drm/omap: use flip-work helper

And simplify how we hold a ref+pin to what is being scanned out by using
fb refcnt'ing.  The previous logic pre-dated fb refcnt, and as a result
was less straightforward than it could have been.  By holding a ref to
the fb, we don't have to care about how many plane's there are and
holding a ref to each color plane's bo.
Signed-off-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

drivers/gpu/drm/omapdrm/omap_drv.h

@@ -203,9 +203,8 @@ struct drm_framebuffer *omap_framebuffer_create(struct drm_device *dev,
 struct drm_framebuffer *omap_framebuffer_init(struct drm_device *dev,
 		struct drm_mode_fb_cmd2 *mode_cmd, struct drm_gem_object **bos);
 struct drm_gem_object *omap_framebuffer_bo(struct drm_framebuffer *fb, int p);
-int omap_framebuffer_replace(struct drm_framebuffer *a,
-		struct drm_framebuffer *b, void *arg,
-		void (*unpin)(void *arg, struct drm_gem_object *bo));
+int omap_framebuffer_pin(struct drm_framebuffer *fb);
+int omap_framebuffer_unpin(struct drm_framebuffer *fb);
 void omap_framebuffer_update_scanout(struct drm_framebuffer *fb,
 		struct omap_drm_window *win, struct omap_overlay_info *info);
 struct drm_connector *omap_framebuffer_get_next_connector(

drivers/gpu/drm/omapdrm/omap_fb.c

@@ -237,55 +237,49 @@ void omap_framebuffer_update_scanout(struct drm_framebuffer *fb,
 	}
 }
 
-/* Call for unpin 'a' (if not NULL), and pin 'b' (if not NULL).  Although
- * buffers to unpin are just pushed to the unpin fifo so that the
- * caller can defer unpin until vblank.
- *
- * Note if this fails (ie. something went very wrong!), all buffers are
- * unpinned, and the caller disables the overlay.  We could have tried
- * to revert back to the previous set of pinned buffers but if things are
- * hosed there is no guarantee that would succeed.
- */
-int omap_framebuffer_replace(struct drm_framebuffer *a,
-		struct drm_framebuffer *b, void *arg,
-		void (*unpin)(void *arg, struct drm_gem_object *bo))
+/* pin, prepare for scanout: */
+int omap_framebuffer_pin(struct drm_framebuffer *fb)
 {
-	int ret = 0, i, na, nb;
-	struct omap_framebuffer *ofba = to_omap_framebuffer(a);
-	struct omap_framebuffer *ofbb = to_omap_framebuffer(b);
-	uint32_t pinned_mask = 0;
+	struct omap_framebuffer *omap_fb = to_omap_framebuffer(fb);
+	int ret, i, n = drm_format_num_planes(fb->pixel_format);
 
-	na = a ? drm_format_num_planes(a->pixel_format) : 0;
-	nb = b ? drm_format_num_planes(b->pixel_format) : 0;
+	for (i = 0; i < n; i++) {
+		struct plane *plane = &omap_fb->planes[i];
+		ret = omap_gem_get_paddr(plane->bo, &plane->paddr, true);
+		if (ret)
+			goto fail;
+		omap_gem_dma_sync(plane->bo, DMA_TO_DEVICE);
+	}
 
-	for (i = 0; i < max(na, nb); i++) {
-		struct plane *pa, *pb;
+	return 0;
 
-		pa = (i < na) ? &ofba->planes[i] : NULL;
-		pb = (i < nb) ? &ofbb->planes[i] : NULL;
+fail:
+	for (i--; i >= 0; i--) {
+		struct plane *plane = &omap_fb->planes[i];
+		omap_gem_put_paddr(plane->bo);
+		plane->paddr = 0;
+	}
 
-		if (pa)
-			unpin(arg, pa->bo);
+	return ret;
+}
 
-		if (pb && !ret) {
-			ret = omap_gem_get_paddr(pb->bo, &pb->paddr, true);
-			if (!ret) {
-				omap_gem_dma_sync(pb->bo, DMA_TO_DEVICE);
-				pinned_mask |= (1 << i);
-			}
-		}
-	}
+/* unpin, no longer being scanned out: */
+int omap_framebuffer_unpin(struct drm_framebuffer *fb)
+{
+	struct omap_framebuffer *omap_fb = to_omap_framebuffer(fb);
+	int ret, i, n = drm_format_num_planes(fb->pixel_format);
 
-	if (ret) {
-		/* something went wrong.. unpin what has been pinned */
-		for (i = 0; i < nb; i++) {
-			if (pinned_mask & (1 << i)) {
-				struct plane *pb = &ofba->planes[i];
-				unpin(arg, pb->bo);
-			}
-		}
+	for (i = 0; i < n; i++) {
+		struct plane *plane = &omap_fb->planes[i];
+		ret = omap_gem_put_paddr(plane->bo);
+		if (ret)
+			goto fail;
+		plane->paddr = 0;
 	}
 
+	return 0;
+
+fail:
 	return ret;
 }
 

drivers/gpu/drm/omapdrm/omap_plane.c

@@ -17,7 +17,7 @@
  * this program.  If not, see <http://www.gnu.org/licenses/>.
  */
 
-#include <linux/kfifo.h>
+#include "drm_flip_work.h"
 
 #include "omap_drv.h"
 #include "omap_dmm_tiler.h"
@@ -58,26 +58,23 @@ struct omap_plane {
 
 	struct omap_drm_irq error_irq;
 
-	/* set of bo's pending unpin until next post_apply() */
-	DECLARE_KFIFO_PTR(unpin_fifo, struct drm_gem_object *);
+	/* for deferring bo unpin's until next post_apply(): */
+	struct drm_flip_work unpin_work;
 
 	// XXX maybe get rid of this and handle vblank in crtc too?
 	struct callback apply_done_cb;
 };
 
-static void unpin(void *arg, struct drm_gem_object *bo)
+static void unpin_worker(struct drm_flip_work *work, void *val)
 {
-	struct drm_plane *plane = arg;
-	struct omap_plane *omap_plane = to_omap_plane(plane);
+	struct omap_plane *omap_plane =
+			container_of(work, struct omap_plane, unpin_work);
+	struct drm_device *dev = omap_plane->base.dev;
 
-	if (kfifo_put(&omap_plane->unpin_fifo,
-			(const struct drm_gem_object **)&bo)) {
-		/* also hold a ref so it isn't free'd while pinned */
-		drm_gem_object_reference(bo);
-	} else {
-		dev_err(plane->dev->dev, "unpin fifo full!\n");
-		omap_gem_put_paddr(bo);
-	}
+	omap_framebuffer_unpin(val);
+	mutex_lock(&dev->mode_config.mutex);
+	drm_framebuffer_unreference(val);
+	mutex_unlock(&dev->mode_config.mutex);
 }
 
 /* update which fb (if any) is pinned for scanout */
@@ -87,22 +84,21 @@ static int update_pin(struct drm_plane *plane, struct drm_framebuffer *fb)
 	struct drm_framebuffer *pinned_fb = omap_plane->pinned_fb;
 
 	if (pinned_fb != fb) {
-		int ret;
+		int ret = 0;
 
 		DBG("%p -> %p", pinned_fb, fb);
 
-		if (fb)
+		if (fb) {
 			drm_framebuffer_reference(fb);
-
-		ret = omap_framebuffer_replace(pinned_fb, fb, plane, unpin);
+			ret = omap_framebuffer_pin(fb);
+		}
 
 		if (pinned_fb)
-			drm_framebuffer_unreference(pinned_fb);
+			drm_flip_work_queue(&omap_plane->unpin_work, pinned_fb);
 
 		if (ret) {
 			dev_err(plane->dev->dev, "could not swap %p -> %p\n",
 					omap_plane->pinned_fb, fb);
-			if (fb)
 			drm_framebuffer_unreference(fb);
 			omap_plane->pinned_fb = NULL;
 			return ret;
@@ -170,17 +166,14 @@ static void omap_plane_post_apply(struct omap_drm_apply *apply)
 	struct omap_plane *omap_plane =
 			container_of(apply, struct omap_plane, apply);
 	struct drm_plane *plane = &omap_plane->base;
+	struct omap_drm_private *priv = plane->dev->dev_private;
 	struct omap_overlay_info *info = &omap_plane->info;
-	struct drm_gem_object *bo = NULL;
 	struct callback cb;
 
 	cb = omap_plane->apply_done_cb;
 	omap_plane->apply_done_cb.fxn = NULL;
 
-	while (kfifo_get(&omap_plane->unpin_fifo, &bo)) {
-		omap_gem_put_paddr(bo);
-		drm_gem_object_unreference_unlocked(bo);
-	}
+	drm_flip_work_commit(&omap_plane->unpin_work, priv->wq);
 
 	if (cb.fxn)
 		cb.fxn(cb.arg);
@@ -277,8 +270,7 @@ static void omap_plane_destroy(struct drm_plane *plane)
 	omap_plane_disable(plane);
 	drm_plane_cleanup(plane);
 
-	WARN_ON(!kfifo_is_empty(&omap_plane->unpin_fifo));
-	kfifo_free(&omap_plane->unpin_fifo);
+	drm_flip_work_cleanup(&omap_plane->unpin_work);
 
 	kfree(omap_plane);
 }
@@ -399,7 +391,8 @@ struct drm_plane *omap_plane_init(struct drm_device *dev,
 	if (!omap_plane)
 		goto fail;
 
-	ret = kfifo_alloc(&omap_plane->unpin_fifo, 16, GFP_KERNEL);
+	ret = drm_flip_work_init(&omap_plane->unpin_work, 16,
+			"unpin", unpin_worker);
 	if (ret) {
 		dev_err(dev->dev, "could not allocate unpin FIFO\n");
 		goto fail;