All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH] drm/i915: Clear scanline waits before disabling the pipe.
@ 2010-08-08 11:01 Chris Wilson
  2010-08-08 18:34 ` Eric Anholt
  0 siblings, 1 reply; 3+ messages in thread
From: Chris Wilson @ 2010-08-08 11:01 UTC (permalink / raw)
  To: intel-gfx

If we disable the pipe and the GPU is currently waiting on a scanline
WAIT_FOR_EVENT, the GPU will hang. Fortunately, there is a magic bit
which we can write on i915+ to break this wait before disabling the
pipe.

References:

  Bug 29252 - [Arrandale] Hung WAIT_FOR_EVENT when running rss-glx-skyrocket
  https://bugs.freedesktop.org/show_bug.cgi?id=29252

  Bug 28964 - [i965gm] GPU infinite MI_WAIT_FOR_EVENT while watching video in Totem
  https://bugs.freedesktop.org/show_bug.cgi?id=28964

and many others.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
---
 drivers/gpu/drm/i915/i915_reg.h      |    2 ++
 drivers/gpu/drm/i915/intel_display.c |   29 +++++++++++++++++++++++++++--
 2 files changed, 29 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index 97a35a4..2953b0d 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -288,6 +288,8 @@
 #define   RING_VALID_MASK	0x00000001
 #define   RING_VALID		0x00000001
 #define   RING_INVALID		0x00000000
+#define   RING_WAIT_I8XX	(1<<0) /* gen2, PRBx_HEAD */
+#define   RING_WAIT		(1<<11) /* gen3+, PRBx_CTL */
 #define PRB1_TAIL	0x02040 /* 915+ only */
 #define PRB1_HEAD	0x02044 /* 915+ only */
 #define PRB1_START	0x02048 /* 915+ only */
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index 2bb3196..696767c 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -2410,6 +2410,28 @@ static int i9xx_crtc_dpms(struct drm_crtc *crtc, int mode)
 	return 0;
 }
 
+/*
+ * When we disable a pipe, we need to clear any pending scanline wait events
+ * to avoid hanging the ring, which we assume we are waiting on.
+ */
+static void intel_clear_scanline_wait(struct drm_device *dev)
+{
+	struct drm_i915_private *dev_priv = dev->dev_private;
+
+	if (IS_GEN2(dev)) {
+		if (I915_READ(PRB0_HEAD) & RING_WAIT_I8XX) {
+			DRM_INFO("Forcing GPU idle to flush scanline wait");
+			i915_gpu_idle(dev);
+		}
+	} else {
+		u32 tmp = I915_READ(PRB0_CTL);
+		if (tmp & RING_WAIT) {
+			I915_WRITE(PRB0_CTL, tmp);
+			POSTING_READ(PRB0_CTL);
+		}
+	}
+}
+
 /**
  * Sets the power management mode of the pipe and plane.
  */
@@ -2429,12 +2451,15 @@ static int intel_crtc_dpms(struct drm_crtc *crtc, int mode)
 	 * with multiple pipes prior to enabling to new pipe.
 	 *
 	 * When switching off the display, make sure the cursor is
-	 * properly hidden prior to disabling the pipe.
+	 * properly hidden and there are no pending waits prior to
+	 * disabling the pipe.
 	 */
 	if (mode == DRM_MODE_DPMS_ON)
 		intel_update_watermarks(dev);
-	else
+	else {
 		intel_crtc_update_cursor(crtc);
+		intel_clear_scanline_wait(dev);
+	}
 
 	dev_priv->display.dpms(crtc, mode);
 
-- 
1.7.1

^ permalink raw reply related	[flat|nested] 3+ messages in thread

* Re: [PATCH] drm/i915: Clear scanline waits before disabling the pipe.
  2010-08-08 11:01 [PATCH] drm/i915: Clear scanline waits before disabling the pipe Chris Wilson
@ 2010-08-08 18:34 ` Eric Anholt
  2010-08-08 18:54   ` Chris Wilson
  0 siblings, 1 reply; 3+ messages in thread
From: Eric Anholt @ 2010-08-08 18:34 UTC (permalink / raw)
  To: Chris Wilson, intel-gfx


[-- Attachment #1.1: Type: text/plain, Size: 991 bytes --]

On Sun,  8 Aug 2010 12:01:38 +0100, Chris Wilson <chris@chris-wilson.co.uk> wrote:
> If we disable the pipe and the GPU is currently waiting on a scanline
> WAIT_FOR_EVENT, the GPU will hang. Fortunately, there is a magic bit
> which we can write on i915+ to break this wait before disabling the
> pipe.
> 
> References:
> 
>   Bug 29252 - [Arrandale] Hung WAIT_FOR_EVENT when running rss-glx-skyrocket
>   https://bugs.freedesktop.org/show_bug.cgi?id=29252
> 
>   Bug 28964 - [i965gm] GPU infinite MI_WAIT_FOR_EVENT while watching video in Totem
>   https://bugs.freedesktop.org/show_bug.cgi?id=28964
> 
> and many others.
> 
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Jesse Barnes <jbarnes@virtuousgeek.org>

At one point I thought we were taking the lock and idling the GPU before
a modeset, which would handle this problem, right?  This doesn't seem
reliable if we aren't, as we may just have not reached the MI_WAIT_FOR_EVENT
in the ring yet.

[-- Attachment #1.2: Type: application/pgp-signature, Size: 197 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH] drm/i915: Clear scanline waits before disabling the pipe.
  2010-08-08 18:34 ` Eric Anholt
@ 2010-08-08 18:54   ` Chris Wilson
  0 siblings, 0 replies; 3+ messages in thread
From: Chris Wilson @ 2010-08-08 18:54 UTC (permalink / raw)
  To: Eric Anholt, intel-gfx

On Sun, 08 Aug 2010 11:34:01 -0700, Eric Anholt <eric@anholt.net> wrote:
> On Sun,  8 Aug 2010 12:01:38 +0100, Chris Wilson <chris@chris-wilson.co.uk> wrote:
> > If we disable the pipe and the GPU is currently waiting on a scanline
> > WAIT_FOR_EVENT, the GPU will hang. Fortunately, there is a magic bit
> > which we can write on i915+ to break this wait before disabling the
> > pipe.
> 
> At one point I thought we were taking the lock and idling the GPU before
> a modeset, which would handle this problem, right?  This doesn't seem
> reliable if we aren't, as we may just have not reached the MI_WAIT_FOR_EVENT
> in the ring yet.

Hear, hear. I do think this is still racy, but Jesse thought the hardware
would do the right thing is could just break the wait and disable the
pipe before it had a chance to stall...

Actually, I'm not even convinced that the hardware does the right thing if
a wait for scanline is executed on a disabled pipe - my headless g45 with
a fake output is broken with page-flipping/video.

Hmm, might be a worthy addition to hangcheck though to just reset the wait.
-- 
Chris Wilson, Intel Open Source Technology Centre

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2010-08-08 18:55 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2010-08-08 11:01 [PATCH] drm/i915: Clear scanline waits before disabling the pipe Chris Wilson
2010-08-08 18:34 ` Eric Anholt
2010-08-08 18:54   ` Chris Wilson

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.