@ void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
  		goto out;
  	}
 
-
  	spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
  	WARN_ON(!dev_priv->uncore.forcewake_count);
 
  	if (--dev_priv->uncore.forcewake_count == 0) {
  		dev_priv->uncore.forcewake_count++;
-		delayed = true;
  		mod_timer_pinned(&dev_priv->uncore.force_wake_timer,
  				 jiffies + 1);
  	}
+
  	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
 
  out:
-	if (!delayed)
-		intel_runtime_pm_put(dev_priv);
+	intel_runtime_pm_put(dev_priv);
  }
 
  void assert_force_wake_inactive(struct drm_i915_private *dev_priv)
@@ -777,12 +772,11 @@ gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
  	    NEEDS_FORCE_WAKE((dev_priv), (reg))) { \

drivers/gpu/drm/i915/intel_uncore.c gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \ · line 777

Quoted 10 lines
drivers/gpu/drm/i915/intel_uncore.c +3 -5
  		dev_priv->uncore.funcs.force_wake_get(dev_priv, \
  						      FORCEWAKE_ALL); \
-		val = __raw_i915_read##x(dev_priv, reg); \
-		dev_priv->uncore.funcs.force_wake_put(dev_priv, \
-						      FORCEWAKE_ALL); \
-	} else { \
-		val = __raw_i915_read##x(dev_priv, reg); \
+		dev_priv->uncore.forcewake_count++; \
+		mod_timer_pinned(&dev_priv->uncore. , \
+				 jiffies + 1); \

why timer, we can do a put after register read right?

Other changes looks fine to me

drivers/gpu/drm/i915/intel_uncore.c gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \ · line 782

Quoted 5 lines
drivers/gpu/drm/i915/intel_uncore.c +1 -0
  	} \
+	val = __raw_i915_read##x(dev_priv, reg); \
  	hsw_unclaimed_reg_debug(dev_priv, reg, true, false); \
  	REG_READ_FOOTER; \
  }

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

#26 lore raw
Quoted 26 lines · Deepak S
Quoted 24 lines · Mika Kuoppala

From: Chris Wilson <[email protected]>

Calling intel_runtime_pm_put() is illegal from a soft-irq context, so revert the crude hack

commit aa0b3b5bb8768c1a6a6788869d9c7015eae7e80c

Author: Paulo Zanoni <[email protected]>
Date:   Tue Apr 1 14:55:07 2014 -0300

drm/i915: don't schedule force_wake_timer at gen6_read

and apply the single line corrective instead.

References: https://bugs.freedesktop.org/show_bug.cgi?id=80913

Cc Paulo Zanoni <[email protected]>
Cc Daniel Vetter <[email protected]>

---

drivers/gpu/drm/i915/intel_uncore.c +0 -2
@@ -777,12 +772,11 @@ gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
 	    NEEDS_FORCE_WAKE((dev_priv), (reg))) { \
 		dev_priv->uncore.funcs.force_wake_get(dev_priv, \
 						      FORCEWAKE_ALL); \
-		val = __raw_i915_read##x(dev_priv, reg); \
-		dev_priv->uncore.funcs.force_wake_put(dev_priv, \
Quoted 8 lines

drivers/gpu/drm/i915/intel_uncore.c gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \ · line 777

Quoted 6 lines
drivers/gpu/drm/i915/intel_uncore.c +3 -3
-						      FORCEWAKE_ALL); \
-	} else { \
-		val = __raw_i915_read##x(dev_priv, reg); \
+		dev_priv->uncore.forcewake_count++; \
+		mod_timer_pinned(&dev_priv->uncore. , \
+				 jiffies + 1); \

why timer, we can do a put after register read right?

The presumption is that we will do another mmio access requiring the forcewake very shortly, and we want to avoid the forcewake clear/ack cycle. So we defer dropping the forcewake until the end of the kernel context on this cpu (the goal being that as the scheduler switches back to the userspace context, we release the wakelock, it would be great if there was an explicit callback for that...).
-Chris

Signature · 5 lines
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#27 lore raw
Quoted 33 lines · Chris Wilson
Quoted 24 lines · Deepak S
Quoted 23 lines · Mika Kuoppala

From: Chris Wilson <[email protected]>

Calling intel_runtime_pm_put() is illegal from a soft-irq context, so revert the crude hack

commit aa0b3b5bb8768c1a6a6788869d9c7015eae7e80c

Author: Paulo Zanoni <[email protected]>
Date:   Tue Apr 1 14:55:07 2014 -0300

drm/i915: don't schedule force_wake_timer at gen6_read

and apply the single line corrective instead.

References: https://bugs.freedesktop.org/show_bug.cgi?id=80913

Cc Paulo Zanoni <[email protected]>
Cc Daniel Vetter <[email protected]>

---

drivers/gpu/drm/i915/intel_uncore.c +0 -1
@@ -777,12 +772,11 @@ gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
  	    NEEDS_FORCE_WAKE((dev_priv), (reg))) { \
  		dev_priv->uncore.funcs.force_wake_get(dev_priv, \
  						      FORCEWAKE_ALL); \
-		val = __raw_i915_read##x(dev_priv, reg); \
Quoted 8 lines

drivers/gpu/drm/i915/intel_uncore.c gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \ · line 777

Quoted 7 lines
drivers/gpu/drm/i915/intel_uncore.c +3 -4
-		dev_priv->uncore.funcs.force_wake_put(dev_priv, \
-						      FORCEWAKE_ALL); \
-	} else { \
-		val = __raw_i915_read##x(dev_priv, reg); \
+		dev_priv->uncore.forcewake_count++; \
+		mod_timer_pinned(&dev_priv->uncore. , \
+				 jiffies + 1); \

why timer, we can do a put after register read right?

Quoted 7 lines

The presumption is that we will do another mmio access requiring the forcewake very shortly, and we want to avoid the forcewake clear/ack cycle. So we defer dropping the forcewake until the end of the kernel context on this cpu (the goal being that as the scheduler switches back to the userspace context, we release the wakelock, it would be great if there was an explicit callback for that...).
-Chris

Thanks Chris. got it :)

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

#28 lore raw
In reply to Chris Wilson
Quoted 36 lines · Chris Wilson
Quoted 26 lines · Deepak S
Quoted 24 lines · Mika Kuoppala

From: Chris Wilson <[email protected]>

Calling intel_runtime_pm_put() is illegal from a soft-irq context, so revert the crude hack

commit aa0b3b5bb8768c1a6a6788869d9c7015eae7e80c

Author: Paulo Zanoni <[email protected]>
Date:   Tue Apr 1 14:55:07 2014 -0300

drm/i915: don't schedule force_wake_timer at gen6_read

and apply the single line corrective instead.

References: https://bugs.freedesktop.org/show_bug.cgi?id=80913

Cc Paulo Zanoni <[email protected]>
Cc Daniel Vetter <[email protected]>

---

drivers/gpu/drm/i915/intel_uncore.c +0 -2
@@ -777,12 +772,11 @@ gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
 	    NEEDS_FORCE_WAKE((dev_priv), (reg))) { \
 		dev_priv->uncore.funcs.force_wake_get(dev_priv, \
 						      FORCEWAKE_ALL); \
-		val = __raw_i915_read##x(dev_priv, reg); \
-		dev_priv->uncore.funcs.force_wake_put(dev_priv, \
Quoted 8 lines

drivers/gpu/drm/i915/intel_uncore.c gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \ · line 777

Quoted 6 lines
drivers/gpu/drm/i915/intel_uncore.c +3 -3
-						      FORCEWAKE_ALL); \
-	} else { \
-		val = __raw_i915_read##x(dev_priv, reg); \
+		dev_priv->uncore.forcewake_count++; \
+		mod_timer_pinned(&dev_priv->uncore. , \
+				 jiffies + 1); \

why timer, we can do a put after register read right?

Quoted 6 lines

The presumption is that we will do another mmio access requiring the forcewake very shortly, and we want to avoid the forcewake clear/ack cycle. So we defer dropping the forcewake until the end of the kernel context on this cpu (the goal being that as the scheduler switches back to the userspace context, we release the wakelock, it would be great if there was an explicit callback for that...).

task_work might be what you're looking for. It's a horribly crude interface though, so we'd need a small wrapper which either picks the task work or timer. Which complicates the logic further since we then have either or both lazy cleanup tasks pending.

Probably not worth the trouble.
-Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#29 lore raw
In reply to Mika Kuoppala

2014-12-08 16:27 GMT-02:00 Mika Kuoppala <[email protected]>:

Quoted 9 lines

From: Chris Wilson <[email protected]>

Calling intel_runtime_pm_put() is illegal from a soft-irq context, so revert the crude hack

commit aa0b3b5bb8768c1a6a6788869d9c7015eae7e80c

Author: Paulo Zanoni <[email protected]>
Date:   Tue Apr 1 14:55:07 2014 -0300
Quoted 8 lines

drm/i915: don't schedule force_wake_timer at gen6_read

and apply the single line corrective instead.

References: https://bugs.freedesktop.org/show_bug.cgi?id=80913

Cc Paulo Zanoni <[email protected]>
Cc Daniel Vetter <[email protected]>

This patch adds tons and tons of new WARNs when running igt/tests/pm_rpm on BDW, including:

WARNING: CPU: 1 PID: 228 at drivers/gpu/drm/i915/intel_uncore.c:623
assert_force_wake_inactive+0x36/0x40 [i915]()

WARN_ON(dev_priv->uncore.forcewake_count > 0)

Quoted 86 lines
---
 drivers/gpu/drm/i915/i915_drv.c     |  1 +
 drivers/gpu/drm/i915/intel_uncore.c | 18 ++++++------------
 2 files changed, 7 insertions(+), 12 deletions(-)
drivers/gpu/drm/i915/i915_drv.c +1 -0
index 71be3c9..706b122 100644
@@ -1402,6 +1402,7 @@ static int intel_runtime_suspend(struct device *device)
        }
 
        del_timer_sync(&dev_priv->gpu_error.hangcheck_timer);
+       intel_uncore_forcewake_reset(dev, false);
        dev_priv->pm.suspended = true;
 
        /*
drivers/gpu/drm/i915/intel_uncore.c +6 -13
index 46de8d7..38ac389 100644
@@ -449,8 +449,6 @@ static void gen6_force_wake_timer(unsigned long arg)
        if (--dev_priv->uncore.forcewake_count == 0)
                dev_priv->uncore.funcs.force_wake_put(dev_priv, FORCEWAKE_ALL);
        spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
-
-       intel_runtime_pm_put(dev_priv);
 }
 
 void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore)
@@ -586,7 +584,6 @@ void gen6_gt_force_wake_get(struct drm_i915_private *dev_priv, int fw_engine)
 void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
 {
        unsigned long irqflags;
-       bool delayed = false;
 
        if (!dev_priv->uncore.funcs.force_wake_put)
                return;
@@ -603,21 +600,19 @@ void gen6_gt_force_wake_put(struct drm_i915_private *dev_priv, int fw_engine)
                goto out;
        }
 
-
        spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
        WARN_ON(!dev_priv->uncore.forcewake_count);
 
        if (--dev_priv->uncore.forcewake_count == 0) {
                dev_priv->uncore.forcewake_count++;
-               delayed = true;
                mod_timer_pinned(&dev_priv->uncore.force_wake_timer,
                                 jiffies + 1);
        }
+
        spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
 
 out:
-       if (!delayed)
-               intel_runtime_pm_put(dev_priv);
+       intel_runtime_pm_put(dev_priv);
 }
 
 void assert_force_wake_inactive(struct drm_i915_private *dev_priv)
@@ -777,12 +772,11 @@ gen6_read##x(struct drm_i915_private *dev_priv, off_t reg, bool trace) { \
            NEEDS_FORCE_WAKE((dev_priv), (reg))) { \
                dev_priv->uncore.funcs.force_wake_get(dev_priv, \
                                                      FORCEWAKE_ALL); \
-               val = __raw_i915_read##x(dev_priv, reg); \
-               dev_priv->uncore.funcs.force_wake_put(dev_priv, \
-                                                     FORCEWAKE_ALL); \
-       } else { \
-               val = __raw_i915_read##x(dev_priv, reg); \
+               dev_priv->uncore.forcewake_count++; \
+               mod_timer_pinned(&dev_priv->uncore.force_wake_timer, \
+                                jiffies + 1); \
        } \
+       val = __raw_i915_read##x(dev_priv, reg); \
        hsw_unclaimed_reg_debug(dev_priv, reg, true, false); \
        REG_READ_FOOTER; \
 }
--

1.9.1

_______________________________________________
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
#30 lore raw
Quoted 19 lines · Paulo Zanoni

2014-12-08 16:27 GMT-02:00 Mika Kuoppala <[email protected]>:

Quoted 9 lines

From: Chris Wilson <[email protected]>

Calling intel_runtime_pm_put() is illegal from a soft-irq context, so revert the crude hack

commit aa0b3b5bb8768c1a6a6788869d9c7015eae7e80c

Author: Paulo Zanoni <[email protected]>
Date:   Tue Apr 1 14:55:07 2014 -0300
Quoted 8 lines

drm/i915: don't schedule force_wake_timer at gen6_read

and apply the single line corrective instead.

References: https://bugs.freedesktop.org/show_bug.cgi?id=80913

Cc Paulo Zanoni <[email protected]>
Cc Daniel Vetter <[email protected]>
Quoted 6 lines

This patch adds tons and tons of new WARNs when running igt/tests/pm_rpm on BDW, including:

WARNING: CPU: 1 PID: 228 at drivers/gpu/drm/i915/intel_uncore.c:623
assert_force_wake_inactive+0x36/0x40 [i915]()

WARN_ON(dev_priv->uncore.forcewake_count > 0)

The assert is in the incorrect place, it should be after the suspend disables forcewake.
-Chris

Signature · 5 lines
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#31 lore raw

cover Prep work patches for GPU scheduler

From: John Harrison <[email protected]>

Assorted patches to make the tree more friendly to the GPU scheduler.

The biggest change is to re-organise the execbuff code path. Basically, the scheduler needs to split the submission path into two sections which are essentially software only (data structure creation, manipulation, etc.) and hardware only (actually writing the commands to the write / execlist). As this is not a small change, any other changes to the execution code path tend to cause merge conflicts and other such headaches for the scheduler. Thus it would be extremely useful to get this preparation work committed to the tree while the scheduler is still being re-worked for upstream acceptability.

The other patches in the series are various minor fixes that were spotted along the way to getting the scheduler working.

[Patches against drm-intel-nightly tree fetched 08/12/2014]

Dave Gordon (3):
drm/i915: Updating assorted register and status page definitions drm/i915: FIFO space query code refactor
drm/i915: Disable 'get seqno' workaround for VLV

John Harrison (7):
drm/i915: Rename 'flags' to 'dispatch_flags' for better code reading drm/i915: Add missing trace point to LRC execbuff code path drm/i915: Add extra add_request calls
drm/i915: Early alloc request
drm/i915: Prelude to splitting i915_gem_do_execbuffer in two drm/i915: Split i915_dem_do_execbuffer() in half drm/i915: Cache ringbuf pointer in request structure

 drivers/gpu/drm/i915/i915_drv.h              |   53 ++++--
 drivers/gpu/drm/i915/i915_gem.c              |   63 +++----
 drivers/gpu/drm/i915/i915_gem_context.c      |    9 +
 drivers/gpu/drm/i915/i915_gem_execbuffer.c   |  237 +++++++++++++++++---------
 drivers/gpu/drm/i915/i915_gem_gtt.c          |    9 +
 drivers/gpu/drm/i915/i915_gem_render_state.c |    2 +-
 drivers/gpu/drm/i915/i915_reg.h              |   30 +++-
 drivers/gpu/drm/i915/intel_display.c         |   23 ++-
 drivers/gpu/drm/i915/intel_lrc.c             |  102 +++++++----
 drivers/gpu/drm/i915/intel_lrc.h             |   12 +-
 drivers/gpu/drm/i915/intel_ringbuffer.c      |   12 +-
 drivers/gpu/drm/i915/intel_ringbuffer.h      |   44 ++++-
 drivers/gpu/drm/i915/intel_uncore.c          |   19 ++-
 13 files changed, 426 insertions(+), 189 deletions(-)
Signature · 6 lines
1.7.9.5

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

PATCH 1/10 drm/i915: Rename 'flags' to 'dispatch_flags' for better code reading

From: John Harrison <[email protected]>

There is a flags word that is passed through the execbuffer code path all the way from initial decoding of the user parameters down to the very final dispatch buffer call. It is simply called 'flags'. Unfortuantely, there are many other flags words floating around in the same blocks of code. Even more once the GPU scheduler arrives.

This patch makes it more obvious exactly which flags word is which by renaming 'flags' to 'dispatch_flags'. Note that the bit definitions for this flags word already have an 'I915_DISPATCH_' prefix on them and so are not quite so ambiguous.

Change-Id: Ibc858930c4a606763f84750c5008af07cf245a90
For: VIZ-1587

---
 drivers/gpu/drm/i915/i915_gem_execbuffer.c |   23 ++++++++++++-----------
 drivers/gpu/drm/i915/intel_lrc.c           |    6 +++---
 drivers/gpu/drm/i915/intel_lrc.h           |    2 +-
 drivers/gpu/drm/i915/intel_ringbuffer.h    |    2 +-
 4 files changed, 17 insertions(+), 16 deletions(-)
drivers/gpu/drm/i915/i915_gem_execbuffer.c +12 -11
index 0c25f62..aa4566e 100644
@@ -1071,7 +1071,7 @@ i915_gem_ringbuffer_submission(struct drm_device *dev, struct drm_file *file,
 			       struct drm_i915_gem_execbuffer2 *args,
 			       struct list_head *vmas,
 			       struct drm_i915_gem_object *batch_obj,
-			       u64 exec_start, u32 flags)
+			       u64 exec_start, u32 dispatch_flags)
 {
 	struct drm_clip_rect *cliprects = NULL;
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -1199,19 +1199,19 @@ i915_gem_ringbuffer_submission(struct drm_device *dev, struct drm_file *file,
 
 			ret = ring->dispatch_execbuffer(ring,
 							exec_start, exec_len,
-							flags);
+							dispatch_flags);
 			if (ret)
 				goto error;
 		}
 	} else {
 		ret = ring->dispatch_execbuffer(ring,
 						exec_start, exec_len,
-						flags);
+						dispatch_flags);
 		if (ret)
 			return ret;
 	}
 
-	trace_i915_gem_ring_dispatch(intel_ring_get_request(ring), flags);
+	trace_i915_gem_ring_dispatch(intel_ring_get_request(ring), dispatch_flags);
 
 	i915_gem_execbuffer_move_to_active(vmas, ring);
 	i915_gem_execbuffer_retire_commands(dev, file, ring, batch_obj);
@@ -1285,7 +1285,7 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 	struct i915_address_space *vm;
 	const u32 ctx_id = i915_execbuffer2_get_context_id(*args);
 	u64 exec_start = args->batch_start_offset;
-	u32 flags;
+	u32 dispatch_flags;
 	int ret;
 	bool need_relocs;
 
@@ -1296,15 +1296,15 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 	if (ret)
 		return ret;
 
-	flags = 0;
+	dispatch_flags = 0;
 	if (args->flags & I915_EXEC_SECURE) {
 		if (!file->is_master || !capable(CAP_SYS_ADMIN))
 		    return -EPERM;
 
-		flags |= I915_DISPATCH_SECURE;
+		dispatch_flags |= I915_DISPATCH_SECURE;
 	}
 	if (args->flags & I915_EXEC_IS_PINNED)
-		flags |= I915_DISPATCH_PINNED;
+		dispatch_flags |= I915_DISPATCH_PINNED;
 
 	if ((args->flags & I915_EXEC_RING_MASK) > LAST_USER_RING) {
 		DRM_DEBUG("execbuf with unknown ring: %d\n",
@@ -1421,7 +1421,7 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 	/* snb/ivb/vlv conflate the "batch in ppgtt" bit with the "non-secure
 	 * batch" bit. Hence we need to pin secure batches into the global gtt.
 	 * hsw should have this fixed, but bdw mucks it up again. */
-	if (flags & I915_DISPATCH_SECURE) {
+	if (dispatch_flags & I915_DISPATCH_SECURE) {
 		/*
 		 * So on first glance it looks freaky that we pin the batch here
 		 * outside of the reservation loop. But:
@@ -1441,7 +1441,8 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 		exec_start += i915_gem_obj_offset(batch_obj, vm);
 
 	ret = dev_priv->gt.do_execbuf(dev, file, ring, ctx, args,
-				      &eb->vmas, batch_obj, exec_start, flags);
+				      &eb->vmas, batch_obj, exec_start,
+				      dispatch_flags);
 
 	/*
 	 * FIXME: We crucially rely upon the active tracking for the (ppgtt)
@@ -1449,7 +1450,7 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 	 * needs to be adjusted to also track the ggtt batch vma properly as
 	 * active.
 	 */
-	if (flags & I915_DISPATCH_SECURE)
+	if (dispatch_flags & I915_DISPATCH_SECURE)
 		i915_gem_object_ggtt_unpin(batch_obj);
 err:
 	/* the request owns the ref now */
drivers/gpu/drm/i915/intel_lrc.c +3 -3
index a82020e..b8baf11 100644
@@ -646,7 +646,7 @@ static int execlists_move_to_gpu(struct intel_ringbuffer *ringbuf,
  * @vmas: list of vmas.
  * @batch_obj: the batchbuffer to submit.
  * @exec_start: batchbuffer start virtual address pointer.
- * @flags: translated execbuffer call flags.
+ * @dispatch_flags: translated execbuffer call flags.
  *
  * This is the evil twin version of i915_gem_ringbuffer_submission. It abstracts
  * away the submission details of the execbuffer ioctl call.
@@ -659,7 +659,7 @@ int intel_execlists_submission(struct drm_device *dev, struct drm_file *file,
 			       struct drm_i915_gem_execbuffer2 *args,
 			       struct list_head *vmas,
 			       struct drm_i915_gem_object *batch_obj,
-			       u64 exec_start, u32 flags)
+			       u64 exec_start, u32 dispatch_flags)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct intel_ringbuffer *ringbuf = ctx->engine[ring->id].ringbuf;
@@ -732,7 +732,7 @@ int intel_execlists_submission(struct drm_device *dev, struct drm_file *file,
 		dev_priv->relative_constants_mode = instp_mode;
 	}
 
-	ret = ring->emit_bb_start(ringbuf, exec_start, flags);
+	ret = ring->emit_bb_start(ringbuf, exec_start, dispatch_flags);
 	if (ret)
 		return ret;
 
drivers/gpu/drm/i915/intel_lrc.h +1 -1
index 14b216b..b8d5364 100644
@@ -80,7 +80,7 @@ int intel_execlists_submission(struct drm_device *dev, struct drm_file *file,
 			       struct drm_i915_gem_execbuffer2 *args,
 			       struct list_head *vmas,
 			       struct drm_i915_gem_object *batch_obj,
-			       u64 exec_start, u32 flags);
+			       u64 exec_start, u32 dispatch_flags);
 u32 intel_execlists_ctx_id(struct drm_i915_gem_object *ctx_obj);
 
 /**
drivers/gpu/drm/i915/intel_ringbuffer.h +1 -1
index 6dbb6f4..2e0fa7b 100644
@@ -169,7 +169,7 @@ struct  intel_engine_cs {
 				     u32 seqno);
 	int		(*dispatch_execbuffer)(struct intel_engine_cs *ring,
 					       u64 offset, u32 length,
-					       unsigned flags);
+					       unsigned dispatch_flags);
 #define I915_DISPATCH_SECURE 0x1
 #define I915_DISPATCH_PINNED 0x2
 	void		(*cleanup)(struct intel_engine_cs *ring);
Signature · 6 lines
1.7.9.5

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

PATCH 2/10 drm/i915: Add missing trace point to LRC execbuff code path

In reply to [email protected]

From: John Harrison <[email protected]>

There is a trace point in the legacy execbuffer execution path that is missing from the execlist path. Trace points are extremely useful for debugging and are used by various automated validation tests. Hence, this patch adds the missing trace point back in.

Change-Id: I8bc80a892e5f7effbe3a0883a58a828d59c10c07
For: VIZ-1587

---
 drivers/gpu/drm/i915/intel_lrc.c |    2 ++
 1 file changed, 2 insertions(+)
drivers/gpu/drm/i915/intel_lrc.c +2 -0
index b8baf11..82a47e3 100644
@@ -736,6 +736,8 @@ int intel_execlists_submission(struct drm_device *dev, struct drm_file *file,
 	if (ret)
 		return ret;
 
+	trace_i915_gem_ring_dispatch(intel_ring_get_request(ring), dispatch_flags);
+
 	i915_gem_execbuffer_move_to_active(vmas, ring);
 	i915_gem_execbuffer_retire_commands(dev, file, ring, batch_obj);
 
Signature · 6 lines
1.7.9.5

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

PATCH 3/10 drm/i915: Updating assorted register and status page definitions

In reply to [email protected]

From: Dave Gordon <[email protected]>

Added various definitions that will be useful for the scheduler in general and pre-emptive context switching in particular.

Change-Id: Ica805b94160426def51f5d520f5ce51c60864a98
For: VIZ-1587

---
 drivers/gpu/drm/i915/i915_reg.h         |   30 ++++++++++++++++++++++-
 drivers/gpu/drm/i915/intel_ringbuffer.h |   40 +++++++++++++++++++++++++++++--
 2 files changed, 67 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/i915_reg.h +29 -1
index 869e5ae..6169720 100644
@@ -216,6 +216,10 @@
 #define  MI_GLOBAL_GTT    (1<<22)
 
 #define MI_NOOP			MI_INSTR(0, 0)
+#define   MI_NOOP_WRITE_ID		(1<<22)
+#define   MI_NOOP_ID_MASK		((1<<22) - 1)
+#define   MI_NOOP_MID(id)		((id) & MI_NOOP_ID_MASK)
+#define MI_NOOP_WITH_ID(id)	MI_INSTR(0, MI_NOOP_WRITE_ID|MI_NOOP_MID(id))
 #define MI_USER_INTERRUPT	MI_INSTR(0x02, 0)
 #define MI_WAIT_FOR_EVENT       MI_INSTR(0x03, 0)
 #define   MI_WAIT_FOR_OVERLAY_FLIP	(1<<16)
@@ -233,6 +237,7 @@
 #define MI_ARB_ON_OFF		MI_INSTR(0x08, 0)
 #define   MI_ARB_ENABLE			(1<<0)
 #define   MI_ARB_DISABLE		(0<<0)
+#define MI_ARB_CHECK		MI_INSTR(0x05, 0)
 #define MI_BATCH_BUFFER_END	MI_INSTR(0x0a, 0)
 #define MI_SUSPEND_FLUSH	MI_INSTR(0x0b, 0)
 #define   MI_SUSPEND_FLUSH_EN	(1<<0)
@@ -281,6 +286,8 @@
 #define   MI_SEMAPHORE_SYNC_INVALID (3<<16)
 #define   MI_SEMAPHORE_SYNC_MASK    (3<<16)
 #define MI_SET_CONTEXT		MI_INSTR(0x18, 0)
+#define   MI_CONTEXT_ADDR_MASK		((~0)<<12)
+#define   MI_SET_CONTEXT_FLAG_MASK	((1<<12)-1)
 #define   MI_MM_SPACE_GTT		(1<<8)
 #define   MI_MM_SPACE_PHYSICAL		(0<<8)
 #define   MI_SAVE_EXT_STATE_EN		(1<<3)
@@ -298,6 +305,10 @@
 #define   MI_USE_GGTT		(1 << 22) /* g4x+ */
 #define MI_STORE_DWORD_INDEX	MI_INSTR(0x21, 1)
 #define   MI_STORE_DWORD_INDEX_SHIFT 2
+#define MI_STORE_REG_MEM	MI_INSTR(0x24, 1)
+#define   MI_STORE_REG_MEM_GTT		(1 << 22)
+#define   MI_STORE_REG_MEM_PREDICATE	(1 << 21)
+
 /* Official intel docs are somewhat sloppy concerning MI_LOAD_REGISTER_IMM:
  * - Always issue a MI_NOOP _before_ the MI_LOAD_REGISTER_IMM - otherwise hw
  *   simply ignores the register load under certain conditions.
@@ -312,7 +323,10 @@
 #define MI_FLUSH_DW		MI_INSTR(0x26, 1) /* for GEN6 */
 #define   MI_FLUSH_DW_STORE_INDEX	(1<<21)
 #define   MI_INVALIDATE_TLB		(1<<18)
+#define   MI_FLUSH_DW_OP_NONE		(0<<14)
 #define   MI_FLUSH_DW_OP_STOREDW	(1<<14)
+#define   MI_FLUSH_DW_OP_RSVD		(2<<14)
+#define   MI_FLUSH_DW_OP_STAMP		(3<<14)
 #define   MI_FLUSH_DW_OP_MASK		(3<<14)
 #define   MI_FLUSH_DW_NOTIFY		(1<<8)
 #define   MI_INVALIDATE_BSD		(1<<7)
@@ -1119,6 +1133,19 @@ enum punit_power_well {
 #define GEN6_VERSYNC	(RING_SYNC_1(VEBOX_RING_BASE))
 #define GEN6_VEVSYNC	(RING_SYNC_2(VEBOX_RING_BASE))
 #define GEN6_NOSYNC 0
+
+/*
+ * Premption-related registers
+ */
+#define RING_UHPTR(base)	((base)+0x134)
+#define   UHPTR_GFX_ADDR_ALIGN		(0x7)
+#define   UHPTR_VALID			(0x1)
+#define RING_PREEMPT_ADDR	0x0214c
+#define   PREEMPT_BATCH_LEVEL_MASK	(0x3)
+#define BB_PREEMPT_ADDR		0x02148
+#define SBB_PREEMPT_ADDR	0x0213c
+#define RS_PREEMPT_STATUS	0x0215c
+
 #define RING_MAX_IDLE(base)	((base)+0x54)
 #define RING_HWS_PGA(base)	((base)+0x80)
 #define RING_HWS_PGA_GEN6(base)	((base)+0x2080)
@@ -5917,7 +5944,8 @@ enum punit_power_well {
 #define  VLV_SPAREG2H				0xA194
 
 #define  GTFIFODBG				0x120000
-#define    GT_FIFO_SBDROPERR			(1<<6)
+#define    GT_FIFO_CPU_ERROR_MASK		0xf
+#define    GT_FIFO_SDDROPERR			(1<<6)
 #define    GT_FIFO_BLOBDROPERR			(1<<5)
 #define    GT_FIFO_SB_READ_ABORTERR		(1<<4)
 #define    GT_FIFO_DROPERR			(1<<3)
drivers/gpu/drm/i915/intel_ringbuffer.h +38 -2
index 2e0fa7b..f15fc46 100644
@@ -47,6 +47,12 @@ struct  intel_hw_status_page {
 #define I915_READ_MODE(ring) I915_READ(RING_MI_MODE((ring)->mmio_base))
 #define I915_WRITE_MODE(ring, val) I915_WRITE(RING_MI_MODE((ring)->mmio_base), val)
 
+#define I915_READ_UHPTR(ring) \
+		I915_READ(RING_UHPTR((ring)->mmio_base))
+#define I915_WRITE_UHPTR(ring, val) \
+		I915_WRITE(RING_UHPTR((ring)->mmio_base), val)
+#define I915_READ_NOPID(ring) I915_READ(RING_NOPID((ring)->mmio_base))
+
 /* seqno size is actually only a uint32, but since we plan to use MI_FLUSH_DW to
  * do the writes, and that must have qw aligned offsets, simply pretend it's 8b.
  */
@@ -377,10 +383,40 @@ intel_write_status_page(struct intel_engine_cs *ring,
  * 0x1f: Last written status offset. (GM45)
  *
  * The area from dword 0x20 to 0x3ff is available for driver usage.
+ *
+ * Note: in general the allocation of these indices is arbitrary, as long
+ * as they are all unique. But a few of them are used with instructions that
+ * have specific alignment requirements, those particular indices must be
+ * chosen carefully to meet those requirements. The list below shows the
+ * currently-known alignment requirements:
+ *
+ *	I915_GEM_SCRATCH_INDEX	    must be EVEN
  */
 #define I915_GEM_HWS_INDEX		0x20
-#define I915_GEM_HWS_SCRATCH_INDEX	0x30
-#define I915_GEM_HWS_SCRATCH_ADDR (I915_GEM_HWS_SCRATCH_INDEX << MI_STORE_DWORD_INDEX_SHIFT)
+
+#define I915_GEM_HWS_SCRATCH_INDEX	0x24  /* QWord */
+#define I915_GEM_HWS_SCRATCH_ADDR	(I915_GEM_HWS_SCRATCH_INDEX << MI_STORE_DWORD_INDEX_SHIFT)
+
+/*
+ * Tracking; these are updated by the GPU at the beginning and/or end of every
+ * batch. One pair for regular buffers, the other for preemptive ones.
+ */
+#define I915_BATCH_DONE_SEQNO		0x30  /* Completed batch seqno        */
+#define I915_BATCH_ACTIVE_SEQNO		0x31  /* In progress batch seqno      */
+#define I915_PREEMPTIVE_DONE_SEQNO	0x32  /* Completed preemptive batch   */
+#define I915_PREEMPTIVE_ACTIVE_SEQNO	0x33  /* In progress preemptive batch */
+
+/*
+ * Preemption; these are used by the GPU to save important registers
+ */
+#define I915_SAVE_PREEMPTED_RING_PTR	0x34  /* HEAD before preemption     */
+#define I915_SAVE_PREEMPTED_BB_PTR	0x35  /* BB ptr before preemption   */
+#define I915_SAVE_PREEMPTED_SBB_PTR	0x36  /* SBB before preemption      */
+#define I915_SAVE_PREEMPTED_UHPTR	0x37  /* UHPTR after preemption     */
+#define I915_SAVE_PREEMPTED_HEAD	0x38  /* HEAD after preemption      */
+#define I915_SAVE_PREEMPTED_TAIL	0x39  /* TAIL after preemption      */
+#define I915_SAVE_PREEMPTED_STATUS	0x3A  /* RS preemption status       */
+#define I915_SAVE_PREEMPTED_NOPID	0x3B  /* Dummy                      */
 
 void intel_unpin_ringbuffer_obj(struct intel_ringbuffer *ringbuf);
 int intel_pin_and_map_ringbuffer_obj(struct drm_device *dev,
Signature · 6 lines
1.7.9.5

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

Re: [PATCH 03/10] drm/i915: Updating assorted register and status page definitions

Quoted 132 lines · On Tue, Dec 09, 2014 at 12:59:06PM +0000, [email protected] wrote:

From: Dave Gordon <[email protected]>

Added various definitions that will be useful for the scheduler in general and pre-emptive context switching in particular.

Change-Id: Ica805b94160426def51f5d520f5ce51c60864a98
For: VIZ-1587

---
 drivers/gpu/drm/i915/i915_reg.h         |   30 ++++++++++++++++++++++-
 drivers/gpu/drm/i915/intel_ringbuffer.h |   40 +++++++++++++++++++++++++++++--
 2 files changed, 67 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/i915_reg.h +29 -1
index 869e5ae..6169720 100644
@@ -216,6 +216,10 @@
 #define  MI_GLOBAL_GTT    (1<<22)
 
 #define MI_NOOP			MI_INSTR(0, 0)
+#define   MI_NOOP_WRITE_ID		(1<<22)
+#define   MI_NOOP_ID_MASK		((1<<22) - 1)
+#define   MI_NOOP_MID(id)		((id) & MI_NOOP_ID_MASK)
+#define MI_NOOP_WITH_ID(id)	MI_INSTR(0, MI_NOOP_WRITE_ID|MI_NOOP_MID(id))
 #define MI_USER_INTERRUPT	MI_INSTR(0x02, 0)
 #define MI_WAIT_FOR_EVENT       MI_INSTR(0x03, 0)
 #define   MI_WAIT_FOR_OVERLAY_FLIP	(1<<16)
@@ -233,6 +237,7 @@
 #define MI_ARB_ON_OFF		MI_INSTR(0x08, 0)
 #define   MI_ARB_ENABLE			(1<<0)
 #define   MI_ARB_DISABLE		(0<<0)
+#define MI_ARB_CHECK		MI_INSTR(0x05, 0)
 #define MI_BATCH_BUFFER_END	MI_INSTR(0x0a, 0)
 #define MI_SUSPEND_FLUSH	MI_INSTR(0x0b, 0)
 #define   MI_SUSPEND_FLUSH_EN	(1<<0)
@@ -281,6 +286,8 @@
 #define   MI_SEMAPHORE_SYNC_INVALID (3<<16)
 #define   MI_SEMAPHORE_SYNC_MASK    (3<<16)
 #define MI_SET_CONTEXT		MI_INSTR(0x18, 0)
+#define   MI_CONTEXT_ADDR_MASK		((~0)<<12)
+#define   MI_SET_CONTEXT_FLAG_MASK	((1<<12)-1)
 #define   MI_MM_SPACE_GTT		(1<<8)
 #define   MI_MM_SPACE_PHYSICAL		(0<<8)
 #define   MI_SAVE_EXT_STATE_EN		(1<<3)
@@ -298,6 +305,10 @@
 #define   MI_USE_GGTT		(1 << 22) /* g4x+ */
 #define MI_STORE_DWORD_INDEX	MI_INSTR(0x21, 1)
 #define   MI_STORE_DWORD_INDEX_SHIFT 2
+#define MI_STORE_REG_MEM	MI_INSTR(0x24, 1)
+#define   MI_STORE_REG_MEM_GTT		(1 << 22)
+#define   MI_STORE_REG_MEM_PREDICATE	(1 << 21)
+
 /* Official intel docs are somewhat sloppy concerning MI_LOAD_REGISTER_IMM:
  * - Always issue a MI_NOOP _before_ the MI_LOAD_REGISTER_IMM - otherwise hw
  *   simply ignores the register load under certain conditions.
@@ -312,7 +323,10 @@
 #define MI_FLUSH_DW		MI_INSTR(0x26, 1) /* for GEN6 */
 #define   MI_FLUSH_DW_STORE_INDEX	(1<<21)
 #define   MI_INVALIDATE_TLB		(1<<18)
+#define   MI_FLUSH_DW_OP_NONE		(0<<14)
 #define   MI_FLUSH_DW_OP_STOREDW	(1<<14)
+#define   MI_FLUSH_DW_OP_RSVD		(2<<14)
+#define   MI_FLUSH_DW_OP_STAMP		(3<<14)
 #define   MI_FLUSH_DW_OP_MASK		(3<<14)
 #define   MI_FLUSH_DW_NOTIFY		(1<<8)
 #define   MI_INVALIDATE_BSD		(1<<7)
@@ -1119,6 +1133,19 @@ enum punit_power_well {
 #define GEN6_VERSYNC	(RING_SYNC_1(VEBOX_RING_BASE))
 #define GEN6_VEVSYNC	(RING_SYNC_2(VEBOX_RING_BASE))
 #define GEN6_NOSYNC 0
+
+/*
+ * Premption-related registers
+ */
+#define RING_UHPTR(base)	((base)+0x134)
+#define   UHPTR_GFX_ADDR_ALIGN		(0x7)
+#define   UHPTR_VALID			(0x1)
+#define RING_PREEMPT_ADDR	0x0214c
+#define   PREEMPT_BATCH_LEVEL_MASK	(0x3)
+#define BB_PREEMPT_ADDR		0x02148
+#define SBB_PREEMPT_ADDR	0x0213c
+#define RS_PREEMPT_STATUS	0x0215c
+
 #define RING_MAX_IDLE(base)	((base)+0x54)
 #define RING_HWS_PGA(base)	((base)+0x80)
 #define RING_HWS_PGA_GEN6(base)	((base)+0x2080)
@@ -5917,7 +5944,8 @@ enum punit_power_well {
 #define  VLV_SPAREG2H				0xA194
 
 #define  GTFIFODBG				0x120000
-#define    GT_FIFO_SBDROPERR			(1<<6)
+#define    GT_FIFO_CPU_ERROR_MASK		0xf
+#define    GT_FIFO_SDDROPERR			(1<<6)
 #define    GT_FIFO_BLOBDROPERR			(1<<5)
 #define    GT_FIFO_SB_READ_ABORTERR		(1<<4)
 #define    GT_FIFO_DROPERR			(1<<3)
drivers/gpu/drm/i915/intel_ringbuffer.h +16 -2
index 2e0fa7b..f15fc46 100644
@@ -47,6 +47,12 @@ struct  intel_hw_status_page {
 #define I915_READ_MODE(ring) I915_READ(RING_MI_MODE((ring)->mmio_base))
 #define I915_WRITE_MODE(ring, val) I915_WRITE(RING_MI_MODE((ring)->mmio_base), val)
 
+#define I915_READ_UHPTR(ring) \
+		I915_READ(RING_UHPTR((ring)->mmio_base))
+#define I915_WRITE_UHPTR(ring, val) \
+		I915_WRITE(RING_UHPTR((ring)->mmio_base), val)
+#define I915_READ_NOPID(ring) I915_READ(RING_NOPID((ring)->mmio_base))
+
 /* seqno size is actually only a uint32, but since we plan to use MI_FLUSH_DW to
  * do the writes, and that must have qw aligned offsets, simply pretend it's 8b.
  */
@@ -377,10 +383,40 @@ intel_write_status_page(struct intel_engine_cs *ring,
  * 0x1f: Last written status offset. (GM45)
  *
  * The area from dword 0x20 to 0x3ff is available for driver usage.
+ *
+ * Note: in general the allocation of these indices is arbitrary, as long
+ * as they are all unique. But a few of them are used with instructions that
+ * have specific alignment requirements, those particular indices must be
+ * chosen carefully to meet those requirements. The list below shows the
+ * currently-known alignment requirements:
+ *
+ *	I915_GEM_SCRATCH_INDEX	    must be EVEN
  */
 #define I915_GEM_HWS_INDEX		0x20
-#define I915_GEM_HWS_SCRATCH_INDEX	0x30
-#define I915_GEM_HWS_SCRATCH_ADDR (I915_GEM_HWS_SCRATCH_INDEX << MI_STORE_DWORD_INDEX_SHIFT)
+
+#define I915_GEM_HWS_SCRATCH_INDEX	0x24  /* QWord */

drivers/gpu/drm/i915/intel_ringbuffer.h intel_write_status_page(struct intel_engine_cs *ring, · line 407

Quoted 10 lines
drivers/gpu/drm/i915/intel_ringbuffer.h +10 -0
+#define I915_GEM_HWS_SCRATCH_ADDR	(I915_GEM_HWS_SCRATCH_INDEX << MI_STORE_DWORD_INDEX_SHIFT)
+
+/*
+ * Tracking; these are updated by the GPU at the beginning and/or end of every
+ * batch. One pair for regular buffers, the other for preemptive ones.
+ */
+#define I915_BATCH_DONE_SEQNO		0x30  /* Completed batch seqno        */
+#define I915_BATCH_ACTIVE_SEQNO		0x31  /* In progress batch seqno      */
+#define I915_PREEMPTIVE_DONE_SEQNO	0x32  /* Completed preemptive batch   */
+#define I915_PREEMPTIVE_ACTIVE_SEQNO	0x33  /* In progress preemptive batch */

This are software define and currently not yet used I think. Imo better to add them together with the scheduler code that uses them. Splitting out #define patches only makes sense if they can be reviewed separately (e.g. registers with Bspec). Just adding software stuff (structs, enums, offsets not enforced by hw) should be added with the actual code using it.

drivers/gpu/drm/i915/intel_ringbuffer.h intel_write_status_page(struct intel_engine_cs *ring, · line 419

Quoted 12 lines
drivers/gpu/drm/i915/intel_ringbuffer.h +12 -0
+
+/*
+ * Preemption; these are used by the GPU to save important registers
+ */
+#define I915_SAVE_PREEMPTED_RING_PTR	0x34  /* HEAD before preemption     */
+#define I915_SAVE_PREEMPTED_BB_PTR	0x35  /* BB ptr before preemption   */
+#define I915_SAVE_PREEMPTED_SBB_PTR	0x36  /* SBB before preemption      */
+#define I915_SAVE_PREEMPTED_UHPTR	0x37  /* UHPTR after preemption     */
+#define I915_SAVE_PREEMPTED_HEAD	0x38  /* HEAD after preemption      */
+#define I915_SAVE_PREEMPTED_TAIL	0x39  /* TAIL after preemption      */
+#define I915_SAVE_PREEMPTED_STATUS	0x3A  /* RS preemption status       */
+#define I915_SAVE_PREEMPTED_NOPID	0x3B  /* Dummy                      */

I guess this is hardcoded in hw? In that case a HWS instead of I915 prefix sounds better to me to make it clear this is not something i915/gem code invented. Also comment should state whether this is execlist or gen8+ or something like that.
-Daniel

Quoted 10 lines
drivers/gpu/drm/i915/intel_ringbuffer.h +0 -0
 void intel_unpin_ringbuffer_obj(struct intel_ringbuffer *ringbuf);
 int intel_pin_and_map_ringbuffer_obj(struct drm_device *dev,
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#36 lore raw

Re: [PATCH 03/10] drm/i915: Updating assorted register and status page definitions

Quoted 13 lines · Daniel Vetter
Quoted 12 lines · On Tue, Dec 09, 2014 at 12:59:06PM +0000, [email protected] wrote:

From: Dave Gordon <[email protected]>

Added various definitions that will be useful for the scheduler in general and pre-emptive context switching in particular.

Change-Id: Ica805b94160426def51f5d520f5ce51c60864a98
For: VIZ-1587

---
 drivers/gpu/drm/i915/i915_reg.h         |   30 ++++++++++++++++++++++-
 drivers/gpu/drm/i915/intel_ringbuffer.h |   40 +++++++++++++++++++++++++++++--
 2 files changed, 67 insertions(+), 3 deletions(-)

[snip]

Quoted 6 lines
Quoted 6 lines
drivers/gpu/drm/i915/intel_ringbuffer.h +6 -0
+/*
+ * Tracking; these are updated by the GPU at the beginning and/or end of every
+ * batch. One pair for regular buffers, the other for preemptive ones.
+ */
+#define I915_BATCH_DONE_SEQNO		0x30  /* Completed batch seqno        */
+#define I915_BATCH_ACTIVE_SEQNO		0x31  /* In progress batch seqno      */
Quoted 8 lines

drivers/gpu/drm/i915/intel_ringbuffer.h intel_write_status_page(struct intel_engine_cs *ring, · line 407

Quoted 2 lines
drivers/gpu/drm/i915/intel_ringbuffer.h +2 -0
+#define I915_PREEMPTIVE_DONE_SEQNO	0x32  /* Completed preemptive batch   */
+#define I915_PREEMPTIVE_ACTIVE_SEQNO	0x33  /* In progress preemptive batch */

This are software define and currently not yet used I think. Imo better to add them together with the scheduler code that uses them. Splitting out #define patches only makes sense if they can be reviewed separately (e.g. registers with Bspec). Just adding software stuff (structs, enums, offsets not enforced by hw) should be added with the actual code using it.

OK

Quoted 10 lines
Quoted 10 lines
drivers/gpu/drm/i915/intel_ringbuffer.h +10 -0
+
+/*
+ * Preemption; these are used by the GPU to save important registers
+ */
+#define I915_SAVE_PREEMPTED_RING_PTR	0x34  /* HEAD before preemption     */
+#define I915_SAVE_PREEMPTED_BB_PTR	0x35  /* BB ptr before preemption   */
+#define I915_SAVE_PREEMPTED_SBB_PTR	0x36  /* SBB before preemption      */
+#define I915_SAVE_PREEMPTED_UHPTR	0x37  /* UHPTR after preemption     */
+#define I915_SAVE_PREEMPTED_HEAD	0x38  /* HEAD after preemption      */
+#define I915_SAVE_PREEMPTED_TAIL	0x39  /* TAIL after preemption      */
Quoted 8 lines

drivers/gpu/drm/i915/intel_ringbuffer.h intel_write_status_page(struct intel_engine_cs *ring, · line 419

Quoted 2 lines
drivers/gpu/drm/i915/intel_ringbuffer.h +2 -0
+#define I915_SAVE_PREEMPTED_STATUS	0x3A  /* RS preemption status       */
+#define I915_SAVE_PREEMPTED_NOPID	0x3B  /* Dummy                      */

I guess this is hardcoded in hw? In that case a HWS instead of I915 prefix sounds better to me to make it clear this is not something i915/gem code invented. Also comment should state whether this is execlist or gen8+ or something like that.
-Daniel

No, these are s/w too, so they can be added later if that's preferred.

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

#37 lore raw

PATCH 4/10 drm/i915: FIFO space query code refactor

In reply to [email protected]

From: Dave Gordon <[email protected]>

When querying the GTFIFOCTL register to check the FIFO space, the read value must be masked. The operation is repeated explicitly in several places. This change refactors the read-and-mask code into a function call.

Change-Id Id1a9f3785cb20b82d4caa330c37b31e4e384a3ef
---
 drivers/gpu/drm/i915/intel_uncore.c |   19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +12 -7
index 46de8d7..4021831 100644
@@ -47,6 +47,13 @@ assert_device_not_suspended(struct drm_i915_private *dev_priv)
 		  "Device suspended\n");
 }
 
+static inline u32 fifo_free_entries(struct drm_i915_private *dev_priv)
+{
+	u32 count = __raw_i915_read32(dev_priv, GTFIFOCTL);
+
+	return count & GT_FIFO_FREE_ENTRIES_MASK;
+}
+
 static void __gen6_gt_wait_for_thread_c0(struct drm_i915_private *dev_priv)
 {
 	/* w/a for a sporadic read returning 0 by waiting for the GT
@@ -154,16 +161,15 @@ static int __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv)
 	/* On VLV, FIFO will be shared by both SW and HW.
 	 * So, we need to read the FREE_ENTRIES everytime */
 	if (IS_VALLEYVIEW(dev_priv->dev))
-		dev_priv->uncore.fifo_count =
-			__raw_i915_read32(dev_priv, GTFIFOCTL) &
-						GT_FIFO_FREE_ENTRIES_MASK;
+		dev_priv->uncore.fifo_count = fifo_free_entries(dev_priv);
 
 	if (dev_priv->uncore.fifo_count < GT_FIFO_NUM_RESERVED_ENTRIES) {
 		int loop = 500;
-		u32 fifo = __raw_i915_read32(dev_priv, GTFIFOCTL) & GT_FIFO_FREE_ENTRIES_MASK;
+		u32 fifo = fifo_free_entries(dev_priv);
+
 		while (fifo <= GT_FIFO_NUM_RESERVED_ENTRIES && loop--) {
 			udelay(10);
-			fifo = __raw_i915_read32(dev_priv, GTFIFOCTL) & GT_FIFO_FREE_ENTRIES_MASK;
+			fifo = fifo_free_entries(dev_priv);
 		}
 		if (WARN_ON(loop < 0 && fifo <= GT_FIFO_NUM_RESERVED_ENTRIES))
 			++ret;
@@ -505,8 +511,7 @@ void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore)
 
 		if (IS_GEN6(dev) || IS_GEN7(dev))
 			dev_priv->uncore.fifo_count =
-				__raw_i915_read32(dev_priv, GTFIFOCTL) &
-				GT_FIFO_FREE_ENTRIES_MASK;
+				fifo_free_entries(dev_priv);
 	}
 
 	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
Signature · 6 lines
1.7.9.5

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

Re: [PATCH 04/10] drm/i915: FIFO space query code refactor

Quoted 8 lines · On Tue, Dec 09, 2014 at 12:59:07PM +0000, [email protected] wrote:

From: Dave Gordon <[email protected]>

When querying the GTFIFOCTL register to check the FIFO space, the read value must be masked. The operation is repeated explicitly in several places. This change refactors the read-and-mask code into a function call.

Change-Id Id1a9f3785cb20b82d4caa330c37b31e4e384a3ef

Looks like an unrelated patch and probably collides with Mika's forcewake refactoring. Please rebase on top of that series if still needed. -Daniel

Quoted 60 lines
---
 drivers/gpu/drm/i915/intel_uncore.c |   19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +12 -7
index 46de8d7..4021831 100644
@@ -47,6 +47,13 @@ assert_device_not_suspended(struct drm_i915_private *dev_priv)
 		  "Device suspended\n");
 }
 
+static inline u32 fifo_free_entries(struct drm_i915_private *dev_priv)
+{
+	u32 count = __raw_i915_read32(dev_priv, GTFIFOCTL);
+
+	return count & GT_FIFO_FREE_ENTRIES_MASK;
+}
+
 static void __gen6_gt_wait_for_thread_c0(struct drm_i915_private *dev_priv)
 {
 	/* w/a for a sporadic read returning 0 by waiting for the GT
@@ -154,16 +161,15 @@ static int __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv)
 	/* On VLV, FIFO will be shared by both SW and HW.
 	 * So, we need to read the FREE_ENTRIES everytime */
 	if (IS_VALLEYVIEW(dev_priv->dev))
-		dev_priv->uncore.fifo_count =
-			__raw_i915_read32(dev_priv, GTFIFOCTL) &
-						GT_FIFO_FREE_ENTRIES_MASK;
+		dev_priv->uncore.fifo_count = fifo_free_entries(dev_priv);
 
 	if (dev_priv->uncore.fifo_count < GT_FIFO_NUM_RESERVED_ENTRIES) {
 		int loop = 500;
-		u32 fifo = __raw_i915_read32(dev_priv, GTFIFOCTL) & GT_FIFO_FREE_ENTRIES_MASK;
+		u32 fifo = fifo_free_entries(dev_priv);
+
 		while (fifo <= GT_FIFO_NUM_RESERVED_ENTRIES && loop--) {
 			udelay(10);
-			fifo = __raw_i915_read32(dev_priv, GTFIFOCTL) & GT_FIFO_FREE_ENTRIES_MASK;
+			fifo = fifo_free_entries(dev_priv);
 		}
 		if (WARN_ON(loop < 0 && fifo <= GT_FIFO_NUM_RESERVED_ENTRIES))
 			++ret;
@@ -505,8 +511,7 @@ void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore)
 
 		if (IS_GEN6(dev) || IS_GEN7(dev))
 			dev_priv->uncore.fifo_count =
-				__raw_i915_read32(dev_priv, GTFIFOCTL) &
-				GT_FIFO_FREE_ENTRIES_MASK;
+				fifo_free_entries(dev_priv);
 	}
 
 	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#39 lore raw

PATCH v2 1/1 drm/i915: FIFO space query code refactor

When querying the GTFIFOCTL register to check the FIFO space, the read value must be masked. The operation is repeated explicitly in several places. This change refactors the read-and-mask code into a function call.

v2: rebased on top of Mika's forcewake patch set, specifically:
[PATCH 8/8] drm/i915: Enum forcewake domains and domain identifiers

Change-Id Id1a9f3785cb20b82d4caa330c37b31e4e384a3ef
---
 drivers/gpu/drm/i915/intel_uncore.c |   19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +12 -7
index 66365e7..a0331a7 100644
@@ -205,6 +205,13 @@ static void fw_domains_put_with_fifo(struct drm_i915_private *dev_priv,
 	gen6_gt_check_fifodbg(dev_priv);
 }
 
+static inline u32 fifo_free_entries(struct drm_i915_private *dev_priv)
+{
+	u32 count = __raw_i915_read32(dev_priv, GTFIFOCTL);
+
+	return count & GT_FIFO_FREE_ENTRIES_MASK;
+}
+
 static int __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv)
 {
 	int ret = 0;
@@ -212,16 +219,15 @@ static int __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv)
 	/* On VLV, FIFO will be shared by both SW and HW.
 	 * So, we need to read the FREE_ENTRIES everytime */
 	if (IS_VALLEYVIEW(dev_priv->dev))
-		dev_priv->uncore.fifo_count =
-			__raw_i915_read32(dev_priv, GTFIFOCTL) &
-						GT_FIFO_FREE_ENTRIES_MASK;
+		dev_priv->uncore.fifo_count = fifo_free_entries(dev_priv);
 
 	if (dev_priv->uncore.fifo_count < GT_FIFO_NUM_RESERVED_ENTRIES) {
 		int loop = 500;
-		u32 fifo = __raw_i915_read32(dev_priv, GTFIFOCTL) & GT_FIFO_FREE_ENTRIES_MASK;
+		u32 fifo = fifo_free_entries(dev_priv);
+
 		while (fifo <= GT_FIFO_NUM_RESERVED_ENTRIES && loop--) {
 			udelay(10);
-			fifo = __raw_i915_read32(dev_priv, GTFIFOCTL) & GT_FIFO_FREE_ENTRIES_MASK;
+			fifo = fifo_free_entries(dev_priv);
 		}
 		if (WARN_ON(loop < 0 && fifo <= GT_FIFO_NUM_RESERVED_ENTRIES))
 			++ret;
@@ -277,8 +283,7 @@ void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore)
 
 		if (IS_GEN6(dev) || IS_GEN7(dev))
 			dev_priv->uncore.fifo_count =
-				__raw_i915_read32(dev_priv, GTFIFOCTL) &
-				GT_FIFO_FREE_ENTRIES_MASK;
+				fifo_free_entries(dev_priv);
 	}
 
 	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
author Reviewed-by on v2 1/1
#40 lore raw

Re: [PATCH v2] drm/i915: FIFO space query code refactor

Quoted 9 lines · Dave Gordon

When querying the GTFIFOCTL register to check the FIFO space, the read value must be masked. The operation is repeated explicitly in several places. This change refactors the read-and-mask code into a function call.

v2: rebased on top of Mika's forcewake patch set, specifically:
[PATCH 8/8] drm/i915: Enum forcewake domains and domain identifiers

Change-Id Id1a9f3785cb20b82d4caa330c37b31e4e384a3ef
Quoted 60 lines
---
 drivers/gpu/drm/i915/intel_uncore.c |   19 ++++++++++++-------
 1 file changed, 12 insertions(+), 7 deletions(-)
drivers/gpu/drm/i915/intel_uncore.c +12 -7
index 66365e7..a0331a7 100644
@@ -205,6 +205,13 @@ static void fw_domains_put_with_fifo(struct drm_i915_private *dev_priv,
 	gen6_gt_check_fifodbg(dev_priv);
 }
 
+static inline u32 fifo_free_entries(struct drm_i915_private *dev_priv)
+{
+	u32 count = __raw_i915_read32(dev_priv, GTFIFOCTL);
+
+	return count & GT_FIFO_FREE_ENTRIES_MASK;
+}
+
 static int __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv)
 {
 	int ret = 0;
@@ -212,16 +219,15 @@ static int __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv)
 	/* On VLV, FIFO will be shared by both SW and HW.
 	 * So, we need to read the FREE_ENTRIES everytime */
 	if (IS_VALLEYVIEW(dev_priv->dev))
-		dev_priv->uncore.fifo_count =
-			__raw_i915_read32(dev_priv, GTFIFOCTL) &
-						GT_FIFO_FREE_ENTRIES_MASK;
+		dev_priv->uncore.fifo_count = fifo_free_entries(dev_priv);
 
 	if (dev_priv->uncore.fifo_count < GT_FIFO_NUM_RESERVED_ENTRIES) {
 		int loop = 500;
-		u32 fifo = __raw_i915_read32(dev_priv, GTFIFOCTL) & GT_FIFO_FREE_ENTRIES_MASK;
+		u32 fifo = fifo_free_entries(dev_priv);
+
 		while (fifo <= GT_FIFO_NUM_RESERVED_ENTRIES && loop--) {
 			udelay(10);
-			fifo = __raw_i915_read32(dev_priv, GTFIFOCTL) & GT_FIFO_FREE_ENTRIES_MASK;
+			fifo = fifo_free_entries(dev_priv);
 		}
 		if (WARN_ON(loop < 0 && fifo <= GT_FIFO_NUM_RESERVED_ENTRIES))
 			++ret;
@@ -277,8 +283,7 @@ void intel_uncore_forcewake_reset(struct drm_device *dev, bool restore)
 
 		if (IS_GEN6(dev) || IS_GEN7(dev))
 			dev_priv->uncore.fifo_count =
-				__raw_i915_read32(dev_priv, GTFIFOCTL) &
-				GT_FIFO_FREE_ENTRIES_MASK;
+				fifo_free_entries(dev_priv);
 	}
 
 	spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
Signature · 6 lines
1.7.9.5

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

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

#41 lore raw

Re: [PATCH v2] drm/i915: FIFO space query code refactor

Quoted 6 lines · Mika Kuoppala
Quoted 4 lines · Dave Gordon

When querying the GTFIFOCTL register to check the FIFO space, the read value must be masked. The operation is repeated explicitly in several places. This change refactors the read-and-mask code into a function call.

Quoted 7 lines
Quoted 5 lines

v2: rebased on top of Mika's forcewake patch set, specifically:
[PATCH 8/8] drm/i915: Enum forcewake domains and domain identifiers

Change-Id Id1a9f3785cb20b82d4caa330c37b31e4e384a3ef

Queued for -next, thanks for the patch.
-Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#42 lore raw

PATCH 5/10 drm/i915: Disable 'get seqno' workaround for VLV

In reply to [email protected]

From: Dave Gordon <[email protected]>

There is a workaround for a hardware bug when reading the seqno from the status page. The bug does not exist on VLV however, the workaround was still being applied.

Change-Id Ic781fdb31e1f794ce1fa8a6d0d5ee379756c5db6
---
 drivers/gpu/drm/i915/intel_ringbuffer.c |    5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_ringbuffer.c +4 -1
index 3887f1a..f990ce4 100644
@@ -2367,7 +2367,10 @@ int intel_init_render_ring_buffer(struct drm_device *dev)
 		ring->irq_get = gen6_ring_get_irq;
 		ring->irq_put = gen6_ring_put_irq;
 		ring->irq_enable_mask = GT_RENDER_USER_INTERRUPT;
-		ring->get_seqno = gen6_ring_get_seqno;
+		if (IS_VALLEYVIEW(dev))
+			ring->get_seqno = ring_get_seqno;
+		else
+			ring->get_seqno = gen6_ring_get_seqno;
 		ring->set_seqno = ring_set_seqno;
 		if (i915_semaphore_is_enabled(dev)) {
 			ring->semaphore.sync_to = gen6_ring_sync;
Signature · 6 lines
1.7.9.5

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

Re: [PATCH 05/10] drm/i915: Disable 'get seqno' workaround for VLV

Quoted 5 lines · On Tue, Dec 09, 2014 at 12:59:08PM +0000, [email protected] wrote:

From: Dave Gordon <[email protected]>

There is a workaround for a hardware bug when reading the seqno from the status page. The bug does not exist on VLV however, the workaround was still being applied.

Given how much trouble the missed seqno fun cause us I'd like more justification. What kind of (stress)-testing has been done here? And you're sure you've never seen the little dmesg notice that the kernel switched to polling?
-Daniel

Quoted 30 lines
Change-Id Ic781fdb31e1f794ce1fa8a6d0d5ee379756c5db6
---
 drivers/gpu/drm/i915/intel_ringbuffer.c |    5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_ringbuffer.c +4 -1
index 3887f1a..f990ce4 100644
@@ -2367,7 +2367,10 @@ int intel_init_render_ring_buffer(struct drm_device *dev)
 		ring->irq_get = gen6_ring_get_irq;
 		ring->irq_put = gen6_ring_put_irq;
 		ring->irq_enable_mask = GT_RENDER_USER_INTERRUPT;
-		ring->get_seqno = gen6_ring_get_seqno;
+		if (IS_VALLEYVIEW(dev))
+			ring->get_seqno = ring_get_seqno;
+		else
+			ring->get_seqno = gen6_ring_get_seqno;
 		ring->set_seqno = ring_set_seqno;
 		if (i915_semaphore_is_enabled(dev)) {
 			ring->semaphore.sync_to = gen6_ring_sync;
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#44 lore raw

Re: [PATCH 05/10] drm/i915: Disable 'get seqno' workaround for VLV

Quoted 7 lines · Daniel Vetter
Quoted 5 lines · On Tue, Dec 09, 2014 at 12:59:08PM +0000, [email protected] wrote:

From: Dave Gordon <[email protected]>

There is a workaround for a hardware bug when reading the seqno from the status page. The bug does not exist on VLV however, the workaround was still being applied.

Quoted 5 lines

Given how much trouble the missed seqno fun cause us I'd like more justification. What kind of (stress)-testing has been done here? And you're sure you've never seen the little dmesg notice that the kernel switched to polling?
-Daniel

This was necessary during VLV scheduler/preemption work, where we stressed the command streamer and IRQ path probably more than any of the usual tests, since we were looking for out-of-sequence anomalies.

The comments in gen6_ring_get_seqno() say it's to fix a bug on ivb/snb, and nothing in the BSpec suggests it's needed on later chips.

I think the problem was really in the fact that the workaround implemented in gen6_ring_get_seqno() was reading a nonshadowed register, therefore triggering forcewake activity, which was pretty buggy; so we took out the gen6-specific code before finding that there were still other problems with forcewake.

[Aside]
The original intention appears to have been to generate a single I/O READ cycle, forcing all pending inbound DMA so be flushed to memory, so that the subsequent memory read would see the latest value. But now, it doesn't just generate a single cycle but a whole flurry of writes and polling reads, because reading ACTHD requires forcewake (unless that was wrong, and has now been fixed). So should gen6_ring_get_seqno() instead read a register that doesn't require forcewake? I wouldn't object to using it in perpetuity if it really only read one register! [/aside]

.Dave.

Quoted 25 lines
Quoted 25 lines
Change-Id Ic781fdb31e1f794ce1fa8a6d0d5ee379756c5db6
---
 drivers/gpu/drm/i915/intel_ringbuffer.c |    5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_ringbuffer.c +4 -1
index 3887f1a..f990ce4 100644
@@ -2367,7 +2367,10 @@ int intel_init_render_ring_buffer(struct drm_device *dev)
 		ring->irq_get = gen6_ring_get_irq;
 		ring->irq_put = gen6_ring_put_irq;
 		ring->irq_enable_mask = GT_RENDER_USER_INTERRUPT;
-		ring->get_seqno = gen6_ring_get_seqno;
+		if (IS_VALLEYVIEW(dev))
+			ring->get_seqno = ring_get_seqno;
+		else
+			ring->get_seqno = gen6_ring_get_seqno;
 		ring->set_seqno = ring_set_seqno;
 		if (i915_semaphore_is_enabled(dev)) {
 			ring->semaphore.sync_to = gen6_ring_sync;
Signature · 1 line
1.7.9.5
Quoted 5 lines
Quoted 4 lines

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

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

#45 lore raw

Re: [PATCH 05/10] drm/i915: Disable 'get seqno' workaround for VLV

Quoted 27 lines · Dave Gordon
Quoted 7 lines · Daniel Vetter
Quoted 5 lines · On Tue, Dec 09, 2014 at 12:59:08PM +0000, [email protected] wrote:

From: Dave Gordon <[email protected]>

There is a workaround for a hardware bug when reading the seqno from the status page. The bug does not exist on VLV however, the workaround was still being applied.

Quoted 5 lines

Given how much trouble the missed seqno fun cause us I'd like more justification. What kind of (stress)-testing has been done here? And you're sure you've never seen the little dmesg notice that the kernel switched to polling?
-Daniel

This was necessary during VLV scheduler/preemption work, where we stressed the command streamer and IRQ path probably more than any of the usual tests, since we were looking for out-of-sequence anomalies.

The comments in gen6_ring_get_seqno() say it's to fix a bug on ivb/snb, and nothing in the BSpec suggests it's needed on later chips.

I think the problem was really in the fact that the workaround implemented in gen6_ring_get_seqno() was reading a nonshadowed register, therefore triggering forcewake activity, which was pretty buggy; so we took out the gen6-specific code before finding that there were still other problems with forcewake.

Quoted 10 lines

[Aside]
The original intention appears to have been to generate a single I/O READ cycle, forcing all pending inbound DMA so be flushed to memory, so that the subsequent memory read would see the latest value. But now, it doesn't just generate a single cycle but a whole flurry of writes and polling reads, because reading ACTHD requires forcewake (unless that was wrong, and has now been fixed). So should gen6_ring_get_seqno() instead read a register that doesn't require forcewake? I wouldn't object to using it in perpetuity if it really only read one register! [/aside]

The original intention was to work around missed interrupts. We still have them on pretty much everything gen6+. The workaround we've done for the performance issues this causes is the lazy_coherency trick where we only do this when waiting. It's all pretty nicely cargo-culted and not in Bspec, but it's also fairly real. We've given up on debugging, current kernels switch all wait_request calls to polling once the first missed interrupt happened. See the logic around gpu_error.missed_irq_rings.

It might be that it's not real on vlv and only on big core machines, but then we don't have a lot of people using vlv on upstream and reporting bugs.

For added insult we don't have any nice way to reproduce this really :( -Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#46 lore raw

PATCH 6/10 drm/i915: Add extra add_request calls

In reply to [email protected]

From: John Harrison <[email protected]>

The scheduler needs to track batch buffers by request without extra, non-batch buffer work being attached to the same request. This means that anywhere which adds work to the ring should explicitly call i915_add_request() when it has finished writing to the ring.

The add_request() function does extra work, such as flushing caches, that does not necessarily want to be done everywhere. Instead, a new i915_add_request_wo_flush() function has been added which skips the cache flush and just tidies up the request structure.

Note, much of this patch was implemented by Naresh Kumar Kachhi for pending power management improvements. However, it is also directly applicable to the scheduler work as noted above.

Change-Id: I66a6861118ee8e7ad7ca6c80c71a3b256db92e18
For: VIZ-1587

---
 drivers/gpu/drm/i915/i915_drv.h              |    9 ++++--
 drivers/gpu/drm/i915/i915_gem.c              |   45 ++++++++++++++++----------
 drivers/gpu/drm/i915/i915_gem_context.c      |    9 ++++++
 drivers/gpu/drm/i915/i915_gem_execbuffer.c   |    4 +--
 drivers/gpu/drm/i915/i915_gem_gtt.c          |    9 ++++++
 drivers/gpu/drm/i915/i915_gem_render_state.c |    2 +-
 drivers/gpu/drm/i915/intel_display.c         |   23 +++++++++----
 drivers/gpu/drm/i915/intel_lrc.c             |    4 +--
 8 files changed, 73 insertions(+), 32 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +6 -3
index 11e85cb..0e280c4 100644
@@ -2553,7 +2553,7 @@ static inline void i915_gem_object_unpin_pages(struct drm_i915_gem_object *obj)
 
 int __must_check i915_mutex_lock_interruptible(struct drm_device *dev);
 int i915_gem_object_sync(struct drm_i915_gem_object *obj,
-			 struct intel_engine_cs *to);
+			 struct intel_engine_cs *to, bool add_request);
 void i915_vma_move_to_active(struct i915_vma *vma,
 			     struct intel_engine_cs *ring);
 int i915_gem_dumb_create(struct drm_file *file_priv,
@@ -2641,9 +2641,12 @@ int __must_check i915_gpu_idle(struct drm_device *dev);
 int __must_check i915_gem_suspend(struct drm_device *dev);
 int __i915_add_request(struct intel_engine_cs *ring,
 		       struct drm_file *file,
-		       struct drm_i915_gem_object *batch_obj);
+		       struct drm_i915_gem_object *batch_obj,
+		       bool flush_caches);
 #define i915_add_request(ring) \
-	__i915_add_request(ring, NULL, NULL)
+	__i915_add_request(ring, NULL, NULL, true)
+#define i915_add_request_no_flush(ring) \
+	__i915_add_request(ring, NULL, NULL, false)
 int __i915_wait_request(struct drm_i915_gem_request *req,
 			unsigned reset_counter,
 			bool interruptible,
drivers/gpu/drm/i915/i915_gem.c +28 -17

101 lines not on this page · raw message

drivers/gpu/drm/i915/i915_gem_context.c +9 -0
index 5cd2b97..b2f1296 100644
@@ -426,6 +426,15 @@ int i915_gem_context_enable(struct drm_i915_private *dev_priv)
 			ret = i915_switch_context(ring, ring->default_context);
 			if (ret)
 				return ret;
+
+			/* Make sure the context switch (if one actually happened)
+			 * gets wrapped up and finished rather than hanging around
+			 * and confusing things later. */
+			if (ring->outstanding_lazy_request) {
+				ret = i915_add_request_no_flush(ring);
+				if (ret)
+					return ret;
+			}
 		}
 
 	return 0;
drivers/gpu/drm/i915/i915_gem_execbuffer.c +2 -2
index aa4566e..1268e89 100644
@@ -832,7 +832,7 @@ i915_gem_execbuffer_move_to_gpu(struct intel_engine_cs *ring,
 
 	list_for_each_entry(vma, vmas, exec_list) {
 		struct drm_i915_gem_object *obj = vma->obj;
-		ret = i915_gem_object_sync(obj, ring);
+		ret = i915_gem_object_sync(obj, ring, false);
 		if (ret)
 			return ret;
 
@@ -993,7 +993,7 @@ i915_gem_execbuffer_retire_commands(struct drm_device *dev,
 	ring->gpu_caches_dirty = true;
 
 	/* Add a breadcrumb for the completion of the batch buffer */
-	(void)__i915_add_request(ring, file, obj);
+	(void)__i915_add_request(ring, file, obj, true);
 }
 
 static int
drivers/gpu/drm/i915/i915_gem_gtt.c +9 -0
index ac03a38..7eead93 100644
@@ -1153,6 +1153,15 @@ int i915_ppgtt_init_hw(struct drm_device *dev)
 			ret = ppgtt->switch_mm(ppgtt, ring);
 			if (ret != 0)
 				return ret;
+
+			/* Make sure the context switch (if one actually happened)
+			 * gets wrapped up and finished rather than hanging around
+			 * and confusing things later. */
+			if (ring->outstanding_lazy_request) {
+				ret = i915_add_request_no_flush(ring);
+				if (ret)
+					return ret;
+			}
 		}
 	}
 
drivers/gpu/drm/i915/i915_gem_render_state.c +1 -1
index 521548a..aba39c3 100644
@@ -173,7 +173,7 @@ int i915_gem_render_state_init(struct intel_engine_cs *ring)
 
 	i915_vma_move_to_active(i915_gem_obj_to_ggtt(so.obj), ring);
 
-	ret = __i915_add_request(ring, NULL, so.obj);
+	ret = __i915_add_request(ring, NULL, so.obj, true);
 	/* __i915_add_request moves object to inactive if it fails */
 out:
 	i915_gem_render_state_fini(&so);
drivers/gpu/drm/i915/intel_display.c +16 -7
index a9bdc12..5b8d4f9 100644
@@ -9116,7 +9116,7 @@ static int intel_gen2_queue_flip(struct drm_device *dev,
 	intel_ring_emit(ring, 0); /* aux display base address, unused */
 
 	intel_mark_page_flip_active(intel_crtc);
-	__intel_ring_advance(ring);
+	i915_add_request_no_flush(ring);
 	return 0;
 }
 
@@ -9148,7 +9148,7 @@ static int intel_gen3_queue_flip(struct drm_device *dev,
 	intel_ring_emit(ring, MI_NOOP);
 
 	intel_mark_page_flip_active(intel_crtc);
-	__intel_ring_advance(ring);
+	i915_add_request_no_flush(ring);
 	return 0;
 }
 
@@ -9187,7 +9187,7 @@ static int intel_gen4_queue_flip(struct drm_device *dev,
 	intel_ring_emit(ring, pf | pipesrc);
 
 	intel_mark_page_flip_active(intel_crtc);
-	__intel_ring_advance(ring);
+	i915_add_request_no_flush(ring);
 	return 0;
 }
 
@@ -9223,7 +9223,7 @@ static int intel_gen6_queue_flip(struct drm_device *dev,
 	intel_ring_emit(ring, pf | pipesrc);
 
 	intel_mark_page_flip_active(intel_crtc);
-	__intel_ring_advance(ring);
+	i915_add_request_no_flush(ring);
 	return 0;
 }
 
@@ -9318,7 +9318,7 @@ static int intel_gen7_queue_flip(struct drm_device *dev,
 	intel_ring_emit(ring, (MI_NOOP));
 
 	intel_mark_page_flip_active(intel_crtc);
-	__intel_ring_advance(ring);
+	i915_add_request_no_flush(ring);
 	return 0;
 }
 
@@ -9528,7 +9528,7 @@ static int intel_gen9_queue_flip(struct drm_device *dev,
 	intel_ring_emit(ring, intel_crtc->unpin_work->gtt_offset);
 
 	intel_mark_page_flip_active(intel_crtc);
-	__intel_ring_advance(ring);
+	i915_add_request_no_flush(ring);
 
 	return 0;
 }
@@ -9734,8 +9734,17 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,
 		if (ret)
 			goto cleanup_unpin;
 
+		/* Borked: need to get the seqno for the request submitted in
+		 * 'queue_flip()' above. However, either the request has been
+		 * posted already and the seqno is gone (q_f calls add_request),
+		 * or the request never gets posted and is merged into whatever
+		 * render comes along next (q_f calls ring_advance).
+		 *
+		 * On the other hand, seqnos are going away soon anyway! So
+		 * hopefully the problem will disappear...
+		 */
 		i915_gem_request_assign(&work->flip_queued_req,
-					intel_ring_get_request(ring));
+					ring->outstanding_lazy_request ? intel_ring_get_request(ring) : NULL);
 	}
 
 	work->flip_queued_vblank = drm_vblank_count(dev, intel_crtc->pipe);
drivers/gpu/drm/i915/intel_lrc.c +2 -2
index 82a47e3..643a56a 100644
@@ -617,7 +617,7 @@ static int execlists_move_to_gpu(struct intel_ringbuffer *ringbuf,
 	list_for_each_entry(vma, vmas, exec_list) {
 		struct drm_i915_gem_object *obj = vma->obj;
 
-		ret = i915_gem_object_sync(obj, ring);
+		ret = i915_gem_object_sync(obj, ring, true);
 		if (ret)
 			return ret;
 
@@ -1630,7 +1630,7 @@ int intel_lr_context_render_state_init(struct intel_engine_cs *ring,
 
 	i915_vma_move_to_active(i915_gem_obj_to_ggtt(so.obj), ring);
 
-	ret = __i915_add_request(ring, file, so.obj);
+	ret = __i915_add_request(ring, file, so.obj, true);
 	/* intel_logical_ring_add_request moves object to inactive if it
 	 * fails */
 out:
Signature · 6 lines
1.7.9.5

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

Re: [PATCH 06/10] drm/i915: Add extra add_request calls

Quoted 12 lines · On Tue, Dec 09, 2014 at 12:59:09PM +0000, [email protected] wrote:

From: John Harrison <[email protected]>

The scheduler needs to track batch buffers by request without extra, non-batch buffer work being attached to the same request. This means that anywhere which adds work to the ring should explicitly call i915_add_request() when it has finished writing to the ring.

The add_request() function does extra work, such as flushing caches, that does not necessarily want to be done everywhere. Instead, a new i915_add_request_wo_flush() function has been added which skips the cache flush and just tidies up the request structure.

Quoted 7 lines

Note, much of this patch was implemented by Naresh Kumar Kachhi for pending power management improvements. However, it is also directly applicable to the scheduler work as noted above.

Change-Id: I66a6861118ee8e7ad7ca6c80c71a3b256db92e18
For: VIZ-1587

For easier review I think it'd be better to split this into the prep patch to rework add_request and then patches to roll it out. Some of the places you're adding an add_requst too look questionable to me. More comments below.

After reading through the patch (see below for comments) I think there's only two places where we need to add requests:
- at the end of pageflips, unconditionally (to catch any semaphores already emitted if something fails)
- after ring init in init_hw (but perhaps with some of the restructuring in-flight applied to make it less messy).

Then a WARN_ON in gpu_idle to make sure no request ever escaped. Should all be separate patches I think (so that we have a nice empty space in the commit message for explanation why this works).

This patch here just sprinkles lots of add_requests all over which ime tends to blow up sooner or later since it's much harder to understand and leaves no room for selfchecks in the code to catch bugs.

Cheers, Daniel

Quoted 114 lines
---
 drivers/gpu/drm/i915/i915_drv.h              |    9 ++++--
 drivers/gpu/drm/i915/i915_gem.c              |   45 ++++++++++++++++----------
 drivers/gpu/drm/i915/i915_gem_context.c      |    9 ++++++
 drivers/gpu/drm/i915/i915_gem_execbuffer.c   |    4 +--
 drivers/gpu/drm/i915/i915_gem_gtt.c          |    9 ++++++
 drivers/gpu/drm/i915/i915_gem_render_state.c |    2 +-
 drivers/gpu/drm/i915/intel_display.c         |   23 +++++++++----
 drivers/gpu/drm/i915/intel_lrc.c             |    4 +--
 8 files changed, 73 insertions(+), 32 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +6 -3
index 11e85cb..0e280c4 100644
@@ -2553,7 +2553,7 @@ static inline void i915_gem_object_unpin_pages(struct drm_i915_gem_object *obj)
 
 int __must_check i915_mutex_lock_interruptible(struct drm_device *dev);
 int i915_gem_object_sync(struct drm_i915_gem_object *obj,
-			 struct intel_engine_cs *to);
+			 struct intel_engine_cs *to, bool add_request);
 void i915_vma_move_to_active(struct i915_vma *vma,
 			     struct intel_engine_cs *ring);
 int i915_gem_dumb_create(struct drm_file *file_priv,
@@ -2641,9 +2641,12 @@ int __must_check i915_gpu_idle(struct drm_device *dev);
 int __must_check i915_gem_suspend(struct drm_device *dev);
 int __i915_add_request(struct intel_engine_cs *ring,
 		       struct drm_file *file,
-		       struct drm_i915_gem_object *batch_obj);
+		       struct drm_i915_gem_object *batch_obj,
+		       bool flush_caches);
 #define i915_add_request(ring) \
-	__i915_add_request(ring, NULL, NULL)
+	__i915_add_request(ring, NULL, NULL, true)
+#define i915_add_request_no_flush(ring) \
+	__i915_add_request(ring, NULL, NULL, false)
 int __i915_wait_request(struct drm_i915_gem_request *req,
 			unsigned reset_counter,
 			bool interruptible,
drivers/gpu/drm/i915/i915_gem.c +14 -16
index de241eb..b022a2d 100644
@@ -2419,7 +2419,8 @@ i915_gem_get_seqno(struct drm_device *dev, u32 *seqno)
 
 int __i915_add_request(struct intel_engine_cs *ring,
 		       struct drm_file *file,
-		       struct drm_i915_gem_object *obj)
+		       struct drm_i915_gem_object *obj,
+		       bool flush_caches)
 {
 	struct drm_i915_private *dev_priv = ring->dev->dev_private;
 	struct drm_i915_gem_request *request;
@@ -2445,12 +2446,11 @@ int __i915_add_request(struct intel_engine_cs *ring,
 	 * is that the flush _must_ happen before the next request, no matter
 	 * what.
 	 */
-	if (i915.enable_execlists) {
-		ret = logical_ring_flush_all_caches(ringbuf);
-		if (ret)
-			return ret;
-	} else {
-		ret = intel_ring_flush_all_caches(ring);
+	if (flush_caches) {
+		if (i915.enable_execlists)
+			ret = logical_ring_flush_all_caches(ringbuf);
+		else
+			ret = intel_ring_flush_all_caches(ring);
 		if (ret)
 			return ret;
 	}
@@ -2462,15 +2462,12 @@ int __i915_add_request(struct intel_engine_cs *ring,
 	 */
 	request_ring_position = intel_ring_get_tail(ringbuf);
 
-	if (i915.enable_execlists) {
+	if (i915.enable_execlists)
 		ret = ring->emit_request(ringbuf);
-		if (ret)
-			return ret;
-	} else {
+	else
 		ret = ring->add_request(ring);
-		if (ret)
-			return ret;
-	}
+	if (ret)
+		return ret;
 
 	request->head = request_start;
 	request->tail = request_ring_position;
@@ -2974,6 +2971,8 @@ out:
  *
  * @obj: object which may be in use on another ring.
  * @to: ring we wish to use the object on. May be NULL.
+ * @add_request: do we need to add a request to track operations
+ *    submitted on ring with sync_to function
  *
  * This code is meant to abstract object synchronization with the GPU.
  * Calling with NULL implies synchronizing the object with the CPU
@@ -2983,7 +2982,7 @@ out:
  */
 int
 i915_gem_object_sync(struct drm_i915_gem_object *obj,
-		     struct intel_engine_cs *to)
+		     struct intel_engine_cs *to, bool add_request)
 {
 	struct intel_engine_cs *from;
 	u32 seqno;
@@ -3011,13 +3010,16 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj,
 
 	trace_i915_gem_ring_sync_to(from, to, obj->last_read_req);
 	ret = to->semaphore.sync_to(to, from, seqno);
-	if (!ret)

drivers/gpu/drm/i915/i915_gem.c i915_gem_object_sync(struct drm_i915_gem_object *obj, · line 3022

Quoted 10 lines
drivers/gpu/drm/i915/i915_gem.c +4 -0
+	if (!ret) {
 		/* We use last_read_req because sync_to()
 		 * might have just caused seqno wrap under
 		 * the radar.
 		 */
 		from->semaphore.sync_seqno[idx] =
 				i915_gem_request_get_seqno(obj->last_read_req);
+		if (add_request)
+			i915_add_request_no_flush(to);
+	}

Is the scheduler really still using hw semaphores for inter-ring sync? I've thought all the depency tracking gets pushed into software? And even then I expect that we'll insert semaphores really late and should treat them as part of the subsequent batch of pageflip.

Quoted 5 lines
drivers/gpu/drm/i915/i915_gem.c +0 -0
 	return ret;
 }
@@ -3126,6 +3128,15 @@ int i915_gpu_idle(struct drm_device *dev)
 				return ret;

drivers/gpu/drm/i915/i915_gem.c int i915_gpu_idle(struct drm_device *dev) · line 3138

Quoted 10 lines
drivers/gpu/drm/i915/i915_gem.c +8 -0
 		}
 
+		/* Make sure the context switch (if one actually happened)
+		 * gets wrapped up and finished rather than hanging around
+		 * and confusing things later. */
+		if (ring->outstanding_lazy_request) {
+			ret = i915_add_request(ring);
+			if (ret)
+				return ret;
+		}

I expect bad interactions with the ilk iommu w/a (see do_idling in i915_gem_gtt.c) due to the recursion this will cause in the execbuf code.

Except for this case we shouldn't ever leak a request with your code anywhere any more, so I think instead this should be a WARN_ON (ofc only when do_idle_maps is false).

drivers/gpu/drm/i915/i915_gem.c i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj, · line 3960

Quoted 10 lines
drivers/gpu/drm/i915/i915_gem.c +2 -1
+
 		ret = intel_ring_idle(ring);
 		if (ret)
 			return ret;
@@ -3946,7 +3957,7 @@ i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj,
 	int ret;
 
 	if (pipelined != i915_gem_request_get_ring(obj->last_read_req)) {
-		ret = i915_gem_object_sync(obj, pipelined);
+		ret = i915_gem_object_sync(obj, pipelined, true);

Async flips should be caught by the flip requests added below. Essentially same comment as above with object_sync.

Quoted 8 lines
drivers/gpu/drm/i915/i915_gem.c +0 -0
 		if (ret)
 			return ret;
 	}
drivers/gpu/drm/i915/i915_gem_context.c +0 -0
index 5cd2b97..b2f1296 100644

drivers/gpu/drm/i915/i915_gem_context.c int i915_gem_context_enable(struct drm_i915_private *dev_priv) · line 434

Quoted 9 lines
drivers/gpu/drm/i915/i915_gem_context.c +6 -0
@@ -426,6 +426,15 @@ int i915_gem_context_enable(struct drm_i915_private *dev_priv)
 			ret = i915_switch_context(ring, ring->default_context);
 			if (ret)
 				return ret;
+
+			/* Make sure the context switch (if one actually happened)
+			 * gets wrapped up and finished rather than hanging around
+			 * and confusing things later. */
+			if (ring->outstanding_lazy_request) {
+				ret = i915_add_request_no_flush(ring);

We have a lot more randome stuff all over when intializing rings (pde loads, l3c remapping, context, golden context, w/a). Imo we should have one central add request for all of them. We've the recently reworked ring init plus some of the patches from Chris (to preload the ring before enabling/submitting the ring) that should be fairly straighforward to implement.

Quoted 167 lines
drivers/gpu/drm/i915/i915_gem_context.c +3 -0
+				if (ret)
+					return ret;
+			}
 		}
 
 	return 0;
drivers/gpu/drm/i915/i915_gem_execbuffer.c +2 -2
index aa4566e..1268e89 100644
@@ -832,7 +832,7 @@ i915_gem_execbuffer_move_to_gpu(struct intel_engine_cs *ring,
 
 	list_for_each_entry(vma, vmas, exec_list) {
 		struct drm_i915_gem_object *obj = vma->obj;
-		ret = i915_gem_object_sync(obj, ring);
+		ret = i915_gem_object_sync(obj, ring, false);
 		if (ret)
 			return ret;
 
@@ -993,7 +993,7 @@ i915_gem_execbuffer_retire_commands(struct drm_device *dev,
 	ring->gpu_caches_dirty = true;
 
 	/* Add a breadcrumb for the completion of the batch buffer */
-	(void)__i915_add_request(ring, file, obj);
+	(void)__i915_add_request(ring, file, obj, true);
 }
 
 static int
drivers/gpu/drm/i915/i915_gem_gtt.c +9 -0
index ac03a38..7eead93 100644
@@ -1153,6 +1153,15 @@ int i915_ppgtt_init_hw(struct drm_device *dev)
 			ret = ppgtt->switch_mm(ppgtt, ring);
 			if (ret != 0)
 				return ret;
+
+			/* Make sure the context switch (if one actually happened)
+			 * gets wrapped up and finished rather than hanging around
+			 * and confusing things later. */
+			if (ring->outstanding_lazy_request) {
+				ret = i915_add_request_no_flush(ring);
+				if (ret)
+					return ret;
+			}
 		}
 	}
 
drivers/gpu/drm/i915/i915_gem_render_state.c +1 -1
index 521548a..aba39c3 100644
@@ -173,7 +173,7 @@ int i915_gem_render_state_init(struct intel_engine_cs *ring)
 
 	i915_vma_move_to_active(i915_gem_obj_to_ggtt(so.obj), ring);
 
-	ret = __i915_add_request(ring, NULL, so.obj);
+	ret = __i915_add_request(ring, NULL, so.obj, true);
 	/* __i915_add_request moves object to inactive if it fails */
 out:
 	i915_gem_render_state_fini(&so);
drivers/gpu/drm/i915/intel_display.c +16 -7
index a9bdc12..5b8d4f9 100644
@@ -9116,7 +9116,7 @@ static int intel_gen2_queue_flip(struct drm_device *dev,
 	intel_ring_emit(ring, 0); /* aux display base address, unused */
 
 	intel_mark_page_flip_active(intel_crtc);
-	__intel_ring_advance(ring);
+	i915_add_request_no_flush(ring);
 	return 0;
 }
 
@@ -9148,7 +9148,7 @@ static int intel_gen3_queue_flip(struct drm_device *dev,
 	intel_ring_emit(ring, MI_NOOP);
 
 	intel_mark_page_flip_active(intel_crtc);
-	__intel_ring_advance(ring);
+	i915_add_request_no_flush(ring);
 	return 0;
 }
 
@@ -9187,7 +9187,7 @@ static int intel_gen4_queue_flip(struct drm_device *dev,
 	intel_ring_emit(ring, pf | pipesrc);
 
 	intel_mark_page_flip_active(intel_crtc);
-	__intel_ring_advance(ring);
+	i915_add_request_no_flush(ring);
 	return 0;
 }
 
@@ -9223,7 +9223,7 @@ static int intel_gen6_queue_flip(struct drm_device *dev,
 	intel_ring_emit(ring, pf | pipesrc);
 
 	intel_mark_page_flip_active(intel_crtc);
-	__intel_ring_advance(ring);
+	i915_add_request_no_flush(ring);
 	return 0;
 }
 
@@ -9318,7 +9318,7 @@ static int intel_gen7_queue_flip(struct drm_device *dev,
 	intel_ring_emit(ring, (MI_NOOP));
 
 	intel_mark_page_flip_active(intel_crtc);
-	__intel_ring_advance(ring);
+	i915_add_request_no_flush(ring);
 	return 0;
 }
 
@@ -9528,7 +9528,7 @@ static int intel_gen9_queue_flip(struct drm_device *dev,
 	intel_ring_emit(ring, intel_crtc->unpin_work->gtt_offset);
 
 	intel_mark_page_flip_active(intel_crtc);
-	__intel_ring_advance(ring);
+	i915_add_request_no_flush(ring);
 
 	return 0;
 }
@@ -9734,8 +9734,17 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,
 		if (ret)
 			goto cleanup_unpin;
 
+		/* Borked: need to get the seqno for the request submitted in
+		 * 'queue_flip()' above. However, either the request has been
+		 * posted already and the seqno is gone (q_f calls add_request),
+		 * or the request never gets posted and is merged into whatever
+		 * render comes along next (q_f calls ring_advance).
+		 *
+		 * On the other hand, seqnos are going away soon anyway! So
+		 * hopefully the problem will disappear...
+		 */
 		i915_gem_request_assign(&work->flip_queued_req,
-					intel_ring_get_request(ring));
+					ring->outstanding_lazy_request ? intel_ring_get_request(ring) : NULL);
 	}
 
 	work->flip_queued_vblank = drm_vblank_count(dev, intel_crtc->pipe);
drivers/gpu/drm/i915/intel_lrc.c +2 -2
index 82a47e3..643a56a 100644
@@ -617,7 +617,7 @@ static int execlists_move_to_gpu(struct intel_ringbuffer *ringbuf,
 	list_for_each_entry(vma, vmas, exec_list) {
 		struct drm_i915_gem_object *obj = vma->obj;
 
-		ret = i915_gem_object_sync(obj, ring);
+		ret = i915_gem_object_sync(obj, ring, true);
 		if (ret)
 			return ret;
 
@@ -1630,7 +1630,7 @@ int intel_lr_context_render_state_init(struct intel_engine_cs *ring,
 
 	i915_vma_move_to_active(i915_gem_obj_to_ggtt(so.obj), ring);
 
-	ret = __i915_add_request(ring, file, so.obj);
+	ret = __i915_add_request(ring, file, so.obj, true);
 	/* intel_logical_ring_add_request moves object to inactive if it
 	 * fails */
 out:
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#48 lore raw

PATCH 7/10 drm/i915: Early alloc request

In reply to [email protected]

From: John Harrison <[email protected]>

The scheduler needs to explicitly allocate a request to track each submitted batch buffer. This must happen a long time before any commands are actually written to the ring.

Change-Id: Id01fbda123bcfaa84531896c38292435270025b1
For: VIZ-1587

---
 drivers/gpu/drm/i915/i915_drv.h            |    2 ++
 drivers/gpu/drm/i915/i915_gem.c            |    2 ++
 drivers/gpu/drm/i915/i915_gem_execbuffer.c |   13 ++++++++++---
 drivers/gpu/drm/i915/intel_lrc.c           |    6 +++---
 drivers/gpu/drm/i915/intel_lrc.h           |    2 ++
 drivers/gpu/drm/i915/intel_ringbuffer.c    |    6 +++---
 drivers/gpu/drm/i915/intel_ringbuffer.h    |    2 ++
 7 files changed, 24 insertions(+), 9 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +2 -0
index 0e280c4..8628a83 100644
@@ -1775,6 +1775,8 @@ struct drm_i915_private {
 
 	/* Abstract the submission mechanism (legacy ringbuffer or execlists) away */
 	struct {
+		int (*alloc_request)(struct intel_engine_cs *ring,
+				     struct intel_context *ctx);
 		int (*do_execbuf)(struct drm_device *dev, struct drm_file *file,
 				  struct intel_engine_cs *ring,
 				  struct intel_context *ctx,
drivers/gpu/drm/i915/i915_gem.c +2 -0
index b022a2d..09332ff 100644
@@ -4870,11 +4870,13 @@ int i915_gem_init(struct drm_device *dev)
 	}
 
 	if (!i915.enable_execlists) {
+		dev_priv->gt.alloc_request = intel_ring_alloc_request;
 		dev_priv->gt.do_execbuf = i915_gem_ringbuffer_submission;
 		dev_priv->gt.init_rings = i915_gem_init_rings;
 		dev_priv->gt.cleanup_ring = intel_cleanup_ring_buffer;
 		dev_priv->gt.stop_ring = intel_stop_ring_buffer;
 	} else {
+		dev_priv->gt.alloc_request = intel_logical_ring_alloc_request;
 		dev_priv->gt.do_execbuf = intel_execlists_submission;
 		dev_priv->gt.init_rings = intel_logical_rings_init;
 		dev_priv->gt.cleanup_ring = intel_logical_ring_cleanup;
drivers/gpu/drm/i915/i915_gem_execbuffer.c +10 -3
index 1268e89..f09501c 100644
@@ -1287,7 +1287,7 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 	u64 exec_start = args->batch_start_offset;
 	u32 dispatch_flags;
 	int ret;
-	bool need_relocs;
+	bool need_relocs, batch_pinned = false;
 
 	if (!i915_gem_check_execbuffer(args))
 		return -EINVAL;
@@ -1436,23 +1436,30 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 		if (ret)
 			goto err;
 
+		batch_pinned = true;
 		exec_start += i915_gem_obj_ggtt_offset(batch_obj);
 	} else
 		exec_start += i915_gem_obj_offset(batch_obj, vm);
 
+	/* Allocate a request for this batch buffer nice and early. */
+	ret = dev_priv->gt.alloc_request(ring, ctx);
+	if (ret)
+		goto err;
+
 	ret = dev_priv->gt.do_execbuf(dev, file, ring, ctx, args,
 				      &eb->vmas, batch_obj, exec_start,
 				      dispatch_flags);
 
+err:
 	/*
 	 * FIXME: We crucially rely upon the active tracking for the (ppgtt)
 	 * batch vma for correctness. For less ugly and less fragility this
 	 * needs to be adjusted to also track the ggtt batch vma properly as
 	 * active.
 	 */
-	if (dispatch_flags & I915_DISPATCH_SECURE)
+	if (batch_pinned)
 		i915_gem_object_ggtt_unpin(batch_obj);
-err:
+
 	/* the request owns the ref now */
 	i915_gem_context_unreference(ctx);
 	eb_destroy(eb);
drivers/gpu/drm/i915/intel_lrc.c +3 -3
index 643a56a..037cbd5 100644
@@ -878,8 +878,8 @@ void intel_lr_context_unpin(struct intel_engine_cs *ring,
 	}
 }
 
-static int logical_ring_alloc_request(struct intel_engine_cs *ring,
-				      struct intel_context *ctx)
+int intel_logical_ring_alloc_request(struct intel_engine_cs *ring,
+				     struct intel_context *ctx)
 {
 	struct drm_i915_gem_request *request;
 	struct drm_i915_private *dev_private = ring->dev->dev_private;
@@ -1082,7 +1082,7 @@ int intel_logical_ring_begin(struct intel_ringbuffer *ringbuf, int num_dwords)
 		return ret;
 
 	/* Preallocate the olr before touching the ring */
-	ret = logical_ring_alloc_request(ring, ringbuf->FIXME_lrc_ctx);
+	ret = intel_logical_ring_alloc_request(ring, ringbuf->FIXME_lrc_ctx);
 	if (ret)
 		return ret;
 
drivers/gpu/drm/i915/intel_lrc.h +2 -0
index b8d5364..2bf868a 100644
@@ -34,6 +34,8 @@
 #define RING_CONTEXT_STATUS_PTR(ring)	((ring)->mmio_base+0x3a0)
 
 /* Logical Rings */
+int __must_check intel_logical_ring_alloc_request(struct intel_engine_cs *ring,
+						  struct intel_context *ctx);
 void intel_logical_ring_stop(struct intel_engine_cs *ring);
 void intel_logical_ring_cleanup(struct intel_engine_cs *ring);
 int intel_logical_rings_init(struct drm_device *dev);
drivers/gpu/drm/i915/intel_ringbuffer.c +3 -3
index f990ce4..65ffdd1 100644
@@ -2023,8 +2023,8 @@ int intel_ring_idle(struct intel_engine_cs *ring)
 	return i915_wait_request(req);
 }
 
-static int
-intel_ring_alloc_request(struct intel_engine_cs *ring)
+int
+intel_ring_alloc_request(struct intel_engine_cs *ring, struct intel_context *ctx)
 {
 	int ret;
 	struct drm_i915_gem_request *request;
@@ -2088,7 +2088,7 @@ int intel_ring_begin(struct intel_engine_cs *ring,
 		return ret;
 
 	/* Preallocate the olr before touching the ring */
-	ret = intel_ring_alloc_request(ring);
+	ret = intel_ring_alloc_request(ring, NULL);
 	if (ret)
 		return ret;
 
drivers/gpu/drm/i915/intel_ringbuffer.h +2 -0
index f15fc46..b313144 100644
@@ -430,6 +430,8 @@ void intel_cleanup_ring_buffer(struct intel_engine_cs *ring);
 
 int __must_check intel_ring_begin(struct intel_engine_cs *ring, int n);
 int __must_check intel_ring_cacheline_align(struct intel_engine_cs *ring);
+int __must_check intel_ring_alloc_request(struct intel_engine_cs *ring,
+					  struct intel_context *ctx);
 static inline void intel_ring_emit(struct intel_engine_cs *ring,
 				   u32 data)
 {
Signature · 6 lines
1.7.9.5

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

PATCH 8/10 drm/i915: Prelude to splitting i915_gem_do_execbuffer in two

In reply to [email protected]

From: John Harrison <[email protected]>

The scheduler decouples the submission of batch buffers to the driver with their submission to the hardware. This basically means splitting the execbuffer() function in half. This change rearranges some code ready for the split to occur.

Change-Id: Icc9c8afaac18821f3eb8a151a49f918f90c068a3
For: VIZ-1587

---
 drivers/gpu/drm/i915/i915_gem_execbuffer.c |   47 +++++++++++++++++-----------
 drivers/gpu/drm/i915/intel_lrc.c           |   17 +++++++---
 2 files changed, 40 insertions(+), 24 deletions(-)
drivers/gpu/drm/i915/i915_gem_execbuffer.c +28 -19
index f09501c..a339556 100644
@@ -848,10 +848,7 @@ i915_gem_execbuffer_move_to_gpu(struct intel_engine_cs *ring,
 	if (flush_domains & I915_GEM_DOMAIN_GTT)
 		wmb();
 
-	/* Unconditionally invalidate gpu caches and ensure that we do flush
-	 * any residual writes from the previous batch.
-	 */
-	return intel_ring_invalidate_all_caches(ring);
+	return 0;
 }
 
 static bool
@@ -1123,14 +1120,6 @@ i915_gem_ringbuffer_submission(struct drm_device *dev, struct drm_file *file,
 		}
 	}
 
-	ret = i915_gem_execbuffer_move_to_gpu(ring, vmas);
-	if (ret)
-		goto error;
-
-	ret = i915_switch_context(ring, ctx);
-	if (ret)
-		goto error;
-
 	instp_mode = args->flags & I915_EXEC_CONSTANTS_MASK;
 	instp_mask = I915_EXEC_CONSTANTS_MASK;
 	switch (instp_mode) {
@@ -1168,6 +1157,28 @@ i915_gem_ringbuffer_submission(struct drm_device *dev, struct drm_file *file,
 		goto error;
 	}
 
+	ret = i915_gem_execbuffer_move_to_gpu(ring, vmas);
+	if (ret)
+		goto error;
+
+	i915_gem_execbuffer_move_to_active(vmas, ring);
+
+	/* To be split into two functions here... */
+
+	intel_runtime_pm_get(dev_priv);
+
+	/* Unconditionally invalidate gpu caches and ensure that we do flush
+	 * any residual writes from the previous batch.
+	 */
+	ret = intel_ring_invalidate_all_caches(ring);
+	if (ret)
+		goto error;
+
+	/* Switch to the correct context for the batch */
+	ret = i915_switch_context(ring, ctx);
+	if (ret)
+		goto error;
+
 	if (ring == &dev_priv->ring[RCS] &&
 			instp_mode != dev_priv->relative_constants_mode) {
 		ret = intel_ring_begin(ring, 4);
@@ -1208,15 +1219,18 @@ i915_gem_ringbuffer_submission(struct drm_device *dev, struct drm_file *file,
 						exec_start, exec_len,
 						dispatch_flags);
 		if (ret)
-			return ret;
+			goto error;
 	}
 
 	trace_i915_gem_ring_dispatch(intel_ring_get_request(ring), dispatch_flags);
 
-	i915_gem_execbuffer_move_to_active(vmas, ring);
 	i915_gem_execbuffer_retire_commands(dev, file, ring, batch_obj);
 
 error:
+	/* intel_gpu_busy should also get a ref, so it will free when the device
+	 * is really idle. */
+	intel_runtime_pm_put(dev_priv);
+
 	kfree(cliprects);
 	return ret;
 }
@@ -1335,8 +1349,6 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 		return -EINVAL;
 	}
 
-	intel_runtime_pm_get(dev_priv);
-
 	ret = i915_mutex_lock_interruptible(dev);
 	if (ret)
 		goto pre_mutex_err;
@@ -1467,9 +1479,6 @@ err:
 	mutex_unlock(&dev->struct_mutex);
 
 pre_mutex_err:
-	/* intel_gpu_busy should also get a ref, so it will free when the device
-	 * is really idle. */
-	intel_runtime_pm_put(dev_priv);
 	return ret;
 }
 
drivers/gpu/drm/i915/intel_lrc.c +12 -5
index 037cbd5..f16b15d 100644
@@ -630,10 +630,7 @@ static int execlists_move_to_gpu(struct intel_ringbuffer *ringbuf,
 	if (flush_domains & I915_GEM_DOMAIN_GTT)
 		wmb();
 
-	/* Unconditionally invalidate gpu caches and ensure that we do flush
-	 * any residual writes from the previous batch.
-	 */
-	return logical_ring_invalidate_all_caches(ringbuf);
+	return 0;
 }
 
 /**
@@ -717,6 +714,17 @@ int intel_execlists_submission(struct drm_device *dev, struct drm_file *file,
 	if (ret)
 		return ret;
 
+	i915_gem_execbuffer_move_to_active(vmas, ring);
+
+	/* To be split into two functions here... */
+
+	/* Unconditionally invalidate gpu caches and ensure that we do flush
+	 * any residual writes from the previous batch.
+	 */
+	ret = logical_ring_invalidate_all_caches(ringbuf);
+	if (ret)
+		return ret;
+
 	if (ring == &dev_priv->ring[RCS] &&
 	    instp_mode != dev_priv->relative_constants_mode) {
 		ret = intel_logical_ring_begin(ringbuf, 4);
@@ -738,7 +746,6 @@ int intel_execlists_submission(struct drm_device *dev, struct drm_file *file,
 
 	trace_i915_gem_ring_dispatch(intel_ring_get_request(ring), dispatch_flags);
 
-	i915_gem_execbuffer_move_to_active(vmas, ring);
 	i915_gem_execbuffer_retire_commands(dev, file, ring, batch_obj);
 
 	return 0;
Signature · 6 lines
1.7.9.5

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

Re: [PATCH 08/10] drm/i915: Prelude to splitting i915_gem_do_execbuffer in two

Quoted 5 lines · On Tue, Dec 09, 2014 at 12:59:11PM +0000, [email protected] wrote:

From: John Harrison <[email protected]>

The scheduler decouples the submission of batch buffers to the driver with their submission to the hardware. This basically means splitting the execbuffer() function in half. This change rearranges some code ready for the split to occur.

Now there's the curios question: Where will the split in top/bottom halves be? Without that I have no idea whether it makes sense and so can't review this patch. At least if the goal is to really prep for the scheduler and not just move a few lines around ;-)
-Daniel

Quoted 158 lines

Change-Id: Icc9c8afaac18821f3eb8a151a49f918f90c068a3
For: VIZ-1587

---
 drivers/gpu/drm/i915/i915_gem_execbuffer.c |   47 +++++++++++++++++-----------
 drivers/gpu/drm/i915/intel_lrc.c           |   17 +++++++---
 2 files changed, 40 insertions(+), 24 deletions(-)
drivers/gpu/drm/i915/i915_gem_execbuffer.c +28 -19
index f09501c..a339556 100644
@@ -848,10 +848,7 @@ i915_gem_execbuffer_move_to_gpu(struct intel_engine_cs *ring,
 	if (flush_domains & I915_GEM_DOMAIN_GTT)
 		wmb();
 
-	/* Unconditionally invalidate gpu caches and ensure that we do flush
-	 * any residual writes from the previous batch.
-	 */
-	return intel_ring_invalidate_all_caches(ring);
+	return 0;
 }
 
 static bool
@@ -1123,14 +1120,6 @@ i915_gem_ringbuffer_submission(struct drm_device *dev, struct drm_file *file,
 		}
 	}
 
-	ret = i915_gem_execbuffer_move_to_gpu(ring, vmas);
-	if (ret)
-		goto error;
-
-	ret = i915_switch_context(ring, ctx);
-	if (ret)
-		goto error;
-
 	instp_mode = args->flags & I915_EXEC_CONSTANTS_MASK;
 	instp_mask = I915_EXEC_CONSTANTS_MASK;
 	switch (instp_mode) {
@@ -1168,6 +1157,28 @@ i915_gem_ringbuffer_submission(struct drm_device *dev, struct drm_file *file,
 		goto error;
 	}
 
+	ret = i915_gem_execbuffer_move_to_gpu(ring, vmas);
+	if (ret)
+		goto error;
+
+	i915_gem_execbuffer_move_to_active(vmas, ring);
+
+	/* To be split into two functions here... */
+
+	intel_runtime_pm_get(dev_priv);
+
+	/* Unconditionally invalidate gpu caches and ensure that we do flush
+	 * any residual writes from the previous batch.
+	 */
+	ret = intel_ring_invalidate_all_caches(ring);
+	if (ret)
+		goto error;
+
+	/* Switch to the correct context for the batch */
+	ret = i915_switch_context(ring, ctx);
+	if (ret)
+		goto error;
+
 	if (ring == &dev_priv->ring[RCS] &&
 			instp_mode != dev_priv->relative_constants_mode) {
 		ret = intel_ring_begin(ring, 4);
@@ -1208,15 +1219,18 @@ i915_gem_ringbuffer_submission(struct drm_device *dev, struct drm_file *file,
 						exec_start, exec_len,
 						dispatch_flags);
 		if (ret)
-			return ret;
+			goto error;
 	}
 
 	trace_i915_gem_ring_dispatch(intel_ring_get_request(ring), dispatch_flags);
 
-	i915_gem_execbuffer_move_to_active(vmas, ring);
 	i915_gem_execbuffer_retire_commands(dev, file, ring, batch_obj);
 
 error:
+	/* intel_gpu_busy should also get a ref, so it will free when the device
+	 * is really idle. */
+	intel_runtime_pm_put(dev_priv);
+
 	kfree(cliprects);
 	return ret;
 }
@@ -1335,8 +1349,6 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data,
 		return -EINVAL;
 	}
 
-	intel_runtime_pm_get(dev_priv);
-
 	ret = i915_mutex_lock_interruptible(dev);
 	if (ret)
 		goto pre_mutex_err;
@@ -1467,9 +1479,6 @@ err:
 	mutex_unlock(&dev->struct_mutex);
 
 pre_mutex_err:
-	/* intel_gpu_busy should also get a ref, so it will free when the device
-	 * is really idle. */
-	intel_runtime_pm_put(dev_priv);
 	return ret;
 }
 
drivers/gpu/drm/i915/intel_lrc.c +12 -5
index 037cbd5..f16b15d 100644
@@ -630,10 +630,7 @@ static int execlists_move_to_gpu(struct intel_ringbuffer *ringbuf,
 	if (flush_domains & I915_GEM_DOMAIN_GTT)
 		wmb();
 
-	/* Unconditionally invalidate gpu caches and ensure that we do flush
-	 * any residual writes from the previous batch.
-	 */
-	return logical_ring_invalidate_all_caches(ringbuf);
+	return 0;
 }
 
 /**
@@ -717,6 +714,17 @@ int intel_execlists_submission(struct drm_device *dev, struct drm_file *file,
 	if (ret)
 		return ret;
 
+	i915_gem_execbuffer_move_to_active(vmas, ring);
+
+	/* To be split into two functions here... */
+
+	/* Unconditionally invalidate gpu caches and ensure that we do flush
+	 * any residual writes from the previous batch.
+	 */
+	ret = logical_ring_invalidate_all_caches(ringbuf);
+	if (ret)
+		return ret;
+
 	if (ring == &dev_priv->ring[RCS] &&
 	    instp_mode != dev_priv->relative_constants_mode) {
 		ret = intel_logical_ring_begin(ringbuf, 4);
@@ -738,7 +746,6 @@ int intel_execlists_submission(struct drm_device *dev, struct drm_file *file,
 
 	trace_i915_gem_ring_dispatch(intel_ring_get_request(ring), dispatch_flags);
 
-	i915_gem_execbuffer_move_to_active(vmas, ring);
 	i915_gem_execbuffer_retire_commands(dev, file, ring, batch_obj);
 
 	return 0;
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#51 lore raw

Re: [PATCH 08/10] drm/i915: Prelude to splitting i915_gem_do_execbuffer in two

Quoted 13 lines · Daniel Vetter
Quoted 5 lines · On Tue, Dec 09, 2014 at 12:59:11PM +0000, [email protected] wrote:

From: John Harrison <[email protected]>

The scheduler decouples the submission of batch buffers to the driver with their submission to the hardware. This basically means splitting the execbuffer() function in half. This change rearranges some code ready for the split to occur.

Now there's the curios question: Where will the split in top/bottom halves be? Without that I have no idea whether it makes sense and so can't review this patch. At least if the goal is to really prep for the scheduler and not just move a few lines around ;-)
-Daniel

[snip]

Quoted 9 lines

drivers/gpu/drm/i915/intel_lrc.c int intel_execlists_submission(struct drm_device *dev, struct drm_file *file, · line 724

Quoted 9 lines
drivers/gpu/drm/i915/intel_lrc.c +8 -0
+	i915_gem_execbuffer_move_to_active(vmas, ring);
+
+	/* To be split into two functions here... */
+
+	/* Unconditionally invalidate gpu caches and ensure that we do flush
+	 * any residual writes from the previous batch.
+	 */
+	ret = logical_ring_invalidate_all_caches(ringbuf);

It'll be where the marker comment is above. Ahead of that point is stuff to do with setting up software state; after that we're talking to h/w. When the scheduler code goes it, it decouples the two by interposing at this point. Then batches go into it with s/w state set up, but don't get to talk to the h/w until they're selected for execution, possibly in a different order.

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

#52 lore raw

Re: [PATCH 08/10] drm/i915: Prelude to splitting i915_gem_do_execbuffer in two

Quoted 16 lines · Dave Gordon
Quoted 13 lines · Daniel Vetter
Quoted 5 lines · On Tue, Dec 09, 2014 at 12:59:11PM +0000, [email protected] wrote:

From: John Harrison <[email protected]>

The scheduler decouples the submission of batch buffers to the driver with their submission to the hardware. This basically means splitting the execbuffer() function in half. This change rearranges some code ready for the split to occur.

Now there's the curios question: Where will the split in top/bottom halves be? Without that I have no idea whether it makes sense and so can't review this patch. At least if the goal is to really prep for the scheduler and not just move a few lines around ;-)
-Daniel

[snip]

Quoted 1 line
Quoted 1 line
Quoted 15 lines
Quoted 8 lines

drivers/gpu/drm/i915/intel_lrc.c int intel_execlists_submission(struct drm_device *dev, struct drm_file *file, · line 724

Quoted 8 lines
drivers/gpu/drm/i915/intel_lrc.c +8 -0
+	i915_gem_execbuffer_move_to_active(vmas, ring);
+
+	/* To be split into two functions here... */
+
+	/* Unconditionally invalidate gpu caches and ensure that we do flush
+	 * any residual writes from the previous batch.
+	 */
+	ret = logical_ring_invalidate_all_caches(ringbuf);

It'll be where the marker comment is above. Ahead of that point is stuff to do with setting up software state; after that we're talking to h/w. When the scheduler code goes it, it decouples the two by interposing at this point. Then batches go into it with s/w state set up, but don't get to talk to the h/w until they're selected for execution, possibly in a different order.

Oops, I guess I need see a doctor to check my eyesight ;-)

Two comments about code that's still in the bottom half:
- There's lots of input validation code still below that cutoff point which needs to be moved above to still be able to return -EINVAL. Test coverage is the critical part here, but I think we've closed most of our gaps. I guess a future patch will address this?

- retire_commands and so add_request are also in the cutoff. For shrinker interactions to work seamlessly we need to have both the objects on the active list and the request in the lists. So I expect more untangling there to separate the request emission to rings from the bookkeeping parts in add_request.

Also move_to_active will fall apart once we start to reorder requests I think. Need to think about this case more, but we definitely need some testcases with depencies and reordering by the scheduler. -Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#53 lore raw

Re: [PATCH 08/10] drm/i915: Prelude to splitting i915_gem_do_execbuffer in two

Quoted 33 lines · Daniel Vetter
Quoted 16 lines · Dave Gordon
Quoted 13 lines · Daniel Vetter
Quoted 5 lines · On Tue, Dec 09, 2014 at 12:59:11PM +0000, [email protected] wrote:

From: John Harrison <[email protected]>

The scheduler decouples the submission of batch buffers to the driver with their submission to the hardware. This basically means splitting the execbuffer() function in half. This change rearranges some code ready for the split to occur.

Now there's the curios question: Where will the split in top/bottom halves be? Without that I have no idea whether it makes sense and so can't review this patch. At least if the goal is to really prep for the scheduler and not just move a few lines around ;-)
-Daniel

[snip]

Quoted 1 line
Quoted 1 line
Quoted 15 lines
Quoted 8 lines

drivers/gpu/drm/i915/intel_lrc.c int intel_execlists_submission(struct drm_device *dev, struct drm_file *file, · line 724

Quoted 8 lines
drivers/gpu/drm/i915/intel_lrc.c +8 -0
+	i915_gem_execbuffer_move_to_active(vmas, ring);
+
+	/* To be split into two functions here... */
+
+	/* Unconditionally invalidate gpu caches and ensure that we do flush
+	 * any residual writes from the previous batch.
+	 */
+	ret = logical_ring_invalidate_all_caches(ringbuf);

It'll be where the marker comment is above. Ahead of that point is stuff to do with setting up software state; after that we're talking to h/w. When the scheduler code goes it, it decouples the two by interposing at this point. Then batches go into it with s/w state set up, but don't get to talk to the h/w until they're selected for execution, possibly in a different order.

Quoted 7 lines

Oops, I guess I need see a doctor to check my eyesight ;-)

Two comments about code that's still in the bottom half:
- There's lots of input validation code still below that cutoff point which needs to be moved above to still be able to return -EINVAL. Test coverage is the critical part here, but I think we've closed most of our gaps. I guess a future patch will address this?

Yes, all validation will end up before the batch gets pushed into the scheduler queue. It had already been shuffled there in an earlier version, but some of it then got relocated into less convenient places during the legacy-vs-execlist split. So for now, the execbuffer path has some common code (including early validation), then a split by submission mechanism (with additional validation that may someday get reconverged into the common section); and back to common code, and /then/ the batches get passed to the scheduler, whereafter there won't be any more EINVAL cases (though there will be other error cases, such as ENOMEM). Then as and when batches are selected for execution, they're passed to the specific backend submission mechanism code (which of course won't contain any further validation code either).

Quoted 10 lines

- retire_commands and so add_request are also in the cutoff. For shrinker interactions to work seamlessly we need to have both the objects on the active list and the request in the lists. So I expect more untangling there to separate the request emission to rings from the bookkeeping parts in add_request.

Also move_to_active will fall apart once we start to reorder requests I think. Need to think about this case more, but we definitely need some testcases with depencies and reordering by the scheduler. -Daniel

I'll leave the rest to John to comment on, as he's still rebasing this section of the scheduler on top of all the other changes that have been going on ...

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

#54 lore raw

Re: [PATCH 08/10] drm/i915: Prelude to splitting i915_gem_do_execbuffer in two

Quoted 42 lines · Dave Gordon
Quoted 33 lines · Daniel Vetter
Quoted 16 lines · Dave Gordon
Quoted 13 lines · Daniel Vetter
Quoted 5 lines · On Tue, Dec 09, 2014 at 12:59:11PM +0000, [email protected] wrote:

From: John Harrison <[email protected]>

The scheduler decouples the submission of batch buffers to the driver with their submission to the hardware. This basically means splitting the execbuffer() function in half. This change rearranges some code ready for the split to occur.

Now there's the curios question: Where will the split in top/bottom halves be? Without that I have no idea whether it makes sense and so can't review this patch. At least if the goal is to really prep for the scheduler and not just move a few lines around ;-)
-Daniel

[snip]

Quoted 1 line
Quoted 1 line
Quoted 15 lines
Quoted 8 lines

drivers/gpu/drm/i915/intel_lrc.c int intel_execlists_submission(struct drm_device *dev, struct drm_file *file, · line 724

Quoted 8 lines
drivers/gpu/drm/i915/intel_lrc.c +8 -0
+	i915_gem_execbuffer_move_to_active(vmas, ring);
+
+	/* To be split into two functions here... */
+
+	/* Unconditionally invalidate gpu caches and ensure that we do flush
+	 * any residual writes from the previous batch.
+	 */
+	ret = logical_ring_invalidate_all_caches(ringbuf);

It'll be where the marker comment is above. Ahead of that point is stuff to do with setting up software state; after that we're talking to h/w. When the scheduler code goes it, it decouples the two by interposing at this point. Then batches go into it with s/w state set up, but don't get to talk to the h/w until they're selected for execution, possibly in a different order.

Quoted 7 lines

Oops, I guess I need see a doctor to check my eyesight ;-)

Two comments about code that's still in the bottom half:
- There's lots of input validation code still below that cutoff point which needs to be moved above to still be able to return -EINVAL. Test coverage is the critical part here, but I think we've closed most of our gaps. I guess a future patch will address this?

Quoted 12 lines

Yes, all validation will end up before the batch gets pushed into the scheduler queue. It had already been shuffled there in an earlier version, but some of it then got relocated into less convenient places during the legacy-vs-execlist split. So for now, the execbuffer path has some common code (including early validation), then a split by submission mechanism (with additional validation that may someday get reconverged into the common section); and back to common code, and /then/ the batches get passed to the scheduler, whereafter there won't be any more EINVAL cases (though there will be other error cases, such as ENOMEM). Then as and when batches are selected for execution, they're passed to the specific backend submission mechanism code (which of course won't contain any further validation code either).

Hm, what can still go ENOMEM in there? We should pin everything we need, the struct is preallocated and everything else should be in place too. Or at least that's how I think it should work out. If the scheduler needs any allocations for its internal tracking the we need to push them all into other places, most likely requests.
-Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#55 lore raw

Re: [PATCH 08/10] drm/i915: Prelude to splitting i915_gem_do_execbuffer in two

Quoted 56 lines · Daniel Vetter
Quoted 42 lines · Dave Gordon
Quoted 33 lines · Daniel Vetter
Quoted 16 lines · Dave Gordon
Quoted 13 lines · Daniel Vetter
Quoted 5 lines · On Tue, Dec 09, 2014 at 12:59:11PM +0000, [email protected] wrote:

From: John Harrison <[email protected]>

The scheduler decouples the submission of batch buffers to the driver with their submission to the hardware. This basically means splitting the execbuffer() function in half. This change rearranges some code ready for the split to occur.

Now there's the curios question: Where will the split in top/bottom halves be? Without that I have no idea whether it makes sense and so can't review this patch. At least if the goal is to really prep for the scheduler and not just move a few lines around ;-)
-Daniel

[snip]

Quoted 1 line
Quoted 1 line
Quoted 15 lines
Quoted 8 lines

drivers/gpu/drm/i915/intel_lrc.c int intel_execlists_submission(struct drm_device *dev, struct drm_file *file, · line 724

Quoted 8 lines
drivers/gpu/drm/i915/intel_lrc.c +8 -0
+	i915_gem_execbuffer_move_to_active(vmas, ring);
+
+	/* To be split into two functions here... */
+
+	/* Unconditionally invalidate gpu caches and ensure that we do flush
+	 * any residual writes from the previous batch.
+	 */
+	ret = logical_ring_invalidate_all_caches(ringbuf);

It'll be where the marker comment is above. Ahead of that point is stuff to do with setting up software state; after that we're talking to h/w. When the scheduler code goes it, it decouples the two by interposing at this point. Then batches go into it with s/w state set up, but don't get to talk to the h/w until they're selected for execution, possibly in a different order.

Quoted 7 lines

Oops, I guess I need see a doctor to check my eyesight ;-)

Two comments about code that's still in the bottom half:
- There's lots of input validation code still below that cutoff point which needs to be moved above to still be able to return -EINVAL. Test coverage is the critical part here, but I think we've closed most of our gaps. I guess a future patch will address this?

Quoted 12 lines

Yes, all validation will end up before the batch gets pushed into the scheduler queue. It had already been shuffled there in an earlier version, but some of it then got relocated into less convenient places during the legacy-vs-execlist split. So for now, the execbuffer path has some common code (including early validation), then a split by submission mechanism (with additional validation that may someday get reconverged into the common section); and back to common code, and /then/ the batches get passed to the scheduler, whereafter there won't be any more EINVAL cases (though there will be other error cases, such as ENOMEM). Then as and when batches are selected for execution, they're passed to the specific backend submission mechanism code (which of course won't contain any further validation code either).

Quoted 6 lines

Hm, what can still go ENOMEM in there? We should pin everything we need, the struct is preallocated and everything else should be in place too. Or at least that's how I think it should work out. If the scheduler needs any allocations for its internal tracking the we need to push them all into other places, most likely requests.
-Daniel

It used to be the case that the scheduler had to allocate some tracking data, but John may well have changed that with the advent of the seqno-request transformation.

I'd like to eventually move all batch-related allocations into a single chunk per request, done after input validation and before acquiring any locks :)

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

#56 lore raw

PATCH 9/10 drm/i915: Split i915_dem_do_execbuffer() in half

In reply to [email protected]

From: John Harrison <[email protected]>

Split the execbuffer() function in half. The first half collects and validates all the information requried to process the batch buffer. It also does all the object pinning, relocations, active list management, etc - basically anything that must be done upfront before the IOCTL returns and allows the user land side to start changing/freeing things. The second half does the actual ring submission.

This change implements the split but leaves the back half being called directly from the end of the front half.

Change-Id: I5e1c77639ce526ab2401b0323186c518bf13da0a
For: VIZ-1587

---
 drivers/gpu/drm/i915/i915_drv.h            |   39 ++++---
 drivers/gpu/drm/i915/i915_gem.c            |    2 +
 drivers/gpu/drm/i915/i915_gem_execbuffer.c |  172 +++++++++++++++++++---------
 drivers/gpu/drm/i915/intel_lrc.c           |   72 ++++++++----
 drivers/gpu/drm/i915/intel_lrc.h           |   10 +-
 5 files changed, 197 insertions(+), 98 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +26 -13
index 8628a83..f885a90 100644
@@ -1525,6 +1525,25 @@ struct i915_workarounds {
 	u32 count;
 };
 
+struct i915_execbuffer_params {
+	struct drm_device               *dev;
+	struct drm_file                 *file;
+	uint32_t                        dispatch_flags;
+	uint32_t                        args_flags;
+	uint32_t                        args_batch_start_offset;
+	uint32_t                        args_batch_len;
+	uint32_t                        args_num_cliprects;
+	uint32_t                        args_DR1;
+	uint32_t                        args_DR4;
+	uint32_t                        batch_obj_vm_offset;
+	struct intel_engine_cs          *ring;
+	struct drm_i915_gem_object      *batch_obj;
+	struct drm_clip_rect            *cliprects;
+	uint32_t                        instp_mask;
+	int                             instp_mode;
+	struct intel_context            *ctx;
+};
+
 struct drm_i915_private {
 	struct drm_device *dev;
 	struct kmem_cache *slab;
@@ -1777,13 +1796,10 @@ struct drm_i915_private {
 	struct {
 		int (*alloc_request)(struct intel_engine_cs *ring,
 				     struct intel_context *ctx);
-		int (*do_execbuf)(struct drm_device *dev, struct drm_file *file,
-				  struct intel_engine_cs *ring,
-				  struct intel_context *ctx,
+		int (*do_execbuf)(struct i915_execbuffer_params *params,
 				  struct drm_i915_gem_execbuffer2 *args,
-				  struct list_head *vmas,
-				  struct drm_i915_gem_object *batch_obj,
-				  u64 exec_start, u32 flags);
+				  struct list_head *vmas);
+		int (*do_execfinal)(struct i915_execbuffer_params *params);
 		int (*init_rings)(struct drm_device *dev);
 		void (*cleanup_ring)(struct intel_engine_cs *ring);
 		void (*stop_ring)(struct intel_engine_cs *ring);
@@ -2464,14 +2480,11 @@ void i915_gem_execbuffer_retire_commands(struct drm_device *dev,
 					 struct drm_file *file,
 					 struct intel_engine_cs *ring,
 					 struct drm_i915_gem_object *obj);
-int i915_gem_ringbuffer_submission(struct drm_device *dev,
-				   struct drm_file *file,
-				   struct intel_engine_cs *ring,
-				   struct intel_context *ctx,
+void i915_gem_execbuff_release_batch_obj(struct drm_i915_gem_object *batch_obj);
+int i915_gem_ringbuffer_submission(struct i915_execbuffer_params *qe,
 				   struct drm_i915_gem_execbuffer2 *args,
-				   struct list_head *vmas,
-				   struct drm_i915_gem_object *batch_obj,
-				   u64 exec_start, u32 flags);
+				   struct list_head *vmas);
+int i915_gem_ringbuffer_submission_final(struct i915_execbuffer_params *params);
 int i915_gem_execbuffer(struct drm_device *dev, void *data,
 			struct drm_file *file_priv);
 int i915_gem_execbuffer2(struct drm_device *dev, void *data,
drivers/gpu/drm/i915/i915_gem.c +2 -0
index 09332ff..031af9d 100644
@@ -4872,12 +4872,14 @@ int i915_gem_init(struct drm_device *dev)
 	if (!i915.enable_execlists) {
 		dev_priv->gt.alloc_request = intel_ring_alloc_request;
 		dev_priv->gt.do_execbuf = i915_gem_ringbuffer_submission;
+		dev_priv->gt.do_execfinal = i915_gem_ringbuffer_submission_final;
 		dev_priv->gt.init_rings = i915_gem_init_rings;
 		dev_priv->gt.cleanup_ring = intel_cleanup_ring_buffer;
 		dev_priv->gt.stop_ring = intel_stop_ring_buffer;
 	} else {
 		dev_priv->gt.alloc_request = intel_logical_ring_alloc_request;
 		dev_priv->gt.do_execbuf = intel_execlists_submission;
+		dev_priv->gt.do_execfinal = intel_execlists_submission_final;
 		dev_priv->gt.init_rings = intel_logical_rings_init;
 		dev_priv->gt.cleanup_ring = intel_logical_ring_cleanup;
 		dev_priv->gt.stop_ring = intel_logical_ring_stop;
drivers/gpu/drm/i915/i915_gem_execbuffer.c +117 -55

316 lines not on this page · raw message

drivers/gpu/drm/i915/intel_lrc.c +48 -24

124 lines not on this page · raw message

drivers/gpu/drm/i915/intel_lrc.h +4 -6
index 2bf868a..ea083d9 100644
@@ -76,13 +76,11 @@ void intel_lr_context_unpin(struct intel_engine_cs *ring,
 
 /* Execlists */
 int intel_sanitize_enable_execlists(struct drm_device *dev, int enable_execlists);
-int intel_execlists_submission(struct drm_device *dev, struct drm_file *file,
-			       struct intel_engine_cs *ring,
-			       struct intel_context *ctx,
+struct i915_execbuffer_params;
+int intel_execlists_submission(struct i915_execbuffer_params *params,
 			       struct drm_i915_gem_execbuffer2 *args,
-			       struct list_head *vmas,
-			       struct drm_i915_gem_object *batch_obj,
-			       u64 exec_start, u32 dispatch_flags);
+			       struct list_head *vmas);
+int intel_execlists_submission_final(struct i915_execbuffer_params *params);
 u32 intel_execlists_ctx_id(struct drm_i915_gem_object *ctx_obj);
 
 /**
Signature · 6 lines
1.7.9.5

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

PATCH 10/10 drm/i915: Cache ringbuf pointer in request structure

In reply to [email protected]

From: John Harrison <[email protected]>

Cache a pointer to the ringbuf so that retire requests does not need to do messy 'are we in execlist mode' decisions.

Change-Id: I75bf6b73dc8a675a5be7bbff1927aa30f393788c
For: VIZ-1587

---
 drivers/gpu/drm/i915/i915_drv.h         |    3 ++-
 drivers/gpu/drm/i915/i915_gem.c         |   14 +-------------
 drivers/gpu/drm/i915/intel_lrc.c        |    3 ++-
 drivers/gpu/drm/i915/intel_ringbuffer.c |    1 +
 4 files changed, 6 insertions(+), 15 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +2 -1
index f885a90..b340455 100644
@@ -2024,8 +2024,9 @@ struct drm_i915_gem_request {
 	/** Position in the ringbuffer of the end of the request */
 	u32 tail;
 
-	/** Context related to this request */
+	/** Context and ring buffer related to this request */
 	struct intel_context *ctx;
+	struct intel_ringbuffer *ringbuf;
 
 	/** Batch buffer related to this request if any */
 	struct drm_i915_gem_object *batch_obj;
drivers/gpu/drm/i915/i915_gem.c +1 -13
index 031af9d..b8ee384 100644
@@ -2762,7 +2762,6 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *ring)
 
 	while (!list_empty(&ring->request_list)) {
 		struct drm_i915_gem_request *request;
-		struct intel_ringbuffer *ringbuf;
 
 		request = list_first_entry(&ring->request_list,
 					   struct drm_i915_gem_request,
@@ -2773,23 +2772,12 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *ring)
 
 		trace_i915_gem_request_retire(request);
 
-		/* This is one of the few common intersection points
-		 * between legacy ringbuffer submission and execlists:
-		 * we need to tell them apart in order to find the correct
-		 * ringbuffer to which the request belongs to.
-		 */
-		if (i915.enable_execlists) {
-			struct intel_context *ctx = request->ctx;
-			ringbuf = ctx->engine[ring->id].ringbuf;
-		} else
-			ringbuf = ring->buffer;
-
 		/* We know the GPU must have read the request to have
 		 * sent us the seqno + interrupt, so use the position
 		 * of tail of the request to update the last known position
 		 * of the GPU head.
 		 */
-		ringbuf->last_retired_head = request->tail;
+		request->ringbuf->last_retired_head = request->tail;
 
 		i915_gem_free_request(request);
 	}
drivers/gpu/drm/i915/intel_lrc.c +2 -1
index 971ba0c..684eb1e 100644
@@ -944,10 +944,11 @@ int intel_logical_ring_alloc_request(struct intel_engine_cs *ring,
 
 	/* Hold a reference to the context this request belongs to
 	 * (we will need it when the time comes to emit/retire the
-	 * request).
+	 * request). Likewise, the ringbuff is useful to keep track of.
 	 */
 	request->ctx = ctx;
 	i915_gem_context_reference(request->ctx);
+	request->ringbuf = ctx->engine[ring->id].ringbuf;
 
 	ring->outstanding_lazy_request = request;
 	return 0;
drivers/gpu/drm/i915/intel_ringbuffer.c +1 -0
index 65ffdd1..f0242cf 100644
@@ -2039,6 +2039,7 @@ intel_ring_alloc_request(struct intel_engine_cs *ring, struct intel_context *ctx
 
 	kref_init(&request->ref);
 	request->ring = ring;
+	request->ringbuf = ring->buffer;
 	request->uniq = dev_private->request_uniq++;
 
 	ret = i915_gem_get_seqno(ring->dev, &request->seqno);
Signature · 6 lines
1.7.9.5

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