Skip to contentSkip to conversation

drm/i915: Add null state batch to active list

PATCH Superseded Mika Kuoppala 5 participants 12 messages May 21, 2014 – May 22, 2014

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

PATCH 1/1 drm/i915: Add null state batch to active list

for proper refcounting to take place as we use i915_add_request() for it.

i915_add_request() also takes the context for the request from ring->last_context so move the null state batch submission after the ring context has been set.

Cc Ville Syrjälä <[email protected]>
Cc Chris Wilson <[email protected]>
Cc Damien Lespiau <[email protected]>
---
 drivers/gpu/drm/i915/i915_drv.h              |    3 +++
 drivers/gpu/drm/i915/i915_gem.c              |    4 ++--
 drivers/gpu/drm/i915/i915_gem_context.c      |   16 ++++++++--------
 drivers/gpu/drm/i915/i915_gem_render_state.c |    7 +++++--
 4 files changed, 18 insertions(+), 12 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +3 -0
index b90ec69..6881f70 100644
@@ -2204,6 +2204,9 @@ static inline void i915_gem_object_unpin_pages(struct drm_i915_gem_object *obj)
 int __must_check i915_mutex_lock_interruptible(struct drm_device *dev);
 int i915_gem_object_sync(struct drm_i915_gem_object *obj,
 			 struct intel_ring_buffer *to);
+void i915_gem_object_move_to_active(struct drm_i915_gem_object *obj,
+				    struct intel_ring_buffer *ring);
+void i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj);
 void i915_vma_move_to_active(struct i915_vma *vma,
 			     struct intel_ring_buffer *ring);
 int i915_gem_dumb_create(struct drm_file *file_priv,
drivers/gpu/drm/i915/i915_gem.c +2 -2
index 440979f..d795800 100644
@@ -2038,7 +2038,7 @@ i915_gem_object_get_pages(struct drm_i915_gem_object *obj)
 	return 0;
 }
 
-static void
+void
 i915_gem_object_move_to_active(struct drm_i915_gem_object *obj,
 			       struct intel_ring_buffer *ring)
 {
@@ -2084,7 +2084,7 @@ void i915_vma_move_to_active(struct i915_vma *vma,
 	return i915_gem_object_move_to_active(vma->obj, ring);
 }
 
-static void
+void
 i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj)
 {
 	struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
drivers/gpu/drm/i915/i915_gem_context.c +8 -8
index f220c94..d3aad5b 100644
@@ -700,21 +700,21 @@ static int do_switch(struct intel_ring_buffer *ring,
 		/* obj is kept alive until the next request by its active ref */
 		i915_gem_object_ggtt_unpin(from->obj);
 		i915_gem_context_unreference(from);
-	} else {
-		if (to->is_initialized == false) {
-			ret = i915_gem_render_state_init(ring);
-			if (ret)
-				DRM_ERROR("init render state: %d\n", ret);
-		}
 	}
 
-	to->is_initialized = true;
-
 done:
 	i915_gem_context_reference(to);
 	ring->last_context = to;
 	to->last_ring = ring;
 
+	if (to->is_initialized == false && from == NULL) {
+		ret = i915_gem_render_state_init(ring);
+		if (ret)
+			DRM_ERROR("init render state: %d\n", ret);
+	}
+
+	to->is_initialized = true;
+
 	return 0;
 
 unpin_out:
drivers/gpu/drm/i915/i915_gem_render_state.c +5 -2
index 392aa7b..d29e6b2 100644
@@ -164,7 +164,6 @@ int i915_gem_render_state_init(struct intel_ring_buffer *ring)
 	const int gen = INTEL_INFO(ring->dev)->gen;
 	struct i915_render_state *so;
 	const struct intel_renderstate_rodata *rodata;
-	u32 seqno;
 	int ret;
 
 	rodata = render_state_get_rodata(ring->dev, gen);
@@ -186,7 +185,11 @@ int i915_gem_render_state_init(struct intel_ring_buffer *ring)
 	if (ret)
 		goto out;
 
-	ret = i915_add_request(ring, &seqno);
+	i915_gem_object_move_to_active(so->obj, ring);
+
+	ret = __i915_add_request(ring, NULL, so->obj, NULL);
+	if (ret)
+		i915_gem_object_move_to_inactive(so->obj);
 
 out:
 	render_state_free(so);
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#1 lore raw
Quoted 72 lines · Mika Kuoppala

for proper refcounting to take place as we use i915_add_request() for it.

i915_add_request() also takes the context for the request from ring->last_context so move the null state batch submission after the ring context has been set.

Cc Ville Syrjälä <[email protected]>
Cc Chris Wilson <[email protected]>
Cc Damien Lespiau <[email protected]>
---
 drivers/gpu/drm/i915/i915_drv.h              |    3 +++
 drivers/gpu/drm/i915/i915_gem.c              |    4 ++--
 drivers/gpu/drm/i915/i915_gem_context.c      |   16 ++++++++--------
 drivers/gpu/drm/i915/i915_gem_render_state.c |    7 +++++--
 4 files changed, 18 insertions(+), 12 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +3 -0
index b90ec69..6881f70 100644
@@ -2204,6 +2204,9 @@ static inline void i915_gem_object_unpin_pages(struct drm_i915_gem_object *obj)
 int __must_check i915_mutex_lock_interruptible(struct drm_device *dev);
 int i915_gem_object_sync(struct drm_i915_gem_object *obj,
 			 struct intel_ring_buffer *to);
+void i915_gem_object_move_to_active(struct drm_i915_gem_object *obj,
+				    struct intel_ring_buffer *ring);
+void i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj);
 void i915_vma_move_to_active(struct i915_vma *vma,
 			     struct intel_ring_buffer *ring);
 int i915_gem_dumb_create(struct drm_file *file_priv,
drivers/gpu/drm/i915/i915_gem.c +2 -2
index 440979f..d795800 100644
@@ -2038,7 +2038,7 @@ i915_gem_object_get_pages(struct drm_i915_gem_object *obj)
 	return 0;
 }
 
-static void
+void
 i915_gem_object_move_to_active(struct drm_i915_gem_object *obj,
 			       struct intel_ring_buffer *ring)
 {
@@ -2084,7 +2084,7 @@ void i915_vma_move_to_active(struct i915_vma *vma,
 	return i915_gem_object_move_to_active(vma->obj, ring);
 }
 
-static void
+void
 i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj)
 {
 	struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
drivers/gpu/drm/i915/i915_gem_context.c +0 -8
index f220c94..d3aad5b 100644
@@ -700,21 +700,21 @@ static int do_switch(struct intel_ring_buffer *ring,
 		/* obj is kept alive until the next request by its active ref */
 		i915_gem_object_ggtt_unpin(from->obj);
 		i915_gem_context_unreference(from);
-	} else {
-		if (to->is_initialized == false) {
-			ret = i915_gem_render_state_init(ring);
-			if (ret)
-				DRM_ERROR("init render state: %d\n", ret);
-		}
 	}
 
-	to->is_initialized = true;
-

drivers/gpu/drm/i915/i915_gem_context.c static int do_switch(struct intel_ring_buffer *ring, · line 714

Quoted 10 lines
drivers/gpu/drm/i915/i915_gem_context.c +5 -0
 done:
 	i915_gem_context_reference(to);
 	ring->last_context = to;
 	to->last_ring = ring;
 
+	if (to->is_initialized == false && from == NULL) {
+		ret = i915_gem_render_state_init(ring);
+		if (ret)
+			DRM_ERROR("init render state: %d\n", ret);
+	}

I think this will end badly on !RCS. So needs a ring check, and maybe throw a WARN into i915_gem_render_state_init() if it gets called with the wrong ring.

Quoted 33 lines
drivers/gpu/drm/i915/i915_gem_context.c +3 -0
+
+	to->is_initialized = true;
+
 	return 0;
 
 unpin_out:
drivers/gpu/drm/i915/i915_gem_render_state.c +5 -2
index 392aa7b..d29e6b2 100644
@@ -164,7 +164,6 @@ int i915_gem_render_state_init(struct intel_ring_buffer *ring)
 	const int gen = INTEL_INFO(ring->dev)->gen;
 	struct i915_render_state *so;
 	const struct intel_renderstate_rodata *rodata;
-	u32 seqno;
 	int ret;
 
 	rodata = render_state_get_rodata(ring->dev, gen);
@@ -186,7 +185,11 @@ int i915_gem_render_state_init(struct intel_ring_buffer *ring)
 	if (ret)
 		goto out;
 
-	ret = i915_add_request(ring, &seqno);
+	i915_gem_object_move_to_active(so->obj, ring);
+
+	ret = __i915_add_request(ring, NULL, so->obj, NULL);
+	if (ret)
+		i915_gem_object_move_to_inactive(so->obj);
 
 out:
 	render_state_free(so);
Signature · 1 line
1.7.9.5
Signature · 2 lines
Ville Syrjälä
Intel OTC
#2 lore raw

PATCH v2 1/1 drm/i915: Add null state batch to active list

for proper refcounting to take place as we use i915_add_request() for it.

i915_add_request() also takes the context for the request from ring->last_context so move the null state batch submission after the ring context has been set.

v2: we need to check for correct ring now (Ville Syrjälä)

Cc Ville Syrjälä <[email protected]>
Cc Chris Wilson <[email protected]>
Cc Damien Lespiau <[email protected]>
---
 drivers/gpu/drm/i915/i915_drv.h              |    3 +++
 drivers/gpu/drm/i915/i915_gem.c              |    4 ++--
 drivers/gpu/drm/i915/i915_gem_context.c      |   16 ++++++++--------
 drivers/gpu/drm/i915/i915_gem_render_state.c |   10 ++++++++--
 4 files changed, 21 insertions(+), 12 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +3 -0
index b90ec69..6881f70 100644
@@ -2204,6 +2204,9 @@ static inline void i915_gem_object_unpin_pages(struct drm_i915_gem_object *obj)
 int __must_check i915_mutex_lock_interruptible(struct drm_device *dev);
 int i915_gem_object_sync(struct drm_i915_gem_object *obj,
 			 struct intel_ring_buffer *to);
+void i915_gem_object_move_to_active(struct drm_i915_gem_object *obj,
+				    struct intel_ring_buffer *ring);
+void i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj);
 void i915_vma_move_to_active(struct i915_vma *vma,
 			     struct intel_ring_buffer *ring);
 int i915_gem_dumb_create(struct drm_file *file_priv,
drivers/gpu/drm/i915/i915_gem.c +2 -2
index 440979f..d795800 100644
@@ -2038,7 +2038,7 @@ i915_gem_object_get_pages(struct drm_i915_gem_object *obj)
 	return 0;
 }
 
-static void
+void
 i915_gem_object_move_to_active(struct drm_i915_gem_object *obj,
 			       struct intel_ring_buffer *ring)
 {
@@ -2084,7 +2084,7 @@ void i915_vma_move_to_active(struct i915_vma *vma,
 	return i915_gem_object_move_to_active(vma->obj, ring);
 }
 
-static void
+void
 i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj)
 {
 	struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
drivers/gpu/drm/i915/i915_gem_context.c +8 -8
index f220c94..6a2d847a 100644
@@ -700,21 +700,21 @@ static int do_switch(struct intel_ring_buffer *ring,
 		/* obj is kept alive until the next request by its active ref */
 		i915_gem_object_ggtt_unpin(from->obj);
 		i915_gem_context_unreference(from);
-	} else {
-		if (to->is_initialized == false) {
-			ret = i915_gem_render_state_init(ring);
-			if (ret)
-				DRM_ERROR("init render state: %d\n", ret);
-		}
 	}
 
-	to->is_initialized = true;
-
 done:
 	i915_gem_context_reference(to);
 	ring->last_context = to;
 	to->last_ring = ring;
 
+	if (ring->id == RCS && !to->is_initialized && from == NULL) {
+		ret = i915_gem_render_state_init(ring);
+		if (ret)
+			DRM_ERROR("init render state: %d\n", ret);
+	}
+
+	to->is_initialized = true;
+
 	return 0;
 
 unpin_out:
drivers/gpu/drm/i915/i915_gem_render_state.c +8 -2
index 392aa7b..463fff1 100644
@@ -164,9 +164,11 @@ int i915_gem_render_state_init(struct intel_ring_buffer *ring)
 	const int gen = INTEL_INFO(ring->dev)->gen;
 	struct i915_render_state *so;
 	const struct intel_renderstate_rodata *rodata;
-	u32 seqno;
 	int ret;
 
+	if (WARN_ON(ring->id != RCS))
+		return -ENOENT;
+
 	rodata = render_state_get_rodata(ring->dev, gen);
 	if (rodata == NULL)
 		return 0;
@@ -186,7 +188,11 @@ int i915_gem_render_state_init(struct intel_ring_buffer *ring)
 	if (ret)
 		goto out;
 
-	ret = i915_add_request(ring, &seqno);
+	i915_gem_object_move_to_active(so->obj, ring);
+
+	ret = __i915_add_request(ring, NULL, so->obj, NULL);
+	if (ret)
+		i915_gem_object_move_to_inactive(so->obj);
 
 out:
 	render_state_free(so);
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#3 lore raw
In reply to Mika Kuoppala
Quoted 13 lines · Mika Kuoppala
drivers/gpu/drm/i915/i915_gem_render_state.c +0 -1
index 392aa7b..d29e6b2 100644
@@ -164,7 +164,6 @@ int i915_gem_render_state_init(struct intel_ring_buffer *ring)
 	const int gen = INTEL_INFO(ring->dev)->gen;
 	struct i915_render_state *so;
 	const struct intel_renderstate_rodata *rodata;
-	u32 seqno;
 	int ret;
 
 	rodata = render_state_get_rodata(ring->dev, gen);

drivers/gpu/drm/i915/i915_gem_render_state.c int i915_gem_render_state_init(struct intel_ring_buffer *ring) · line 188

Quoted 5 lines
drivers/gpu/drm/i915/i915_gem_render_state.c +1 -1
@@ -186,7 +185,11 @@ int i915_gem_render_state_init(struct intel_ring_buffer *ring)
 	if (ret)
 		goto out;
 
-	ret = i915_add_request(ring, &seqno);
+	i915_gem_object_move_to_active(so->obj, ring);

You have the vma, use it rather than exposing this function. -Chris

Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
#4 lore raw

PATCH v3 1/1 drm/i915: Add null state batch to active list

for proper refcounting to take place as we use i915_add_request() for it.

i915_add_request() also takes the context for the request from ring->last_context so move the null state batch submission after the ring context has been set.

v2: we need to check for correct ring now (Ville Syrjälä)
v3: no need to expose i915_gem_move_object_to_active (Chris Wilson)

Cc Ville Syrjälä <[email protected]>
Cc Chris Wilson <[email protected]>
Cc Damien Lespiau <[email protected]>
---
 drivers/gpu/drm/i915/i915_drv.h              |    1 +
 drivers/gpu/drm/i915/i915_gem.c              |    2 +-
 drivers/gpu/drm/i915/i915_gem_context.c      |   16 ++++++++--------
 drivers/gpu/drm/i915/i915_gem_render_state.c |   17 +++++++++++++++--
 4 files changed, 25 insertions(+), 11 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +1 -0
index b90ec69..0bf7bfb 100644
@@ -2206,6 +2206,7 @@ int i915_gem_object_sync(struct drm_i915_gem_object *obj,
 			 struct intel_ring_buffer *to);
 void i915_vma_move_to_active(struct i915_vma *vma,
 			     struct intel_ring_buffer *ring);
+void i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj);
 int i915_gem_dumb_create(struct drm_file *file_priv,
 			 struct drm_device *dev,
 			 struct drm_mode_create_dumb *args);
drivers/gpu/drm/i915/i915_gem.c +1 -1
index 440979f..d9bf694 100644
@@ -2084,7 +2084,7 @@ void i915_vma_move_to_active(struct i915_vma *vma,
 	return i915_gem_object_move_to_active(vma->obj, ring);
 }
 
-static void
+void
 i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj)
 {
 	struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
drivers/gpu/drm/i915/i915_gem_context.c +8 -8
index f220c94..6a2d847a 100644
@@ -700,21 +700,21 @@ static int do_switch(struct intel_ring_buffer *ring,
 		/* obj is kept alive until the next request by its active ref */
 		i915_gem_object_ggtt_unpin(from->obj);
 		i915_gem_context_unreference(from);
-	} else {
-		if (to->is_initialized == false) {
-			ret = i915_gem_render_state_init(ring);
-			if (ret)
-				DRM_ERROR("init render state: %d\n", ret);
-		}
 	}
 
-	to->is_initialized = true;
-
 done:
 	i915_gem_context_reference(to);
 	ring->last_context = to;
 	to->last_ring = ring;
 
+	if (ring->id == RCS && !to->is_initialized && from == NULL) {
+		ret = i915_gem_render_state_init(ring);
+		if (ret)
+			DRM_ERROR("init render state: %d\n", ret);
+	}
+
+	to->is_initialized = true;
+
 	return 0;
 
 unpin_out:
drivers/gpu/drm/i915/i915_gem_render_state.c +15 -2
index 392aa7b..82abe1e 100644
@@ -164,9 +164,12 @@ int i915_gem_render_state_init(struct intel_ring_buffer *ring)
 	const int gen = INTEL_INFO(ring->dev)->gen;
 	struct i915_render_state *so;
 	const struct intel_renderstate_rodata *rodata;
-	u32 seqno;
+	struct i915_vma *vma;
 	int ret;
 
+	if (WARN_ON(ring->id != RCS))
+		return -ENOENT;
+
 	rodata = render_state_get_rodata(ring->dev, gen);
 	if (rodata == NULL)
 		return 0;
@@ -186,7 +189,17 @@ int i915_gem_render_state_init(struct intel_ring_buffer *ring)
 	if (ret)
 		goto out;
 
-	ret = i915_add_request(ring, &seqno);
+	vma = i915_gem_obj_to_ggtt(so->obj);
+	if (vma == NULL) {
+		ret = -ENOSPC;
+		goto out;
+	}
+
+	i915_vma_move_to_active(vma, ring);
+
+	ret = __i915_add_request(ring, NULL, so->obj, NULL);
+	if (ret)
+		i915_gem_object_move_to_inactive(so->obj);
 
 out:
 	render_state_free(so);
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#5 lore raw
Quoted 7 lines · Mika Kuoppala

for proper refcounting to take place as we use i915_add_request() for it.

i915_add_request() also takes the context for the request from ring->last_context so move the null state batch submission after the ring context has been set.

Quoted 7 lines

v2: we need to check for correct ring now (Ville Syrjälä)
v3: no need to expose i915_gem_move_object_to_active (Chris Wilson)

Cc Ville Syrjälä <[email protected]>
Cc Chris Wilson <[email protected]>
Cc Damien Lespiau <[email protected]>

Merged with Ville's irc r-b, thanks for the quick fix. -Daniel

Quoted 110 lines
---
 drivers/gpu/drm/i915/i915_drv.h              |    1 +
 drivers/gpu/drm/i915/i915_gem.c              |    2 +-
 drivers/gpu/drm/i915/i915_gem_context.c      |   16 ++++++++--------
 drivers/gpu/drm/i915/i915_gem_render_state.c |   17 +++++++++++++++--
 4 files changed, 25 insertions(+), 11 deletions(-)
drivers/gpu/drm/i915/i915_drv.h +1 -0
index b90ec69..0bf7bfb 100644
@@ -2206,6 +2206,7 @@ int i915_gem_object_sync(struct drm_i915_gem_object *obj,
 			 struct intel_ring_buffer *to);
 void i915_vma_move_to_active(struct i915_vma *vma,
 			     struct intel_ring_buffer *ring);
+void i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj);
 int i915_gem_dumb_create(struct drm_file *file_priv,
 			 struct drm_device *dev,
 			 struct drm_mode_create_dumb *args);
drivers/gpu/drm/i915/i915_gem.c +1 -1
index 440979f..d9bf694 100644
@@ -2084,7 +2084,7 @@ void i915_vma_move_to_active(struct i915_vma *vma,
 	return i915_gem_object_move_to_active(vma->obj, ring);
 }
 
-static void
+void
 i915_gem_object_move_to_inactive(struct drm_i915_gem_object *obj)
 {
 	struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
drivers/gpu/drm/i915/i915_gem_context.c +8 -8
index f220c94..6a2d847a 100644
@@ -700,21 +700,21 @@ static int do_switch(struct intel_ring_buffer *ring,
 		/* obj is kept alive until the next request by its active ref */
 		i915_gem_object_ggtt_unpin(from->obj);
 		i915_gem_context_unreference(from);
-	} else {
-		if (to->is_initialized == false) {
-			ret = i915_gem_render_state_init(ring);
-			if (ret)
-				DRM_ERROR("init render state: %d\n", ret);
-		}
 	}
 
-	to->is_initialized = true;
-
 done:
 	i915_gem_context_reference(to);
 	ring->last_context = to;
 	to->last_ring = ring;
 
+	if (ring->id == RCS && !to->is_initialized && from == NULL) {
+		ret = i915_gem_render_state_init(ring);
+		if (ret)
+			DRM_ERROR("init render state: %d\n", ret);
+	}
+
+	to->is_initialized = true;
+
 	return 0;
 
 unpin_out:
drivers/gpu/drm/i915/i915_gem_render_state.c +15 -2
index 392aa7b..82abe1e 100644
@@ -164,9 +164,12 @@ int i915_gem_render_state_init(struct intel_ring_buffer *ring)
 	const int gen = INTEL_INFO(ring->dev)->gen;
 	struct i915_render_state *so;
 	const struct intel_renderstate_rodata *rodata;
-	u32 seqno;
+	struct i915_vma *vma;
 	int ret;
 
+	if (WARN_ON(ring->id != RCS))
+		return -ENOENT;
+
 	rodata = render_state_get_rodata(ring->dev, gen);
 	if (rodata == NULL)
 		return 0;
@@ -186,7 +189,17 @@ int i915_gem_render_state_init(struct intel_ring_buffer *ring)
 	if (ret)
 		goto out;
 
-	ret = i915_add_request(ring, &seqno);
+	vma = i915_gem_obj_to_ggtt(so->obj);
+	if (vma == NULL) {
+		ret = -ENOSPC;
+		goto out;
+	}
+
+	i915_vma_move_to_active(vma, ring);
+
+	ret = __i915_add_request(ring, NULL, so->obj, NULL);
+	if (ret)
+		i915_gem_object_move_to_inactive(so->obj);
 
 out:
 	render_state_free(so);
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
#6 lore raw
Quoted 11 lines · Daniel Vetter
Quoted 10 lines · Mika Kuoppala

for proper refcounting to take place as we use i915_add_request() for it.

i915_add_request() also takes the context for the request from ring->last_context so move the null state batch submission after the ring context has been set.

v2: we need to check for correct ring now (Ville Syrjälä)
v3: no need to expose i915_gem_move_object_to_active (Chris Wilson)

Quoted 6 lines
Quoted 4 lines
Cc Ville Syrjälä <[email protected]>
Cc Chris Wilson <[email protected]>
Cc Damien Lespiau <[email protected]>

Merged with Ville's irc r-b, thanks for the quick fix.

Pity, it still contains some code that I'd rather not start cargo-culting.

Quoted 6 lines

drivers/gpu/drm/i915/i915_gem_render_state.c int i915_gem_render_state_init(struct intel_ring_buffer *ring) · line 196

Quoted 6 lines
drivers/gpu/drm/i915/i915_gem_render_state.c +5 -1
-	ret = i915_add_request(ring, &seqno);
+	vma = i915_gem_obj_to_ggtt(so->obj);
+	if (vma == NULL) {
+		ret = -ENOSPC;
+		goto out;
+	}

We use the GGTT vma much earlier, so this is suspect.

Quoted 6 lines

drivers/gpu/drm/i915/i915_gem_render_state.c int i915_gem_render_state_init(struct intel_ring_buffer *ring) · line 202

Quoted 6 lines
drivers/gpu/drm/i915/i915_gem_render_state.c +6 -0
+
+	i915_vma_move_to_active(vma, ring);
+
+	ret = __i915_add_request(ring, NULL, so->obj, NULL);
+	if (ret)
+		i915_gem_object_move_to_inactive(so->obj);

As is move-to-inactive here. The only way add-request can fail is via an EIO, and that will have triggered the move-to-inactive already - which should nicely catch a BUG or two.
-Chris

Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
#7 lore raw
Quoted 17 lines · Chris Wilson
Quoted 16 lines · Daniel Vetter
Quoted 14 lines · Mika Kuoppala

for proper refcounting to take place as we use i915_add_request() for it.

i915_add_request() also takes the context for the request from ring->last_context so move the null state batch submission after the ring context has been set.

v2: we need to check for correct ring now (Ville Syrjälä)
v3: no need to expose i915_gem_move_object_to_active (Chris Wilson)

Cc Ville Syrjälä <[email protected]>
Cc Chris Wilson <[email protected]>
Cc Damien Lespiau <[email protected]>
Quoted 4 lines
Quoted 1 line

Merged with Ville's irc r-b, thanks for the quick fix.

Pity, it still contains some code that I'd rather not start cargo-culting.

Using vmas more as first-class objects I support. Mika can you please resend? I'll keep v3 for now to avoid blocking people. -Daniel

Quoted 23 lines
Quoted 6 lines

drivers/gpu/drm/i915/i915_gem_render_state.c int i915_gem_render_state_init(struct intel_ring_buffer *ring) · line 196

Quoted 6 lines
drivers/gpu/drm/i915/i915_gem_render_state.c +5 -1
-	ret = i915_add_request(ring, &seqno);
+	vma = i915_gem_obj_to_ggtt(so->obj);
+	if (vma == NULL) {
+		ret = -ENOSPC;
+		goto out;
+	}

We use the GGTT vma much earlier, so this is suspect.

Quoted 6 lines

drivers/gpu/drm/i915/i915_gem_render_state.c int i915_gem_render_state_init(struct intel_ring_buffer *ring) · line 202

Quoted 6 lines
drivers/gpu/drm/i915/i915_gem_render_state.c +6 -0
+
+	i915_vma_move_to_active(vma, ring);
+
+	ret = __i915_add_request(ring, NULL, so->obj, NULL);
+	if (ret)
+		i915_gem_object_move_to_inactive(so->obj);

As is move-to-inactive here. The only way add-request can fail is via an EIO, and that will have triggered the move-to-inactive already - which should nicely catch a BUG or two.
-Chris

Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
#8 lore raw

PATCH v4 1/1 drm/i915: Add null state batch to active list

From: Mika Kuoppala <[email protected]>

for proper refcounting to take place as we use
i915_add_request() for it.

i915_add_request() also takes the context for the request from ring->last_context so move the null state batch submission after the ring context has been set.

v2: we need to check for correct ring now (Ville Syrjälä)
v3: no need to expose i915_gem_move_object_to_active (Chris Wilson)
v4: cargoculted vma/active/inactive error handling removed (Chris Wilson)

Cc Ville Syrjälä <[email protected]>
Cc Chris Wilson <[email protected]>
Cc Damien Lespiau <[email protected]>
---
 drivers/gpu/drm/i915/i915_gem_context.c      |   16 ++++++++--------
 drivers/gpu/drm/i915/i915_gem_render_state.c |    8 ++++++--
 2 files changed, 14 insertions(+), 10 deletions(-)
drivers/gpu/drm/i915/i915_gem_context.c +8 -8
index f220c94..6a2d847a 100644
@@ -700,21 +700,21 @@ static int do_switch(struct intel_ring_buffer *ring,
 		/* obj is kept alive until the next request by its active ref */
 		i915_gem_object_ggtt_unpin(from->obj);
 		i915_gem_context_unreference(from);
-	} else {
-		if (to->is_initialized == false) {
-			ret = i915_gem_render_state_init(ring);
-			if (ret)
-				DRM_ERROR("init render state: %d\n", ret);
-		}
 	}
 
-	to->is_initialized = true;
-
 done:
 	i915_gem_context_reference(to);
 	ring->last_context = to;
 	to->last_ring = ring;
 
+	if (ring->id == RCS && !to->is_initialized && from == NULL) {
+		ret = i915_gem_render_state_init(ring);
+		if (ret)
+			DRM_ERROR("init render state: %d\n", ret);
+	}
+
+	to->is_initialized = true;
+
 	return 0;
 
 unpin_out:
drivers/gpu/drm/i915/i915_gem_render_state.c +6 -2
index 392aa7b..cfbf6fc 100644
@@ -164,9 +164,11 @@ int i915_gem_render_state_init(struct intel_ring_buffer *ring)
 	const int gen = INTEL_INFO(ring->dev)->gen;
 	struct i915_render_state *so;
 	const struct intel_renderstate_rodata *rodata;
-	u32 seqno;
 	int ret;
 
+	if (WARN_ON(ring->id != RCS))
+		return -ENOENT;
+
 	rodata = render_state_get_rodata(ring->dev, gen);
 	if (rodata == NULL)
 		return 0;
@@ -186,8 +188,10 @@ int i915_gem_render_state_init(struct intel_ring_buffer *ring)
 	if (ret)
 		goto out;
 
-	ret = i915_add_request(ring, &seqno);
+	i915_vma_move_to_active(i915_gem_obj_to_ggtt(so->obj), ring);
 
+	ret = __i915_add_request(ring, NULL, so->obj, NULL);
+	/* __i915_add_request moves object to inactive if it fails */
 out:
 	render_state_free(so);
 	return ret;
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#9 lore raw
Quoted 13 lines · Mika Kuoppala

From: Mika Kuoppala <[email protected]>

for proper refcounting to take place as we use
i915_add_request() for it.

i915_add_request() also takes the context for the request from ring->last_context so move the null state batch submission after the ring context has been set.

v2: we need to check for correct ring now (Ville Syrjälä)
v3: no need to expose i915_gem_move_object_to_active (Chris Wilson)
v4: cargoculted vma/active/inactive error handling removed (Chris Wilson)

Quoted 5 lines
Cc Ville Syrjälä <[email protected]>
Cc Chris Wilson <[email protected]>
Cc Damien Lespiau <[email protected]>

Ok, merged instead of v3 with Chris' irc r-b.

Thanks, Daniel

Quoted 75 lines
---
 drivers/gpu/drm/i915/i915_gem_context.c      |   16 ++++++++--------
 drivers/gpu/drm/i915/i915_gem_render_state.c |    8 ++++++--
 2 files changed, 14 insertions(+), 10 deletions(-)
drivers/gpu/drm/i915/i915_gem_context.c +8 -8
index f220c94..6a2d847a 100644
@@ -700,21 +700,21 @@ static int do_switch(struct intel_ring_buffer *ring,
 		/* obj is kept alive until the next request by its active ref */
 		i915_gem_object_ggtt_unpin(from->obj);
 		i915_gem_context_unreference(from);
-	} else {
-		if (to->is_initialized == false) {
-			ret = i915_gem_render_state_init(ring);
-			if (ret)
-				DRM_ERROR("init render state: %d\n", ret);
-		}
 	}
 
-	to->is_initialized = true;
-
 done:
 	i915_gem_context_reference(to);
 	ring->last_context = to;
 	to->last_ring = ring;
 
+	if (ring->id == RCS && !to->is_initialized && from == NULL) {
+		ret = i915_gem_render_state_init(ring);
+		if (ret)
+			DRM_ERROR("init render state: %d\n", ret);
+	}
+
+	to->is_initialized = true;
+
 	return 0;
 
 unpin_out:
drivers/gpu/drm/i915/i915_gem_render_state.c +6 -2
index 392aa7b..cfbf6fc 100644
@@ -164,9 +164,11 @@ int i915_gem_render_state_init(struct intel_ring_buffer *ring)
 	const int gen = INTEL_INFO(ring->dev)->gen;
 	struct i915_render_state *so;
 	const struct intel_renderstate_rodata *rodata;
-	u32 seqno;
 	int ret;
 
+	if (WARN_ON(ring->id != RCS))
+		return -ENOENT;
+
 	rodata = render_state_get_rodata(ring->dev, gen);
 	if (rodata == NULL)
 		return 0;
@@ -186,8 +188,10 @@ int i915_gem_render_state_init(struct intel_ring_buffer *ring)
 	if (ret)
 		goto out;
 
-	ret = i915_add_request(ring, &seqno);
+	i915_vma_move_to_active(i915_gem_obj_to_ggtt(so->obj), ring);
 
+	ret = __i915_add_request(ring, NULL, so->obj, NULL);
+	/* __i915_add_request moves object to inactive if it fails */
 out:
 	render_state_free(so);
 	return ret;
Signature · 6 lines
1.7.9.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
#10 lore raw
In reply to Mika Kuoppala

drivers/gpu/drm/i915/i915_gem_context.c static int do_switch(struct intel_ring_buffer *ring, · line 714

Quoted 5 lines · Mika Kuoppala
drivers/gpu/drm/i915/i915_gem_context.c +5 -0
+	if (ring->id == RCS && !to->is_initialized && from == NULL) {
+		ret = i915_gem_render_state_init(ring);
+		if (ret)
+			DRM_ERROR("init render state: %d\n", ret);
+	}

Apologies if this has already been discussed, but why do we have the 'from == NULL' check? Shouldn't we initialize all uninitialized RCS contexts? Otherwise I thought we'll inherit whatever state 'from' left behind.

The hw state should be valid in either case (and so I expect would fix the rc6 issue either way), it's just the difference between initializing every context to a specific valid state or initializing every context to _some_ valid state. The commit message on the first render state patch seemed to indicate the former while the implementation looks like the latter. Just want to understand which we intended.

Thanks,
Brad

#11 lore raw
Quoted 12 lines · Bradley D

drivers/gpu/drm/i915/i915_gem_context.c static int do_switch(struct intel_ring_buffer *ring, · line 714

Quoted 5 lines · Mika Kuoppala
drivers/gpu/drm/i915/i915_gem_context.c +5 -0
+	if (ring->id == RCS && !to->is_initialized && from == NULL) {
+		ret = i915_gem_render_state_init(ring);
+		if (ret)
+			DRM_ERROR("init render state: %d\n", ret);
+	}

Apologies if this has already been discussed, but why do we have the 'from == NULL' check? Shouldn't we initialize all uninitialized RCS contexts? Otherwise I thought we'll inherit whatever state 'from' left behind.

Quoted 6 lines

The hw state should be valid in either case (and so I expect would fix the rc6 issue either way), it's just the difference between initializing every context to a specific valid state or initializing every context to _some_ valid state. The commit message on the first render state patch seemed to indicate the former while the implementation looks like the latter. Just want to understand which we intended.

It seems that the intentions changed and I forgot to update the commit message. For now this is just to push some state and hoping that we can get in/out from rc6 without symptomps.

The idea that we would restore/initialize to a specific (golden) state for each new context makes me think that we would get rid of some transient bugs we are seeing. As how I understand things are now, app might inherit some parts from previous ctx and then have lacking initialization by itself and then see a hang..sometimes.

I guess the main opponent here is the performance implications. And I lack the experience from user/application side to estimate the impact.

I hope that other devs with more experience on this topic will join the discussion.

Thanks,
-Mika