Skip to contentSkip to conversation

drm/i915: Fix execlist cleanup bug

PATCH In review Dave Gordon 5 participants 8 messages Oct 30, 2014 – Nov 3, 2014

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 drm/i915: Fix execlist cleanup bug

From: John Harrison <[email protected]>

Check whether each engine exists before trying to clean up the corresponding logical ring.

Change-Id I31b1ed941824db2d6bd7233360dbce05671979a8
---
 drivers/gpu/drm/i915/intel_lrc.c |   12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)
drivers/gpu/drm/i915/intel_lrc.c +8 -4
index cd74e5c..7a7d30a 100644
@@ -1444,13 +1444,17 @@ int intel_logical_rings_init(struct drm_device *dev)
 	return 0;
 
 cleanup_bsd2_ring:
-	intel_logical_ring_cleanup(&dev_priv->ring[VCS2]);
+	if (HAS_BSD2(dev))
+		intel_logical_ring_cleanup(&dev_priv->ring[VCS2]);
 cleanup_vebox_ring:
-	intel_logical_ring_cleanup(&dev_priv->ring[VECS]);
+	if (HAS_VEBOX(dev))
+		intel_logical_ring_cleanup(&dev_priv->ring[VECS]);
 cleanup_blt_ring:
-	intel_logical_ring_cleanup(&dev_priv->ring[BCS]);
+	if (HAS_BLT(dev))
+		intel_logical_ring_cleanup(&dev_priv->ring[BCS]);
 cleanup_bsd_ring:
-	intel_logical_ring_cleanup(&dev_priv->ring[VCS]);
+	if (HAS_BSD(dev))
+		intel_logical_ring_cleanup(&dev_priv->ring[VCS]);
 cleanup_render_ring:
 	intel_logical_ring_cleanup(&dev_priv->ring[RCS]);
 
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#1 lore raw

This is an old version. I have a new and improved patch that also fixes the legacy ring buffer case (which was equally broken). Will post shortly...

Quoted 27 lines · Dave Gordon

From: John Harrison <[email protected]>

Check whether each engine exists before trying to clean up the corresponding logical ring.

Change-Id I31b1ed941824db2d6bd7233360dbce05671979a8
---
  drivers/gpu/drm/i915/intel_lrc.c |   12 ++++++++----
  1 file changed, 8 insertions(+), 4 deletions(-)
drivers/gpu/drm/i915/intel_lrc.c +4 -2
index cd74e5c..7a7d30a 100644
@@ -1444,13 +1444,17 @@ int intel_logical_rings_init(struct drm_device *dev)
  	return 0;
 
  cleanup_bsd2_ring:
-	intel_logical_ring_cleanup(&dev_priv->ring[VCS2]);
+	if (HAS_BSD2(dev))
+		intel_logical_ring_cleanup(&dev_priv->ring[VCS2]);
  cleanup_vebox_ring:
-	intel_logical_ring_cleanup(&dev_priv->ring[VECS]);
+	if (HAS_VEBOX(dev))
+		intel_logical_ring_cleanup(&dev_priv->ring[VECS]);
  cleanup_blt_ring:

drivers/gpu/drm/i915/intel_lrc.c int intel_logical_rings_init(struct drm_device *dev) · line 1460

Quoted 10 lines
drivers/gpu/drm/i915/intel_lrc.c +4 -2
-	intel_logical_ring_cleanup(&dev_priv->ring[BCS]);
+	if (HAS_BLT(dev))
+		intel_logical_ring_cleanup(&dev_priv->ring[BCS]);
  cleanup_bsd_ring:
-	intel_logical_ring_cleanup(&dev_priv->ring[VCS]);
+	if (HAS_BSD(dev))
+		intel_logical_ring_cleanup(&dev_priv->ring[VCS]);
  cleanup_render_ring:
  	intel_logical_ring_cleanup(&dev_priv->ring[RCS]);
 

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#2 lore raw

PATCH 1/1 drm/i915: Fix null pointer dereference in ring cleanup code

From: John Harrison <[email protected]>

If a ring failed to initialise for any reason then the error path would try to clean up all rings including those that had not yet been allocated. The ring clean up code did a check that the ring was valid before starting its work. Unfortunately, that was after it had already dereferenced the ring to obtain a dev_private pointer.

---
 drivers/gpu/drm/i915/intel_lrc.c        |    4 +++-
 drivers/gpu/drm/i915/intel_ringbuffer.c |    7 +++++--
 2 files changed, 8 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/intel_lrc.c +3 -1
index cd74e5c..76776fa 100644
@@ -1214,11 +1214,13 @@ static int gen8_emit_request(struct intel_ringbuffer *ringbuf)
  */
 void intel_logical_ring_cleanup(struct intel_engine_cs *ring)
 {
-	struct drm_i915_private *dev_priv = ring->dev->dev_private;
+	struct drm_i915_private *dev_priv;
 
 	if (!intel_ring_initialized(ring))
 		return;
 
+	dev_priv = ring->dev->dev_private;
+
 	intel_logical_ring_stop(ring);
 	WARN_ON((I915_READ_MODE(ring) & MODE_IDLE) == 0);
 	ring->preallocated_lazy_request = NULL;
drivers/gpu/drm/i915/intel_ringbuffer.c +5 -2
index a8f72e8..f457146 100644
@@ -1845,12 +1845,15 @@ error:
 
 void intel_cleanup_ring_buffer(struct intel_engine_cs *ring)
 {
-	struct drm_i915_private *dev_priv = to_i915(ring->dev);
-	struct intel_ringbuffer *ringbuf = ring->buffer;
+	struct drm_i915_private *dev_priv;
+	struct intel_ringbuffer *ringbuf;
 
 	if (!intel_ring_initialized(ring))
 		return;
 
+	dev_priv = to_i915(ring->dev);
+	ringbuf = ring->buffer;
+
 	intel_stop_ring_buffer(ring);
 	WARN_ON(!IS_GEN2(ring->dev) && (I915_READ_MODE(ring) & MODE_IDLE) == 0);
 
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Reviewed-by on v1 1/1
#3 lore raw

Re: [PATCH] drm/i915: Fix null pointer dereference in ring cleanup code

Quoted 9 lines · On Fri, Oct 31, 2014 at 12:00:26PM +0000, [email protected] wrote:

From: John Harrison <[email protected]>

If a ring failed to initialise for any reason then the error path would try to clean up all rings including those that had not yet been allocated. The ring clean up code did a check that the ring was valid before starting its work. Unfortunately, that was after it had already dereferenced the ring to obtain a dev_private pointer.

This looks good to me.

Signature · 1 line
Damien
Quoted 53 lines
---
 drivers/gpu/drm/i915/intel_lrc.c        |    4 +++-
 drivers/gpu/drm/i915/intel_ringbuffer.c |    7 +++++--
 2 files changed, 8 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/intel_lrc.c +3 -1
index cd74e5c..76776fa 100644
@@ -1214,11 +1214,13 @@ static int gen8_emit_request(struct intel_ringbuffer *ringbuf)
  */
 void intel_logical_ring_cleanup(struct intel_engine_cs *ring)
 {
-	struct drm_i915_private *dev_priv = ring->dev->dev_private;
+	struct drm_i915_private *dev_priv;
 
 	if (!intel_ring_initialized(ring))
 		return;
 
+	dev_priv = ring->dev->dev_private;
+
 	intel_logical_ring_stop(ring);
 	WARN_ON((I915_READ_MODE(ring) & MODE_IDLE) == 0);
 	ring->preallocated_lazy_request = NULL;
drivers/gpu/drm/i915/intel_ringbuffer.c +5 -2
index a8f72e8..f457146 100644
@@ -1845,12 +1845,15 @@ error:
 
 void intel_cleanup_ring_buffer(struct intel_engine_cs *ring)
 {
-	struct drm_i915_private *dev_priv = to_i915(ring->dev);
-	struct intel_ringbuffer *ringbuf = ring->buffer;
+	struct drm_i915_private *dev_priv;
+	struct intel_ringbuffer *ringbuf;
 
 	if (!intel_ring_initialized(ring))
 		return;
 
+	dev_priv = to_i915(ring->dev);
+	ringbuf = ring->buffer;
+
 	intel_stop_ring_buffer(ring);
 	WARN_ON(!IS_GEN2(ring->dev) && (I915_READ_MODE(ring) & MODE_IDLE) == 0);
 
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#4 lore raw

Re: [PATCH] drm/i915: Fix null pointer dereference in ring cleanup code

Quoted 9 lines · Damien Lespiau
Quoted 8 lines · On Fri, Oct 31, 2014 at 12:00:26PM +0000, [email protected] wrote:

From: John Harrison <[email protected]>

If a ring failed to initialise for any reason then the error path would try to clean up all rings including those that had not yet been allocated. The ring clean up code did a check that the ring was valid before starting its work. Unfortunately, that was after it had already dereferenced the ring to obtain a dev_private pointer.

Quoted 3 lines
Quoted 1 line

This looks good to me.

Really? These functions(!!!) are only called under controlled conditions... I would have been happy to see this follow my suggestion I made to fix this bug months ago.
-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
#5 lore raw

Re: [PATCH] drm/i915: Fix null pointer dereference in ring cleanup code

Quoted 10 lines · Chris Wilson
Quoted 9 lines · Damien Lespiau
Quoted 8 lines · On Fri, Oct 31, 2014 at 12:00:26PM +0000, [email protected] wrote:

From: John Harrison <[email protected]>

If a ring failed to initialise for any reason then the error path would try to clean up all rings including those that had not yet been allocated. The ring clean up code did a check that the ring was valid before starting its work. Unfortunately, that was after it had already dereferenced the ring to obtain a dev_private pointer.

Quoted 7 lines
Quoted 3 lines
Quoted 1 line

This looks good to me.

Really? These functions(!!!) are only called under controlled conditions... I would have been happy to see this follow my suggestion I made to fix this bug months ago.

Hm, do you mean to shuffle the ring_initialized checks into callers? Or something else?

John's patch didn't look offensive really, so merged it for now. -Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#7 lore raw

Re: [PATCH] drm/i915: Fix null pointer dereference in ring cleanup code

In reply to Damien Lespiau
Quoted 9 lines · Damien Lespiau
Quoted 8 lines · On Fri, Oct 31, 2014 at 12:00:26PM +0000, [email protected] wrote:

From: John Harrison <[email protected]>

If a ring failed to initialise for any reason then the error path would try to clean up all rings including those that had not yet been allocated. The ring clean up code did a check that the ring was valid before starting its work. Unfortunately, that was after it had already dereferenced the ring to obtain a dev_private pointer.

Quoted 5 lines
Quoted 1 line

This looks good to me.

And simpler than the version I previously posted, as that would have had to have another test added for each new ring in future hardware. However I think the description above is slightly misleading, as the problem wasn't dereferencing "ring" but "ring->dev". "ring" is always non-NULL (it's the address of a member of an array inside dev_priv), but the backpointer "ring->dev" is only filled in during ring initialisation.

.Dave.
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#6 lore raw

Re: [PATCH] drm/i915: Fix null pointer dereference in ring cleanup code

In reply to Daniel Vetter
Quoted 13 lines · Daniel Vetter
Quoted 12 lines · Chris Wilson
Quoted 11 lines · Damien Lespiau
Quoted 9 lines · On Fri, Oct 31, 2014 at 12:00:26PM +0000, [email protected] wrote:

From: John Harrison <[email protected]>

If a ring failed to initialise for any reason then the error path would try to clean up all rings including those that had not yet been allocated. The ring clean up code did a check that the ring was valid before starting its work. Unfortunately, that was after it had already dereferenced the ring to obtain a dev_private pointer.

Quoted 8 lines
Quoted 5 lines
Quoted 1 line

This looks good to me.

Really? These functions(!!!) are only called under controlled conditions... I would have been happy to see this follow my suggestion I made to fix this bug months ago.

Hm, do you mean to shuffle the ring_initialized checks into callers? Or something else?

i915_gem_cleanup_engines()

{
	/* Not the regular for_each_engine so we can cleanup a failed setup */
	for (int i =0 ; i < I915_NUM_ENGINES; i++)
		intel_engine_cleanup(&to_i915(dev)->engine[i]);
}

And that is callable then from an incomplete setup as well as normal teardown.

And intel_engine_cleanup() need just be:
void intel_engine_cleanup(struct intel_engine_cs *engine)

{
	WARN_ON(engine->last_request);

	if (engine->cleanup)
		engine->cleanup(engine);
}

Remove bugs by removing code, win.
-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