diff --git a/shared-core/nouveau_drm.h b/shared-core/nouveau_drm.h
index bbb51bc4..4b5869ad 100644
--- a/shared-core/nouveau_drm.h
+++ b/shared-core/nouveau_drm.h
@@ -25,7 +25,7 @@
 #ifndef __NOUVEAU_DRM_H__
 #define __NOUVEAU_DRM_H__
 
-#define NOUVEAU_DRM_HEADER_PATCHLEVEL 10
+#define NOUVEAU_DRM_HEADER_PATCHLEVEL 11
 
 struct drm_nouveau_channel_alloc {
 	uint32_t     fb_ctxdma_handle;
@@ -85,10 +85,12 @@ struct drm_nouveau_gpuobj_free {
 #define NOUVEAU_MEM_PINNED		0x00000040
 #define NOUVEAU_MEM_USER_BACKED		0x00000080
 #define NOUVEAU_MEM_MAPPED		0x00000100
-#define NOUVEAU_MEM_INSTANCE		0x00000200 /* internal */
-#define NOUVEAU_MEM_NOTIFIER            0x00000400 /* internal */
-#define NOUVEAU_MEM_NOVM		0x00000800 /* internal */
-#define NOUVEAU_MEM_USER		0x00001000 /* internal */
+#define NOUVEAU_MEM_TILE		0x00000200
+#define NOUVEAU_MEM_TILE_ZETA		0x00000400
+#define NOUVEAU_MEM_INSTANCE		0x01000000 /* internal */
+#define NOUVEAU_MEM_NOTIFIER            0x02000000 /* internal */
+#define NOUVEAU_MEM_NOVM		0x04000000 /* internal */
+#define NOUVEAU_MEM_USER		0x08000000 /* internal */
 #define NOUVEAU_MEM_INTERNAL (NOUVEAU_MEM_INSTANCE | \
 			      NOUVEAU_MEM_NOTIFIER | \
 			      NOUVEAU_MEM_NOVM | \
@@ -107,6 +109,13 @@ struct drm_nouveau_mem_free {
 	int flags;
 };
 
+struct drm_nouveau_mem_tile {
+	uint64_t offset;
+	uint64_t delta;
+	uint64_t size;
+	int flags;
+};
+
 /* FIXME : maybe unify {GET,SET}PARAMs */
 #define NOUVEAU_GETPARAM_PCI_VENDOR      3
 #define NOUVEAU_GETPARAM_PCI_DEVICE      4
@@ -168,5 +177,6 @@ struct drm_nouveau_sarea {
 #define DRM_NOUVEAU_GPUOBJ_FREE        0x07
 #define DRM_NOUVEAU_MEM_ALLOC          0x08
 #define DRM_NOUVEAU_MEM_FREE           0x09
+#define DRM_NOUVEAU_MEM_TILE           0x0a
 
 #endif /* __NOUVEAU_DRM_H__ */
diff --git a/shared-core/nouveau_drv.h b/shared-core/nouveau_drv.h
index 1aaf1d74..cd5f9cf8 100644
--- a/shared-core/nouveau_drv.h
+++ b/shared-core/nouveau_drv.h
@@ -34,7 +34,7 @@
 
 #define DRIVER_MAJOR		0
 #define DRIVER_MINOR		0
-#define DRIVER_PATCHLEVEL	10
+#define DRIVER_PATCHLEVEL	11
 
 #define NOUVEAU_FAMILY   0x0000FFFF
 #define NOUVEAU_FLAGS    0xFFFF0000
@@ -359,6 +359,8 @@ extern int  nouveau_ioctl_mem_alloc(struct drm_device *, void *data,
 				    struct drm_file *);
 extern int  nouveau_ioctl_mem_free(struct drm_device *, void *data,
 				   struct drm_file *);
+extern int  nouveau_ioctl_mem_tile(struct drm_device *, void *data,
+				   struct drm_file *);
 extern struct mem_block* nouveau_mem_alloc(struct drm_device *,
 					   int alignment, uint64_t size,
 					   int flags, struct drm_file *);
diff --git a/shared-core/nouveau_fifo.c b/shared-core/nouveau_fifo.c
index d8fda277..085336af 100644
--- a/shared-core/nouveau_fifo.c
+++ b/shared-core/nouveau_fifo.c
@@ -593,6 +593,7 @@ struct drm_ioctl_desc nouveau_ioctls[] = {
 	DRM_IOCTL_DEF(DRM_NOUVEAU_GPUOBJ_FREE, nouveau_ioctl_gpuobj_free, DRM_AUTH),
 	DRM_IOCTL_DEF(DRM_NOUVEAU_MEM_ALLOC, nouveau_ioctl_mem_alloc, DRM_AUTH),
 	DRM_IOCTL_DEF(DRM_NOUVEAU_MEM_FREE, nouveau_ioctl_mem_free, DRM_AUTH),
+	DRM_IOCTL_DEF(DRM_NOUVEAU_MEM_TILE, nouveau_ioctl_mem_tile, DRM_AUTH),
 };
 
 int nouveau_max_ioctl = DRM_ARRAY_SIZE(nouveau_ioctls);
diff --git a/shared-core/nouveau_mem.c b/shared-core/nouveau_mem.c
index 58942829..d79c1a52 100644
--- a/shared-core/nouveau_mem.c
+++ b/shared-core/nouveau_mem.c
@@ -594,7 +594,7 @@ nouveau_mem_alloc(struct drm_device *dev, int alignment, uint64_t size,
 	 * page size in the GPU VM.
 	 */
 	if (flags & NOUVEAU_MEM_FB && dev_priv->card_type >= NV_50) {
-		size = (size + (64 * 1024)) & ~((64 * 1024) - 1);
+		size = (size + 65535) & ~65535;
 		if (alignment < 16)
 			alignment = 16;
 	}
@@ -659,6 +659,7 @@ alloc_ok:
 		struct nouveau_gpuobj *pt = dev_priv->vm_vram_pt;
 		unsigned offset = block->start;
 		unsigned count = block->size / 65536;
+		unsigned tile = 0;
 
 		if (!pt) {
 			DRM_ERROR("vm alloc without vm pt\n");
@@ -666,11 +667,22 @@ alloc_ok:
 			return NULL;
 		}
 
+		/* The tiling stuff is *not* what NVIDIA does - but both the
+		 * 2D and 3D engines seem happy with this simpler method.
+		 * Should look into why NVIDIA do what they do at some point.
+		 */
+		if (flags & NOUVEAU_MEM_TILE) {
+			if (flags & NOUVEAU_MEM_TILE_ZETA)
+				tile = 0x00002800;
+			else
+				tile = 0x00007000;
+		}
+
 		while (count--) {
 			unsigned pte = offset / 65536;
 
 			INSTANCE_WR(pt, (pte * 2) + 0, offset | 1);
-			INSTANCE_WR(pt, (pte * 2) + 1, 0x00000000);
+			INSTANCE_WR(pt, (pte * 2) + 1, 0x00000000 | tile);
 			offset += 65536;
 		}
 	} else {
@@ -808,3 +820,53 @@ nouveau_ioctl_mem_free(struct drm_device *dev, void *data,
 	nouveau_mem_free(dev, block);
 	return 0;
 }
+
+int
+nouveau_ioctl_mem_tile(struct drm_device *dev, void *data,
+		       struct drm_file *file_priv)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct drm_nouveau_mem_tile *memtile = data;
+	struct mem_block *block = NULL;
+
+	NOUVEAU_CHECK_INITIALISED_WITH_RETURN;
+
+	if (dev_priv->card_type < NV_50)
+		return -EINVAL;
+	
+	if (memtile->flags & NOUVEAU_MEM_FB) {
+		memtile->offset -= 512*1024*1024;
+		block = find_block(dev_priv->fb_heap, memtile->offset);
+	}
+
+	if (!block)
+		return -EINVAL;
+
+	if (block->file_priv != file_priv)
+		return -EPERM;
+
+	{
+		struct nouveau_gpuobj *pt = dev_priv->vm_vram_pt;
+		unsigned offset = block->start + memtile->delta;
+		unsigned count = memtile->size / 65536;
+		unsigned tile = 0;
+
+		if (memtile->flags & NOUVEAU_MEM_TILE) {
+			if (memtile->flags & NOUVEAU_MEM_TILE_ZETA)
+				tile = 0x00002800;
+			else
+				tile = 0x00007000;
+		}
+
+		while (count--) {
+			unsigned pte = offset / 65536;
+
+			INSTANCE_WR(pt, (pte * 2) + 0, offset | 1);
+			INSTANCE_WR(pt, (pte * 2) + 1, 0x00000000 | tile);
+			offset += 65536;
+		}
+	}
+
+	return 0;
+}
+