[v2,06/16] drm/i915: track ring progression using seqnos

Commit Message

Mika Kuoppala March 14, 2013, 3:52 p.m. UTC

Instead of relying in acthd, track ring seqno progression
to detect if ring has hung.

Signed-off-by: Mika Kuoppala <[e-mail]>
---
 drivers/gpu/drm/i915/i915_drv.h         |    2 --
 drivers/gpu/drm/i915/i915_irq.c         |   30 +++++++++++++-----------------
 drivers/gpu/drm/i915/intel_ringbuffer.h |    2 ++
 3 files changed, 15 insertions(+), 19 deletions(-)

Comments

Chris Wilson March 15, 2013, 9:48 a.m. UTC | #1

On Thu, Mar 14, 2013 at 05:52:07PM +0200, Mika Kuoppala wrote:
> index d66208c..9599c56 100644
> --- a/drivers/gpu/drm/i915/intel_ringbuffer.h
> +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
> @@ -137,6 +137,8 @@ struct  intel_ring_buffer {
>  	struct i915_hw_context *default_context;
>  	struct drm_i915_gem_object *last_context_obj;
>  
> +	u32 hangcheck_seqno;

We end up with quite a few hangheck_* variables. An indication that we
should sort those out into their own struct.
-Chris

Patch

diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 79cd27f..534e673 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -797,8 +797,6 @@  struct i915_gpu_error {
 #define DRM_I915_HANGCHECK_JIFFIES msecs_to_jiffies(DRM_I915_HANGCHECK_PERIOD)
 	struct timer_list hangcheck_timer;
 	int hangcheck_count;
-	uint32_t last_acthd[I915_NUM_RINGS];
-	uint32_t prev_instdone[I915_NUM_INSTDONE_REG];
 
 	/* For reset and error_state handling. */
 	spinlock_t lock;
diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
index ed8ac1f..6594c97 100644
--- a/drivers/gpu/drm/i915/i915_irq.c
+++ b/drivers/gpu/drm/i915/i915_irq.c
@@ -1880,22 +1880,19 @@  void i915_hangcheck_elapsed(unsigned long data)
 {
 	struct drm_device *dev = (struct drm_device *)data;
 	drm_i915_private_t *dev_priv = dev->dev_private;
-	uint32_t acthd[I915_NUM_RINGS], instdone[I915_NUM_INSTDONE_REG];
 	struct intel_ring_buffer *ring;
 	bool err = false, idle;
 	int i;
+	u32 seqno[I915_NUM_RINGS];
+	bool work_done;
 
 	if (!i915_enable_hangcheck)
 		return;
 
-	memset(acthd, 0, sizeof(acthd));
 	idle = true;
 	for_each_ring(ring, dev_priv, i) {
-		u32 seqno;
-
-		seqno = ring->get_seqno(ring, false);
-		idle &= i915_hangcheck_ring_idle(ring, seqno, &err);
-	    acthd[i] = intel_ring_get_active_head(ring);
+		seqno[i] = ring->get_seqno(ring, false);
+		idle &= i915_hangcheck_ring_idle(ring, seqno[i], &err);
 	}
 
 	/* If all work is done then ACTHD clearly hasn't advanced. */
@@ -1911,20 +1908,19 @@  void i915_hangcheck_elapsed(unsigned long data)
 		return;
 	}
 
-	i915_get_extra_instdone(dev, instdone);
-	if (memcmp(dev_priv->gpu_error.last_acthd, acthd,
-		   sizeof(acthd)) == 0 &&
-	    memcmp(dev_priv->gpu_error.prev_instdone, instdone,
-		   sizeof(instdone)) == 0) {
+	work_done = false;
+	for_each_ring(ring, dev_priv, i) {
+		if (ring->hangcheck_seqno != seqno[i]) {
+			work_done = true;
+			ring->hangcheck_seqno = seqno[i];
+		}
+	}
+
+	if (!work_done) {
 		if (i915_hangcheck_hung(dev))
 			return;
 	} else {
 		dev_priv->gpu_error.hangcheck_count = 0;
-
-		memcpy(dev_priv->gpu_error.last_acthd, acthd,
-		       sizeof(acthd));
-		memcpy(dev_priv->gpu_error.prev_instdone, instdone,
-		       sizeof(instdone));
 	}
 
 repeat:
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
index d66208c..9599c56 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.h
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -137,6 +137,8 @@  struct  intel_ring_buffer {
 	struct i915_hw_context *default_context;
 	struct drm_i915_gem_object *last_context_obj;
 
+	u32 hangcheck_seqno;
+
 	void *private;
 };