diff mbox

[10/28] vmwgfx: Major KMS refactoring / cleanup in preparation of screen targets

Message ID 1439444597-4664-11-git-send-email-thellstrom@vmware.com (mailing list archive)
State New, archived
Headers show

Commit Message

Thomas Hellstrom Aug. 13, 2015, 5:42 a.m. UTC
From: Sinclair Yeh <syeh@vmware.com>

Signed-off-by: Sinclair Yeh <syeh@vmware.com>
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
---
 drivers/gpu/drm/vmwgfx/vmwgfx_drv.h     |  22 +-
 drivers/gpu/drm/vmwgfx/vmwgfx_fifo.c    |   4 +-
 drivers/gpu/drm/vmwgfx/vmwgfx_kms.c     | 562 +++++---------------------------
 drivers/gpu/drm/vmwgfx/vmwgfx_kms.h     |  80 +++--
 drivers/gpu/drm/vmwgfx/vmwgfx_ldu.c     |  45 ++-
 drivers/gpu/drm/vmwgfx/vmwgfx_overlay.c |   6 +-
 drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c    | 424 ++++++++++++++++++++++--
 7 files changed, 597 insertions(+), 546 deletions(-)
diff mbox

Patch

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
index c300a0a..b65eb02 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
@@ -1,6 +1,6 @@ 
 /**************************************************************************
  *
- * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * Copyright © 2009-2014 VMware, Inc., Palo Alto, CA., USA
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -330,6 +330,17 @@  struct vmw_ctx_binding_state {
 	struct vmw_ctx_binding shaders[SVGA3D_SHADERTYPE_MAX];
 };
 
+
+/*
+ * enum vmw_display_unit_type - Describes the display unit
+ */
+enum vmw_display_unit_type {
+	vmw_du_invalid = 0,
+	vmw_du_legacy,
+	vmw_du_screen_object
+};
+
+
 struct vmw_sw_context{
 	struct drm_open_hash res_ht;
 	bool res_ht_initialized;
@@ -421,6 +432,7 @@  struct vmw_private {
 	 */
 
 	void *fb_info;
+	enum vmw_display_unit_type active_display_unit;
 	struct vmw_legacy_display *ldu_priv;
 	struct vmw_screen_object_display *sou_priv;
 	struct vmw_overlay *overlay_priv;
@@ -844,8 +856,8 @@  extern void vmw_execbuf_copy_fence_user(struct vmw_private *dev_priv,
 
 extern irqreturn_t vmw_irq_handler(int irq, void *arg);
 extern int vmw_wait_seqno(struct vmw_private *dev_priv, bool lazy,
-			     uint32_t seqno, bool interruptible,
-			     unsigned long timeout);
+			  uint32_t seqno, bool interruptible,
+			  unsigned long timeout);
 extern void vmw_irq_preinstall(struct drm_device *dev);
 extern int vmw_irq_postinstall(struct drm_device *dev);
 extern void vmw_irq_uninstall(struct drm_device *dev);
@@ -876,9 +888,9 @@  extern void vmw_generic_waiter_remove(struct vmw_private *dev_priv,
 extern void vmw_marker_queue_init(struct vmw_marker_queue *queue);
 extern void vmw_marker_queue_takedown(struct vmw_marker_queue *queue);
 extern int vmw_marker_push(struct vmw_marker_queue *queue,
-			  uint32_t seqno);
+			   uint32_t seqno);
 extern int vmw_marker_pull(struct vmw_marker_queue *queue,
-			  uint32_t signaled_seqno);
+			   uint32_t signaled_seqno);
 extern int vmw_wait_lag(struct vmw_private *dev_priv,
 			struct vmw_marker_queue *queue, uint32_t us);
 
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_fifo.c b/drivers/gpu/drm/vmwgfx/vmwgfx_fifo.c
index 189102d..239815c 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_fifo.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_fifo.c
@@ -71,8 +71,8 @@  bool vmw_fifo_have_3d(struct vmw_private *dev_priv)
 	if (hwversion < SVGA3D_HWVERSION_WS8_B1)
 		return false;
 
-	/* Non-Screen Object path does not support surfaces */
-	if (!dev_priv->sou_priv)
+	/* Legacy Display Unit does not support surfaces */
+	if (dev_priv->active_display_unit == vmw_du_legacy)
 		return false;
 
 	return true;
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c
index b5632c2..cac17c2 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c
@@ -1,6 +1,6 @@ 
 /**************************************************************************
  *
- * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * Copyright © 2009-2014 VMware, Inc., Palo Alto, CA., USA
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -32,15 +32,12 @@ 
 #define VMWGFX_PRESENT_RATE ((HZ / 60 > 0) ? HZ / 60 : 1)
 
 
-struct vmw_clip_rect {
-	int x1, x2, y1, y2;
-};
 
 /**
  * Clip @num_rects number of @rects against @clip storing the
  * results in @out_rects and the number of passed rects in @out_num.
  */
-static void vmw_clip_cliprects(struct drm_clip_rect *rects,
+void vmw_clip_cliprects(struct drm_clip_rect *rects,
 			int num_rects,
 			struct vmw_clip_rect clip,
 			SVGASignedRect *out_rects,
@@ -69,7 +66,7 @@  static void vmw_clip_cliprects(struct drm_clip_rect *rects,
 	*out_num = k;
 }
 
-void vmw_display_unit_cleanup(struct vmw_display_unit *du)
+void vmw_du_cleanup(struct vmw_display_unit *du)
 {
 	if (du->cursor_surface)
 		vmw_surface_unreference(&du->cursor_surface);
@@ -367,15 +364,6 @@  void vmw_kms_cursor_snoop(struct vmw_surface *srf,
 
 	srf->snooper.age++;
 
-	/* we can't call this function from this function since execbuf has
-	 * reserved fifo space.
-	 *
-	 * if (srf->snooper.crtc)
-	 *	vmw_ldu_crtc_cursor_update_image(dev_priv,
-	 *					 srf->snooper.image, 64, 64,
-	 *					 du->hotspot_x, du->hotspot_y);
-	 */
-
 	ttm_bo_kunmap(&map);
 err_unreserve:
 	ttm_bo_unreserve(bo);
@@ -412,17 +400,6 @@  void vmw_kms_cursor_post_execbuf(struct vmw_private *dev_priv)
  * Surface framebuffer code
  */
 
-#define vmw_framebuffer_to_vfbs(x) \
-	container_of(x, struct vmw_framebuffer_surface, base.base)
-
-struct vmw_framebuffer_surface {
-	struct vmw_framebuffer base;
-	struct vmw_surface *surface;
-	struct vmw_dma_buffer *buffer;
-	struct list_head head;
-	struct drm_master *master;
-};
-
 static void vmw_framebuffer_surface_destroy(struct drm_framebuffer *framebuffer)
 {
 	struct vmw_framebuffer_surface *vfbs =
@@ -442,153 +419,6 @@  static void vmw_framebuffer_surface_destroy(struct drm_framebuffer *framebuffer)
 	kfree(vfbs);
 }
 
-static int do_surface_dirty_sou(struct vmw_private *dev_priv,
-				struct drm_file *file_priv,
-				struct vmw_framebuffer *framebuffer,
-				unsigned flags, unsigned color,
-				struct drm_clip_rect *clips,
-				unsigned num_clips, int inc,
-				struct vmw_fence_obj **out_fence)
-{
-	struct vmw_display_unit *units[VMWGFX_NUM_DISPLAY_UNITS];
-	struct drm_clip_rect *clips_ptr;
-	struct drm_clip_rect *tmp;
-	struct drm_crtc *crtc;
-	size_t fifo_size;
-	int i, num_units;
-	int ret = 0; /* silence warning */
-	int left, right, top, bottom;
-
-	struct {
-		SVGA3dCmdHeader header;
-		SVGA3dCmdBlitSurfaceToScreen body;
-	} *cmd;
-	SVGASignedRect *blits;
-
-	num_units = 0;
-	list_for_each_entry(crtc, &dev_priv->dev->mode_config.crtc_list,
-			    head) {
-		if (crtc->primary->fb != &framebuffer->base)
-			continue;
-		units[num_units++] = vmw_crtc_to_du(crtc);
-	}
-
-	BUG_ON(!clips || !num_clips);
-
-	tmp = kzalloc(sizeof(*tmp) * num_clips, GFP_KERNEL);
-	if (unlikely(tmp == NULL)) {
-		DRM_ERROR("Temporary cliprect memory alloc failed.\n");
-		return -ENOMEM;
-	}
-
-	fifo_size = sizeof(*cmd) + sizeof(SVGASignedRect) * num_clips;
-	cmd = kzalloc(fifo_size, GFP_KERNEL);
-	if (unlikely(cmd == NULL)) {
-		DRM_ERROR("Temporary fifo memory alloc failed.\n");
-		ret = -ENOMEM;
-		goto out_free_tmp;
-	}
-
-	/* setup blits pointer */
-	blits = (SVGASignedRect *)&cmd[1];
-
-	/* initial clip region */
-	left = clips->x1;
-	right = clips->x2;
-	top = clips->y1;
-	bottom = clips->y2;
-
-	/* skip the first clip rect */
-	for (i = 1, clips_ptr = clips + inc;
-	     i < num_clips; i++, clips_ptr += inc) {
-		left = min_t(int, left, (int)clips_ptr->x1);
-		right = max_t(int, right, (int)clips_ptr->x2);
-		top = min_t(int, top, (int)clips_ptr->y1);
-		bottom = max_t(int, bottom, (int)clips_ptr->y2);
-	}
-
-	/* only need to do this once */
-	cmd->header.id = cpu_to_le32(SVGA_3D_CMD_BLIT_SURFACE_TO_SCREEN);
-	cmd->header.size = cpu_to_le32(fifo_size - sizeof(cmd->header));
-
-	cmd->body.srcRect.left = left;
-	cmd->body.srcRect.right = right;
-	cmd->body.srcRect.top = top;
-	cmd->body.srcRect.bottom = bottom;
-
-	clips_ptr = clips;
-	for (i = 0; i < num_clips; i++, clips_ptr += inc) {
-		tmp[i].x1 = clips_ptr->x1 - left;
-		tmp[i].x2 = clips_ptr->x2 - left;
-		tmp[i].y1 = clips_ptr->y1 - top;
-		tmp[i].y2 = clips_ptr->y2 - top;
-	}
-
-	/* do per unit writing, reuse fifo for each */
-	for (i = 0; i < num_units; i++) {
-		struct vmw_display_unit *unit = units[i];
-		struct vmw_clip_rect clip;
-		int num;
-
-		clip.x1 = left - unit->crtc.x;
-		clip.y1 = top - unit->crtc.y;
-		clip.x2 = right - unit->crtc.x;
-		clip.y2 = bottom - unit->crtc.y;
-
-		/* skip any crtcs that misses the clip region */
-		if (clip.x1 >= unit->crtc.mode.hdisplay ||
-		    clip.y1 >= unit->crtc.mode.vdisplay ||
-		    clip.x2 <= 0 || clip.y2 <= 0)
-			continue;
-
-		/*
-		 * In order for the clip rects to be correctly scaled
-		 * the src and dest rects needs to be the same size.
-		 */
-		cmd->body.destRect.left = clip.x1;
-		cmd->body.destRect.right = clip.x2;
-		cmd->body.destRect.top = clip.y1;
-		cmd->body.destRect.bottom = clip.y2;
-
-		/* create a clip rect of the crtc in dest coords */
-		clip.x2 = unit->crtc.mode.hdisplay - clip.x1;
-		clip.y2 = unit->crtc.mode.vdisplay - clip.y1;
-		clip.x1 = 0 - clip.x1;
-		clip.y1 = 0 - clip.y1;
-
-		/* need to reset sid as it is changed by execbuf */
-		cmd->body.srcImage.sid = cpu_to_le32(framebuffer->user_handle);
-		cmd->body.destScreenId = unit->unit;
-
-		/* clip and write blits to cmd stream */
-		vmw_clip_cliprects(tmp, num_clips, clip, blits, &num);
-
-		/* if no cliprects hit skip this */
-		if (num == 0)
-			continue;
-
-		/* only return the last fence */
-		if (out_fence && *out_fence)
-			vmw_fence_obj_unreference(out_fence);
-
-		/* recalculate package length */
-		fifo_size = sizeof(*cmd) + sizeof(SVGASignedRect) * num;
-		cmd->header.size = cpu_to_le32(fifo_size - sizeof(cmd->header));
-		ret = vmw_execbuf_process(file_priv, dev_priv, NULL, cmd,
-					  fifo_size, 0, NULL, out_fence);
-
-		if (unlikely(ret != 0))
-			break;
-	}
-
-
-	kfree(cmd);
-out_free_tmp:
-	kfree(tmp);
-
-	return ret;
-}
-
 static int vmw_framebuffer_surface_dirty(struct drm_framebuffer *framebuffer,
 				  struct drm_file *file_priv,
 				  unsigned flags, unsigned color,
@@ -604,8 +434,8 @@  static int vmw_framebuffer_surface_dirty(struct drm_framebuffer *framebuffer,
 	if (unlikely(vfbs->master != file_priv->master))
 		return -EINVAL;
 
-	/* Require ScreenObject support for 3D */
-	if (!dev_priv->sou_priv)
+	/* Legacy Display Unit does not support 3D */
+	if (dev_priv->active_display_unit == vmw_du_legacy)
 		return -EINVAL;
 
 	drm_modeset_lock_all(dev_priv->dev);
@@ -627,9 +457,12 @@  static int vmw_framebuffer_surface_dirty(struct drm_framebuffer *framebuffer,
 		inc = 2; /* skip source rects */
 	}
 
-	ret = do_surface_dirty_sou(dev_priv, file_priv, &vfbs->base,
-				   flags, color,
-				   clips, num_clips, inc, NULL);
+	if (dev_priv->active_display_unit == vmw_du_screen_object)
+		ret = vmw_kms_sou_do_surface_dirty(dev_priv, file_priv,
+						   &vfbs->base,
+						   flags, color,
+						   clips, num_clips,
+						   inc, NULL);
 
 	vmw_fifo_flush(dev_priv, false);
 	ttm_read_unlock(&dev_priv->reservation_sem);
@@ -658,8 +491,8 @@  static int vmw_kms_new_framebuffer_surface(struct vmw_private *dev_priv,
 	struct vmw_master *vmaster = vmw_master(file_priv->master);
 	int ret;
 
-	/* 3D is only supported on HWv8 hosts which supports screen objects */
-	if (!dev_priv->sou_priv)
+	/* 3D is only supported on HWv8 and newer hosts */
+	if (dev_priv->active_display_unit == vmw_du_legacy)
 		return -ENOSYS;
 
 	/*
@@ -693,9 +526,6 @@  static int vmw_kms_new_framebuffer_surface(struct vmw_private *dev_priv,
 	case 15:
 		format = SVGA3D_A1R5G5B5;
 		break;
-	case 8:
-		format = SVGA3D_LUMINANCE8;
-		break;
 	default:
 		DRM_ERROR("Invalid color depth: %d\n", mode_cmd->depth);
 		return -EINVAL;
@@ -753,14 +583,6 @@  out_err1:
  * Dmabuf framebuffer code
  */
 
-#define vmw_framebuffer_to_vfbd(x) \
-	container_of(x, struct vmw_framebuffer_dmabuf, base.base)
-
-struct vmw_framebuffer_dmabuf {
-	struct vmw_framebuffer base;
-	struct vmw_dma_buffer *buffer;
-};
-
 static void vmw_framebuffer_dmabuf_destroy(struct drm_framebuffer *framebuffer)
 {
 	struct vmw_framebuffer_dmabuf *vfbd =
@@ -773,180 +595,6 @@  static void vmw_framebuffer_dmabuf_destroy(struct drm_framebuffer *framebuffer)
 	kfree(vfbd);
 }
 
-static int do_dmabuf_dirty_ldu(struct vmw_private *dev_priv,
-			       struct vmw_framebuffer *framebuffer,
-			       unsigned flags, unsigned color,
-			       struct drm_clip_rect *clips,
-			       unsigned num_clips, int increment)
-{
-	size_t fifo_size;
-	int i;
-
-	struct {
-		uint32_t header;
-		SVGAFifoCmdUpdate body;
-	} *cmd;
-
-	fifo_size = sizeof(*cmd) * num_clips;
-	cmd = vmw_fifo_reserve(dev_priv, fifo_size);
-	if (unlikely(cmd == NULL)) {
-		DRM_ERROR("Fifo reserve failed.\n");
-		return -ENOMEM;
-	}
-
-	memset(cmd, 0, fifo_size);
-	for (i = 0; i < num_clips; i++, clips += increment) {
-		cmd[i].header = cpu_to_le32(SVGA_CMD_UPDATE);
-		cmd[i].body.x = cpu_to_le32(clips->x1);
-		cmd[i].body.y = cpu_to_le32(clips->y1);
-		cmd[i].body.width = cpu_to_le32(clips->x2 - clips->x1);
-		cmd[i].body.height = cpu_to_le32(clips->y2 - clips->y1);
-	}
-
-	vmw_fifo_commit(dev_priv, fifo_size);
-	return 0;
-}
-
-static int do_dmabuf_define_gmrfb(struct drm_file *file_priv,
-				  struct vmw_private *dev_priv,
-				  struct vmw_framebuffer *framebuffer)
-{
-	int depth = framebuffer->base.depth;
-	size_t fifo_size;
-	int ret;
-
-	struct {
-		uint32_t header;
-		SVGAFifoCmdDefineGMRFB body;
-	} *cmd;
-
-	/* Emulate RGBA support, contrary to svga_reg.h this is not
-	 * supported by hosts. This is only a problem if we are reading
-	 * this value later and expecting what we uploaded back.
-	 */
-	if (depth == 32)
-		depth = 24;
-
-	fifo_size = sizeof(*cmd);
-	cmd = kmalloc(fifo_size, GFP_KERNEL);
-	if (unlikely(cmd == NULL)) {
-		DRM_ERROR("Failed to allocate temporary cmd buffer.\n");
-		return -ENOMEM;
-	}
-
-	memset(cmd, 0, fifo_size);
-	cmd->header = SVGA_CMD_DEFINE_GMRFB;
-	cmd->body.format.bitsPerPixel = framebuffer->base.bits_per_pixel;
-	cmd->body.format.colorDepth = depth;
-	cmd->body.format.reserved = 0;
-	cmd->body.bytesPerLine = framebuffer->base.pitches[0];
-	cmd->body.ptr.gmrId = framebuffer->user_handle;
-	cmd->body.ptr.offset = 0;
-
-	ret = vmw_execbuf_process(file_priv, dev_priv, NULL, cmd,
-				  fifo_size, 0, NULL, NULL);
-
-	kfree(cmd);
-
-	return ret;
-}
-
-static int do_dmabuf_dirty_sou(struct drm_file *file_priv,
-			       struct vmw_private *dev_priv,
-			       struct vmw_framebuffer *framebuffer,
-			       unsigned flags, unsigned color,
-			       struct drm_clip_rect *clips,
-			       unsigned num_clips, int increment,
-			       struct vmw_fence_obj **out_fence)
-{
-	struct vmw_display_unit *units[VMWGFX_NUM_DISPLAY_UNITS];
-	struct drm_clip_rect *clips_ptr;
-	int i, k, num_units, ret;
-	struct drm_crtc *crtc;
-	size_t fifo_size;
-
-	struct {
-		uint32_t header;
-		SVGAFifoCmdBlitGMRFBToScreen body;
-	} *blits;
-
-	ret = do_dmabuf_define_gmrfb(file_priv, dev_priv, framebuffer);
-	if (unlikely(ret != 0))
-		return ret; /* define_gmrfb prints warnings */
-
-	fifo_size = sizeof(*blits) * num_clips;
-	blits = kmalloc(fifo_size, GFP_KERNEL);
-	if (unlikely(blits == NULL)) {
-		DRM_ERROR("Failed to allocate temporary cmd buffer.\n");
-		return -ENOMEM;
-	}
-
-	num_units = 0;
-	list_for_each_entry(crtc, &dev_priv->dev->mode_config.crtc_list, head) {
-		if (crtc->primary->fb != &framebuffer->base)
-			continue;
-		units[num_units++] = vmw_crtc_to_du(crtc);
-	}
-
-	for (k = 0; k < num_units; k++) {
-		struct vmw_display_unit *unit = units[k];
-		int hit_num = 0;
-
-		clips_ptr = clips;
-		for (i = 0; i < num_clips; i++, clips_ptr += increment) {
-			int clip_x1 = clips_ptr->x1 - unit->crtc.x;
-			int clip_y1 = clips_ptr->y1 - unit->crtc.y;
-			int clip_x2 = clips_ptr->x2 - unit->crtc.x;
-			int clip_y2 = clips_ptr->y2 - unit->crtc.y;
-			int move_x, move_y;
-
-			/* skip any crtcs that misses the clip region */
-			if (clip_x1 >= unit->crtc.mode.hdisplay ||
-			    clip_y1 >= unit->crtc.mode.vdisplay ||
-			    clip_x2 <= 0 || clip_y2 <= 0)
-				continue;
-
-			/* clip size to crtc size */
-			clip_x2 = min_t(int, clip_x2, unit->crtc.mode.hdisplay);
-			clip_y2 = min_t(int, clip_y2, unit->crtc.mode.vdisplay);
-
-			/* translate both src and dest to bring clip into screen */
-			move_x = min_t(int, clip_x1, 0);
-			move_y = min_t(int, clip_y1, 0);
-
-			/* actual translate done here */
-			blits[hit_num].header = SVGA_CMD_BLIT_GMRFB_TO_SCREEN;
-			blits[hit_num].body.destScreenId = unit->unit;
-			blits[hit_num].body.srcOrigin.x = clips_ptr->x1 - move_x;
-			blits[hit_num].body.srcOrigin.y = clips_ptr->y1 - move_y;
-			blits[hit_num].body.destRect.left = clip_x1 - move_x;
-			blits[hit_num].body.destRect.top = clip_y1 - move_y;
-			blits[hit_num].body.destRect.right = clip_x2;
-			blits[hit_num].body.destRect.bottom = clip_y2;
-			hit_num++;
-		}
-
-		/* no clips hit the crtc */
-		if (hit_num == 0)
-			continue;
-
-		/* only return the last fence */
-		if (out_fence && *out_fence)
-			vmw_fence_obj_unreference(out_fence);
-
-		fifo_size = sizeof(*blits) * hit_num;
-		ret = vmw_execbuf_process(file_priv, dev_priv, NULL, blits,
-					  fifo_size, 0, NULL, out_fence);
-
-		if (unlikely(ret != 0))
-			break;
-	}
-
-	kfree(blits);
-
-	return ret;
-}
-
 static int vmw_framebuffer_dmabuf_dirty(struct drm_framebuffer *framebuffer,
 				 struct drm_file *file_priv,
 				 unsigned flags, unsigned color,
@@ -979,13 +627,15 @@  static int vmw_framebuffer_dmabuf_dirty(struct drm_framebuffer *framebuffer,
 	}
 
 	if (dev_priv->ldu_priv) {
-		ret = do_dmabuf_dirty_ldu(dev_priv, &vfbd->base,
-					  flags, color,
-					  clips, num_clips, increment);
-	} else {
-		ret = do_dmabuf_dirty_sou(file_priv, dev_priv, &vfbd->base,
-					  flags, color,
-					  clips, num_clips, increment, NULL);
+		ret = vmw_kms_ldu_do_dmabuf_dirty(dev_priv, &vfbd->base,
+						  flags, color,
+						  clips, num_clips, increment);
+	} else if (dev_priv->active_display_unit == vmw_du_screen_object) {
+		ret = vmw_kms_sou_do_dmabuf_dirty(file_priv, dev_priv,
+						  &vfbd->base,
+						  flags, color,
+						  clips, num_clips, increment,
+						  NULL);
 	}
 
 	vmw_fifo_flush(dev_priv, false);
@@ -1011,8 +661,8 @@  static int vmw_framebuffer_dmabuf_pin(struct vmw_framebuffer *vfb)
 		vmw_framebuffer_to_vfbd(&vfb->base);
 	int ret;
 
-	/* This code should not be used with screen objects */
-	BUG_ON(dev_priv->sou_priv);
+	/* This code should only be used with Legacy Display Unit */
+	BUG_ON(dev_priv->active_display_unit != vmw_du_legacy);
 
 	vmw_overlay_pause_all(dev_priv);
 
@@ -1059,7 +709,7 @@  static int vmw_kms_new_framebuffer_dmabuf(struct vmw_private *dev_priv,
 	}
 
 	/* Limited framebuffer color depth support for screen objects */
-	if (dev_priv->sou_priv) {
+	if (dev_priv->active_display_unit == vmw_du_screen_object) {
 		switch (mode_cmd->depth) {
 		case 32:
 		case 24:
@@ -1102,7 +752,7 @@  static int vmw_kms_new_framebuffer_dmabuf(struct vmw_private *dev_priv,
 	vfbd->base.base.depth = mode_cmd->depth;
 	vfbd->base.base.width = mode_cmd->width;
 	vfbd->base.base.height = mode_cmd->height;
-	if (!dev_priv->sou_priv) {
+	if (dev_priv->active_display_unit == vmw_du_legacy) {
 		vfbd->base.pin = vmw_framebuffer_dmabuf_pin;
 		vfbd->base.unpin = vmw_framebuffer_dmabuf_unpin;
 	}
@@ -1159,7 +809,7 @@  static struct drm_framebuffer *vmw_kms_fb_create(struct drm_device *dev,
 	if (!vmw_kms_validate_mode_vram(dev_priv,
 					mode_cmd.pitch,
 					mode_cmd.height)) {
-		DRM_ERROR("VRAM size is too small for requested mode.\n");
+		DRM_ERROR("Requested mode exceed bounding box limit.\n");
 		return ERR_PTR(-ENOMEM);
 	}
 
@@ -1220,7 +870,7 @@  static const struct drm_mode_config_funcs vmw_kms_funcs = {
 	.fb_create = vmw_kms_fb_create,
 };
 
-int vmw_kms_present(struct vmw_private *dev_priv,
+int vmw_kms_generic_present(struct vmw_private *dev_priv,
 		    struct drm_file *file_priv,
 		    struct vmw_framebuffer *vfb,
 		    struct vmw_surface *surface,
@@ -1358,6 +1008,19 @@  out_free_tmp:
 	return ret;
 }
 
+int vmw_kms_present(struct vmw_private *dev_priv,
+		    struct drm_file *file_priv,
+		    struct vmw_framebuffer *vfb,
+		    struct vmw_surface *surface,
+		    uint32_t sid,
+		    int32_t destX, int32_t destY,
+		    struct drm_vmw_rect *clips,
+		    uint32_t num_clips)
+{
+	return vmw_kms_generic_present(dev_priv, file_priv, vfb, surface, sid,
+				       destX, destY, clips, num_clips);
+}
+
 int vmw_kms_readback(struct vmw_private *dev_priv,
 		     struct drm_file *file_priv,
 		     struct vmw_framebuffer *vfb,
@@ -1478,26 +1141,29 @@  int vmw_kms_init(struct vmw_private *dev_priv)
 	dev->mode_config.max_width = 8192;
 	dev->mode_config.max_height = 8192;
 
-	ret = vmw_kms_init_screen_object_display(dev_priv);
+	ret = vmw_kms_sou_init_display(dev_priv);
 	if (ret) /* Fallback */
-		(void)vmw_kms_init_legacy_display_system(dev_priv);
+		ret = vmw_kms_ldu_init_display(dev_priv);
 
-	return 0;
+	return ret;
 }
 
 int vmw_kms_close(struct vmw_private *dev_priv)
 {
+	int ret;
+
 	/*
 	 * Docs says we should take the lock before calling this function
 	 * but since it destroys encoders and our destructor calls
 	 * drm_encoder_cleanup which takes the lock we deadlock.
 	 */
 	drm_mode_config_cleanup(dev_priv->dev);
-	if (dev_priv->sou_priv)
-		vmw_kms_close_screen_object_display(dev_priv);
+	if (dev_priv->active_display_unit == vmw_du_screen_object)
+		ret = vmw_kms_sou_close_display(dev_priv);
 	else
-		vmw_kms_close_legacy_display_system(dev_priv);
-	return 0;
+		ret = vmw_kms_ldu_close_display(dev_priv);
+
+	return ret;
 }
 
 int vmw_kms_cursor_bypass_ioctl(struct drm_device *dev, void *data,
@@ -1573,7 +1239,7 @@  int vmw_kms_save_vga(struct vmw_private *vmw_priv)
 		  vmw_read(vmw_priv, SVGA_REG_PITCHLOCK);
 	else if (vmw_fifo_have_pitchlock(vmw_priv))
 		vmw_priv->vga_pitchlock = ioread32(vmw_priv->mmio_virt +
-						       SVGA_FIFO_PITCHLOCK);
+						   SVGA_FIFO_PITCHLOCK);
 
 	if (!(vmw_priv->capabilities & SVGA_CAP_DISPLAY_TOPOLOGY))
 		return 0;
@@ -1719,75 +1385,6 @@  static int vmw_du_update_layout(struct vmw_private *dev_priv, unsigned num,
 	return 0;
 }
 
-int vmw_du_page_flip(struct drm_crtc *crtc,
-		     struct drm_framebuffer *fb,
-		     struct drm_pending_vblank_event *event,
-		     uint32_t page_flip_flags)
-{
-	struct vmw_private *dev_priv = vmw_priv(crtc->dev);
-	struct drm_framebuffer *old_fb = crtc->primary->fb;
-	struct vmw_framebuffer *vfb = vmw_framebuffer_to_vfb(fb);
-	struct drm_file *file_priv ;
-	struct vmw_fence_obj *fence = NULL;
-	struct drm_clip_rect clips;
-	int ret;
-
-	if (event == NULL)
-		return -EINVAL;
-
-	/* require ScreenObject support for page flipping */
-	if (!dev_priv->sou_priv)
-		return -ENOSYS;
-
-	file_priv = event->base.file_priv;
-	if (!vmw_kms_screen_object_flippable(dev_priv, crtc))
-		return -EINVAL;
-
-	crtc->primary->fb = fb;
-
-	/* do a full screen dirty update */
-	clips.x1 = clips.y1 = 0;
-	clips.x2 = fb->width;
-	clips.y2 = fb->height;
-
-	if (vfb->dmabuf)
-		ret = do_dmabuf_dirty_sou(file_priv, dev_priv, vfb,
-					  0, 0, &clips, 1, 1, &fence);
-	else
-		ret = do_surface_dirty_sou(dev_priv, file_priv, vfb,
-					   0, 0, &clips, 1, 1, &fence);
-
-
-	if (ret != 0)
-		goto out_no_fence;
-	if (!fence) {
-		ret = -EINVAL;
-		goto out_no_fence;
-	}
-
-	ret = vmw_event_fence_action_queue(file_priv, fence,
-					   &event->base,
-					   &event->event.tv_sec,
-					   &event->event.tv_usec,
-					   true);
-
-	/*
-	 * No need to hold on to this now. The only cleanup
-	 * we need to do if we fail is unref the fence.
-	 */
-	vmw_fence_obj_unreference(&fence);
-
-	if (vmw_crtc_to_du(crtc)->is_implicit)
-		vmw_kms_screen_object_update_implicit_fb(dev_priv, crtc);
-
-	return ret;
-
-out_no_fence:
-	crtc->primary->fb = old_fb;
-	return ret;
-}
-
-
 void vmw_du_crtc_save(struct drm_crtc *crtc)
 {
 }
@@ -1958,36 +1555,34 @@  int vmw_du_connector_fill_modes(struct drm_connector *connector,
 	 * If using screen objects, then assume 32-bpp because that's what the
 	 * SVGA device is assuming
 	 */
-	if (dev_priv->sou_priv)
+	if (dev_priv->active_display_unit == vmw_du_screen_object)
 		assumed_bpp = 4;
 
 	/* Add preferred mode */
-	{
-		mode = drm_mode_duplicate(dev, &prefmode);
-		if (!mode)
-			return 0;
-		mode->hdisplay = du->pref_width;
-		mode->vdisplay = du->pref_height;
-		vmw_guess_mode_timing(mode);
-
-		if (vmw_kms_validate_mode_vram(dev_priv,
-						mode->hdisplay * assumed_bpp,
-						mode->vdisplay)) {
-			drm_mode_probed_add(connector, mode);
-		} else {
-			drm_mode_destroy(dev, mode);
-			mode = NULL;
-		}
+	mode = drm_mode_duplicate(dev, &prefmode);
+	if (!mode)
+		return 0;
+	mode->hdisplay = du->pref_width;
+	mode->vdisplay = du->pref_height;
+	vmw_guess_mode_timing(mode);
 
-		if (du->pref_mode) {
-			list_del_init(&du->pref_mode->head);
-			drm_mode_destroy(dev, du->pref_mode);
-		}
+	if (vmw_kms_validate_mode_vram(dev_priv,
+					mode->hdisplay * assumed_bpp,
+					mode->vdisplay)) {
+		drm_mode_probed_add(connector, mode);
+	} else {
+		drm_mode_destroy(dev, mode);
+		mode = NULL;
+	}
 
-		/* mode might be null here, this is intended */
-		du->pref_mode = mode;
+	if (du->pref_mode) {
+		list_del_init(&du->pref_mode->head);
+		drm_mode_destroy(dev, du->pref_mode);
 	}
 
+	/* mode might be null here, this is intended */
+	du->pref_mode = mode;
+
 	for (i = 0; vmw_kms_connector_builtin[i].type != 0; i++) {
 		bmode = &vmw_kms_connector_builtin[i];
 		if (bmode->hdisplay > max_width ||
@@ -2036,6 +1631,7 @@  int vmw_kms_update_layout_ioctl(struct drm_device *dev, void *data,
 	int ret;
 	int i;
 	struct drm_mode_config *mode_config = &dev->mode_config;
+	struct drm_vmw_rect bounding_box = {0};
 
 	if (!arg->num_outputs) {
 		struct drm_vmw_rect def_rect = {0, 0, 800, 600};
@@ -2066,6 +1662,16 @@  int vmw_kms_update_layout_ioctl(struct drm_device *dev, void *data,
 			ret = -EINVAL;
 			goto out_free;
 		}
+
+		/*
+		 * bounding_box.w and bunding_box.h are used as
+		 * lower-right coordinates
+		 */
+		if (rects[i].x + rects[i].w > bounding_box.w)
+			bounding_box.w = rects[i].x + rects[i].w;
+
+		if (rects[i].y + rects[i].h > bounding_box.h)
+			bounding_box.h = rects[i].y + rects[i].h;
 	}
 
 	vmw_du_update_layout(dev_priv, arg->num_outputs, rects);
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.h b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.h
index 8d038c3..0f2c291 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.h
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.h
@@ -1,6 +1,6 @@ 
 /**************************************************************************
  *
- * Copyright © 2009 VMware, Inc., Palo Alto, CA., USA
+ * Copyright © 2009-2014 VMware, Inc., Palo Alto, CA., USA
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -32,11 +32,17 @@ 
 #include <drm/drm_crtc_helper.h>
 #include "vmwgfx_drv.h"
 
+
+
 #define VMWGFX_NUM_DISPLAY_UNITS 8
 
 
 #define vmw_framebuffer_to_vfb(x) \
 	container_of(x, struct vmw_framebuffer, base)
+#define vmw_framebuffer_to_vfbs(x) \
+	container_of(x, struct vmw_framebuffer_surface, base.base)
+#define vmw_framebuffer_to_vfbd(x) \
+	container_of(x, struct vmw_framebuffer_dmabuf, base.base)
 
 /**
  * Base class for framebuffers
@@ -53,9 +59,36 @@  struct vmw_framebuffer {
 	uint32_t user_handle;
 };
 
+/*
+ * Clip rectangle
+ */
+struct vmw_clip_rect {
+	int x1, x2, y1, y2;
+};
+
+struct vmw_framebuffer_surface {
+	struct vmw_framebuffer base;
+	struct vmw_surface *surface;
+	struct vmw_dma_buffer *buffer;
+	struct list_head head;
+	struct drm_master *master;
+};
 
-#define vmw_crtc_to_du(x) \
-	container_of(x, struct vmw_display_unit, crtc)
+
+struct vmw_framebuffer_dmabuf {
+	struct vmw_framebuffer base;
+	struct vmw_dma_buffer *buffer;
+};
+
+
+/*
+ * Basic clip rect manipulation
+ */
+void vmw_clip_cliprects(struct drm_clip_rect *rects,
+			int num_rects,
+			struct vmw_clip_rect clip,
+			SVGASignedRect *out_rects,
+			int *out_num);
 
 /*
  * Basic cursor manipulation
@@ -120,11 +153,7 @@  struct vmw_display_unit {
 /*
  * Shared display unit functions - vmwgfx_kms.c
  */
-void vmw_display_unit_cleanup(struct vmw_display_unit *du);
-int vmw_du_page_flip(struct drm_crtc *crtc,
-		     struct drm_framebuffer *fb,
-		     struct drm_pending_vblank_event *event,
-		     uint32_t page_flip_flags);
+void vmw_du_cleanup(struct vmw_display_unit *du);
 void vmw_du_crtc_save(struct drm_crtc *crtc);
 void vmw_du_crtc_restore(struct drm_crtc *crtc);
 void vmw_du_crtc_gamma_set(struct drm_crtc *crtc,
@@ -148,20 +177,31 @@  int vmw_du_connector_set_property(struct drm_connector *connector,
 /*
  * Legacy display unit functions - vmwgfx_ldu.c
  */
-int vmw_kms_init_legacy_display_system(struct vmw_private *dev_priv);
-int vmw_kms_close_legacy_display_system(struct vmw_private *dev_priv);
+int vmw_kms_ldu_init_display(struct vmw_private *dev_priv);
+int vmw_kms_ldu_close_display(struct vmw_private *dev_priv);
+int vmw_kms_ldu_do_dmabuf_dirty(struct vmw_private *dev_priv,
+				struct vmw_framebuffer *framebuffer,
+				unsigned flags, unsigned color,
+				struct drm_clip_rect *clips,
+				unsigned num_clips, int increment);
 
 /*
  * Screen Objects display functions - vmwgfx_scrn.c
  */
-int vmw_kms_init_screen_object_display(struct vmw_private *dev_priv);
-int vmw_kms_close_screen_object_display(struct vmw_private *dev_priv);
-int vmw_kms_sou_update_layout(struct vmw_private *dev_priv, unsigned num,
-			      struct drm_vmw_rect *rects);
-bool vmw_kms_screen_object_flippable(struct vmw_private *dev_priv,
-				     struct drm_crtc *crtc);
-void vmw_kms_screen_object_update_implicit_fb(struct vmw_private *dev_priv,
-					      struct drm_crtc *crtc);
-
-
+int vmw_kms_sou_init_display(struct vmw_private *dev_priv);
+int vmw_kms_sou_close_display(struct vmw_private *dev_priv);
+int vmw_kms_sou_do_surface_dirty(struct vmw_private *dev_priv,
+				 struct drm_file *file_priv,
+				 struct vmw_framebuffer *framebuffer,
+				 unsigned flags, unsigned color,
+				 struct drm_clip_rect *clips,
+				 unsigned num_clips, int inc,
+				 struct vmw_fence_obj **out_fence);
+int vmw_kms_sou_do_dmabuf_dirty(struct drm_file *file_priv,
+				struct vmw_private *dev_priv,
+				struct vmw_framebuffer *framebuffer,
+				unsigned flags, unsigned color,
+				struct drm_clip_rect *clips,
+				unsigned num_clips, int increment,
+				struct vmw_fence_obj **out_fence);
 #endif
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ldu.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ldu.c
index 53579f2..f0fd565 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ldu.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ldu.c
@@ -57,7 +57,7 @@  struct vmw_legacy_display_unit {
 static void vmw_ldu_destroy(struct vmw_legacy_display_unit *ldu)
 {
 	list_del_init(&ldu->active);
-	vmw_display_unit_cleanup(&ldu->base);
+	vmw_du_cleanup(&ldu->base);
 	kfree(ldu);
 }
 
@@ -386,7 +386,7 @@  static int vmw_ldu_init(struct vmw_private *dev_priv, unsigned unit)
 	return 0;
 }
 
-int vmw_kms_init_legacy_display_system(struct vmw_private *dev_priv)
+int vmw_kms_ldu_init_display(struct vmw_private *dev_priv)
 {
 	struct drm_device *dev = dev_priv->dev;
 	int i, ret;
@@ -423,6 +423,10 @@  int vmw_kms_init_legacy_display_system(struct vmw_private *dev_priv)
 	else
 		vmw_ldu_init(dev_priv, 0);
 
+	dev_priv->active_display_unit = vmw_du_legacy;
+
+	DRM_INFO("Legacy Display Unit initialized\n");
+
 	return 0;
 
 err_vblank_cleanup:
@@ -433,7 +437,7 @@  err_free:
 	return ret;
 }
 
-int vmw_kms_close_legacy_display_system(struct vmw_private *dev_priv)
+int vmw_kms_ldu_close_display(struct vmw_private *dev_priv)
 {
 	struct drm_device *dev = dev_priv->dev;
 
@@ -448,3 +452,38 @@  int vmw_kms_close_legacy_display_system(struct vmw_private *dev_priv)
 
 	return 0;
 }
+
+
+int vmw_kms_ldu_do_dmabuf_dirty(struct vmw_private *dev_priv,
+				struct vmw_framebuffer *framebuffer,
+				unsigned flags, unsigned color,
+				struct drm_clip_rect *clips,
+				unsigned num_clips, int increment)
+{
+	size_t fifo_size;
+	int i;
+
+	struct {
+		uint32_t header;
+		SVGAFifoCmdUpdate body;
+	} *cmd;
+
+	fifo_size = sizeof(*cmd) * num_clips;
+	cmd = vmw_fifo_reserve(dev_priv, fifo_size);
+	if (unlikely(cmd == NULL)) {
+		DRM_ERROR("Fifo reserve failed.\n");
+		return -ENOMEM;
+	}
+
+	memset(cmd, 0, fifo_size);
+	for (i = 0; i < num_clips; i++, clips += increment) {
+		cmd[i].header = cpu_to_le32(SVGA_CMD_UPDATE);
+		cmd[i].body.x = cpu_to_le32(clips->x1);
+		cmd[i].body.y = cpu_to_le32(clips->y1);
+		cmd[i].body.width = cpu_to_le32(clips->x2 - clips->x1);
+		cmd[i].body.height = cpu_to_le32(clips->y2 - clips->y1);
+	}
+
+	vmw_fifo_commit(dev_priv, fifo_size);
+	return 0;
+}
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_overlay.c b/drivers/gpu/drm/vmwgfx/vmwgfx_overlay.c
index 87e39f6..7f4b2f0 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_overlay.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_overlay.c
@@ -100,7 +100,7 @@  static int vmw_overlay_send_put(struct vmw_private *dev_priv,
 {
 	struct vmw_escape_video_flush *flush;
 	size_t fifo_size;
-	bool have_so = dev_priv->sou_priv ? true : false;
+	bool have_so = (dev_priv->active_display_unit == vmw_du_screen_object);
 	int i, num_items;
 	SVGAGuestPtr ptr;
 
@@ -231,7 +231,7 @@  static int vmw_overlay_move_buffer(struct vmw_private *dev_priv,
 	if (!pin)
 		return vmw_dmabuf_unpin(dev_priv, buf, inter);
 
-	if (!dev_priv->sou_priv)
+	if (dev_priv->active_display_unit == vmw_du_legacy)
 		return vmw_dmabuf_to_vram(dev_priv, buf, true, inter);
 
 	return vmw_dmabuf_to_vram_or_gmr(dev_priv, buf, true, inter);
@@ -453,7 +453,7 @@  int vmw_overlay_pause_all(struct vmw_private *dev_priv)
 
 static bool vmw_overlay_available(const struct vmw_private *dev_priv)
 {
-	return (dev_priv->overlay_priv != NULL && 
+	return (dev_priv->overlay_priv != NULL &&
 		((dev_priv->fifo.capabilities & VMW_OVERLAY_CAP_MASK) ==
 		 VMW_OVERLAY_CAP_MASK));
 }
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c b/drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c
index 9e8eb36..807fc87 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c
@@ -1,6 +1,6 @@ 
 /**************************************************************************
  *
- * Copyright © 2011 VMware, Inc., Palo Alto, CA., USA
+ * Copyright © 2011-2014 VMware, Inc., Palo Alto, CA., USA
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -57,7 +57,7 @@  struct vmw_screen_object_unit {
 
 static void vmw_sou_destroy(struct vmw_screen_object_unit *sou)
 {
-	vmw_display_unit_cleanup(&sou->base);
+	vmw_du_cleanup(&sou->base);
 	kfree(sou);
 }
 
@@ -72,7 +72,7 @@  static void vmw_sou_crtc_destroy(struct drm_crtc *crtc)
 }
 
 static void vmw_sou_del_active(struct vmw_private *vmw_priv,
-			      struct vmw_screen_object_unit *sou)
+			       struct vmw_screen_object_unit *sou)
 {
 	struct vmw_screen_object_display *ld = vmw_priv->sou_priv;
 
@@ -84,8 +84,8 @@  static void vmw_sou_del_active(struct vmw_private *vmw_priv,
 }
 
 static void vmw_sou_add_active(struct vmw_private *vmw_priv,
-			      struct vmw_screen_object_unit *sou,
-			      struct vmw_framebuffer *vfb)
+			       struct vmw_screen_object_unit *sou,
+			       struct vmw_framebuffer *vfb)
 {
 	struct vmw_screen_object_display *ld = vmw_priv->sou_priv;
 
@@ -274,13 +274,13 @@  static int vmw_sou_crtc_set_config(struct drm_mode_set *set)
 	dev_priv = vmw_priv(crtc->dev);
 
 	if (set->num_connectors > 1) {
-		DRM_ERROR("to many connectors\n");
+		DRM_ERROR("Too many connectors\n");
 		return -EINVAL;
 	}
 
 	if (set->num_connectors == 1 &&
 	    set->connectors[0] != &sou->base.connector) {
-		DRM_ERROR("connector doesn't match %p %p\n",
+		DRM_ERROR("Connector doesn't match %p %p\n",
 			set->connectors[0], &sou->base.connector);
 		return -EINVAL;
 	}
@@ -391,6 +391,250 @@  static int vmw_sou_crtc_set_config(struct drm_mode_set *set)
 	return 0;
 }
 
+/**
+ * Returns if this unit can be page flipped.
+ * Must be called with the mode_config mutex held.
+ */
+static bool vmw_sou_screen_object_flippable(struct vmw_private *dev_priv,
+					    struct drm_crtc *crtc)
+{
+	struct vmw_screen_object_unit *sou = vmw_crtc_to_sou(crtc);
+
+	if (!sou->base.is_implicit)
+		return true;
+
+	if (dev_priv->sou_priv->num_implicit != 1)
+		return false;
+
+	return true;
+}
+
+/**
+ * Update the implicit fb to the current fb of this crtc.
+ * Must be called with the mode_config mutex held.
+ */
+void vmw_sou_update_implicit_fb(struct vmw_private *dev_priv,
+				struct drm_crtc *crtc)
+{
+	struct vmw_screen_object_unit *sou = vmw_crtc_to_sou(crtc);
+
+	BUG_ON(!sou->base.is_implicit);
+
+	dev_priv->sou_priv->implicit_fb =
+		vmw_framebuffer_to_vfb(sou->base.crtc.primary->fb);
+}
+
+static int vmw_sou_crtc_page_flip(struct drm_crtc *crtc,
+				  struct drm_framebuffer *fb,
+				  struct drm_pending_vblank_event *event,
+				  uint32_t flags)
+{
+	struct vmw_private *dev_priv = vmw_priv(crtc->dev);
+	struct drm_framebuffer *old_fb = crtc->primary->fb;
+	struct vmw_framebuffer *vfb = vmw_framebuffer_to_vfb(fb);
+	struct drm_file *file_priv = event->base.file_priv;
+	struct vmw_fence_obj *fence = NULL;
+	struct drm_clip_rect clips;
+	int ret;
+
+	/* require ScreenObject support for page flipping */
+	if (!dev_priv->sou_priv)
+		return -ENOSYS;
+
+	if (!vmw_sou_screen_object_flippable(dev_priv, crtc))
+		return -EINVAL;
+
+	crtc->primary->fb = fb;
+
+	/* do a full screen dirty update */
+	clips.x1 = clips.y1 = 0;
+	clips.x2 = fb->width;
+	clips.y2 = fb->height;
+
+	if (vfb->dmabuf)
+		ret = vmw_kms_sou_do_dmabuf_dirty(file_priv, dev_priv, vfb,
+						  0, 0, &clips, 1, 1, &fence);
+	else
+		ret = vmw_kms_sou_do_surface_dirty(dev_priv, file_priv, vfb,
+						   0, 0, &clips, 1, 1, &fence);
+
+
+	if (ret != 0)
+		goto out_no_fence;
+	if (!fence) {
+		ret = -EINVAL;
+		goto out_no_fence;
+	}
+
+	ret = vmw_event_fence_action_queue(file_priv, fence,
+					   &event->base,
+					   &event->event.tv_sec,
+					   &event->event.tv_usec,
+					   true);
+
+	/*
+	 * No need to hold on to this now. The only cleanup
+	 * we need to do if we fail is unref the fence.
+	 */
+	vmw_fence_obj_unreference(&fence);
+
+	if (vmw_crtc_to_du(crtc)->is_implicit)
+		vmw_sou_update_implicit_fb(dev_priv, crtc);
+
+	return ret;
+
+out_no_fence:
+	crtc->primary->fb = old_fb;
+	return ret;
+}
+
+int vmw_kms_sou_do_surface_dirty(struct vmw_private *dev_priv,
+				 struct drm_file *file_priv,
+				 struct vmw_framebuffer *framebuffer,
+				 unsigned flags, unsigned color,
+				 struct drm_clip_rect *clips,
+				 unsigned num_clips, int inc,
+				 struct vmw_fence_obj **out_fence)
+{
+	struct vmw_display_unit *units[VMWGFX_NUM_DISPLAY_UNITS];
+	struct drm_clip_rect *clips_ptr;
+	struct drm_clip_rect *tmp;
+	struct drm_crtc *crtc;
+	size_t fifo_size;
+	int i, num_units;
+	int ret = 0; /* silence warning */
+	int left, right, top, bottom;
+
+	struct {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdBlitSurfaceToScreen body;
+	} *cmd;
+	SVGASignedRect *blits;
+
+	num_units = 0;
+	list_for_each_entry(crtc, &dev_priv->dev->mode_config.crtc_list,
+			    head) {
+		if (crtc->primary->fb != &framebuffer->base)
+			continue;
+		units[num_units++] = vmw_crtc_to_du(crtc);
+	}
+
+	BUG_ON(!clips || !num_clips);
+
+	tmp = kzalloc(sizeof(*tmp) * num_clips, GFP_KERNEL);
+	if (unlikely(tmp == NULL)) {
+		DRM_ERROR("Temporary cliprect memory alloc failed.\n");
+		return -ENOMEM;
+	}
+
+	fifo_size = sizeof(*cmd) + sizeof(SVGASignedRect) * num_clips;
+	cmd = kzalloc(fifo_size, GFP_KERNEL);
+	if (unlikely(cmd == NULL)) {
+		DRM_ERROR("Temporary fifo memory alloc failed.\n");
+		ret = -ENOMEM;
+		goto out_free_tmp;
+	}
+
+	/* setup blits pointer */
+	blits = (SVGASignedRect *)&cmd[1];
+
+	/* initial clip region */
+	left = clips->x1;
+	right = clips->x2;
+	top = clips->y1;
+	bottom = clips->y2;
+
+	/* skip the first clip rect */
+	for (i = 1, clips_ptr = clips + inc;
+	     i < num_clips; i++, clips_ptr += inc) {
+		left = min_t(int, left, (int)clips_ptr->x1);
+		right = max_t(int, right, (int)clips_ptr->x2);
+		top = min_t(int, top, (int)clips_ptr->y1);
+		bottom = max_t(int, bottom, (int)clips_ptr->y2);
+	}
+
+	/* only need to do this once */
+	cmd->header.id = cpu_to_le32(SVGA_3D_CMD_BLIT_SURFACE_TO_SCREEN);
+	cmd->header.size = cpu_to_le32(fifo_size - sizeof(cmd->header));
+
+	cmd->body.srcRect.left = left;
+	cmd->body.srcRect.right = right;
+	cmd->body.srcRect.top = top;
+	cmd->body.srcRect.bottom = bottom;
+
+	clips_ptr = clips;
+	for (i = 0; i < num_clips; i++, clips_ptr += inc) {
+		tmp[i].x1 = clips_ptr->x1 - left;
+		tmp[i].x2 = clips_ptr->x2 - left;
+		tmp[i].y1 = clips_ptr->y1 - top;
+		tmp[i].y2 = clips_ptr->y2 - top;
+	}
+
+	/* do per unit writing, reuse fifo for each */
+	for (i = 0; i < num_units; i++) {
+		struct vmw_display_unit *unit = units[i];
+		struct vmw_clip_rect clip;
+		int num;
+
+		clip.x1 = left - unit->crtc.x;
+		clip.y1 = top - unit->crtc.y;
+		clip.x2 = right - unit->crtc.x;
+		clip.y2 = bottom - unit->crtc.y;
+
+		/* skip any crtcs that misses the clip region */
+		if (clip.x1 >= unit->crtc.mode.hdisplay ||
+		    clip.y1 >= unit->crtc.mode.vdisplay ||
+		    clip.x2 <= 0 || clip.y2 <= 0)
+			continue;
+
+		/*
+		 * In order for the clip rects to be correctly scaled
+		 * the src and dest rects needs to be the same size.
+		 */
+		cmd->body.destRect.left = clip.x1;
+		cmd->body.destRect.right = clip.x2;
+		cmd->body.destRect.top = clip.y1;
+		cmd->body.destRect.bottom = clip.y2;
+
+		/* create a clip rect of the crtc in dest coords */
+		clip.x2 = unit->crtc.mode.hdisplay - clip.x1;
+		clip.y2 = unit->crtc.mode.vdisplay - clip.y1;
+		clip.x1 = 0 - clip.x1;
+		clip.y1 = 0 - clip.y1;
+
+		/* need to reset sid as it is changed by execbuf */
+		cmd->body.srcImage.sid = cpu_to_le32(framebuffer->user_handle);
+		cmd->body.destScreenId = unit->unit;
+
+		/* clip and write blits to cmd stream */
+		vmw_clip_cliprects(tmp, num_clips, clip, blits, &num);
+
+		/* if no cliprects hit skip this */
+		if (num == 0)
+			continue;
+
+		/* only return the last fence */
+		if (out_fence && *out_fence)
+			vmw_fence_obj_unreference(out_fence);
+
+		/* recalculate package length */
+		fifo_size = sizeof(*cmd) + sizeof(SVGASignedRect) * num;
+		cmd->header.size = cpu_to_le32(fifo_size - sizeof(cmd->header));
+		ret = vmw_execbuf_process(file_priv, dev_priv, NULL, cmd,
+					  fifo_size, 0, NULL, out_fence);
+
+		if (unlikely(ret != 0))
+			break;
+	}
+
+
+	kfree(cmd);
+out_free_tmp:
+	kfree(tmp);
+
+	return ret;
+}
+
 static struct drm_crtc_funcs vmw_screen_object_crtc_funcs = {
 	.save = vmw_du_crtc_save,
 	.restore = vmw_du_crtc_restore,
@@ -399,7 +643,7 @@  static struct drm_crtc_funcs vmw_screen_object_crtc_funcs = {
 	.gamma_set = vmw_du_crtc_gamma_set,
 	.destroy = vmw_sou_crtc_destroy,
 	.set_config = vmw_sou_crtc_set_config,
-	.page_flip = vmw_du_page_flip,
+	.page_flip = vmw_sou_crtc_page_flip,
 };
 
 /*
@@ -424,7 +668,7 @@  static void vmw_sou_connector_destroy(struct drm_connector *connector)
 	vmw_sou_destroy(vmw_connector_to_sou(connector));
 }
 
-static struct drm_connector_funcs vmw_legacy_connector_funcs = {
+static struct drm_connector_funcs vmw_sou_connector_funcs = {
 	.dpms = vmw_du_connector_dpms,
 	.save = vmw_du_connector_save,
 	.restore = vmw_du_connector_restore,
@@ -459,7 +703,7 @@  static int vmw_sou_init(struct vmw_private *dev_priv, unsigned unit)
 	sou->base.pref_mode = NULL;
 	sou->base.is_implicit = true;
 
-	drm_connector_init(dev, connector, &vmw_legacy_connector_funcs,
+	drm_connector_init(dev, connector, &vmw_sou_connector_funcs,
 			   DRM_MODE_CONNECTOR_VIRTUAL);
 	connector->status = vmw_du_connector_detect(connector, true);
 
@@ -482,7 +726,7 @@  static int vmw_sou_init(struct vmw_private *dev_priv, unsigned unit)
 	return 0;
 }
 
-int vmw_kms_init_screen_object_display(struct vmw_private *dev_priv)
+int vmw_kms_sou_init_display(struct vmw_private *dev_priv)
 {
 	struct drm_device *dev = dev_priv->dev;
 	int i, ret;
@@ -517,7 +761,9 @@  int vmw_kms_init_screen_object_display(struct vmw_private *dev_priv)
 	for (i = 0; i < VMWGFX_NUM_DISPLAY_UNITS; ++i)
 		vmw_sou_init(dev_priv, i);
 
-	DRM_INFO("Screen objects system initialized\n");
+	dev_priv->active_display_unit = vmw_du_screen_object;
+
+	DRM_INFO("Screen Objects Display Unit initialized\n");
 
 	return 0;
 
@@ -530,7 +776,7 @@  err_no_mem:
 	return ret;
 }
 
-int vmw_kms_close_screen_object_display(struct vmw_private *dev_priv)
+int vmw_kms_sou_close_display(struct vmw_private *dev_priv)
 {
 	struct drm_device *dev = dev_priv->dev;
 
@@ -544,35 +790,143 @@  int vmw_kms_close_screen_object_display(struct vmw_private *dev_priv)
 	return 0;
 }
 
-/**
- * Returns if this unit can be page flipped.
- * Must be called with the mode_config mutex held.
- */
-bool vmw_kms_screen_object_flippable(struct vmw_private *dev_priv,
-				     struct drm_crtc *crtc)
+static int do_dmabuf_define_gmrfb(struct drm_file *file_priv,
+				  struct vmw_private *dev_priv,
+				  struct vmw_framebuffer *framebuffer)
 {
-	struct vmw_screen_object_unit *sou = vmw_crtc_to_sou(crtc);
+	int depth = framebuffer->base.depth;
+	size_t fifo_size;
+	int ret;
 
-	if (!sou->base.is_implicit)
-		return true;
+	struct {
+		uint32_t header;
+		SVGAFifoCmdDefineGMRFB body;
+	} *cmd;
 
-	if (dev_priv->sou_priv->num_implicit != 1)
-		return false;
+	/* Emulate RGBA support, contrary to svga_reg.h this is not
+	 * supported by hosts. This is only a problem if we are reading
+	 * this value later and expecting what we uploaded back.
+	 */
+	if (depth == 32)
+		depth = 24;
 
-	return true;
+	fifo_size = sizeof(*cmd);
+	cmd = kmalloc(fifo_size, GFP_KERNEL);
+	if (unlikely(cmd == NULL)) {
+		DRM_ERROR("Failed to allocate temporary cmd buffer.\n");
+		return -ENOMEM;
+	}
+
+	memset(cmd, 0, fifo_size);
+	cmd->header = SVGA_CMD_DEFINE_GMRFB;
+	cmd->body.format.bitsPerPixel = framebuffer->base.bits_per_pixel;
+	cmd->body.format.colorDepth = depth;
+	cmd->body.format.reserved = 0;
+	cmd->body.bytesPerLine = framebuffer->base.pitches[0];
+	cmd->body.ptr.gmrId = framebuffer->user_handle;
+	cmd->body.ptr.offset = 0;
+
+	ret = vmw_execbuf_process(file_priv, dev_priv, NULL, cmd,
+				  fifo_size, 0, NULL, NULL);
+
+	kfree(cmd);
+
+	return ret;
 }
 
-/**
- * Update the implicit fb to the current fb of this crtc.
- * Must be called with the mode_config mutex held.
- */
-void vmw_kms_screen_object_update_implicit_fb(struct vmw_private *dev_priv,
-					      struct drm_crtc *crtc)
+int vmw_kms_sou_do_dmabuf_dirty(struct drm_file *file_priv,
+				struct vmw_private *dev_priv,
+				struct vmw_framebuffer *framebuffer,
+				unsigned flags, unsigned color,
+				struct drm_clip_rect *clips,
+				unsigned num_clips, int increment,
+				struct vmw_fence_obj **out_fence)
 {
-	struct vmw_screen_object_unit *sou = vmw_crtc_to_sou(crtc);
+	struct vmw_display_unit *units[VMWGFX_NUM_DISPLAY_UNITS];
+	struct drm_clip_rect *clips_ptr;
+	int i, k, num_units, ret;
+	struct drm_crtc *crtc;
+	size_t fifo_size;
 
-	BUG_ON(!sou->base.is_implicit);
+	struct {
+		uint32_t header;
+		SVGAFifoCmdBlitGMRFBToScreen body;
+	} *blits;
 
-	dev_priv->sou_priv->implicit_fb =
-		vmw_framebuffer_to_vfb(sou->base.crtc.primary->fb);
+	ret = do_dmabuf_define_gmrfb(file_priv, dev_priv, framebuffer);
+	if (unlikely(ret != 0))
+		return ret; /* define_gmrfb prints warnings */
+
+	fifo_size = sizeof(*blits) * num_clips;
+	blits = kmalloc(fifo_size, GFP_KERNEL);
+	if (unlikely(blits == NULL)) {
+		DRM_ERROR("Failed to allocate temporary cmd buffer.\n");
+		return -ENOMEM;
+	}
+
+	num_units = 0;
+	list_for_each_entry(crtc, &dev_priv->dev->mode_config.crtc_list, head) {
+		if (crtc->primary->fb != &framebuffer->base)
+			continue;
+		units[num_units++] = vmw_crtc_to_du(crtc);
+	}
+
+	for (k = 0; k < num_units; k++) {
+		struct vmw_display_unit *unit = units[k];
+		int hit_num = 0;
+
+		clips_ptr = clips;
+		for (i = 0; i < num_clips; i++, clips_ptr += increment) {
+			int clip_x1 = clips_ptr->x1 - unit->crtc.x;
+			int clip_y1 = clips_ptr->y1 - unit->crtc.y;
+			int clip_x2 = clips_ptr->x2 - unit->crtc.x;
+			int clip_y2 = clips_ptr->y2 - unit->crtc.y;
+			int move_x, move_y;
+
+			/* skip any crtcs that misses the clip region */
+			if (clip_x1 >= unit->crtc.mode.hdisplay ||
+			    clip_y1 >= unit->crtc.mode.vdisplay ||
+			    clip_x2 <= 0 || clip_y2 <= 0)
+				continue;
+
+			/* clip size to crtc size */
+			clip_x2 = min_t(int, clip_x2, unit->crtc.mode.hdisplay);
+			clip_y2 = min_t(int, clip_y2, unit->crtc.mode.vdisplay);
+
+			/* translate both src and dest to bring clip into screen */
+			move_x = min_t(int, clip_x1, 0);
+			move_y = min_t(int, clip_y1, 0);
+
+			/* actual translate done here */
+			blits[hit_num].header = SVGA_CMD_BLIT_GMRFB_TO_SCREEN;
+			blits[hit_num].body.destScreenId = unit->unit;
+			blits[hit_num].body.srcOrigin.x = clips_ptr->x1 - move_x;
+			blits[hit_num].body.srcOrigin.y = clips_ptr->y1 - move_y;
+			blits[hit_num].body.destRect.left = clip_x1 - move_x;
+			blits[hit_num].body.destRect.top = clip_y1 - move_y;
+			blits[hit_num].body.destRect.right = clip_x2;
+			blits[hit_num].body.destRect.bottom = clip_y2;
+			hit_num++;
+		}
+
+		/* no clips hit the crtc */
+		if (hit_num == 0)
+			continue;
+
+		/* only return the last fence */
+		if (out_fence && *out_fence)
+			vmw_fence_obj_unreference(out_fence);
+
+		fifo_size = sizeof(*blits) * hit_num;
+		ret = vmw_execbuf_process(file_priv, dev_priv, NULL, blits,
+					  fifo_size, 0, NULL, out_fence);
+
+		if (unlikely(ret != 0))
+			break;
+	}
+
+	kfree(blits);
+
+	return ret;
 }
+