[PATCH 0/6] make gem_object embedable and convert i915 driver

[PATCH 0/6] make gem_object embedable and convert i915 driver

[Daniel Vetter]

Hi all,

As promised here's my patch series to make struct drm_gem_object embedable.
Also converts drm/i915 as a proof of concept.

I've already looked at the radeon and nouveau modules and a straightforward
replament of the gem_object pointer looks simple. But I think embedding it
into the ttm_bo is better long-term. Stuff like duplicated ref-counting
between the ttm object and the gem one just doesn't make too much sense.

Tested on my i945GM. The changes to radeon and nouveau in patch 2 are only
compile-tested, but the identical change to the i915 survived testing.

Comments on the patches and my future plans highly welcome.

Yours, Daniel

Daniel Vetter (6):
  drm: extract drm_gem_object_init
  drm: free core gem object from driver callbacks
  drm/i915: introduce i915_gem_alloc_object
  drm/i915: embed the gem object into drm_i915_gem_object
  drm/i915: don't use ->driver_private anymore
  drm/i915: drop pointer to drm_gem_object

 drivers/gpu/drm/drm_gem.c             |   49 +++++++++++++-------
 drivers/gpu/drm/i915/i915_debugfs.c   |   15 +++---
 drivers/gpu/drm/i915/i915_drv.h       |    6 ++-
 drivers/gpu/drm/i915/i915_gem.c       |   80 ++++++++++++++++++---------------
 drivers/gpu/drm/i915/i915_gem_debug.c |    2 +-
 drivers/gpu/drm/i915/i915_irq.c       |    4 +-
 drivers/gpu/drm/i915/intel_display.c  |    2 +-
 drivers/gpu/drm/i915/intel_fb.c       |    2 +-
 drivers/gpu/drm/i915/intel_overlay.c  |    8 ++--
 drivers/gpu/drm/nouveau/nouveau_gem.c |    3 +
 drivers/gpu/drm/radeon/radeon_gem.c   |    3 +
 include/drm/drmP.h                    |    3 +
 12 files changed, 105 insertions(+), 72 deletions(-)


------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--

[PATCH 1/6] drm: extract drm_gem_object_init

[Daniel Vetter]

This function can be used by drivers who allocate the drm gem object
on their own. No functional change in here, just preparation.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/drm_gem.c |   39 +++++++++++++++++++++++++++++----------
 include/drm/drmP.h        |    2 ++
 2 files changed, 31 insertions(+), 10 deletions(-)

diff --git a/drivers/gpu/drm/drm_gem.c b/drivers/gpu/drm/drm_gem.c
index aa89d4b..3b64d0e 100644
--- a/drivers/gpu/drm/drm_gem.c
+++ b/drivers/gpu/drm/drm_gem.c
@@ -124,6 +124,31 @@ drm_gem_destroy(struct drm_device *dev)
 }
 
 /**
+ * Initialize an already allocate GEM object of the specified size with
+ * shmfs backing store.
+ */
+int drm_gem_object_init(struct drm_device *dev,
+			struct drm_gem_object *obj, size_t size)
+{
+	BUG_ON((size & (PAGE_SIZE - 1)) != 0);
+
+	obj->dev = dev;
+	obj->filp = shmem_file_setup("drm mm object", size, VM_NORESERVE);
+	if (IS_ERR(obj->filp))
+		return -ENOMEM;
+
+	kref_init(&obj->refcount);
+	kref_init(&obj->handlecount);
+	obj->size = size;
+
+	atomic_inc(&dev->object_count);
+	atomic_add(obj->size, &dev->object_memory);
+
+	return 0;
+}
+EXPORT_SYMBOL(drm_gem_object_init);
+
+/**
  * Allocate a GEM object of the specified size with shmfs backing store
  */
 struct drm_gem_object *
@@ -131,28 +156,22 @@ drm_gem_object_alloc(struct drm_device *dev, size_t size)
 {
 	struct drm_gem_object *obj;
 
-	BUG_ON((size & (PAGE_SIZE - 1)) != 0);
-
 	obj = kzalloc(sizeof(*obj), GFP_KERNEL);
 	if (!obj)
 		goto free;
 
-	obj->dev = dev;
-	obj->filp = shmem_file_setup("drm mm object", size, VM_NORESERVE);
-	if (IS_ERR(obj->filp))
+	if (drm_gem_object_init(dev, obj, size) != 0)
 		goto free;
 
-	kref_init(&obj->refcount);
-	kref_init(&obj->handlecount);
-	obj->size = size;
 	if (dev->driver->gem_init_object != NULL &&
 	    dev->driver->gem_init_object(obj) != 0) {
 		goto fput;
 	}
-	atomic_inc(&dev->object_count);
-	atomic_add(obj->size, &dev->object_memory);
 	return obj;
 fput:
+	/* Object_init mangles the global counters - readjust them. */
+	atomic_dec(&dev->object_count);
+	atomic_sub(obj->size, &dev->object_memory);
 	fput(obj->filp);
 free:
 	kfree(obj);
diff --git a/include/drm/drmP.h b/include/drm/drmP.h
index 2f3b3a0..b3b57b5 100644
--- a/include/drm/drmP.h
+++ b/include/drm/drmP.h
@@ -1432,6 +1432,8 @@ void drm_gem_object_free(struct kref *kref);
 void drm_gem_object_free_unlocked(struct kref *kref);
 struct drm_gem_object *drm_gem_object_alloc(struct drm_device *dev,
 					    size_t size);
+int drm_gem_object_init(struct drm_device *dev,
+			struct drm_gem_object *obj, size_t size);
 void drm_gem_object_handle_free(struct kref *kref);
 void drm_gem_vm_open(struct vm_area_struct *vma);
 void drm_gem_vm_close(struct vm_area_struct *vma);
-- 
1.6.6.1


------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--

[PATCH 2/6] drm: free core gem object from driver callbacks

[Daniel Vetter]

When drivers embed the core gem object into their own structures,
they'll have to do this. Temporarily this results in an ugly

kfree(gem_obj);

in every gem driver.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/drm_gem.c             |   10 +++-------
 drivers/gpu/drm/i915/i915_gem.c       |    3 +++
 drivers/gpu/drm/nouveau/nouveau_gem.c |    3 +++
 drivers/gpu/drm/radeon/radeon_gem.c   |    3 +++
 include/drm/drmP.h                    |    1 +
 5 files changed, 13 insertions(+), 7 deletions(-)

diff --git a/drivers/gpu/drm/drm_gem.c b/drivers/gpu/drm/drm_gem.c
index 3b64d0e..33dad3f 100644
--- a/drivers/gpu/drm/drm_gem.c
+++ b/drivers/gpu/drm/drm_gem.c
@@ -422,15 +422,15 @@ drm_gem_release(struct drm_device *dev, struct drm_file *file_private)
 	idr_destroy(&file_private->object_idr);
 }
 
-static void
-drm_gem_object_free_common(struct drm_gem_object *obj)
+void
+drm_gem_object_release(struct drm_gem_object *obj)
 {
 	struct drm_device *dev = obj->dev;
 	fput(obj->filp);
 	atomic_dec(&dev->object_count);
 	atomic_sub(obj->size, &dev->object_memory);
-	kfree(obj);
 }
+EXPORT_SYMBOL(drm_gem_object_release);
 
 /**
  * Called after the last reference to the object has been lost.
@@ -448,8 +448,6 @@ drm_gem_object_free(struct kref *kref)
 
 	if (dev->driver->gem_free_object != NULL)
 		dev->driver->gem_free_object(obj);
-
-	drm_gem_object_free_common(obj);
 }
 EXPORT_SYMBOL(drm_gem_object_free);
 
@@ -472,8 +470,6 @@ drm_gem_object_free_unlocked(struct kref *kref)
 		dev->driver->gem_free_object(obj);
 		mutex_unlock(&dev->struct_mutex);
 	}
-
-	drm_gem_object_free_common(obj);
 }
 EXPORT_SYMBOL(drm_gem_object_free_unlocked);
 
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 80871c6..d4ea909 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -4474,6 +4474,9 @@ void i915_gem_free_object(struct drm_gem_object *obj)
 	kfree(obj_priv->page_cpu_valid);
 	kfree(obj_priv->bit_17);
 	kfree(obj->driver_private);
+
+	drm_gem_object_release(obj);
+	kfree(obj);
 }
 
 /** Unbinds all inactive objects. */
diff --git a/drivers/gpu/drm/nouveau/nouveau_gem.c b/drivers/gpu/drm/nouveau/nouveau_gem.c
index 0d22f66..cc6b191 100644
--- a/drivers/gpu/drm/nouveau/nouveau_gem.c
+++ b/drivers/gpu/drm/nouveau/nouveau_gem.c
@@ -57,6 +57,9 @@ nouveau_gem_object_del(struct drm_gem_object *gem)
 	}
 
 	ttm_bo_unref(&bo);
+
+	drm_gem_object_release(gem);
+	kfree(gem);
 }
 
 int
diff --git a/drivers/gpu/drm/radeon/radeon_gem.c b/drivers/gpu/drm/radeon/radeon_gem.c
index ef92d14..833454a 100644
--- a/drivers/gpu/drm/radeon/radeon_gem.c
+++ b/drivers/gpu/drm/radeon/radeon_gem.c
@@ -44,6 +44,9 @@ void radeon_gem_object_free(struct drm_gem_object *gobj)
 	if (robj) {
 		radeon_bo_unref(&robj);
 	}
+
+	drm_gem_object_release(gobj);
+	kfree(gobj);
 }
 
 int radeon_gem_object_create(struct radeon_device *rdev, int size,
diff --git a/include/drm/drmP.h b/include/drm/drmP.h
index b3b57b5..c1b9871 100644
--- a/include/drm/drmP.h
+++ b/include/drm/drmP.h
@@ -1428,6 +1428,7 @@ extern void drm_sysfs_connector_remove(struct drm_connector *connector);
 /* Graphics Execution Manager library functions (drm_gem.c) */
 int drm_gem_init(struct drm_device *dev);
 void drm_gem_destroy(struct drm_device *dev);
+void drm_gem_object_release(struct drm_gem_object *obj);
 void drm_gem_object_free(struct kref *kref);
 void drm_gem_object_free_unlocked(struct kref *kref);
 struct drm_gem_object *drm_gem_object_alloc(struct drm_device *dev,
-- 
1.6.6.1


------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--

[PATCH 3/6] drm/i915: introduce i915_gem_alloc_object

[Daniel Vetter]

Just preparation, no functional change.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/i915/i915_drv.h      |    2 ++
 drivers/gpu/drm/i915/i915_gem.c      |   12 +++++++++---
 drivers/gpu/drm/i915/intel_display.c |    2 +-
 drivers/gpu/drm/i915/intel_fb.c      |    2 +-
 drivers/gpu/drm/i915/intel_overlay.c |    2 +-
 5 files changed, 14 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index b7cb4aa..3d4e135 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -893,6 +893,8 @@ int i915_gem_get_aperture_ioctl(struct drm_device *dev, void *data,
 				struct drm_file *file_priv);
 void i915_gem_load(struct drm_device *dev);
 int i915_gem_init_object(struct drm_gem_object *obj);
+struct drm_gem_object * i915_gem_alloc_object(struct drm_device *dev,
+					      size_t size);
 void i915_gem_free_object(struct drm_gem_object *obj);
 int i915_gem_object_pin(struct drm_gem_object *obj, uint32_t alignment);
 void i915_gem_object_unpin(struct drm_gem_object *obj);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index d4ea909..92dd522 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -124,7 +124,7 @@ i915_gem_create_ioctl(struct drm_device *dev, void *data,
 	args->size = roundup(args->size, PAGE_SIZE);
 
 	/* Allocate the new object */
-	obj = drm_gem_object_alloc(dev, args->size);
+	obj = i915_gem_alloc_object(dev, args->size);
 	if (obj == NULL)
 		return -ENOMEM;
 
@@ -4421,6 +4421,12 @@ i915_gem_madvise_ioctl(struct drm_device *dev, void *data,
 	return 0;
 }
 
+struct drm_gem_object * i915_gem_alloc_object(struct drm_device *dev,
+					      size_t size)
+{
+	return drm_gem_object_alloc(dev, size);
+}
+
 int i915_gem_init_object(struct drm_gem_object *obj)
 {
 	struct drm_i915_gem_object *obj_priv;
@@ -4563,7 +4569,7 @@ i915_gem_init_hws(struct drm_device *dev)
 	if (!I915_NEED_GFX_HWS(dev))
 		return 0;
 
-	obj = drm_gem_object_alloc(dev, 4096);
+	obj = i915_gem_alloc_object(dev, 4096);
 	if (obj == NULL) {
 		DRM_ERROR("Failed to allocate status page\n");
 		return -ENOMEM;
@@ -4640,7 +4646,7 @@ i915_gem_init_ringbuffer(struct drm_device *dev)
 	if (ret != 0)
 		return ret;
 
-	obj = drm_gem_object_alloc(dev, 128 * 1024);
+	obj = i915_gem_alloc_object(dev, 128 * 1024);
 	if (obj == NULL) {
 		DRM_ERROR("Failed to allocate ringbuffer\n");
 		i915_gem_cleanup_hws(dev);
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index eb9d825..bc775b6 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -4598,7 +4598,7 @@ intel_alloc_power_context(struct drm_device *dev)
 	struct drm_gem_object *pwrctx;
 	int ret;
 
-	pwrctx = drm_gem_object_alloc(dev, 4096);
+	pwrctx = i915_gem_alloc_object(dev, 4096);
 	if (!pwrctx) {
 		DRM_DEBUG("failed to alloc power context, RC6 disabled\n");
 		return NULL;
diff --git a/drivers/gpu/drm/i915/intel_fb.c b/drivers/gpu/drm/i915/intel_fb.c
index 8a0b3bc..c7af7e1 100644
--- a/drivers/gpu/drm/i915/intel_fb.c
+++ b/drivers/gpu/drm/i915/intel_fb.c
@@ -138,7 +138,7 @@ static int intelfb_create(struct drm_device *dev, uint32_t fb_width,
 
 	size = mode_cmd.pitch * mode_cmd.height;
 	size = ALIGN(size, PAGE_SIZE);
-	fbo = drm_gem_object_alloc(dev, size);
+	fbo = i915_gem_alloc_object(dev, size);
 	if (!fbo) {
 		DRM_ERROR("failed to allocate framebuffer\n");
 		ret = -ENOMEM;
diff --git a/drivers/gpu/drm/i915/intel_overlay.c b/drivers/gpu/drm/i915/intel_overlay.c
index 6d524a1..bc3721a 100644
--- a/drivers/gpu/drm/i915/intel_overlay.c
+++ b/drivers/gpu/drm/i915/intel_overlay.c
@@ -1341,7 +1341,7 @@ void intel_setup_overlay(struct drm_device *dev)
 		return;
 	overlay->dev = dev;
 
-	reg_bo = drm_gem_object_alloc(dev, PAGE_SIZE);
+	reg_bo = i915_gem_alloc_object(dev, PAGE_SIZE);
 	if (!reg_bo)
 		goto out_free;
 	overlay->reg_bo = to_intel_bo(reg_bo);
-- 
1.6.6.1


------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--

[PATCH 4/6] drm/i915: embed the gem object into drm_i915_gem_object

[Daniel Vetter]

Just embed it and adjust the pointers, No other changes (that's
for later patches).

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/i915/i915_drv.h |    1 +
 drivers/gpu/drm/i915/i915_gem.c |   58 +++++++++++++++++++-------------------
 2 files changed, 30 insertions(+), 29 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 3d4e135..3d59e21 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -632,6 +632,7 @@ typedef struct drm_i915_private {
 
 /** driver private structure attached to each drm_gem_object */
 struct drm_i915_gem_object {
+	struct drm_gem_object base;
 	struct drm_gem_object *obj;
 
 	/** Current space allocated to this object in the GTT, if any. */
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 92dd522..23c67e0 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -4424,37 +4424,38 @@ i915_gem_madvise_ioctl(struct drm_device *dev, void *data,
 struct drm_gem_object * i915_gem_alloc_object(struct drm_device *dev,
 					      size_t size)
 {
-	return drm_gem_object_alloc(dev, size);
-}
+	struct drm_i915_gem_object *obj;
 
-int i915_gem_init_object(struct drm_gem_object *obj)
-{
-	struct drm_i915_gem_object *obj_priv;
+	obj = kzalloc(sizeof(*obj), GFP_KERNEL);
+	if (obj == NULL)
+		return NULL;
 
-	obj_priv = kzalloc(sizeof(*obj_priv), GFP_KERNEL);
-	if (obj_priv == NULL)
-		return -ENOMEM;
+	if (drm_gem_object_init(dev, &obj->base, size) != 0) {
+		kfree(obj);
+		return NULL;
+	}
 
-	/*
-	 * We've just allocated pages from the kernel,
-	 * so they've just been written by the CPU with
-	 * zeros. They'll need to be clflushed before we
-	 * use them with the GPU.
-	 */
-	obj->write_domain = I915_GEM_DOMAIN_CPU;
-	obj->read_domains = I915_GEM_DOMAIN_CPU;
+	obj->base.write_domain = I915_GEM_DOMAIN_CPU;
+	obj->base.read_domains = I915_GEM_DOMAIN_CPU;
 
-	obj_priv->agp_type = AGP_USER_MEMORY;
+	obj->agp_type = AGP_USER_MEMORY;
 
-	obj->driver_private = obj_priv;
-	obj_priv->obj = obj;
-	obj_priv->fence_reg = I915_FENCE_REG_NONE;
-	INIT_LIST_HEAD(&obj_priv->list);
-	INIT_LIST_HEAD(&obj_priv->gpu_write_list);
-	INIT_LIST_HEAD(&obj_priv->fence_list);
-	obj_priv->madv = I915_MADV_WILLNEED;
+	obj->base.driver_private = obj;
+	obj->obj = &obj->base;
+	obj->fence_reg = I915_FENCE_REG_NONE;
+	INIT_LIST_HEAD(&obj->list);
+	INIT_LIST_HEAD(&obj->gpu_write_list);
+	INIT_LIST_HEAD(&obj->fence_list);
+	obj->madv = I915_MADV_WILLNEED;
 
-	trace_i915_gem_object_create(obj);
+	trace_i915_gem_object_create(&obj->base);
+
+	return &obj->base;
+}
+
+int i915_gem_init_object(struct drm_gem_object *obj)
+{
+	BUG();
 
 	return 0;
 }
@@ -4477,12 +4478,11 @@ void i915_gem_free_object(struct drm_gem_object *obj)
 	if (obj_priv->mmap_offset)
 		i915_gem_free_mmap_offset(obj);
 
+	drm_gem_object_release(obj);
+
 	kfree(obj_priv->page_cpu_valid);
 	kfree(obj_priv->bit_17);
-	kfree(obj->driver_private);
-
-	drm_gem_object_release(obj);
-	kfree(obj);
+	kfree(obj_priv);
 }
 
 /** Unbinds all inactive objects. */
-- 
1.6.6.1


------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--

[PATCH 5/6] drm/i915: don't use ->driver_private anymore

[Daniel Vetter]

Thanks to the to_intel_bo helper, this change is rather trivial.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/i915/i915_drv.h |    2 +-
 drivers/gpu/drm/i915/i915_gem.c |    2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 3d59e21..c5c7ce0 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -732,7 +732,7 @@ struct drm_i915_gem_object {
 	atomic_t pending_flip;
 };
 
-#define to_intel_bo(x) ((struct drm_i915_gem_object *) (x)->driver_private)
+#define to_intel_bo(x) container_of(x, struct drm_i915_gem_object, base)
 
 /**
  * Request queue structure.
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 23c67e0..7c8c01b 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -4440,7 +4440,7 @@ struct drm_gem_object * i915_gem_alloc_object(struct drm_device *dev,
 
 	obj->agp_type = AGP_USER_MEMORY;
 
-	obj->base.driver_private = obj;
+	obj->base.driver_private = NULL;
 	obj->obj = &obj->base;
 	obj->fence_reg = I915_FENCE_REG_NONE;
 	INIT_LIST_HEAD(&obj->list);
-- 
1.6.6.1


------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--

[PATCH 6/6] drm/i915: drop pointer to drm_gem_object

[Daniel Vetter]

Luckily the change is quite a little bit less invasive than I've
feared.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/i915/i915_debugfs.c   |   15 +++++++--------
 drivers/gpu/drm/i915/i915_drv.h       |    1 -
 drivers/gpu/drm/i915/i915_gem.c       |   21 ++++++++++-----------
 drivers/gpu/drm/i915/i915_gem_debug.c |    2 +-
 drivers/gpu/drm/i915/i915_irq.c       |    4 ++--
 drivers/gpu/drm/i915/intel_overlay.c  |    6 +++---
 6 files changed, 23 insertions(+), 26 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index a0b8447..213aa3f 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -96,19 +96,18 @@ static int i915_gem_object_list_info(struct seq_file *m, void *data)
 		spin_lock(lock);
 	list_for_each_entry(obj_priv, head, list)
 	{
-		struct drm_gem_object *obj = obj_priv->obj;
-
 		seq_printf(m, "    %p: %s %8zd %08x %08x %d%s%s",
-			   obj,
+			   &obj_priv->base,
 			   get_pin_flag(obj_priv),
-			   obj->size,
-			   obj->read_domains, obj->write_domain,
+			   obj_priv->base.size,
+			   obj_priv->base.read_domains,
+			   obj_priv->base.write_domain,
 			   obj_priv->last_rendering_seqno,
 			   obj_priv->dirty ? " dirty" : "",
 			   obj_priv->madv == I915_MADV_DONTNEED ? " purgeable" : "");
 
-		if (obj->name)
-			seq_printf(m, " (name: %d)", obj->name);
+		if (obj_priv->base.name)
+			seq_printf(m, " (name: %d)", obj_priv->base.name);
 		if (obj_priv->fence_reg != I915_FENCE_REG_NONE)
 			seq_printf(m, " (fence: %d)", obj_priv->fence_reg);
 		if (obj_priv->gtt_space != NULL)
@@ -289,7 +288,7 @@ static int i915_batchbuffer_info(struct seq_file *m, void *data)
 	spin_lock(&dev_priv->mm.active_list_lock);
 
 	list_for_each_entry(obj_priv, &dev_priv->mm.active_list, list) {
-		obj = obj_priv->obj;
+		obj = &obj_priv->base;
 		if (obj->read_domains & I915_GEM_DOMAIN_COMMAND) {
 		    ret = i915_gem_object_get_pages(obj, 0);
 		    if (ret) {
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index c5c7ce0..1b44ca7 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -633,7 +633,6 @@ typedef struct drm_i915_private {
 /** driver private structure attached to each drm_gem_object */
 struct drm_i915_gem_object {
 	struct drm_gem_object base;
-	struct drm_gem_object *obj;
 
 	/** Current space allocated to this object in the GTT, if any. */
 	struct drm_mm_node *gtt_space;
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 7c8c01b..47c46ed 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1566,7 +1566,7 @@ i915_gem_process_flushing_list(struct drm_device *dev,
 	list_for_each_entry_safe(obj_priv, next,
 				 &dev_priv->mm.gpu_write_list,
 				 gpu_write_list) {
-		struct drm_gem_object *obj = obj_priv->obj;
+		struct drm_gem_object *obj = &obj_priv->base;
 
 		if ((obj->write_domain & flush_domains) ==
 		    obj->write_domain) {
@@ -1704,7 +1704,7 @@ i915_gem_retire_request(struct drm_device *dev,
 		obj_priv = list_first_entry(&dev_priv->mm.active_list,
 					    struct drm_i915_gem_object,
 					    list);
-		obj = obj_priv->obj;
+		obj = &obj_priv->base;
 
 		/* If the seqno being retired doesn't match the oldest in the
 		 * list, then the oldest in the list must still be newer than
@@ -2075,7 +2075,7 @@ i915_gem_find_inactive_object(struct drm_device *dev, int min_size)
 
 	/* Try to find the smallest clean object */
 	list_for_each_entry(obj_priv, &dev_priv->mm.inactive_list, list) {
-		struct drm_gem_object *obj = obj_priv->obj;
+		struct drm_gem_object *obj = &obj_priv->base;
 		if (obj->size >= min_size) {
 			if ((!obj_priv->dirty ||
 			     i915_gem_object_is_purgeable(obj_priv)) &&
@@ -2209,7 +2209,7 @@ i915_gem_evict_something(struct drm_device *dev, int min_size)
 
 			/* Find an object that we can immediately reuse */
 			list_for_each_entry(obj_priv, &dev_priv->mm.flushing_list, list) {
-				obj = obj_priv->obj;
+				obj = &obj_priv->base;
 				if (obj->size >= min_size)
 					break;
 
@@ -2437,7 +2437,7 @@ static int i915_find_fence_reg(struct drm_device *dev)
 	i = I915_FENCE_REG_NONE;
 	list_for_each_entry(obj_priv, &dev_priv->mm.fence_list,
 			    fence_list) {
-		obj = obj_priv->obj;
+		obj = &obj_priv->base;
 
 		if (obj_priv->pin_count)
 			continue;
@@ -4441,7 +4441,6 @@ struct drm_gem_object * i915_gem_alloc_object(struct drm_device *dev,
 	obj->agp_type = AGP_USER_MEMORY;
 
 	obj->base.driver_private = NULL;
-	obj->obj = &obj->base;
 	obj->fence_reg = I915_FENCE_REG_NONE;
 	INIT_LIST_HEAD(&obj->list);
 	INIT_LIST_HEAD(&obj->gpu_write_list);
@@ -4495,9 +4494,9 @@ i915_gem_evict_from_inactive_list(struct drm_device *dev)
 		struct drm_gem_object *obj;
 		int ret;
 
-		obj = list_first_entry(&dev_priv->mm.inactive_list,
-				       struct drm_i915_gem_object,
-				       list)->obj;
+		obj = &list_first_entry(&dev_priv->mm.inactive_list,
+					struct drm_i915_gem_object,
+					list)->base;
 
 		ret = i915_gem_object_unbind(obj);
 		if (ret != 0) {
@@ -5111,7 +5110,7 @@ i915_gem_shrink(int nr_to_scan, gfp_t gfp_mask)
 					 &dev_priv->mm.inactive_list,
 					 list) {
 			if (i915_gem_object_is_purgeable(obj_priv)) {
-				i915_gem_object_unbind(obj_priv->obj);
+				i915_gem_object_unbind(&obj_priv->base);
 				if (--nr_to_scan <= 0)
 					break;
 			}
@@ -5140,7 +5139,7 @@ i915_gem_shrink(int nr_to_scan, gfp_t gfp_mask)
 					 &dev_priv->mm.inactive_list,
 					 list) {
 			if (nr_to_scan > 0) {
-				i915_gem_object_unbind(obj_priv->obj);
+				i915_gem_object_unbind(&obj_priv->base);
 				nr_to_scan--;
 			} else
 				cnt++;
diff --git a/drivers/gpu/drm/i915/i915_gem_debug.c b/drivers/gpu/drm/i915/i915_gem_debug.c
index 35507cf..80f380b 100644
--- a/drivers/gpu/drm/i915/i915_gem_debug.c
+++ b/drivers/gpu/drm/i915/i915_gem_debug.c
@@ -39,7 +39,7 @@ i915_verify_inactive(struct drm_device *dev, char *file, int line)
 	struct drm_i915_gem_object *obj_priv;
 
 	list_for_each_entry(obj_priv, &dev_priv->mm.inactive_list, list) {
-		obj = obj_priv->obj;
+		obj = &obj_priv->base;
 		if (obj_priv->pin_count || obj_priv->active ||
 		    (obj->write_domain & ~(I915_GEM_DOMAIN_CPU |
 					   I915_GEM_DOMAIN_GTT)))
diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
index a21c5d4..02b2d6d 100644
--- a/drivers/gpu/drm/i915/i915_irq.c
+++ b/drivers/gpu/drm/i915/i915_irq.c
@@ -608,7 +608,7 @@ static void i915_capture_error_state(struct drm_device *dev)
 	batchbuffer[1] = NULL;
 	count = 0;
 	list_for_each_entry(obj_priv, &dev_priv->mm.active_list, list) {
-		struct drm_gem_object *obj = obj_priv->obj;
+		struct drm_gem_object *obj = &obj_priv->base;
 
 		if (batchbuffer[0] == NULL &&
 		    bbaddr >= obj_priv->gtt_offset &&
@@ -644,7 +644,7 @@ static void i915_capture_error_state(struct drm_device *dev)
 	if (error->active_bo) {
 		int i = 0;
 		list_for_each_entry(obj_priv, &dev_priv->mm.active_list, list) {
-			struct drm_gem_object *obj = obj_priv->obj;
+			struct drm_gem_object *obj = &obj_priv->base;
 
 			error->active_bo[i].size = obj->size;
 			error->active_bo[i].name = obj->name;
diff --git a/drivers/gpu/drm/i915/intel_overlay.c b/drivers/gpu/drm/i915/intel_overlay.c
index bc3721a..b0e17b0 100644
--- a/drivers/gpu/drm/i915/intel_overlay.c
+++ b/drivers/gpu/drm/i915/intel_overlay.c
@@ -373,7 +373,7 @@ static void intel_overlay_off_tail(struct intel_overlay *overlay)
 
 	/* never have the overlay hw on without showing a frame */
 	BUG_ON(!overlay->vid_bo);
-	obj = overlay->vid_bo->obj;
+	obj = &overlay->vid_bo->base;
 
 	i915_gem_object_unpin(obj);
 	drm_gem_object_unreference(obj);
@@ -411,7 +411,7 @@ int intel_overlay_recover_from_interrupt(struct intel_overlay *overlay,
 
 	switch (overlay->hw_wedged) {
 		case RELEASE_OLD_VID:
-			obj = overlay->old_vid_bo->obj;
+			obj = &overlay->old_vid_bo->base;
 			i915_gem_object_unpin(obj);
 			drm_gem_object_unreference(obj);
 			overlay->old_vid_bo = NULL;
@@ -467,7 +467,7 @@ static int intel_overlay_release_old_vid(struct intel_overlay *overlay)
 	if (ret != 0)
 		return ret;
 
-	obj = overlay->old_vid_bo->obj;
+	obj = &overlay->old_vid_bo->base;
 	i915_gem_object_unpin(obj);
 	drm_gem_object_unreference(obj);
 	overlay->old_vid_bo = NULL;
-- 
1.6.6.1


------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--

Re: [PATCH 0/6] make gem_object embedable and convert i915 driver

[Eric Anholt]

[-- Attachment #1.1: Type: text/plain, Size: 1157 bytes --]

On Fri,  9 Apr 2010 21:05:03 +0200, Daniel Vetter <daniel.vetter@ffwll.ch> wrote:
> Hi all,
> 
> As promised here's my patch series to make struct drm_gem_object embedable.
> Also converts drm/i915 as a proof of concept.
> 
> I've already looked at the radeon and nouveau modules and a straightforward
> replament of the gem_object pointer looks simple. But I think embedding it
> into the ttm_bo is better long-term. Stuff like duplicated ref-counting
> between the ttm object and the gem one just doesn't make too much sense.
> 
> Tested on my i945GM. The changes to radeon and nouveau in patch 2 are only
> compile-tested, but the identical change to the i915 survived testing.
> 
> Comments on the patches and my future plans highly welcome.
> 
> Yours, Daniel
> 
> Daniel Vetter (6):
>   drm: extract drm_gem_object_init
>   drm: free core gem object from driver callbacks
>   drm/i915: introduce i915_gem_alloc_object
>   drm/i915: embed the gem object into drm_i915_gem_object
>   drm/i915: don't use ->driver_private anymore
>   drm/i915: drop pointer to drm_gem_object

I like this series.  Dave, should I pull this one?

[-- Attachment #1.2: Type: application/pgp-signature, Size: 197 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: [PATCH 0/6] make gem_object embedable and convert i915 driver

[Daniel Vetter]

On Mon, Apr 12, 2010 at 10:51:20AM -0700, Eric Anholt wrote:
> On Fri,  9 Apr 2010 21:05:03 +0200, Daniel Vetter <daniel.vetter@ffwll.ch> wrote:
> > Daniel Vetter (6):
> >   drm: extract drm_gem_object_init
> >   drm: free core gem object from driver callbacks
> >   drm/i915: introduce i915_gem_alloc_object
> >   drm/i915: embed the gem object into drm_i915_gem_object
> >   drm/i915: don't use ->driver_private anymore
> >   drm/i915: drop pointer to drm_gem_object
> 
> I like this series.  Dave, should I pull this one?

Cool. wrt merging I'd prefer if Dave could take the first two via drm-core.
That way round I could start working on the radeon/nouveau stuff
independently of the i915 stuff. That'd stall i915 slightly but i915 is the
easiest conversion (that's why I did it first) so I can quickly rebase in
case of conflicts.

-Daniel
-- 
Daniel Vetter
Mail: daniel@ffwll.ch
Mobile: +41 (0)79 365 57 48

Re: [PATCH 0/6] make gem_object embedable and convert i915 driver

[Dave Airlie]

On Tue, Apr 13, 2010 at 5:19 AM, Daniel Vetter <daniel@ffwll.ch> wrote:
> On Mon, Apr 12, 2010 at 10:51:20AM -0700, Eric Anholt wrote:
>> On Fri,  9 Apr 2010 21:05:03 +0200, Daniel Vetter <daniel.vetter@ffwll.ch> wrote:
>> > Daniel Vetter (6):
>> >   drm: extract drm_gem_object_init
>> >   drm: free core gem object from driver callbacks
>> >   drm/i915: introduce i915_gem_alloc_object
>> >   drm/i915: embed the gem object into drm_i915_gem_object
>> >   drm/i915: don't use ->driver_private anymore
>> >   drm/i915: drop pointer to drm_gem_object
>>
>> I like this series.  Dave, should I pull this one?
>
> Cool. wrt merging I'd prefer if Dave could take the first two via drm-core.
> That way round I could start working on the radeon/nouveau stuff
> independently of the i915 stuff. That'd stall i915 slightly but i915 is the
> easiest conversion (that's why I did it first) so I can quickly rebase in
> case of conflicts

I'll take these via my tree, Eric just let me know if I can assume your ack
on the i915 ones and even the main one. I'll try and review them over the next
couple of days.

Dave.

Re: [PATCH 0/6] make gem_object embedable and convert i915 driver

[Eric Anholt]

[-- Attachment #1.1: Type: text/plain, Size: 1402 bytes --]

On Tue, 13 Apr 2010 15:33:05 +1000, Dave Airlie <airlied@gmail.com> wrote:
> On Tue, Apr 13, 2010 at 5:19 AM, Daniel Vetter <daniel@ffwll.ch> wrote:
> > On Mon, Apr 12, 2010 at 10:51:20AM -0700, Eric Anholt wrote:
> >> On Fri,  9 Apr 2010 21:05:03 +0200, Daniel Vetter <daniel.vetter@ffwll.ch> wrote:
> >> > Daniel Vetter (6):
> >> >   drm: extract drm_gem_object_init
> >> >   drm: free core gem object from driver callbacks
> >> >   drm/i915: introduce i915_gem_alloc_object
> >> >   drm/i915: embed the gem object into drm_i915_gem_object
> >> >   drm/i915: don't use ->driver_private anymore
> >> >   drm/i915: drop pointer to drm_gem_object
> >>
> >> I like this series.  Dave, should I pull this one?
> >
> > Cool. wrt merging I'd prefer if Dave could take the first two via drm-core.
> > That way round I could start working on the radeon/nouveau stuff
> > independently of the i915 stuff. That'd stall i915 slightly but i915 is the
> > easiest conversion (that's why I did it first) so I can quickly rebase in
> > case of conflicts
> 
> I'll take these via my tree, Eric just let me know if I can assume your ack
> on the i915 ones and even the main one. I'll try and review them over the next
> couple of days.

OK.  And if you felt like pulling my -next at the same time, that would
be awesome as then I can continue merging there based off of this
series.

[-- Attachment #1.2: Type: application/pgp-signature, Size: 197 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx