Skip to contentSkip to conversation

drm/i915: Snapshot seqno of most recently submitted request.

PATCH In review Tomas Elf 3 participants 3 messages Jul 9, 2015

Subsystems (2)
lore
Share & tools
lore

Tomas Elf

author
#0 lore raw

PATCH v2 1/1 drm/i915: Snapshot seqno of most recently submitted request.

The hang checker needs to inspect whether or not the ring request list is empty as well as if the given engine has reached or passed the most recently submitted request. The problem with this is that the hang checker cannot grab the struct_mutex, which is required in order to safely inspect requests since requests might be deallocated during inspection. In the past we've had kernel panics due to this very unsynchronized access in the hang checker.

One solution to this problem is to not inspect the requests directly since we're only interested in the seqno of the most recently submitted request - not the request itself. Instead the seqno of the most recently submitted request is stored separately, which the hang checker then inspects, circumventing the issue of synchronization from the hang checker entirely.

v2 (Chris Wilson):
- Pass current engine seqno to ring_idle() from i915_hangcheck_elapsed() rather than compute it over again.
- Remove extra whitespace.

Issue: VIZ-5998

---
 drivers/gpu/drm/i915/i915_gem.c         |  1 +
 drivers/gpu/drm/i915/i915_irq.c         | 13 +++----------
 drivers/gpu/drm/i915/intel_ringbuffer.h |  7 +++++++
 3 files changed, 11 insertions(+), 10 deletions(-)
drivers/gpu/drm/i915/i915_gem.c +1 -0
index a0bff41..a871be4 100644
@@ -2549,6 +2549,7 @@ void __i915_add_request(struct drm_i915_gem_request *request,
 	request->batch_obj = obj;
 
 	request->emitted_jiffies = jiffies;
+	ring->last_submitted_seqno = request->seqno;
 	list_add_tail(&request->list, &ring->request_list);
 
 	trace_i915_gem_request_add(request);
drivers/gpu/drm/i915/i915_irq.c +3 -10
index a6fbe64..0d0741f 100644
@@ -2484,18 +2484,11 @@ static void gen8_disable_vblank(struct drm_device *dev, int pipe)
 	spin_unlock_irqrestore(&dev_priv->irq_lock, irqflags);
 }
 
-static struct drm_i915_gem_request *
-ring_last_request(struct intel_engine_cs *ring)
-{
-	return list_entry(ring->request_list.prev,
-			  struct drm_i915_gem_request, list);
-}
-
 static bool
-ring_idle(struct intel_engine_cs *ring)
+ring_idle(struct intel_engine_cs *ring, u32 seqno)
 {
 	return (list_empty(&ring->request_list) ||
-		i915_gem_request_completed(ring_last_request(ring), false));
+		i915_seqno_passed(seqno, ring->last_submitted_seqno));
 }
 
 static bool
@@ -2717,7 +2710,7 @@ static void i915_hangcheck_elapsed(struct work_struct *work)
 		acthd = intel_ring_get_active_head(ring);
 
 		if (ring->hangcheck.seqno == seqno) {
-			if (ring_idle(ring)) {
+			if (ring_idle(ring, seqno)) {
 				ring->hangcheck.action = HANGCHECK_IDLE;
 
 				if (waitqueue_active(&ring->irq_queue)) {
drivers/gpu/drm/i915/intel_ringbuffer.h +7 -0
index 0ea89ea..2e85fda 100644
@@ -292,6 +292,13 @@ struct  intel_engine_cs {
 	 */
 	struct list_head request_list;
 
+	/**
+	 * Seqno of request most recently submitted to request_list.
+	 * Used exclusively by hang checker to avoid grabbing lock while
+	 * inspecting request list.
+	 */
+	u32 last_submitted_seqno;
+
 	bool gpu_caches_dirty;
 
 	wait_queue_head_t irq_queue;
Signature · 6 lines
1.9.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Reviewed-by on v2 1/1
#1 lore raw
Quoted 13 lines · Tomas Elf

The hang checker needs to inspect whether or not the ring request list is empty as well as if the given engine has reached or passed the most recently submitted request. The problem with this is that the hang checker cannot grab the struct_mutex, which is required in order to safely inspect requests since requests might be deallocated during inspection. In the past we've had kernel panics due to this very unsynchronized access in the hang checker.

One solution to this problem is to not inspect the requests directly since we're only interested in the seqno of the most recently submitted request - not the request itself. Instead the seqno of the most recently submitted request is stored separately, which the hang checker then inspects, circumventing the issue of synchronization from the hang checker entirely.

Quoted 7 lines

v2 (Chris Wilson):
- Pass current engine seqno to ring_idle() from i915_hangcheck_elapsed() rather than compute it over again.
- Remove extra whitespace.

Issue: VIZ-5998

Yup, that is a nice simple fix by partially reverting the s/seqno/requests/ change (and improving upon it in the process).

We should mention

Fixes commit 44cdd6d219bc64f6810b8ed0023a4d4db9e0fe68

Author: John Harrison <[email protected]>
Date:   Mon Nov 24 18:49:40 2014 +0000

drm/i915: Convert 'ring_idle()' to use requests not seqnos

and

-Chris

Signature · 5 lines
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#2 lore raw
Quoted 28 lines · Chris Wilson
Quoted 13 lines · Tomas Elf

The hang checker needs to inspect whether or not the ring request list is empty as well as if the given engine has reached or passed the most recently submitted request. The problem with this is that the hang checker cannot grab the struct_mutex, which is required in order to safely inspect requests since requests might be deallocated during inspection. In the past we've had kernel panics due to this very unsynchronized access in the hang checker.

One solution to this problem is to not inspect the requests directly since we're only interested in the seqno of the most recently submitted request - not the request itself. Instead the seqno of the most recently submitted request is stored separately, which the hang checker then inspects, circumventing the issue of synchronization from the hang checker entirely.

Quoted 7 lines

v2 (Chris Wilson):
- Pass current engine seqno to ring_idle() from i915_hangcheck_elapsed() rather than compute it over again.
- Remove extra whitespace.

Issue: VIZ-5998

Yup, that is a nice simple fix by partially reverting the s/seqno/requests/ change (and improving upon it in the process).

We should mention

Fixes commit 44cdd6d219bc64f6810b8ed0023a4d4db9e0fe68

Quoted 8 lines
Author: John Harrison <[email protected]>
Date:   Mon Nov 24 18:49:40 2014 +0000

drm/i915: Convert 'ring_idle()' to use requests not seqnos

and

Added and applied to -fixes, thanks.
-Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx