Skip to contentSkip to conversation

drm/i915: fix HW lockup due to missing RPS IRQ workaround on GEN6

PATCH In review Imre Deak 3 participants 3 messages Dec 19, 2014

Subsystems (2)
lore
Share & tools
lore

Imre Deak

author
#0 lore raw

PATCH 1/1 drm/i915: fix HW lockup due to missing RPS IRQ workaround on GEN6

In

commit dbea3cea69508e9d548ed4a6be13de35492e5d15

Author: Imre Deak <[email protected]>
Date:   Mon Dec 15 18:59:28 2014 +0200

drm/i915: sanitize RPS resetting during GPU reset

we disable RPS interrupts during GPU resetting, but don't apply the necessary GEN6 HW workaround. This leads to a HW lockup during a subsequent "looping batchbuffer" workload. This is triggered by the testcase that submits exactly this kind of workload after a simulated GPU reset. I'm not sure how likely the bug would have triggered otherwise, since we would have applied the workaround anyway shortly after the GPU reset, when enabling GT powersaving from the deferred work.

This may also fix unrelated issues, since during driver loading / suspending we also disable RPS interrupts and so we also had a short window during the rest of the loading / resuming where a similar workload could run without the workaround applied.

At the same time also keep the IRQ routing to display enabled at all times and everywhere on GEN8+, since there isn't ever a reason to route them to GT. For instance this way we could catch spurious interrupts on the CPU instead of allowing some undefined behaviour on the GT (due to the same spurious interrupt routed there).

Testcase: igt/gem_reset_stats/ban-ctx-render
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=87429

---
 drivers/gpu/drm/i915/i915_irq.c  | 17 +++++++++++++++--
 drivers/gpu/drm/i915/intel_drv.h |  1 +
 drivers/gpu/drm/i915/intel_pm.c  | 11 +----------
 3 files changed, 17 insertions(+), 12 deletions(-)
drivers/gpu/drm/i915/i915_irq.c +15 -2
index aa3180c..12acc40 100644
@@ -296,6 +296,20 @@ void gen6_enable_rps_interrupts(struct drm_device *dev)
 	spin_unlock_irq(&dev_priv->irq_lock);
 }
 
+u32 gen6_sanitize_rps_pm_mask(struct drm_i915_private *dev_priv, u32 mask)
+{
+	/* IVB and SNB hard hangs on looping batchbuffer
+	 * if GEN6_PM_UP_EI_EXPIRED is masked.
+	 */
+	if (INTEL_INFO(dev_priv->dev)->gen <= 7 && !IS_HASWELL(dev_priv->dev))
+		mask &= ~GEN6_PM_RP_UP_EI_EXPIRED;
+
+	if (INTEL_INFO(dev_priv->dev)->gen >= 8)
+		mask &= ~GEN8_PMINTR_REDIRECT_TO_NON_DISP;
+
+	return mask;
+}
+
 void gen6_disable_rps_interrupts(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -308,8 +322,7 @@ void gen6_disable_rps_interrupts(struct drm_device *dev)
 
 	spin_lock_irq(&dev_priv->irq_lock);
 
-	I915_WRITE(GEN6_PMINTRMSK, INTEL_INFO(dev_priv)->gen >= 8 ?
-		   ~GEN8_PMINTR_REDIRECT_TO_NON_DISP : ~0);
+	I915_WRITE(GEN6_PMINTRMSK, gen6_sanitize_rps_pm_mask(dev_priv, ~0));
 
 	__gen6_disable_pm_irq(dev_priv, dev_priv->pm_rps_events);
 	I915_WRITE(gen6_pm_ier(dev_priv), I915_READ(gen6_pm_ier(dev_priv)) &
drivers/gpu/drm/i915/intel_drv.h +1 -0
index 588b618..bb871f3 100644
@@ -795,6 +795,7 @@ void gen6_disable_pm_irq(struct drm_i915_private *dev_priv, uint32_t mask);
 void gen6_reset_rps_interrupts(struct drm_device *dev);
 void gen6_enable_rps_interrupts(struct drm_device *dev);
 void gen6_disable_rps_interrupts(struct drm_device *dev);
+u32 gen6_sanitize_rps_pm_mask(struct drm_i915_private *dev_priv, u32 mask);
 void intel_runtime_pm_disable_interrupts(struct drm_i915_private *dev_priv);
 void intel_runtime_pm_enable_interrupts(struct drm_i915_private *dev_priv);
 static inline bool intel_irqs_enabled(struct drm_i915_private *dev_priv)
drivers/gpu/drm/i915/intel_pm.c +1 -10
index a3ebaa8..4bd1b8b 100644
@@ -3745,16 +3745,7 @@ static u32 gen6_rps_pm_mask(struct drm_i915_private *dev_priv, u8 val)
 	mask |= dev_priv->pm_rps_events & (GEN6_PM_RP_DOWN_EI_EXPIRED | GEN6_PM_RP_UP_EI_EXPIRED);
 	mask &= dev_priv->pm_rps_events;
 
-	/* IVB and SNB hard hangs on looping batchbuffer
-	 * if GEN6_PM_UP_EI_EXPIRED is masked.
-	 */
-	if (INTEL_INFO(dev_priv->dev)->gen <= 7 && !IS_HASWELL(dev_priv->dev))
-		mask |= GEN6_PM_RP_UP_EI_EXPIRED;
-
-	if (IS_GEN8(dev_priv->dev))
-		mask |= GEN8_PMINTR_REDIRECT_TO_NON_DISP;
-
-	return ~mask;
+	return gen6_sanitize_rps_pm_mask(dev_priv, ~mask);
 }
 
 /* gen6_set_rps is called to update the frequency request, but should also be
Signature · 6 lines
1.8.4

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

In

commit dbea3cea69508e9d548ed4a6be13de35492e5d15

Author: Imre Deak <[email protected]>
Date:   Mon Dec 15 18:59:28 2014 +0200

drm/i915: sanitize RPS resetting during GPU reset

we disable RPS interrupts during GPU resetting, but don't apply the necessary GEN6 HW workaround. This leads to a HW lockup during a subsequent "looping batchbuffer" workload. This is triggered by the testcase that submits exactly this kind of workload after a simulated GPU reset. I'm not sure how likely the bug would have triggered otherwise, since we would have applied the workaround anyway shortly after the GPU reset, when enabling GT powersaving from the deferred work.

Quoted 4 lines

This may also fix unrelated issues, since during driver loading / suspending we also disable RPS interrupts and so we also had a short window during the rest of the loading / resuming where a similar workload could run without the workaround applied.

Hm this might explain why the reset tests randomly hard-crashed snb every once in a while. So symptoms matches. Mika, can you please give this a spin?

Quoted 5 lines

At the same time also keep the IRQ routing to display enabled at all times and everywhere on GEN8+, since there isn't ever a reason to route them to GT. For instance this way we could catch spurious interrupts on the CPU instead of allowing some undefined behaviour on the GT (due to the same spurious interrupt routed there).

I don't really grok that part, but it does feel like a separate patch (especially since you also change from IS_GEN8 to gen >= 8). Or maybe I'm just really confused.
-Daniel

Quoted 86 lines

Testcase: igt/gem_reset_stats/ban-ctx-render
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=87429

---
 drivers/gpu/drm/i915/i915_irq.c  | 17 +++++++++++++++--
 drivers/gpu/drm/i915/intel_drv.h |  1 +
 drivers/gpu/drm/i915/intel_pm.c  | 11 +----------
 3 files changed, 17 insertions(+), 12 deletions(-)
drivers/gpu/drm/i915/i915_irq.c +15 -2
index aa3180c..12acc40 100644
@@ -296,6 +296,20 @@ void gen6_enable_rps_interrupts(struct drm_device *dev)
 	spin_unlock_irq(&dev_priv->irq_lock);
 }
 
+u32 gen6_sanitize_rps_pm_mask(struct drm_i915_private *dev_priv, u32 mask)
+{
+	/* IVB and SNB hard hangs on looping batchbuffer
+	 * if GEN6_PM_UP_EI_EXPIRED is masked.
+	 */
+	if (INTEL_INFO(dev_priv->dev)->gen <= 7 && !IS_HASWELL(dev_priv->dev))
+		mask &= ~GEN6_PM_RP_UP_EI_EXPIRED;
+
+	if (INTEL_INFO(dev_priv->dev)->gen >= 8)
+		mask &= ~GEN8_PMINTR_REDIRECT_TO_NON_DISP;
+
+	return mask;
+}
+
 void gen6_disable_rps_interrupts(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -308,8 +322,7 @@ void gen6_disable_rps_interrupts(struct drm_device *dev)
 
 	spin_lock_irq(&dev_priv->irq_lock);
 
-	I915_WRITE(GEN6_PMINTRMSK, INTEL_INFO(dev_priv)->gen >= 8 ?
-		   ~GEN8_PMINTR_REDIRECT_TO_NON_DISP : ~0);
+	I915_WRITE(GEN6_PMINTRMSK, gen6_sanitize_rps_pm_mask(dev_priv, ~0));
 
 	__gen6_disable_pm_irq(dev_priv, dev_priv->pm_rps_events);
 	I915_WRITE(gen6_pm_ier(dev_priv), I915_READ(gen6_pm_ier(dev_priv)) &
drivers/gpu/drm/i915/intel_drv.h +1 -0
index 588b618..bb871f3 100644
@@ -795,6 +795,7 @@ void gen6_disable_pm_irq(struct drm_i915_private *dev_priv, uint32_t mask);
 void gen6_reset_rps_interrupts(struct drm_device *dev);
 void gen6_enable_rps_interrupts(struct drm_device *dev);
 void gen6_disable_rps_interrupts(struct drm_device *dev);
+u32 gen6_sanitize_rps_pm_mask(struct drm_i915_private *dev_priv, u32 mask);
 void intel_runtime_pm_disable_interrupts(struct drm_i915_private *dev_priv);
 void intel_runtime_pm_enable_interrupts(struct drm_i915_private *dev_priv);
 static inline bool intel_irqs_enabled(struct drm_i915_private *dev_priv)
drivers/gpu/drm/i915/intel_pm.c +1 -10
index a3ebaa8..4bd1b8b 100644
@@ -3745,16 +3745,7 @@ static u32 gen6_rps_pm_mask(struct drm_i915_private *dev_priv, u8 val)
 	mask |= dev_priv->pm_rps_events & (GEN6_PM_RP_DOWN_EI_EXPIRED | GEN6_PM_RP_UP_EI_EXPIRED);
 	mask &= dev_priv->pm_rps_events;
 
-	/* IVB and SNB hard hangs on looping batchbuffer
-	 * if GEN6_PM_UP_EI_EXPIRED is masked.
-	 */
-	if (INTEL_INFO(dev_priv->dev)->gen <= 7 && !IS_HASWELL(dev_priv->dev))
-		mask |= GEN6_PM_RP_UP_EI_EXPIRED;
-
-	if (IS_GEN8(dev_priv->dev))
-		mask |= GEN8_PMINTR_REDIRECT_TO_NON_DISP;
-
-	return ~mask;
+	return gen6_sanitize_rps_pm_mask(dev_priv, ~mask);
 }
 
 /* gen6_set_rps is called to update the frequency request, but should also be
Signature · 6 lines
1.8.4

_______________________________________________
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
#2 lore raw
Quoted 18 lines · Daniel Vetter
Quoted 17 lines · Imre Deak

In

commit dbea3cea69508e9d548ed4a6be13de35492e5d15

Author: Imre Deak <[email protected]>
Date:   Mon Dec 15 18:59:28 2014 +0200

drm/i915: sanitize RPS resetting during GPU reset

we disable RPS interrupts during GPU resetting, but don't apply the necessary GEN6 HW workaround. This leads to a HW lockup during a subsequent "looping batchbuffer" workload. This is triggered by the testcase that submits exactly this kind of workload after a simulated GPU reset. I'm not sure how likely the bug would have triggered otherwise, since we would have applied the workaround anyway shortly after the GPU reset, when enabling GT powersaving from the deferred work.

Quoted 8 lines
Quoted 4 lines

This may also fix unrelated issues, since during driver loading / suspending we also disable RPS interrupts and so we also had a short window during the rest of the loading / resuming where a similar workload could run without the workaround applied.

Hm this might explain why the reset tests randomly hard-crashed snb every once in a while. So symptoms matches. Mika, can you please give this a spin?

307 loops of gem_reset_stats/ban-ctx-render on snb without issues.

-Mika

Quoted 89 lines
Quoted 5 lines

At the same time also keep the IRQ routing to display enabled at all times and everywhere on GEN8+, since there isn't ever a reason to route them to GT. For instance this way we could catch spurious interrupts on the CPU instead of allowing some undefined behaviour on the GT (due to the same spurious interrupt routed there).

I don't really grok that part, but it does feel like a separate patch (especially since you also change from IS_GEN8 to gen >= 8). Or maybe I'm just really confused.
-Daniel

Quoted 78 lines

Testcase: igt/gem_reset_stats/ban-ctx-render
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=87429

---
 drivers/gpu/drm/i915/i915_irq.c  | 17 +++++++++++++++--
 drivers/gpu/drm/i915/intel_drv.h |  1 +
 drivers/gpu/drm/i915/intel_pm.c  | 11 +----------
 3 files changed, 17 insertions(+), 12 deletions(-)
drivers/gpu/drm/i915/i915_irq.c +15 -2
index aa3180c..12acc40 100644
@@ -296,6 +296,20 @@ void gen6_enable_rps_interrupts(struct drm_device *dev)
 	spin_unlock_irq(&dev_priv->irq_lock);
 }
 
+u32 gen6_sanitize_rps_pm_mask(struct drm_i915_private *dev_priv, u32 mask)
+{
+	/* IVB and SNB hard hangs on looping batchbuffer
+	 * if GEN6_PM_UP_EI_EXPIRED is masked.
+	 */
+	if (INTEL_INFO(dev_priv->dev)->gen <= 7 && !IS_HASWELL(dev_priv->dev))
+		mask &= ~GEN6_PM_RP_UP_EI_EXPIRED;
+
+	if (INTEL_INFO(dev_priv->dev)->gen >= 8)
+		mask &= ~GEN8_PMINTR_REDIRECT_TO_NON_DISP;
+
+	return mask;
+}
+
 void gen6_disable_rps_interrupts(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
@@ -308,8 +322,7 @@ void gen6_disable_rps_interrupts(struct drm_device *dev)
 
 	spin_lock_irq(&dev_priv->irq_lock);
 
-	I915_WRITE(GEN6_PMINTRMSK, INTEL_INFO(dev_priv)->gen >= 8 ?
-		   ~GEN8_PMINTR_REDIRECT_TO_NON_DISP : ~0);
+	I915_WRITE(GEN6_PMINTRMSK, gen6_sanitize_rps_pm_mask(dev_priv, ~0));
 
 	__gen6_disable_pm_irq(dev_priv, dev_priv->pm_rps_events);
 	I915_WRITE(gen6_pm_ier(dev_priv), I915_READ(gen6_pm_ier(dev_priv)) &
drivers/gpu/drm/i915/intel_drv.h +1 -0
index 588b618..bb871f3 100644
@@ -795,6 +795,7 @@ void gen6_disable_pm_irq(struct drm_i915_private *dev_priv, uint32_t mask);
 void gen6_reset_rps_interrupts(struct drm_device *dev);
 void gen6_enable_rps_interrupts(struct drm_device *dev);
 void gen6_disable_rps_interrupts(struct drm_device *dev);
+u32 gen6_sanitize_rps_pm_mask(struct drm_i915_private *dev_priv, u32 mask);
 void intel_runtime_pm_disable_interrupts(struct drm_i915_private *dev_priv);
 void intel_runtime_pm_enable_interrupts(struct drm_i915_private *dev_priv);
 static inline bool intel_irqs_enabled(struct drm_i915_private *dev_priv)
drivers/gpu/drm/i915/intel_pm.c +1 -10
index a3ebaa8..4bd1b8b 100644
@@ -3745,16 +3745,7 @@ static u32 gen6_rps_pm_mask(struct drm_i915_private *dev_priv, u8 val)
 	mask |= dev_priv->pm_rps_events & (GEN6_PM_RP_DOWN_EI_EXPIRED | GEN6_PM_RP_UP_EI_EXPIRED);
 	mask &= dev_priv->pm_rps_events;
 
-	/* IVB and SNB hard hangs on looping batchbuffer
-	 * if GEN6_PM_UP_EI_EXPIRED is masked.
-	 */
-	if (INTEL_INFO(dev_priv->dev)->gen <= 7 && !IS_HASWELL(dev_priv->dev))
-		mask |= GEN6_PM_RP_UP_EI_EXPIRED;
-
-	if (IS_GEN8(dev_priv->dev))
-		mask |= GEN8_PMINTR_REDIRECT_TO_NON_DISP;
-
-	return ~mask;
+	return gen6_sanitize_rps_pm_mask(dev_priv, ~mask);
 }
 
Quoted 13 lines
Quoted 8 lines
drivers/gpu/drm/i915/intel_pm.c +0 -0
 /* gen6_set_rps is called to update the frequency request, but should also be
Signature · 6 lines
1.8.4

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 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