[Intel-gfx] [PATCH 1/3] drm/i915: Support for pre-populating the object with system pages

[ankitprasad.r.sharma at intel.com]

From: Ankitprasad Sharma <ankitprasad.r.sharma at intel.com>

This patch provides support for the User to populate the object
with system pages at its creation time. Since this can be safely
performed without holding the 'struct_mutex', it would help to reduce
the time 'struct_mutex' is kept locked especially during the exec-buffer
path, where it is generally held for the longest time.

v2: Corrected error handling on pre-populate failure (Arun),
wrap list_add_tail() in __i915_gem_object_get_pages__tail_locked(),
moved error handling to the caller (Chris)

Signed-off-by: Ankitprasad Sharma <ankitprasad.r.sharma at intel.com>
---
 drivers/gpu/drm/i915/i915_dma.c |  2 +-
 drivers/gpu/drm/i915/i915_gem.c | 50 +++++++++++++++++++++++++++++++++++------
 include/uapi/drm/i915_drm.h     | 11 +++++----
 3 files changed, 49 insertions(+), 14 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
index 8319e07..955aa16 100644
--- a/drivers/gpu/drm/i915/i915_dma.c
+++ b/drivers/gpu/drm/i915/i915_dma.c
@@ -171,7 +171,7 @@ static int i915_getparam(struct drm_device *dev, void *data,
 		value = HAS_RESOURCE_STREAMER(dev);
 		break;
 	case I915_PARAM_CREATE_VERSION:
-		value = 2;
+		value = 3;
 		break;
 	default:
 		DRM_DEBUG("Unknown parameter %d\n", param->param);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index c44bd05..4430128 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -46,6 +46,9 @@ static void
 i915_gem_object_retire__write(struct drm_i915_gem_object *obj);
 static void
 i915_gem_object_retire__read(struct drm_i915_gem_object *obj, int ring);
+static int __i915_gem_object_get_pages(struct drm_i915_gem_object *obj);
+static void
+__i915_gem_object_get_pages__tail_locked(struct drm_i915_gem_object *obj);
 
 static bool cpu_cache_is_coherent(struct drm_device *dev,
 				  enum i915_cache_level level)
@@ -414,6 +417,20 @@ i915_gem_create(struct drm_file *file,
 	if (obj == NULL)
 		return -ENOMEM;
 
+	if (flags & I915_CREATE_POPULATE) {
+		struct drm_i915_private *dev_priv = dev->dev_private;
+
+		ret = __i915_gem_object_get_pages(obj);
+		if (ret) {
+			drm_gem_object_unreference_unlocked(&obj->base);
+			return ret;
+		}
+
+		mutex_lock(&dev->struct_mutex);
+		__i915_gem_object_get_pages__tail_locked(obj);
+		mutex_unlock(&dev->struct_mutex);
+	}
+
 	ret = drm_gem_handle_create(file, &obj->base, &handle);
 	/* drop reference from allocate - handle holds it now */
 	drm_gem_object_unreference_unlocked(&obj->base);
@@ -2328,6 +2345,30 @@ err_pages:
 	return ret;
 }
 
+static int
+__i915_gem_object_get_pages(struct drm_i915_gem_object *obj)
+{
+	const struct drm_i915_gem_object_ops *ops = obj->ops;
+	int ret;
+
+	ret = ops->get_pages(obj);
+	if (ret)
+		return ret;
+
+	obj->get_page.sg = obj->pages->sgl;
+	obj->get_page.last = 0;
+
+	return 0;
+}
+
+static void
+__i915_gem_object_get_pages__tail_locked(struct drm_i915_gem_object *obj)
+{
+	struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
+
+	list_add_tail(&obj->global_list, &dev_priv->mm.unbound_list);
+}
+
 /* Ensure that the associated pages are gathered from the backing storage
  * and pinned into our object. i915_gem_object_get_pages() may be called
  * multiple times before they are released by a single call to
@@ -2338,8 +2379,6 @@ err_pages:
 int
 i915_gem_object_get_pages(struct drm_i915_gem_object *obj)
 {
-	struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
-	const struct drm_i915_gem_object_ops *ops = obj->ops;
 	int ret;
 
 	if (obj->pages)
@@ -2352,14 +2391,11 @@ i915_gem_object_get_pages(struct drm_i915_gem_object *obj)
 
 	BUG_ON(obj->pages_pin_count);
 
-	ret = ops->get_pages(obj);
+	ret = __i915_gem_object_get_pages(obj);
 	if (ret)
 		return ret;
 
-	list_add_tail(&obj->global_list, &dev_priv->mm.unbound_list);
-
-	obj->get_page.sg = obj->pages->sgl;
-	obj->get_page.last = 0;
+	__i915_gem_object_get_pages__tail_locked(obj);
 
 	return 0;
 }
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index f71f75c..26ea715 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -457,20 +457,19 @@ struct drm_i915_gem_create {
 	__u32 handle;
 	__u32 pad;
 	/**
-	 * Requested flags (currently used for placement
-	 * (which memory domain))
+	 * Requested flags
 	 *
 	 * You can request that the object be created from special memory
 	 * rather than regular system pages using this parameter. Such
 	 * irregular objects may have certain restrictions (such as CPU
 	 * access to a stolen object is verboten).
-	 *
-	 * This can be used in the future for other purposes too
-	 * e.g. specifying tiling/caching/madvise
+	 * Also using this parameter object can be pre-populated with system
+	 * pages.
 	 */
 	__u32 flags;
 #define I915_CREATE_PLACEMENT_STOLEN 	(1<<0) /* Cannot use CPU mmaps */
-#define __I915_CREATE_UNKNOWN_FLAGS	-(I915_CREATE_PLACEMENT_STOLEN << 1)
+#define I915_CREATE_POPULATE		(1<<1) /* Pre-populate object pages */
+#define __I915_CREATE_UNKNOWN_FLAGS	-(I915_CREATE_POPULATE << 1)
 };
 
 struct drm_i915_gem_pread {
-- 
1.9.1