drm/i915: Use hangcheck score to find guilty context

With full ppgtt using acthd is not enough to find guilty
batch buffer. We get multiple false positives as acthd is
per vm.

Instead of scanning which vm was running on a ring,
to find corressponding context, use a different, simpler,
strategy of finding batches that caused gpu hang:

If hangcheck has declared ring to be hung, find first non complete
request on that ring and claim it was guilty.

v2: Rebase

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=73652
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net> (v1)
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

3 files changed, 32 insertions, 15 deletions

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 08331e105f8..37c2ea45f6d 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2332,9 +2332,10 @@ static bool i915_context_is_banned(struct drm_device *dev,
 
 static void i915_set_reset_status(struct intel_ring_buffer *ring,
 				  struct drm_i915_gem_request *request,
-				  u32 acthd)
+				  const bool guilty)
 {
-	bool inside, guilty;
+	const u32 acthd = intel_ring_get_active_head(ring);
+	bool inside;
 	unsigned long offset = 0;
 	struct i915_hw_context *ctx = request->ctx;
 	struct i915_ctx_hang_stats *hs;
@@ -2342,14 +2343,11 @@ static void i915_set_reset_status(struct intel_ring_buffer *ring,
 	if (WARN_ON(!ctx))
 		return;
 
-	/* Innocent until proven guilty */
-	guilty = false;
-
 	if (request->batch_obj)
 		offset = i915_gem_obj_offset(request->batch_obj,
 					     request_to_vm(request));
 
-	if (ring->hangcheck.action != HANGCHECK_WAIT &&
+	if (guilty &&
 	    i915_request_guilty(request, acthd, &inside)) {
 		DRM_DEBUG("%s hung %s bo (0x%lx ctx %d) at 0x%x\n",
 			  ring->name,
@@ -2357,8 +2355,6 @@ static void i915_set_reset_status(struct intel_ring_buffer *ring,
 			  offset,
 			  ctx->id,
 			  acthd);
-
-		guilty = true;
 	}
 
 	WARN_ON(!ctx->last_ring);
@@ -2385,19 +2381,39 @@ static void i915_gem_free_request(struct drm_i915_gem_request *request)
 	kfree(request);
 }
 
-static void i915_gem_reset_ring_status(struct drm_i915_private *dev_priv,
-				       struct intel_ring_buffer *ring)
+static struct drm_i915_gem_request *
+i915_gem_find_first_non_complete(struct intel_ring_buffer *ring)
 {
-	u32 completed_seqno = ring->get_seqno(ring, false);
-	u32 acthd = intel_ring_get_active_head(ring);
 	struct drm_i915_gem_request *request;
+	const u32 completed_seqno = ring->get_seqno(ring, false);
 
 	list_for_each_entry(request, &ring->request_list, list) {
 		if (i915_seqno_passed(completed_seqno, request->seqno))
 			continue;
 
-		i915_set_reset_status(ring, request, acthd);
+		return request;
 	}
+
+	return NULL;
+}
+
+static void i915_gem_reset_ring_status(struct drm_i915_private *dev_priv,
+				       struct intel_ring_buffer *ring)
+{
+	struct drm_i915_gem_request *request;
+	bool ring_hung;
+
+	request = i915_gem_find_first_non_complete(ring);
+
+	if (request == NULL)
+		return;
+
+	ring_hung = ring->hangcheck.score >= HANGCHECK_SCORE_RING_HUNG;
+
+	i915_set_reset_status(ring, request, ring_hung);
+
+	list_for_each_entry_continue(request, &ring->request_list, list)
+		i915_set_reset_status(ring, request, false);
 }
 
 static void i915_gem_reset_ring_cleanup(struct drm_i915_private *dev_priv,
diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
index b226ae67464..ec9eec4ce95 100644
--- a/drivers/gpu/drm/i915/i915_irq.c
+++ b/drivers/gpu/drm/i915/i915_irq.c
@@ -2532,7 +2532,6 @@ static void i915_hangcheck_elapsed(unsigned long data)
 #define BUSY 1
 #define KICK 5
 #define HUNG 20
-#define FIRE 30
 
 	if (!i915.enable_hangcheck)
 		return;
@@ -2616,7 +2615,7 @@ static void i915_hangcheck_elapsed(unsigned long data)
 	}
 
 	for_each_ring(ring, dev_priv, i) {
-		if (ring->hangcheck.score > FIRE) {
+		if (ring->hangcheck.score >= HANGCHECK_SCORE_RING_HUNG) {
 			DRM_INFO("%s on %s\n",
 				 stuck[i] ? "stuck" : "no progress",
 				 ring->name);
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
index 71a73f4fe25..38c757e136d 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.h
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -41,6 +41,8 @@ enum intel_ring_hangcheck_action {
 	HANGCHECK_HUNG,
 };
 
+#define HANGCHECK_SCORE_RING_HUNG 31
+
 struct intel_ring_hangcheck {
 	bool deadlock;
 	u32 seqno;