Skip to contentSkip to conversation

drm/i915/bdw: PIPE_[BC] I[ME]R moved to powerwell

PATCH In review Ben Widawsky 3 participants 6 messages Nov 8, 2013 – Nov 12, 2013

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

PATCH 1/1 drm/i915/bdw: PIPE_[BC] I[ME]R moved to powerwell

The pipe B and pipe C interrupt mask and enable registers are now part of the pipe, so disabling the pipe power wells will lost the contests of the registers.

Art totally debugged this one!

Cc Art Runyan <[email protected]>
Cc Paulo Zanoni <[email protected]>
---
 drivers/gpu/drm/i915/intel_pm.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)
drivers/gpu/drm/i915/intel_pm.c +12 -0
index 0a07d7c..d68cec4 100644
@@ -5701,6 +5701,18 @@ static void __intel_set_power_well(struct drm_device *dev, bool enable)
 				      HSW_PWR_WELL_STATE_ENABLED), 20))
 				DRM_ERROR("Timeout enabling power well\n");
 		}
+
+		if (IS_BROADWELL(dev)) {
+			I915_WRITE(GEN8_DE_PIPE_IMR(PIPE_B),
+				   dev_priv->de_irq_mask[PIPE_B]);
+			I915_WRITE(GEN8_DE_PIPE_IER(PIPE_B),
+				   ~dev_priv->de_irq_mask[PIPE_B] | GEN8_PIPE_VBLANK);
+			I915_WRITE(GEN8_DE_PIPE_IMR(PIPE_C),
+				   dev_priv->de_irq_mask[PIPE_C]);
+			I915_WRITE(GEN8_DE_PIPE_IER(PIPE_C),
+				   ~dev_priv->de_irq_mask[PIPE_C] | GEN8_PIPE_VBLANK);
+			POSTING_READ(GEN8_DE_PIPE_IER(PIPE_C));
+		}
 	} else {
 		if (enable_requested) {
 			unsigned long irqflags;
Signature · 1 line
1.8.4.2
#1 lore raw
Quoted 22 lines · Ben Widawsky

The pipe B and pipe C interrupt mask and enable registers are now part of the pipe, so disabling the pipe power wells will lost the contests of the registers.

Art totally debugged this one!

Cc Art Runyan <[email protected]>
Cc Paulo Zanoni <[email protected]>
---
 drivers/gpu/drm/i915/intel_pm.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)
drivers/gpu/drm/i915/intel_pm.c +1 -0
index 0a07d7c..d68cec4 100644
@@ -5701,6 +5701,18 @@ static void __intel_set_power_well(struct drm_device *dev, bool enable)
 				      HSW_PWR_WELL_STATE_ENABLED), 20))
 				DRM_ERROR("Timeout enabling power well\n");
 		}
+

drivers/gpu/drm/i915/intel_pm.c static void __intel_set_power_well(struct drm_device *dev, bool enable) · line 5714

Quoted 10 lines
drivers/gpu/drm/i915/intel_pm.c +10 -0
+		if (IS_BROADWELL(dev)) {
+			I915_WRITE(GEN8_DE_PIPE_IMR(PIPE_B),
+				   dev_priv->de_irq_mask[PIPE_B]);
+			I915_WRITE(GEN8_DE_PIPE_IER(PIPE_B),
+				   ~dev_priv->de_irq_mask[PIPE_B] | GEN8_PIPE_VBLANK);
+			I915_WRITE(GEN8_DE_PIPE_IMR(PIPE_C),
+				   dev_priv->de_irq_mask[PIPE_C]);
+			I915_WRITE(GEN8_DE_PIPE_IER(PIPE_C),
+				   ~dev_priv->de_irq_mask[PIPE_C] | GEN8_PIPE_VBLANK);
+			POSTING_READ(GEN8_DE_PIPE_IER(PIPE_C));

This needs to be protected by the dev_priv->irq_lock in case we see concurrent changes from elseplace. It shouldn't be possible since if a pipe is off we shouldn't ever ask for vblank or pageflip interrupts. But given the mess that is drm_irq.c I wouldn't trust that.

Hm, actually if we'd shovel this into ->crtc_enable hook before we update crtc->active we'd see this guarantee a bit more clearly. But imo ok to do here, too.
-Daniel

Quoted 11 lines
drivers/gpu/drm/i915/intel_pm.c +1 -0
+		}
 	} else {
 		if (enable_requested) {
 			unsigned long irqflags;
Signature · 6 lines
1.8.4.2

_______________________________________________
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
#2 lore raw
Quoted 30 lines · Daniel Vetter
Quoted 29 lines · Ben Widawsky

The pipe B and pipe C interrupt mask and enable registers are now part of the pipe, so disabling the pipe power wells will lost the contests of the registers.

Art totally debugged this one!

Cc Art Runyan <[email protected]>
Cc Paulo Zanoni <[email protected]>
---
 drivers/gpu/drm/i915/intel_pm.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)
drivers/gpu/drm/i915/intel_pm.c +8 -0
index 0a07d7c..d68cec4 100644
@@ -5701,6 +5701,18 @@ static void __intel_set_power_well(struct drm_device *dev, bool enable)
 				      HSW_PWR_WELL_STATE_ENABLED), 20))
 				DRM_ERROR("Timeout enabling power well\n");
 		}
+
+		if (IS_BROADWELL(dev)) {
+			I915_WRITE(GEN8_DE_PIPE_IMR(PIPE_B),
+				   dev_priv->de_irq_mask[PIPE_B]);
+			I915_WRITE(GEN8_DE_PIPE_IER(PIPE_B),
+				   ~dev_priv->de_irq_mask[PIPE_B] | GEN8_PIPE_VBLANK);
+			I915_WRITE(GEN8_DE_PIPE_IMR(PIPE_C),
+				   dev_priv->de_irq_mask[PIPE_C]);
Quoted 9 lines

drivers/gpu/drm/i915/intel_pm.c static void __intel_set_power_well(struct drm_device *dev, bool enable) · line 5714

Quoted 3 lines
drivers/gpu/drm/i915/intel_pm.c +3 -0
+			I915_WRITE(GEN8_DE_PIPE_IER(PIPE_C),
+				   ~dev_priv->de_irq_mask[PIPE_C] | GEN8_PIPE_VBLANK);
+			POSTING_READ(GEN8_DE_PIPE_IER(PIPE_C));

This needs to be protected by the dev_priv->irq_lock in case we see concurrent changes from elseplace. It shouldn't be possible since if a pipe is off we shouldn't ever ask for vblank or pageflip interrupts. But given the mess that is drm_irq.c I wouldn't trust that.

I have no issue resending with the lock - assuming you've already thought about lock ordering rules, and it will just work. For posterity, the concern you have is impossible to hit, not, "shouldn't be posible." I agree that adding an uncontested lock for, at present, code clarity, and who knows what in the future, is a good idea.

Quoted 3 lines

Hm, actually if we'd shovel this into ->crtc_enable hook before we update crtc->active we'd see this guarantee a bit more clearly. But imo ok to do here, too.

I'll wait for Paulo's input for this. To me it seems to make the most sense in its current place, unless you want to get in the business of always enabling/disabling around crtc enable/disable. I also don't know this part of the code, or HW well enough to assert we'll never want some interrupt before we enable the crtc.

In either case, if Paulo doesn't feel really strongly that it should be moved, I'll simply resubmit with the lock addition.

Quoted 18 lines

-Daniel

Quoted 11 lines
drivers/gpu/drm/i915/intel_pm.c +1 -0
+		}
 	} else {
 		if (enable_requested) {
 			unsigned long irqflags;
Signature · 6 lines
1.8.4.2

_______________________________________________
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
Signature · 1 line
Ben Widawsky, Intel Open Source Technology Center
#3 lore raw
Quoted 41 lines · Ben Widawsky
Quoted 30 lines · Daniel Vetter
Quoted 29 lines · Ben Widawsky

The pipe B and pipe C interrupt mask and enable registers are now part of the pipe, so disabling the pipe power wells will lost the contests of the registers.

Art totally debugged this one!

Cc Art Runyan <[email protected]>
Cc Paulo Zanoni <[email protected]>
---
 drivers/gpu/drm/i915/intel_pm.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)
drivers/gpu/drm/i915/intel_pm.c +8 -0
index 0a07d7c..d68cec4 100644
@@ -5701,6 +5701,18 @@ static void __intel_set_power_well(struct drm_device *dev, bool enable)
                                  HSW_PWR_WELL_STATE_ENABLED), 20))
                            DRM_ERROR("Timeout enabling power well\n");
            }
+
+           if (IS_BROADWELL(dev)) {
+                   I915_WRITE(GEN8_DE_PIPE_IMR(PIPE_B),
+                              dev_priv->de_irq_mask[PIPE_B]);
+                   I915_WRITE(GEN8_DE_PIPE_IER(PIPE_B),
+                              ~dev_priv->de_irq_mask[PIPE_B] | GEN8_PIPE_VBLANK);
+                   I915_WRITE(GEN8_DE_PIPE_IMR(PIPE_C),
+                              dev_priv->de_irq_mask[PIPE_C]);
Quoted 9 lines

drivers/gpu/drm/i915/intel_pm.c static void __intel_set_power_well(struct drm_device *dev, bool enable) · line 5714

Quoted 3 lines
drivers/gpu/drm/i915/intel_pm.c +3 -0
+                   I915_WRITE(GEN8_DE_PIPE_IER(PIPE_C),
+                              ~dev_priv->de_irq_mask[PIPE_C] | GEN8_PIPE_VBLANK);
+                   POSTING_READ(GEN8_DE_PIPE_IER(PIPE_C));

This needs to be protected by the dev_priv->irq_lock in case we see concurrent changes from elseplace. It shouldn't be possible since if a pipe is off we shouldn't ever ask for vblank or pageflip interrupts. But given the mess that is drm_irq.c I wouldn't trust that.

Quoted 5 lines

I have no issue resending with the lock - assuming you've already thought about lock ordering rules, and it will just work. For posterity, the concern you have is impossible to hit, not, "shouldn't be posible." I agree that adding an uncontested lock for, at present, code clarity, and who knows what in the future, is a good idea.

Yeah, it's fairly impossible. But if we refactor things a bit for bdw to have helpers handling the pipe imr stuff we'll probably need it for consistency. I think add a little comment (like we have in all the postinstall hooks when grabbing the ->irq_lock) together with the spinlock would be nice. And I'm still not convinced that a racing vblank ioctl couldn't wreak havoc with the current locking from ums days we have in drm_irq.c.

Quoted 4 lines
Quoted 3 lines

Hm, actually if we'd shovel this into ->crtc_enable hook before we update crtc->active we'd see this guarantee a bit more clearly. But imo ok to do here, too.

Quoted 8 lines

I'll wait for Paulo's input for this. To me it seems to make the most sense in its current place, unless you want to get in the business of always enabling/disabling around crtc enable/disable. I also don't know this part of the code, or HW well enough to assert we'll never want some interrupt before we enable the crtc.

In either case, if Paulo doesn't feel really strongly that it should be moved, I'll simply resubmit with the lock addition.

Like I've said it's ok here to, so just let it be. I guess we'll see in 1-2 platform generations what looks best and whether this theme continues or not. So better to bikeshed it later.

Cheers,
Daniel

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

PATCH 1/1 drm/i915/bdw: PIPE_[BC] I[ME]R moved to powerwell

The pipe B and pipe C interrupt mask and enable registers are now part of the pipe, so disabling the pipe power wells will lost the contests of the registers.

Art totally debugged this one!

v2: Use the irq_lock to clarify code, and prevent future bugs (Daniel)

Cc Art Runyan <[email protected]>
Cc Paulo Zanoni <[email protected]>
---
 drivers/gpu/drm/i915/intel_pm.c | 16 +++++++++++++++-
 1 file changed, 15 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_pm.c +15 -1
index 0a07d7c..38915dc 100644
@@ -5684,6 +5684,7 @@ static void __intel_set_power_well(struct drm_device *dev, bool enable)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	bool is_enabled, enable_requested;
+	unsigned long irqflags;
 	uint32_t tmp;
 
 	tmp = I915_READ(HSW_PWR_WELL_DRIVER);
@@ -5701,9 +5702,22 @@ static void __intel_set_power_well(struct drm_device *dev, bool enable)
 				      HSW_PWR_WELL_STATE_ENABLED), 20))
 				DRM_ERROR("Timeout enabling power well\n");
 		}
+
+		if (IS_BROADWELL(dev)) {
+			spin_lock_irqsave(&dev_priv->irq_lock, irqflags);
+			I915_WRITE(GEN8_DE_PIPE_IMR(PIPE_B),
+				   dev_priv->de_irq_mask[PIPE_B]);
+			I915_WRITE(GEN8_DE_PIPE_IER(PIPE_B),
+				   ~dev_priv->de_irq_mask[PIPE_B] | GEN8_PIPE_VBLANK);
+			I915_WRITE(GEN8_DE_PIPE_IMR(PIPE_C),
+				   dev_priv->de_irq_mask[PIPE_C]);
+			I915_WRITE(GEN8_DE_PIPE_IER(PIPE_C),
+				   ~dev_priv->de_irq_mask[PIPE_C] | GEN8_PIPE_VBLANK);
+			POSTING_READ(GEN8_DE_PIPE_IER(PIPE_C));
+			spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags);
+		}
 	} else {
 		if (enable_requested) {
-			unsigned long irqflags;
 			enum pipe p;
 
 			I915_WRITE(HSW_PWR_WELL_DRIVER, 0);
Signature · 1 line
1.8.4.2
#5 lore raw
Quoted 11 lines · Ben Widawsky

The pipe B and pipe C interrupt mask and enable registers are now part of the pipe, so disabling the pipe power wells will lost the contests of the registers.

Art totally debugged this one!

v2: Use the irq_lock to clarify code, and prevent future bugs (Daniel)

Cc Art Runyan <[email protected]>
Cc Paulo Zanoni <[email protected]>

Merged to bdw-fixes, thanks.
-Daniel

Quoted 47 lines
---
 drivers/gpu/drm/i915/intel_pm.c | 16 +++++++++++++++-
 1 file changed, 15 insertions(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_pm.c +15 -1
index 0a07d7c..38915dc 100644
@@ -5684,6 +5684,7 @@ static void __intel_set_power_well(struct drm_device *dev, bool enable)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	bool is_enabled, enable_requested;
+	unsigned long irqflags;
 	uint32_t tmp;
 
 	tmp = I915_READ(HSW_PWR_WELL_DRIVER);
@@ -5701,9 +5702,22 @@ static void __intel_set_power_well(struct drm_device *dev, bool enable)
 				      HSW_PWR_WELL_STATE_ENABLED), 20))
 				DRM_ERROR("Timeout enabling power well\n");
 		}
+
+		if (IS_BROADWELL(dev)) {
+			spin_lock_irqsave(&dev_priv->irq_lock, irqflags);
+			I915_WRITE(GEN8_DE_PIPE_IMR(PIPE_B),
+				   dev_priv->de_irq_mask[PIPE_B]);
+			I915_WRITE(GEN8_DE_PIPE_IER(PIPE_B),
+				   ~dev_priv->de_irq_mask[PIPE_B] | GEN8_PIPE_VBLANK);
+			I915_WRITE(GEN8_DE_PIPE_IMR(PIPE_C),
+				   dev_priv->de_irq_mask[PIPE_C]);
+			I915_WRITE(GEN8_DE_PIPE_IER(PIPE_C),
+				   ~dev_priv->de_irq_mask[PIPE_C] | GEN8_PIPE_VBLANK);
+			POSTING_READ(GEN8_DE_PIPE_IER(PIPE_C));
+			spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags);
+		}
 	} else {
 		if (enable_requested) {
-			unsigned long irqflags;
 			enum pipe p;
 
 			I915_WRITE(HSW_PWR_WELL_DRIVER, 0);
Signature · 6 lines
1.8.4.2

_______________________________________________
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