drm/i915: execbuf2 support

This patch adds a new execbuf ioctl, execbuf2, for use by clients that
want to control fence register allocation more finely.  The buffer
passed in to the new ioctl includes a new relocation type to indicate
whether a given object needs a fence register assigned for the command
buffer in question.

Compatibility with the existing execbuf ioctl is implemented in terms
of the new code, preserving the assumption that fence registers are
required for pre-965 rendering commands.
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
[ickle: Remove pre-emptive clear_fence_reg()]
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
[anholt: Removed dmesg spam]
Signed-off-by: Eric Anholt <eric@anholt.net>

drivers/gpu/drm/i915/i915_dma.c

@@ -813,9 +813,13 @@ static int i915_getparam(struct drm_device *dev, void *data,
 	case I915_PARAM_HAS_PAGEFLIPPING:
 		value = 1;
 		break;
+	case I915_PARAM_HAS_EXECBUF2:
+		/* depends on GEM */
+		value = dev_priv->has_gem;
+		break;
 	default:
 		DRM_DEBUG_DRIVER("Unknown parameter %d\n",
-					param->param);
+				 param->param);
 		return -EINVAL;
 	}
 
@@ -1646,6 +1650,7 @@ struct drm_ioctl_desc i915_ioctls[] = {
 	DRM_IOCTL_DEF(DRM_I915_HWS_ADDR, i915_set_status_page, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
 	DRM_IOCTL_DEF(DRM_I915_GEM_INIT, i915_gem_init_ioctl, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
 	DRM_IOCTL_DEF(DRM_I915_GEM_EXECBUFFER, i915_gem_execbuffer, DRM_AUTH),
+	DRM_IOCTL_DEF(DRM_I915_GEM_EXECBUFFER2, i915_gem_execbuffer2, DRM_AUTH),
 	DRM_IOCTL_DEF(DRM_I915_GEM_PIN, i915_gem_pin_ioctl, DRM_AUTH|DRM_ROOT_ONLY),
 	DRM_IOCTL_DEF(DRM_I915_GEM_UNPIN, i915_gem_unpin_ioctl, DRM_AUTH|DRM_ROOT_ONLY),
 	DRM_IOCTL_DEF(DRM_I915_GEM_BUSY, i915_gem_busy_ioctl, DRM_AUTH),

drivers/gpu/drm/i915/i915_drv.h

@@ -815,6 +815,8 @@ int i915_gem_sw_finish_ioctl(struct drm_device *dev, void *data,
 			     struct drm_file *file_priv);
 int i915_gem_execbuffer(struct drm_device *dev, void *data,
 			struct drm_file *file_priv);
+int i915_gem_execbuffer2(struct drm_device *dev, void *data,
+			 struct drm_file *file_priv);
 int i915_gem_pin_ioctl(struct drm_device *dev, void *data,
 		       struct drm_file *file_priv);
 int i915_gem_unpin_ioctl(struct drm_device *dev, void *data,
@@ -881,6 +883,9 @@ void i915_gem_shrinker_exit(void);
 void i915_gem_detect_bit_6_swizzle(struct drm_device *dev);
 void i915_gem_object_do_bit_17_swizzle(struct drm_gem_object *obj);
 void i915_gem_object_save_bit_17_swizzle(struct drm_gem_object *obj);
+bool i915_tiling_ok(struct drm_device *dev, int stride, int size,
+		    int tiling_mode);
+bool i915_obj_fenceable(struct drm_device *dev, struct drm_gem_object *obj);
 
 /* i915_gem_debug.c */
 void i915_gem_dump_object(struct drm_gem_object *obj, int len,

drivers/gpu/drm/i915/i915_gem_tiling.c

@@ -304,35 +304,39 @@ i915_gem_detect_bit_6_swizzle(struct drm_device *dev)
 
 
 /**
- * Returns the size of the fence for a tiled object of the given size.
+ * Returns whether an object is currently fenceable.  If not, it may need
+ * to be unbound and have its pitch adjusted.
  */
-static int
-i915_get_fence_size(struct drm_device *dev, int size)
+bool
+i915_obj_fenceable(struct drm_device *dev, struct drm_gem_object *obj)
 {
-	int i;
-	int start;
+	struct drm_i915_gem_object *obj_priv = obj->driver_private;
 
 	if (IS_I965G(dev)) {
 		/* The 965 can have fences at any page boundary. */
-		return ALIGN(size, 4096);
+		if (obj->size & 4095)
+			return false;
+		return true;
+	} else if (IS_I9XX(dev)) {
+		if (obj_priv->gtt_offset & ~I915_FENCE_START_MASK)
+			return false;
 	} else {
-		/* Align the size to a power of two greater than the smallest
-		 * fence size.
-		 */
-		if (IS_I9XX(dev))
-			start = 1024 * 1024;
-		else
-			start = 512 * 1024;
+		if (obj_priv->gtt_offset & ~I830_FENCE_START_MASK)
+			return false;
+	}
 
-		for (i = start; i < size; i <<= 1)
-			;
+	/* Power of two sized... */
+	if (obj->size & (obj->size - 1))
+		return false;
 
-		return i;
-	}
+	/* Objects must be size aligned as well */
+	if (obj_priv->gtt_offset & (obj->size - 1))
+		return false;
+	return true;
 }
 
 /* Check pitch constriants for all chips & tiling formats */
-static bool
+bool
 i915_tiling_ok(struct drm_device *dev, int stride, int size, int tiling_mode)
 {
 	int tile_width;
@@ -384,12 +388,6 @@ i915_tiling_ok(struct drm_device *dev, int stride, int size, int tiling_mode)
 	if (stride & (stride - 1))
 		return false;
 
-	/* We don't 0handle the aperture area covered by the fence being bigger
-	 * than the object size.
-	 */
-	if (i915_get_fence_size(dev, size) != size)
-		return false;
-
 	return true;
 }
 

include/drm/i915_drm.h

@@ -188,6 +188,7 @@ typedef struct _drm_i915_sarea {
 #define DRM_I915_GEM_MADVISE	0x26
 #define DRM_I915_OVERLAY_PUT_IMAGE	0x27
 #define DRM_I915_OVERLAY_ATTRS	0x28
+#define DRM_I915_GEM_EXECBUFFER2	0x29
 
 #define DRM_IOCTL_I915_INIT		DRM_IOW( DRM_COMMAND_BASE + DRM_I915_INIT, drm_i915_init_t)
 #define DRM_IOCTL_I915_FLUSH		DRM_IO ( DRM_COMMAND_BASE + DRM_I915_FLUSH)
@@ -207,6 +208,7 @@ typedef struct _drm_i915_sarea {
 #define DRM_IOCTL_I915_VBLANK_SWAP	DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_VBLANK_SWAP, drm_i915_vblank_swap_t)
 #define DRM_IOCTL_I915_GEM_INIT		DRM_IOW(DRM_COMMAND_BASE + DRM_I915_GEM_INIT, struct drm_i915_gem_init)
 #define DRM_IOCTL_I915_GEM_EXECBUFFER	DRM_IOW(DRM_COMMAND_BASE + DRM_I915_GEM_EXECBUFFER, struct drm_i915_gem_execbuffer)
+#define DRM_IOCTL_I915_GEM_EXECBUFFER2	DRM_IOW(DRM_COMMAND_BASE + DRM_I915_GEM_EXECBUFFER2, struct drm_i915_gem_execbuffer2)
 #define DRM_IOCTL_I915_GEM_PIN		DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_GEM_PIN, struct drm_i915_gem_pin)
 #define DRM_IOCTL_I915_GEM_UNPIN	DRM_IOW(DRM_COMMAND_BASE + DRM_I915_GEM_UNPIN, struct drm_i915_gem_unpin)
 #define DRM_IOCTL_I915_GEM_BUSY		DRM_IOWR(DRM_COMMAND_BASE + DRM_I915_GEM_BUSY, struct drm_i915_gem_busy)
@@ -272,6 +274,7 @@ typedef struct drm_i915_irq_wait {
 #define I915_PARAM_NUM_FENCES_AVAIL      6
 #define I915_PARAM_HAS_OVERLAY           7
 #define I915_PARAM_HAS_PAGEFLIPPING	 8
+#define I915_PARAM_HAS_EXECBUF2          9
 
 typedef struct drm_i915_getparam {
 	int param;
@@ -567,6 +570,57 @@ struct drm_i915_gem_execbuffer {
 	__u64 cliprects_ptr;
 };
 
+struct drm_i915_gem_exec_object2 {
+	/**
+	 * User's handle for a buffer to be bound into the GTT for this
+	 * operation.
+	 */
+	__u32 handle;
+
+	/** Number of relocations to be performed on this buffer */
+	__u32 relocation_count;
+	/**
+	 * Pointer to array of struct drm_i915_gem_relocation_entry containing
+	 * the relocations to be performed in this buffer.
+	 */
+	__u64 relocs_ptr;
+
+	/** Required alignment in graphics aperture */
+	__u64 alignment;
+
+	/**
+	 * Returned value of the updated offset of the object, for future
+	 * presumed_offset writes.
+	 */
+	__u64 offset;
+
+#define EXEC_OBJECT_NEEDS_FENCE (1<<0)
+	__u64 flags;
+	__u64 rsvd1;
+	__u64 rsvd2;
+};
+
+struct drm_i915_gem_execbuffer2 {
+	/**
+	 * List of gem_exec_object2 structs
+	 */
+	__u64 buffers_ptr;
+	__u32 buffer_count;
+
+	/** Offset in the batchbuffer to start execution from. */
+	__u32 batch_start_offset;
+	/** Bytes used in batchbuffer from batch_start_offset */
+	__u32 batch_len;
+	__u32 DR1;
+	__u32 DR4;
+	__u32 num_cliprects;
+	/** This is a struct drm_clip_rect *cliprects */
+	__u64 cliprects_ptr;
+	__u64 flags; /* currently unused */
+	__u64 rsvd1;
+	__u64 rsvd2;
+};
+
 struct drm_i915_gem_pin {
 	/** Handle of the buffer to be pinned. */
 	__u32 handle;