[Intel-gfx] Linux 3.16-rc2

[Thomas Meyer]

Am Montag, den 30.06.2014, 11:09 +0100 schrieb Chris Wilson:
> On Mon, Jun 30, 2014 at 12:02:20PM +0200, Pavel Machek wrote:
> > On Tue 2014-06-24 13:27:37, Chris Wilson wrote:
> > > On Tue, Jun 24, 2014 at 02:24:30PM +0200, Thomas Meyer wrote:
> > > > Am Dienstag, den 24.06.2014, 12:57 +0100 schrieb Chris Wilson:
> > > > > On Tue, Jun 24, 2014 at 02:06:24PM +0300, Jani Nikula wrote:
> > > > > > On Tue, 24 Jun 2014, Thomas Meyer <thomas at m3y3r.de> wrote:
> > > > > > > the i915 driver is still broken in 3.16-rc2. Resume from ram crashes the
> > > > > > > X server.
> > > > > > 
> > > > > > This is not new to 3.16-rc2; apparently we've had it since v3.15-rc4
> > > > > > [1]. Also related [2].
> > > > > > 
> > > > > > Chris, any fresh ideas?
> > > > > 
> > > > > Nope. The bug is https://bugs.freedesktop.org/show_bug.cgi?id=76554
> > > > > everything we know and have tried is there. Which is not much more than
> > > > > at time of the original incarnation:
> > > > > 
> > > > > commit 50aa253d820ad4577e2231202f2c8fd89f9dc4e6
> > > > > Author: Keith Packard <keithp at keithp.com>
> > > > > Date:   Tue Oct 14 17:20:35 2008 -0700
> > > > > 
> > > > >     i915: Fix up ring initialization to cover G45 oddities
> > > > >     
> > > > >     G45 appears quite sensitive to ring initialization register writes,
> > > > >     sometimes leaving the HEAD register with the START register contents. Check
> > > > >     to make sure HEAD is reset correctly when START is written, and fix it up,
> > > > >     screaming loudly.
> > > > > -Chris
> > > > > 
> > > > 
> > > > Hi,
> > > > 
> > > > so why not revert 78f2975eec9faff353a6194e854d3d39907bab68 (drm/i915:
> > > > Move all ring resets before setting the HWS page) ?
> > > 
> > > Because that patch was in response to a boot time regression.
> > 
> > It seems we are in a fairly ugly "fix one board, it breaks another" iterations, right?
> > 
> > (BTW, if you apply patch to fix this bug, could you Cc me? I have strange feeling
> > that it will break my setup... Actually, it probably makes sense to Cc all the people
> > who reported problems with ring initialization...
> > 
> > What patch caused the boot time regression you are talking about? We need to get 
> > list of commits involved in this, and revert the original one...
> 
> commit 9991ae787a0c87fe7c783b4b6f4754c3cdbb6213
> Author: Chris Wilson <chris at chris-wilson.co.uk>
> Date:   Wed Apr 2 16:36:07 2014 +0100
> 
>     drm/i915: Move all ring resets before setting the HWS page
>     
>     In commit a51435a3137ad8ae75c288c39bd2d8b2696bae8f
>     Author: Naresh Kumar Kachhi <naresh.kumar.kachhi at intel.com>
>     Date:   Wed Mar 12 16:39:40 2014 +0530
>     
>         drm/i915: disable rings before HW status page setup
>     
>     we reordered stopping the rings to do so before we set the HWS register.
>     However, there is an extra workaround for g45 to reset the rings twice,
>     and for consistency we should apply that workaround before setting the
>     HWS to be sure that the rings are truly stopped.
>     
>     Cc: Naresh Kumar Kachhi <naresh.kumar.kachhi at intel.com>
>     Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
>     Reviewed-by: Jesse Barnes <jbarnes at virtuousgeek.org>
>     Signed-off-by: Daniel Vetter <daniel.vetter at ffwll.ch>
> 
> 
> commit a51435a3137ad8ae75c288c39bd2d8b2696bae8f
> Author: Naresh Kumar Kachhi <naresh.kumar.kachhi at intel.com>
> Date:   Wed Mar 12 16:39:40 2014 +0530
> 
>     drm/i915: disable rings before HW status page setup
>     
>     Rings should be idle before issuing sync_flush
>     (in intel_ring_setup_status_page). This patch moves the ring
>     disabling before doing the HW status page setup.
>     
>     Signed-off-by: Naresh Kumar Kachhi <naresh.kumar.kachhi at intel.com>
>     Reviewed-by: Chris Wilson <chris at chris-wilson.co.uk>
>     Signed-off-by: Daniel Vetter <daniel.vetter at ffwll.ch>
> 
> 

Hi,

this patch on top of v3.16-rc3-62-gd92a333 makes the resume from ram
regression go away on my machine:

diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index 279488a..b896ac8 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -459,22 +459,25 @@ static bool stop_ring(struct intel_engine_cs *ring)
 {
 	struct drm_i915_private *dev_priv = to_i915(ring->dev);
 
-	if (!IS_GEN2(ring->dev)) {
-		I915_WRITE_MODE(ring, _MASKED_BIT_ENABLE(STOP_RING));
-		if (wait_for_atomic((I915_READ_MODE(ring) & MODE_IDLE) != 0, 1000)) {
-			DRM_ERROR("%s :timed out trying to stop ring\n", ring->name);
-			return false;
-		}
-	}
-
+//	if (!IS_GEN2(ring->dev)) {
+//		I915_WRITE_MODE(ring, _MASKED_BIT_ENABLE(STOP_RING));
+//		if (wait_for_atomic((I915_READ_MODE(ring) & MODE_IDLE) != 0, 1000)) {
+//			DRM_ERROR("%s :timed out trying to stop ring1\n", ring->name);
+//			return false;
+//		}
+//	}
+
+	/* Stop the ring if it's running. */
 	I915_WRITE_CTL(ring, 0);
 	I915_WRITE_HEAD(ring, 0);
 	ring->write_tail(ring, 0);
+	if (wait_for_atomic((I915_READ_MODE(ring) & MODE_IDLE) != 0, 1000))
+		DRM_ERROR("%s :timed out trying to stop ring2\n", ring->name);
 
-	if (!IS_GEN2(ring->dev)) {
-		(void)I915_READ_CTL(ring);
-		I915_WRITE_MODE(ring, _MASKED_BIT_DISABLE(STOP_RING));
-	}
+//	if (!IS_GEN2(ring->dev)) {
+//		(void)I915_READ_CTL(ring);
+//		I915_WRITE_MODE(ring, _MASKED_BIT_DISABLE(STOP_RING));
+//	}
 
 	return (I915_READ_HEAD(ring) & HEAD_ADDR) == 0;
 }

Chris, any ideas why explicitly stopping the ring before reset, results
in this kind of misbehaviour on my machine on resume from ram?

with kind regards
thomas