drm: move to kref per-master structures.

This is step one towards having multiple masters sharing a drm
device in order to get fast-user-switching to work.

It splits out the information associated with the drm master
into a separate kref counted structure, and allocates this when
a master opens the device node. It also allows the current master
to abdicate (say while VT switched), and a new master to take over
the hardware.

It moves the Intel and radeon drivers to using the sarea from
within the new master structures.

Signed-off-by: Dave Airlie <airlied@redhat.com>

18 files changed, 563 insertions, 332 deletions

diff --git a/drivers/gpu/drm/drm_auth.c b/drivers/gpu/drm/drm_auth.c
index a73462723d2d..ca7a9ef5007b 100644
--- a/drivers/gpu/drm/drm_auth.c
+++ b/drivers/gpu/drm/drm_auth.c
@@ -45,14 +45,15 @@
  * the one with matching magic number, while holding the drm_device::struct_mutex
  * lock.
  */
-static struct drm_file *drm_find_file(struct drm_device * dev, drm_magic_t magic)
+static struct drm_file *drm_find_file(struct drm_master *master, drm_magic_t magic)
 {
 	struct drm_file *retval = NULL;
 	struct drm_magic_entry *pt;
 	struct drm_hash_item *hash;
+	struct drm_device *dev = master->minor->dev;
 
 	mutex_lock(&dev->struct_mutex);
-	if (!drm_ht_find_item(&dev->magiclist, (unsigned long)magic, &hash)) {
+	if (!drm_ht_find_item(&master->magiclist, (unsigned long)magic, &hash)) {
 		pt = drm_hash_entry(hash, struct drm_magic_entry, hash_item);
 		retval = pt->priv;
 	}
@@ -71,11 +72,11 @@ static struct drm_file *drm_find_file(struct drm_device * dev, drm_magic_t magic
  * associated the magic number hash key in drm_device::magiclist, while holding
  * the drm_device::struct_mutex lock.
  */
-static int drm_add_magic(struct drm_device * dev, struct drm_file * priv,
+static int drm_add_magic(struct drm_master *master, struct drm_file *priv,
 			 drm_magic_t magic)
 {
 	struct drm_magic_entry *entry;
-
+	struct drm_device *dev = master->minor->dev;
 	DRM_DEBUG("%d\n", magic);
 
 	entry = drm_alloc(sizeof(*entry), DRM_MEM_MAGIC);
@@ -83,11 +84,10 @@ static int drm_add_magic(struct drm_device * dev, struct drm_file * priv,
 		return -ENOMEM;
 	memset(entry, 0, sizeof(*entry));
 	entry->priv = priv;
-
 	entry->hash_item.key = (unsigned long)magic;
 	mutex_lock(&dev->struct_mutex);
-	drm_ht_insert_item(&dev->magiclist, &entry->hash_item);
-	list_add_tail(&entry->head, &dev->magicfree);
+	drm_ht_insert_item(&master->magiclist, &entry->hash_item);
+	list_add_tail(&entry->head, &master->magicfree);
 	mutex_unlock(&dev->struct_mutex);
 
 	return 0;
@@ -102,20 +102,21 @@ static int drm_add_magic(struct drm_device * dev, struct drm_file * priv,
  * Searches and unlinks the entry in drm_device::magiclist with the magic
  * number hash key, while holding the drm_device::struct_mutex lock.
  */
-static int drm_remove_magic(struct drm_device * dev, drm_magic_t magic)
+static int drm_remove_magic(struct drm_master *master, drm_magic_t magic)
 {
 	struct drm_magic_entry *pt;
 	struct drm_hash_item *hash;
+	struct drm_device *dev = master->minor->dev;
 
 	DRM_DEBUG("%d\n", magic);
 
 	mutex_lock(&dev->struct_mutex);
-	if (drm_ht_find_item(&dev->magiclist, (unsigned long)magic, &hash)) {
+	if (drm_ht_find_item(&master->magiclist, (unsigned long)magic, &hash)) {
 		mutex_unlock(&dev->struct_mutex);
 		return -EINVAL;
 	}
 	pt = drm_hash_entry(hash, struct drm_magic_entry, hash_item);
-	drm_ht_remove_item(&dev->magiclist, hash);
+	drm_ht_remove_item(&master->magiclist, hash);
 	list_del(&pt->head);
 	mutex_unlock(&dev->struct_mutex);
 
@@ -153,9 +154,9 @@ int drm_getmagic(struct drm_device *dev, void *data, struct drm_file *file_priv)
 				++sequence;	/* reserve 0 */
 			auth->magic = sequence++;
 			spin_unlock(&lock);
-		} while (drm_find_file(dev, auth->magic));
+		} while (drm_find_file(file_priv->master, auth->magic));
 		file_priv->magic = auth->magic;
-		drm_add_magic(dev, file_priv, auth->magic);
+		drm_add_magic(file_priv->master, file_priv, auth->magic);
 	}
 
 	DRM_DEBUG("%u\n", auth->magic);
@@ -181,9 +182,9 @@ int drm_authmagic(struct drm_device *dev, void *data,
 	struct drm_file *file;
 
 	DRM_DEBUG("%u\n", auth->magic);
-	if ((file = drm_find_file(dev, auth->magic))) {
+	if ((file = drm_find_file(file_priv->master, auth->magic))) {
 		file->authenticated = 1;
-		drm_remove_magic(dev, auth->magic);
+		drm_remove_magic(file_priv->master, auth->magic);
 		return 0;
 	}
 	return -EINVAL;
diff --git a/drivers/gpu/drm/drm_bufs.c b/drivers/gpu/drm/drm_bufs.c
index bde64b84166e..dc3ce3e0a0a4 100644
--- a/drivers/gpu/drm/drm_bufs.c
+++ b/drivers/gpu/drm/drm_bufs.c
@@ -54,9 +54,9 @@ static struct drm_map_list *drm_find_matching_map(struct drm_device *dev,
 {
 	struct drm_map_list *entry;
 	list_for_each_entry(entry, &dev->maplist, head) {
-		if (entry->map && map->type == entry->map->type &&
+		if (entry->map && (entry->master == dev->primary->master) && (map->type == entry->map->type) &&
 		    ((entry->map->offset == map->offset) ||
-		     (map->type == _DRM_SHM && map->flags==_DRM_CONTAINS_LOCK))) {
+		     ((map->type == _DRM_SHM) && (map->flags&_DRM_CONTAINS_LOCK)))) {
 			return entry;
 		}
 	}
@@ -210,12 +210,12 @@ static int drm_addmap_core(struct drm_device * dev, unsigned int offset,
 		map->offset = (unsigned long)map->handle;
 		if (map->flags & _DRM_CONTAINS_LOCK) {
 			/* Prevent a 2nd X Server from creating a 2nd lock */
-			if (dev->lock.hw_lock != NULL) {
+			if (dev->primary->master->lock.hw_lock != NULL) {
 				vfree(map->handle);
 				drm_free(map, sizeof(*map), DRM_MEM_MAPS);
 				return -EBUSY;
 			}
-			dev->sigdata.lock = dev->lock.hw_lock = map->handle;	/* Pointer to lock */
+			dev->sigdata.lock = dev->primary->master->lock.hw_lock = map->handle;	/* Pointer to lock */
 		}
 		break;
 	case _DRM_AGP: {
@@ -319,6 +319,7 @@ static int drm_addmap_core(struct drm_device * dev, unsigned int offset,
 	list->user_token = list->hash.key << PAGE_SHIFT;
 	mutex_unlock(&dev->struct_mutex);
 
+	list->master = dev->primary->master;
 	*maplist = list;
 	return 0;
 	}
@@ -345,7 +346,7 @@ int drm_addmap_ioctl(struct drm_device *dev, void *data,
 	struct drm_map_list *maplist;
 	int err;
 
-	if (!(capable(CAP_SYS_ADMIN) || map->type == _DRM_AGP))
+	if (!(capable(CAP_SYS_ADMIN) || map->type == _DRM_AGP || map->type == _DRM_SHM))
 		return -EPERM;
 
 	err = drm_addmap_core(dev, map->offset, map->size, map->type,
@@ -380,10 +381,12 @@ int drm_rmmap_locked(struct drm_device *dev, drm_local_map_t *map)
 	struct drm_map_list *r_list = NULL, *list_t;
 	drm_dma_handle_t dmah;
 	int found = 0;
+	struct drm_master *master;
 
 	/* Find the list entry for the map and remove it */
 	list_for_each_entry_safe(r_list, list_t, &dev->maplist, head) {
 		if (r_list->map == map) {
+			master = r_list->master;
 			list_del(&r_list->head);
 			drm_ht_remove_key(&dev->map_hash,
 					  r_list->user_token >> PAGE_SHIFT);
@@ -409,6 +412,13 @@ int drm_rmmap_locked(struct drm_device *dev, drm_local_map_t *map)
 		break;
 	case _DRM_SHM:
 		vfree(map->handle);
+		if (master) {
+			if (dev->sigdata.lock == master->lock.hw_lock)
+				dev->sigdata.lock = NULL;
+			master->lock.hw_lock = NULL;   /* SHM removed */
+			master->lock.file_priv = NULL;
+			wake_up_interruptible(&master->lock.lock_queue);
+		}
 		break;
 	case _DRM_AGP:
 	case _DRM_SCATTER_GATHER:
diff --git a/drivers/gpu/drm/drm_context.c b/drivers/gpu/drm/drm_context.c
index d505f695421f..809ec0f03452 100644
--- a/drivers/gpu/drm/drm_context.c
+++ b/drivers/gpu/drm/drm_context.c
@@ -256,12 +256,13 @@ static int drm_context_switch(struct drm_device * dev, int old, int new)
  * hardware lock is held, clears the drm_device::context_flag and wakes up
  * drm_device::context_wait.
  */
-static int drm_context_switch_complete(struct drm_device * dev, int new)
+static int drm_context_switch_complete(struct drm_device *dev,
+				       struct drm_file *file_priv, int new)
 {
 	dev->last_context = new;	/* PRE/POST: This is the _only_ writer. */
 	dev->last_switch = jiffies;
 
-	if (!_DRM_LOCK_IS_HELD(dev->lock.hw_lock->lock)) {
+	if (!_DRM_LOCK_IS_HELD(file_priv->master->lock.hw_lock->lock)) {
 		DRM_ERROR("Lock isn't held after context switch\n");
 	}
 
@@ -420,7 +421,7 @@ int drm_newctx(struct drm_device *dev, void *data,
 	struct drm_ctx *ctx = data;
 
 	DRM_DEBUG("%d\n", ctx->handle);
-	drm_context_switch_complete(dev, ctx->handle);
+	drm_context_switch_complete(dev, file_priv, ctx->handle);
 
 	return 0;
 }
@@ -442,9 +443,6 @@ int drm_rmctx(struct drm_device *dev, void *data,
 	struct drm_ctx *ctx = data;
 
 	DRM_DEBUG("%d\n", ctx->handle);
-	if (ctx->handle == DRM_KERNEL_CONTEXT + 1) {
-		file_priv->remove_auth_on_close = 1;
-	}
 	if (ctx->handle != DRM_KERNEL_CONTEXT) {
 		if (dev->driver->context_dtor)
 			dev->driver->context_dtor(dev, ctx->handle);
diff --git a/drivers/gpu/drm/drm_drv.c b/drivers/gpu/drm/drm_drv.c
index 3cb87a932b33..9f04ca37df6d 100644
--- a/drivers/gpu/drm/drm_drv.c
+++ b/drivers/gpu/drm/drm_drv.c
@@ -74,6 +74,9 @@ static struct drm_ioctl_desc drm_ioctls[] = {
 	DRM_IOCTL_DEF(DRM_IOCTL_SET_SAREA_CTX, drm_setsareactx, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
 	DRM_IOCTL_DEF(DRM_IOCTL_GET_SAREA_CTX, drm_getsareactx, DRM_AUTH),
 
+	DRM_IOCTL_DEF(DRM_IOCTL_SET_MASTER, drm_setmaster_ioctl, DRM_ROOT_ONLY),
+	DRM_IOCTL_DEF(DRM_IOCTL_DROP_MASTER, drm_dropmaster_ioctl, DRM_ROOT_ONLY),
+
 	DRM_IOCTL_DEF(DRM_IOCTL_ADD_CTX, drm_addctx, DRM_AUTH|DRM_ROOT_ONLY),
 	DRM_IOCTL_DEF(DRM_IOCTL_RM_CTX, drm_rmctx, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
 	DRM_IOCTL_DEF(DRM_IOCTL_MOD_CTX, drm_modctx, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
@@ -138,8 +141,6 @@ static struct drm_ioctl_desc drm_ioctls[] = {
  */
 int drm_lastclose(struct drm_device * dev)
 {
-	struct drm_magic_entry *pt, *next;
-	struct drm_map_list *r_list, *list_t;
 	struct drm_vma_entry *vma, *vma_temp;
 	int i;
 
@@ -149,12 +150,6 @@ int drm_lastclose(struct drm_device * dev)
 		dev->driver->lastclose(dev);
 	DRM_DEBUG("driver lastclose completed\n");
 
-	if (dev->unique) {
-		drm_free(dev->unique, strlen(dev->unique) + 1, DRM_MEM_DRIVER);
-		dev->unique = NULL;
-		dev->unique_len = 0;
-	}
-
 	if (dev->irq_enabled)
 		drm_irq_uninstall(dev);
 
@@ -164,16 +159,6 @@ int drm_lastclose(struct drm_device * dev)
 	drm_drawable_free_all(dev);
 	del_timer(&dev->timer);
 
-	/* Clear pid list */
-	if (dev->magicfree.next) {
-		list_for_each_entry_safe(pt, next, &dev->magicfree, head) {
-			list_del(&pt->head);
-			drm_ht_remove_item(&dev->magiclist, &pt->hash_item);
-			drm_free(pt, sizeof(*pt), DRM_MEM_MAGIC);
-		}
-		drm_ht_remove(&dev->magiclist);
-	}
-
 	/* Clear AGP information */
 	if (drm_core_has_AGP(dev) && dev->agp) {
 		struct drm_agp_mem *entry, *tempe;
@@ -205,13 +190,6 @@ int drm_lastclose(struct drm_device * dev)
 		drm_free(vma, sizeof(*vma), DRM_MEM_VMAS);
 	}
 
-	list_for_each_entry_safe(r_list, list_t, &dev->maplist, head) {
-		if (!(r_list->map->flags & _DRM_DRIVER)) {
-			drm_rmmap_locked(dev, r_list->map);
-			r_list = NULL;
-		}
-	}
-
 	if (drm_core_check_feature(dev, DRIVER_DMA_QUEUE) && dev->queuelist) {
 		for (i = 0; i < dev->queue_count; i++) {
 			if (dev->queuelist[i]) {
@@ -231,11 +209,6 @@ int drm_lastclose(struct drm_device * dev)
 	if (drm_core_check_feature(dev, DRIVER_HAVE_DMA))
 		drm_dma_takedown(dev);
 
-	if (dev->lock.hw_lock) {
-		dev->sigdata.lock = dev->lock.hw_lock = NULL;	/* SHM removed */
-		dev->lock.file_priv = NULL;
-		wake_up_interruptible(&dev->lock.lock_queue);
-	}
 	mutex_unlock(&dev->struct_mutex);
 
 	DRM_DEBUG("lastclose completed\n");
diff --git a/drivers/gpu/drm/drm_fops.c b/drivers/gpu/drm/drm_fops.c
index 78eeed5caaff..f2285237df49 100644
--- a/drivers/gpu/drm/drm_fops.c
+++ b/drivers/gpu/drm/drm_fops.c
@@ -44,10 +44,8 @@ static int drm_open_helper(struct inode *inode, struct file *filp,
 
 static int drm_setup(struct drm_device * dev)
 {
-	drm_local_map_t *map;
 	int i;
 	int ret;
-	u32 sareapage;
 
 	if (dev->driver->firstopen) {
 		ret = dev->driver->firstopen(dev);
@@ -55,14 +53,6 @@ static int drm_setup(struct drm_device * dev)
 			return ret;
 	}
 
-	dev->magicfree.next = NULL;
-
-	/* prebuild the SAREA */
-	sareapage = max_t(unsigned, SAREA_MAX, PAGE_SIZE);
-	i = drm_addmap(dev, 0, sareapage, _DRM_SHM, _DRM_CONTAINS_LOCK, &map);
-	if (i != 0)
-		return i;
-
 	atomic_set(&dev->ioctl_count, 0);
 	atomic_set(&dev->vma_count, 0);
 	dev->buf_use = 0;
@@ -77,16 +67,12 @@ static int drm_setup(struct drm_device * dev)
 	for (i = 0; i < ARRAY_SIZE(dev->counts); i++)
 		atomic_set(&dev->counts[i], 0);
 
-	drm_ht_create(&dev->magiclist, DRM_MAGIC_HASH_ORDER);
-	INIT_LIST_HEAD(&dev->magicfree);
-
 	dev->sigdata.lock = NULL;
-	init_waitqueue_head(&dev->lock.lock_queue);
+
 	dev->queue_count = 0;
 	dev->queue_reserved = 0;
 	dev->queue_slots = 0;
 	dev->queuelist = NULL;
-	dev->irq_enabled = 0;
 	dev->context_flag = 0;
 	dev->interrupt_flag = 0;
 	dev->dma_flag = 0;
@@ -265,10 +251,42 @@ static int drm_open_helper(struct inode *inode, struct file *filp,
 			goto out_free;
 	}
 
+
+	/* if there is no current master make this fd it */
 	mutex_lock(&dev->struct_mutex);
-	if (list_empty(&dev->filelist))
-		priv->master = 1;
+	if (!priv->minor->master) {
+		/* create a new master */
+		priv->minor->master = drm_master_create(priv->minor);
+		if (!priv->minor->master) {
+			ret = -ENOMEM;
+			goto out_free;
+		}
+
+		priv->is_master = 1;
+		/* take another reference for the copy in the local file priv */
+		priv->master = drm_master_get(priv->minor->master);
+
+		priv->authenticated = 1;
+
+		mutex_unlock(&dev->struct_mutex);
+		if (dev->driver->master_create) {
+			ret = dev->driver->master_create(dev, priv->master);
+			if (ret) {
+				mutex_lock(&dev->struct_mutex);
+				/* drop both references if this fails */
+				drm_master_put(&priv->minor->master);
+				drm_master_put(&priv->master);
+				mutex_unlock(&dev->struct_mutex);
+				goto out_free;
+			}
+		}
+	} else {
+		/* get a reference to the master */
+		priv->master = drm_master_get(priv->minor->master);
+		mutex_unlock(&dev->struct_mutex);
+	}
 
+	mutex_lock(&dev->struct_mutex);
 	list_add(&priv->lhead, &dev->filelist);
 	mutex_unlock(&dev->struct_mutex);
 
@@ -314,6 +332,74 @@ int drm_fasync(int fd, struct file *filp, int on)
 }
 EXPORT_SYMBOL(drm_fasync);
 
+/*
+ * Reclaim locked buffers; note that this may be a bad idea if the current
+ * context doesn't have the hw lock...
+ */
+static void drm_reclaim_locked_buffers(struct drm_device *dev, struct file *f)
+{
+	struct drm_file *file_priv = f->private_data;
+
+	if (drm_i_have_hw_lock(dev, file_priv)) {
+		dev->driver->reclaim_buffers_locked(dev, file_priv);
+	} else {
+		unsigned long _end = jiffies + 3 * DRM_HZ;
+		int locked = 0;
+
+		drm_idlelock_take(&file_priv->master->lock);
+
+		/*
+		 * Wait for a while.
+		 */
+		do {
+			spin_lock_bh(&file_priv->master->lock.spinlock);
+			locked = file_priv->master->lock.idle_has_lock;
+			spin_unlock_bh(&file_priv->master->lock.spinlock);
+			if (locked)
+				break;
+			schedule();
+		} while (!time_after_eq(jiffies, _end));
+
+		if (!locked) {
+			DRM_ERROR("reclaim_buffers_locked() deadlock. Please rework this\n"
+				  "\tdriver to use reclaim_buffers_idlelocked() instead.\n"
+				  "\tI will go on reclaiming the buffers anyway.\n");
+		}
+
+		dev->driver->reclaim_buffers_locked(dev, file_priv);
+		drm_idlelock_release(&file_priv->master->lock);
+	}
+}
+
+static void drm_master_release(struct drm_device *dev, struct file *filp)
+{
+	struct drm_file *file_priv = filp->private_data;
+
+	if (dev->driver->reclaim_buffers_locked &&
+	    file_priv->master->lock.hw_lock)
+		drm_reclaim_locked_buffers(dev, filp);
+
+	if (dev->driver->reclaim_buffers_idlelocked &&
+	    file_priv->master->lock.hw_lock) {
+		drm_idlelock_take(&file_priv->master->lock);
+		dev->driver->reclaim_buffers_idlelocked(dev, file_priv);
+		drm_idlelock_release(&file_priv->master->lock);
+	}
+
+
+	if (drm_i_have_hw_lock(dev, file_priv)) {
+		DRM_DEBUG("File %p released, freeing lock for context %d\n",
+			  filp, _DRM_LOCKING_CONTEXT(file_priv->master->lock.hw_lock->lock));
+		drm_lock_free(&file_priv->master->lock,
+			      _DRM_LOCKING_CONTEXT(file_priv->master->lock.hw_lock->lock));
+	}
+
+	if (drm_core_check_feature(dev, DRIVER_HAVE_DMA) &&
+	    !dev->driver->reclaim_buffers_locked) {
+		dev->driver->reclaim_buffers(dev, file_priv);
+	}
+}
+
 /**
  * Release file.
  *
@@ -348,60 +434,9 @@ int drm_release(struct inode *inode, struct file *filp)
 		  (long)old_encode_dev(file_priv->minor->device),
 		  dev->open_count);
 
-	if (dev->driver->reclaim_buffers_locked && dev->lock.hw_lock) {
-		if (drm_i_have_hw_lock(dev, file_priv)) {
-			dev->driver->reclaim_buffers_locked(dev, file_priv);
-		} else {
-			unsigned long endtime = jiffies + 3 * DRM_HZ;
-			int locked = 0;
-
-			drm_idlelock_take(&dev->lock);
-
-			/*
-			 * Wait for a while.
-			 */
-
-			do{
-				spin_lock_bh(&dev->lock.spinlock);
-				locked = dev->lock.idle_has_lock;
-				spin_unlock_bh(&dev->lock.spinlock);
-				if (locked)
-					break;
-				schedule();
-			} while (!time_after_eq(jiffies, endtime));
-
-			if (!locked) {
-				DRM_ERROR("reclaim_buffers_locked() deadlock. Please rework this\n"
-					  "\tdriver to use reclaim_buffers_idlelocked() instead.\n"
-					  "\tI will go on reclaiming the buffers anyway.\n");
-			}
-
-			dev->driver->reclaim_buffers_locked(dev, file_priv);
-			drm_idlelock_release(&dev->lock);
-		}
-	}
-
-	if (dev->driver->reclaim_buffers_idlelocked && dev->lock.hw_lock) {
-
-		drm_idlelock_take(&dev->lock);
-		dev->driver->reclaim_buffers_idlelocked(dev, file_priv);
-		drm_idlelock_release(&dev->lock);
-
-	}
-
-	if (drm_i_have_hw_lock(dev, file_priv)) {
-		DRM_DEBUG("File %p released, freeing lock for context %d\n",
-			  filp, _DRM_LOCKING_CONTEXT(dev->lock.hw_lock->lock));
-
-		drm_lock_free(&dev->lock,
-			      _DRM_LOCKING_CONTEXT(dev->lock.hw_lock->lock));
-	}
-
-
-	if (drm_core_check_feature(dev, DRIVER_HAVE_DMA) &&
-	    !dev->driver->reclaim_buffers_locked) {
-		dev->driver->reclaim_buffers(dev, file_priv);
-	}
+	/* if the master has gone away we can't do anything with the lock */
+	if (file_priv->minor->master)
+		drm_master_release(dev, filp);
 
 	if (dev->driver->driver_features & DRIVER_GEM)
 		drm_gem_release(dev, file_priv);
@@ -428,12 +463,24 @@ int drm_release(struct inode *inode, struct file *filp)
 	mutex_unlock(&dev->ctxlist_mutex);
 
 	mutex_lock(&dev->struct_mutex);
-	if (file_priv->remove_auth_on_close == 1) {
+
+	if (file_priv->is_master) {
 		struct drm_file *temp;
+		list_for_each_entry(temp, &dev->filelist, lhead) {
+			if ((temp->master == file_priv->master) &&
+			    (temp != file_priv))
+				temp->authenticated = 0;
+		}
 
-		list_for_each_entry(temp, &dev->filelist, lhead)
-			temp->authenticated = 0;
+		if (file_priv->minor->master == file_priv->master) {
+			/* drop the reference held my the minor */
+			drm_master_put(&file_priv->minor->master);
+		}
 	}
+
+	/* drop the reference held my the file priv */
+	drm_master_put(&file_priv->master);
+	file_priv->is_master = 0;
 	list_del(&file_priv->lhead);
 	mutex_unlock(&dev->struct_mutex);
 
@@ -448,9 +495,9 @@ int drm_release(struct inode *inode, struct file *filp)
 	atomic_inc(&dev->counts[_DRM_STAT_CLOSES]);
 	spin_lock(&dev->count_lock);
 	if (!--dev->open_count) {
-		if (atomic_read(&dev->ioctl_count) || dev->blocked) {
-			DRM_ERROR("Device busy: %d %d\n",
-				  atomic_read(&dev->ioctl_count), dev->blocked);
+		if (atomic_read(&dev->ioctl_count)) {
+			DRM_ERROR("Device busy: %d\n",
+				  atomic_read(&dev->ioctl_count));
 			spin_unlock(&dev->count_lock);
 			unlock_kernel();
 			return -EBUSY;
diff --git a/drivers/gpu/drm/drm_ioctl.c b/drivers/gpu/drm/drm_ioctl.c
index 16829fb3089d..e35126a35093 100644
--- a/drivers/gpu/drm/drm_ioctl.c
+++ b/drivers/gpu/drm/drm_ioctl.c
@@ -53,12 +53,13 @@ int drm_getunique(struct drm_device *dev, void *data,
 		  struct drm_file *file_priv)
 {
 	struct drm_unique *u = data;
+	struct drm_master *master = file_priv->master;
 
-	if (u->unique_len >= dev->unique_len) {
-		if (copy_to_user(u->unique, dev->unique, dev->unique_len))
+	if (u->unique_len >= master->unique_len) {
+		if (copy_to_user(u->unique, master->unique, master->unique_len))
 			return -EFAULT;
 	}
-	u->unique_len = dev->unique_len;
+	u->unique_len = master->unique_len;
 
 	return 0;
 }
@@ -81,36 +82,37 @@ int drm_setunique(struct drm_device *dev, void *data,
 		  struct drm_file *file_priv)
 {
 	struct drm_unique *u = data;
+	struct drm_master *master = file_priv->master;
 	int domain, bus, slot, func, ret;
 
-	if (dev->unique_len || dev->unique)
+	if (master->unique_len || master->unique)
 		return -EBUSY;
 
 	if (!u->unique_len || u->unique_len > 1024)
 		return -EINVAL;
 
-	dev->unique_len = u->unique_len;
-	dev->unique = drm_alloc(u->unique_len + 1, DRM_MEM_DRIVER);
-	if (!dev->unique)
+	master->unique_len = u->unique_len;
+	master->unique = drm_alloc(u->unique_len + 1, DRM_MEM_DRIVER);
+	if (!master->unique)
 		return -ENOMEM;
-	if (copy_from_user(dev->unique, u->unique, dev->unique_len))
+	if (copy_from_user(master->unique, u->unique, master->unique_len))
 		return -EFAULT;
 
-	dev->unique[dev->unique_len] = '\0';
+	master->unique[master->unique_len] = '\0';
 
 	dev->devname =
 	    drm_alloc(strlen(dev->driver->pci_driver.name) +
-		      strlen(dev->unique) + 2, DRM_MEM_DRIVER);
+		      strlen(master->unique) + 2, DRM_MEM_DRIVER);
 	if (!dev->devname)
 		return -ENOMEM;
 
 	sprintf(dev->devname, "%s@%s", dev->driver->pci_driver.name,
-		dev->unique);
+		master->unique);
 
 	/* Return error if the busid submitted doesn't match the device's actual
 	 * busid.
 	 */
-	ret = sscanf(dev->unique, "PCI:%d:%d:%d", &bus, &slot, &func);
+	ret = sscanf(master->unique, "PCI:%d:%d:%d", &bus, &slot, &func);
 	if (ret != 3)
 		return -EINVAL;
 	domain = bus >> 8;
@@ -125,34 +127,35 @@ int drm_setunique(struct drm_device *dev, void *data,
 	return 0;
 }
 
-static int drm_set_busid(struct drm_device * dev)
+static int drm_set_busid(struct drm_device *dev, struct drm_file *file_priv)
 {
+	struct drm_master *master = file_priv->master;
 	int len;
 
-	if (dev->unique != NULL)
-		return 0;
+	if (master->unique != NULL)
+		return -EBUSY;
 
-	dev->unique_len = 40;
-	dev->unique = drm_alloc(dev->unique_len + 1, DRM_MEM_DRIVER);
-	if (dev->unique == NULL)
+	master->unique_len = 40;
+	master->unique = drm_alloc(master->unique_len + 1, DRM_MEM_DRIVER);
+	if (master->unique == NULL)
 		return -ENOMEM;
 
-	len = snprintf(dev->unique, dev->unique_len, "pci:%04x:%02x:%02x.%d",
-		       drm_get_pci_domain(dev), dev->pdev->bus->number,
+	len = snprintf(master->unique, master->unique_len, "pci:%04x:%02x:%02x.%d",
+		       drm_get_pci_domain(dev),
+		       dev->pdev->bus->number,
 		       PCI_SLOT(dev->pdev->devfn),
 		       PCI_FUNC(dev->pdev->devfn));
-
-	if (len > dev->unique_len)
-		DRM_ERROR("Unique buffer overflowed\n");
+	if (len > master->unique_len)
+		DRM_ERROR("buffer overflow");
 
 	dev->devname =
-	    drm_alloc(strlen(dev->driver->pci_driver.name) + dev->unique_len +
+	    drm_alloc(strlen(dev->driver->pci_driver.name) + master->unique_len +
 		      2, DRM_MEM_DRIVER);
 	if (dev->devname == NULL)
 		return -ENOMEM;
 
 	sprintf(dev->devname, "%s@%s", dev->driver->pci_driver.name,
-		dev->unique);
+		master->unique);
 
 	return 0;
 }
@@ -276,7 +279,7 @@ int drm_getstats(struct drm_device *dev, void *data,
 	for (i = 0; i < dev->counters; i++) {
 		if (dev->types[i] == _DRM_STAT_LOCK)
 			stats->data[i].value =
-			    (dev->lock.hw_lock ? dev->lock.hw_lock->lock : 0);
+			    (file_priv->master->lock.hw_lock ? file_priv->master->lock.hw_lock->lock : 0);
 		else
 			stats->data[i].value = atomic_read(&dev->counts[i]);
 		stats->data[i].type = dev->types[i];
@@ -318,7 +321,7 @@ int drm_setversion(struct drm_device *dev, void *data, struct drm_file *file_pri
 			/*
 			 * Version 1.1 includes tying of DRM to specific device
 			 */
-			drm_set_busid(dev);
+			drm_set_busid(dev, file_priv);
 		}
 	}
 
diff --git a/drivers/gpu/drm/drm_lock.c b/drivers/gpu/drm/drm_lock.c
index 1cfa72031f8f..46e7b28f0707 100644
--- a/drivers/gpu/drm/drm_lock.c
+++ b/drivers/gpu/drm/drm_lock.c
@@ -52,6 +52,7 @@ int drm_lock(struct drm_device *dev, void *data, struct drm_file *file_priv)
 {
 	DECLARE_WAITQUEUE(entry, current);
 	struct drm_lock *lock = data;
+	struct drm_master *master = file_priv->master;
 	int ret = 0;
 
 	++file_priv->lock_count;
@@ -64,26 +65,27 @@ int drm_lock(struct drm_device *dev, void *data, struct drm_file *file_priv)
 
 	DRM_DEBUG("%d (pid %d) requests lock (0x%08x), flags = 0x%08x\n",
 		  lock->context, task_pid_nr(current),
-		  dev->lock.hw_lock->lock, lock->flags);
+		  master->lock.hw_lock->lock, lock->flags);
 
 	if (drm_core_check_feature(dev, DRIVER_DMA_QUEUE))
 		if (lock->context < 0)
 			return -EINVAL;
 
-	add_wait_queue(&dev->lock.lock_queue, &entry);
-	spin_lock_bh(&dev->lock.spinlock);
-	dev->lock.user_waiters++;
-	spin_unlock_bh(&dev->lock.spinlock);
+	add_wait_queue(&master->lock.lock_queue, &entry);
+	spin_lock_bh(&master->lock.spinlock);
+	master->lock.user_waiters++;
+	spin_unlock_bh(&master->lock.spinlock);
+
 	for (;;) {
 		__set_current_state(TASK_INTERRUPTIBLE);
-		if (!dev->lock.hw_lock) {
+		if (!master->lock.hw_lock) {
 			/* Device has been unregistered */
 			ret = -EINTR;
 			break;
 		}
-		if (drm_lock_take(&dev->lock, lock->context)) {
-			dev->lock.file_priv = file_priv;
-			dev->lock.lock_time = jiffies;
+		if (drm_lock_take(&master->lock, lock->context)) {
+			master->lock.file_priv = file_priv;
+			master->lock.lock_time = jiffies;
 			atomic_inc(&dev->counts[_DRM_STAT_LOCKS]);
 			break;	/* Got lock */
 		}
@@ -95,11 +97,11 @@ int drm_lock(struct drm_device *dev, void *data, struct drm_file *file_priv)
 			break;
 		}
 	}
-	spin_lock_bh(&dev->lock.spinlock);
-	dev->lock.user_waiters--;
-	spin_unlock_bh(&dev->lock.spinlock);
+	spin_lock_bh(&master->lock.spinlock);
+	master->lock.user_waiters--;
+	spin_unlock_bh(&master->lock.spinlock);
 	__set_current_state(TASK_RUNNING);
-	remove_wait_queue(&dev->lock.lock_queue, &entry);
+	remove_wait_queue(&master->lock.lock_queue, &entry);
 
 	DRM_DEBUG("%d %s\n", lock->context,
 		  ret ? "interrupted" : "has lock");
@@ -108,14 +110,14 @@ int drm_lock(struct drm_device *dev, void *data, struct drm_file *file_priv)
 	/* don't set the block all signals on the master process for now 
 	 * really probably not the correct answer but lets us debug xkb
  	 * xserver for now */
-	if (!file_priv->master) {
+	if (!file_priv->is_master) {
 		sigemptyset(&dev->sigmask);
 		sigaddset(&dev->sigmask, SIGSTOP);
 		sigaddset(&dev->sigmask, SIGTSTP);
 		sigaddset(&dev->sigmask, SIGTTIN);
 		sigaddset(&dev->sigmask, SIGTTOU);
 		dev->sigdata.context = lock->context;
-		dev->sigdata.lock = dev->lock.hw_lock;
+		dev->sigdata.lock = master->lock.hw_lock;
 		block_all_signals(drm_notifier, &dev->sigdata, &dev->sigmask);
 	}
 
@@ -154,6 +156,7 @@ int drm_lock(struct drm_device *dev, void *data, struct drm_file *file_priv)
 int drm_unlock(struct drm_device *dev, void *data, struct drm_file *file_priv)
 {
 	struct drm_lock *lock = data;
+	struct drm_master *master = file_priv->master;
 
 	if (lock->context == DRM_KERNEL_CONTEXT) {
 		DRM_ERROR("Process %d using kernel context %d\n",
@@ -169,7 +172,7 @@ int drm_unlock(struct drm_device *dev, void *data, struct drm_file *file_priv)
 	if (dev->driver->kernel_context_switch_unlock)
 		dev->driver->kernel_context_switch_unlock(dev);
 	else {
-		if (drm_lock_free(&dev->lock,lock->context)) {
+		if (drm_lock_free(&master->lock, lock->context)) {
 			/* FIXME: Should really bail out here. */
 		}
 	}
@@ -379,9 +382,10 @@ EXPORT_SYMBOL(drm_idlelock_release);
 
 int drm_i_have_hw_lock(struct drm_device *dev, struct drm_file *file_priv)
 {
-	return (file_priv->lock_count && dev->lock.hw_lock &&
-		_DRM_LOCK_IS_HELD(dev->lock.hw_lock->lock) &&
-		dev->lock.file_priv == file_priv);
+	struct drm_master *master = file_priv->master;
+	return (file_priv->lock_count && master->lock.hw_lock &&
+		_DRM_LOCK_IS_HELD(master->lock.hw_lock->lock) &&
+		master->lock.file_priv == file_priv);
 }
 
 EXPORT_SYMBOL(drm_i_have_hw_lock);
diff --git a/drivers/gpu/drm/drm_proc.c b/drivers/gpu/drm/drm_proc.c
index ae73b7f7249a..7dbaa1a19ea4 100644
--- a/drivers/gpu/drm/drm_proc.c
+++ b/drivers/gpu/drm/drm_proc.c
@@ -195,6 +195,7 @@ static int drm_name_info(char *buf, char **start, off_t offset, int request,
 			 int *eof, void *data)
 {
 	struct drm_minor *minor = (struct drm_minor *) data;
+	struct drm_master *master = minor->master;
 	struct drm_device *dev = minor->dev;
 	int len = 0;
 
@@ -203,13 +204,16 @@ static int drm_name_info(char *buf, char **start, off_t offset, int request,
 		return 0;
 	}
 
+	if (!master)
+		return 0;
+
 	*start = &buf[offset];
 	*eof = 0;
 
-	if (dev->unique) {
+	if (master->unique) {
 		DRM_PROC_PRINT("%s %s %s\n",
 			       dev->driver->pci_driver.name,
-			       pci_name(dev->pdev), dev->unique);
+			       pci_name(dev->pdev), master->unique);
 	} else {
 		DRM_PROC_PRINT("%s %s\n", dev->driver->pci_driver.name,
 			       pci_name(dev->pdev));
diff --git a/drivers/gpu/drm/drm_stub.c b/drivers/gpu/drm/drm_stub.c
index 849c0a9fe7fd..0f24c2dcd517 100644
--- a/drivers/gpu/drm/drm_stub.c
+++ b/drivers/gpu/drm/drm_stub.c
@@ -79,6 +79,104 @@ again:
 	return new_id;
 }
 
+struct drm_master *drm_master_create(struct drm_minor *minor)
+{
+	struct drm_master *master;
+
+	master = drm_calloc(1, sizeof(*master), DRM_MEM_DRIVER);
+	if (!master)
+		return NULL;
+
+	kref_init(&master->refcount);
+	spin_lock_init(&master->lock.spinlock);
+	init_waitqueue_head(&master->lock.lock_queue);
+	drm_ht_create(&master->magiclist, DRM_MAGIC_HASH_ORDER);
+	INIT_LIST_HEAD(&master->magicfree);
+	master->minor = minor;
+
+	list_add_tail(&master->head, &minor->master_list);
+
+	return master;
+}
+
+struct drm_master *drm_master_get(struct drm_master *master)
+{
+	kref_get(&master->refcount);
+	return master;
+}
+
+static void drm_master_destroy(struct kref *kref)
+{
+	struct drm_master *master = container_of(kref, struct drm_master, refcount);
+	struct drm_magic_entry *pt, *next;
+	struct drm_device *dev = master->minor->dev;
+
+	list_del(&master->head);
+
+	if (dev->driver->master_destroy)
+		dev->driver->master_destroy(dev, master);
+
+	if (master->unique) {
+		drm_free(master->unique, strlen(master->unique) + 1, DRM_MEM_DRIVER);
+		master->unique = NULL;
+		master->unique_len = 0;
+	}
+
+	list_for_each_entry_safe(pt, next, &master->magicfree, head) {
+		list_del(&pt->head);
+		drm_ht_remove_item(&master->magiclist, &pt->hash_item);
+		drm_free(pt, sizeof(*pt), DRM_MEM_MAGIC);
+	}
+
+	drm_ht_remove(&master->magiclist);
+
+	if (master->lock.hw_lock) {
+		if (dev->sigdata.lock == master->lock.hw_lock)
+			dev->sigdata.lock = NULL;
+		master->lock.hw_lock = NULL;
+		master->lock.file_priv = NULL;
+		wake_up_interruptible(&master->lock.lock_queue);
+	}
+
+	drm_free(master, sizeof(*master), DRM_MEM_DRIVER);
+}
+
+void drm_master_put(struct drm_master **master)
+{
+	kref_put(&(*master)->refcount, drm_master_destroy);
+	*master = NULL;
+}
+
+int drm_setmaster_ioctl(struct drm_device *dev, void *data,
+			struct drm_file *file_priv)
+{
+	if (file_priv->minor->master && file_priv->minor->master != file_priv->master)
+		return -EINVAL;
+
+	if (!file_priv->master)
+		return -EINVAL;
+
+	if (!file_priv->minor->master &&
+	    file_priv->minor->master != file_priv->master) {
+		mutex_lock(&dev->struct_mutex);
+		file_priv->minor->master = drm_master_get(file_priv->master);
+		mutex_lock(&dev->struct_mutex);
+	}
+
+	return 0;
+}
+
+int drm_dropmaster_ioctl(struct drm_device *dev, void *data,
+			 struct drm_file *file_priv)
+{
+	if (!file_priv->master)
+		return -EINVAL;
+	mutex_lock(&dev->struct_mutex);
+	drm_master_put(&file_priv->minor->master);
+	mutex_unlock(&dev->struct_mutex);
+	return 0;
+}
+
 static int drm_fill_in_dev(struct drm_device * dev, struct pci_dev *pdev,
 			   const struct pci_device_id *ent,
 			   struct drm_driver *driver)
@@ -92,7 +190,6 @@ static int drm_fill_in_dev(struct drm_device * dev, struct pci_dev *pdev,
 
 	spin_lock_init(&dev->count_lock);
 	spin_lock_init(&dev->drw_lock);
-	spin_lock_init(&dev->lock.spinlock);
 	init_timer(&dev->timer);
 	mutex_init(&dev->struct_mutex);
 	mutex_init(&dev->ctxlist_mutex);
@@ -200,6 +297,7 @@ static int drm_get_minor(struct drm_device *dev, struct drm_minor **minor, int t
 	new_minor->device = MKDEV(DRM_MAJOR, minor_id);
 	new_minor->dev = dev;
 	new_minor->index = minor_id;
+	INIT_LIST_HEAD(&new_minor->master_list);
 
 	idr_replace(&drm_minors_idr, new_minor, minor_id);
 
@@ -299,11 +397,6 @@ int drm_put_dev(struct drm_device * dev)
 {
 	DRM_DEBUG("release primary %s\n", dev->driver->pci_driver.name);
 
-	if (dev->unique) {
-		drm_free(dev->unique, strlen(dev->unique) + 1, DRM_MEM_DRIVER);
-		dev->unique = NULL;
-		dev->unique_len = 0;
-	}
 	if (dev->devname) {
 		drm_free(dev->devname, strlen(dev->devname) + 1,
 			 DRM_MEM_DRIVER);
diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
index afa8a12cd009..dacdf3c577cb 100644
--- a/drivers/gpu/drm/i915/i915_dma.c
+++ b/drivers/gpu/drm/i915/i915_dma.c
@@ -39,6 +39,7 @@
 int i915_wait_ring(struct drm_device * dev, int n, const char *caller)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
+	struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv;
 	drm_i915_ring_buffer_t *ring = &(dev_priv->ring);
 	u32 acthd_reg = IS_I965G(dev) ? ACTHD_I965 : ACTHD;
 	u32 last_acthd = I915_READ(acthd_reg);
@@ -55,8 +56,8 @@ int i915_wait_ring(struct drm_device * dev, int n, const char *caller)
 		if (ring->space >= n)
 			return 0;
 
-		if (dev_priv->sarea_priv)
-			dev_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT;
+		if (master_priv->sarea_priv)
+			master_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT;
 
 		if (ring->head != last_head)
 			i = 0;
@@ -121,6 +122,7 @@ static void i915_free_hws(struct drm_device *dev)
 void i915_kernel_lost_context(struct drm_device * dev)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
+	struct drm_i915_master_private *master_priv;
 	drm_i915_ring_buffer_t *ring = &(dev_priv->ring);
 
 	ring->head = I915_READ(PRB0_HEAD) & HEAD_ADDR;
@@ -129,8 +131,12 @@ void i915_kernel_lost_context(struct drm_device * dev)
 	if (ring->space < 0)
 		ring->space += ring->Size;
 
-	if (ring->head == ring->tail && dev_priv->sarea_priv)
-		dev_priv->sarea_priv->perf_boxes |= I915_BOX_RING_EMPTY;
+	if (!dev->primary->master)
+		return;
+
+	master_priv = dev->primary->master->driver_priv;
+	if (ring->head == ring->tail && master_priv->sarea_priv)
+		master_priv->sarea_priv->perf_boxes |= I915_BOX_RING_EMPTY;
 }
 
 static int i915_dma_cleanup(struct drm_device * dev)
@@ -154,25 +160,13 @@ static int i915_dma_cleanup(struct drm_device * dev)
 	if (I915_NEED_GFX_HWS(dev))
 		i915_free_hws(dev);
 
-	dev_priv->sarea = NULL;
-	dev_priv->sarea_priv = NULL;
-
 	return 0;
 }
 
 static int i915_initialize(struct drm_device * dev, drm_i915_init_t * init)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
-
-	dev_priv->sarea = drm_getsarea(dev);
-	if (!dev_priv->sarea) {
-		DRM_ERROR("can not find sarea!\n");
-		i915_dma_cleanup(dev);
-		return -EINVAL;
-	}
-
-	dev_priv->sarea_priv = (drm_i915_sarea_t *)
-	    ((u8 *) dev_priv->sarea->handle + init->sarea_priv_offset);
+	struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv;
 
 	if (init->ring_size != 0) {
 		if (dev_priv->ring.ring_obj != NULL) {
@@ -207,7 +201,8 @@ static int i915_initialize(struct drm_device * dev, drm_i915_init_t * init)
 	dev_priv->back_offset = init->back_offset;
 	dev_priv->front_offset = init->front_offset;
 	dev_priv->current_page = 0;
-	dev_priv->sarea_priv->pf_current_page = dev_priv->current_page;
+	if (master_priv->sarea_priv)
+		master_priv->sarea_priv->pf_current_page = 0;
 
 	/* Allow hardware batchbuffers unless told otherwise.
 	 */
@@ -222,11 +217,6 @@ static int i915_dma_resume(struct drm_device * dev)
 
 	DRM_DEBUG("%s\n", __func__);
 
-	if (!dev_priv->sarea) {
-		DRM_ERROR("can not find sarea!\n");
-		return -EINVAL;
-	}
-
 	if (dev_priv->ring.map.handle == NULL) {
 		DRM_ERROR("can not ioremap virtual address for"
 			  " ring buffer\n");
@@ -435,13 +425,14 @@ i915_emit_box(struct drm_device *dev,
 static void i915_emit_breadcrumb(struct drm_device *dev)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
+	struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv;
 	RING_LOCALS;
 
 	dev_priv->counter++;
 	if (dev_priv->counter > 0x7FFFFFFFUL)
 		dev_priv->counter = 0;
-	if (dev_priv->sarea_priv)
-		dev_priv->sarea_priv->last_enqueue = dev_priv->counter;
+	if (master_priv->sarea_priv)
+		master_priv->sarea_priv->last_enqueue = dev_priv->counter;
 
 	BEGIN_LP_RING(4);
 	OUT_RING(MI_STORE_DWORD_INDEX);
@@ -537,15 +528,17 @@ static int i915_dispatch_batchbuffer(struct drm_device * dev,
 static int i915_dispatch_flip(struct drm_device * dev)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
+	struct drm_i915_master_private *master_priv =
+		dev->primary->master->driver_priv;
 	RING_LOCALS;
 
-	if (!dev_priv->sarea_priv)
+	if (!master_priv->sarea_priv)
 		return -EINVAL;
 
 	DRM_DEBUG("%s: page=%d pfCurrentPage=%d\n",
 		  __func__,
 		  dev_priv->current_page,
-		  dev_priv->sarea_priv->pf_current_page);
+		  master_priv->sarea_priv->pf_current_page);
 
 	i915_kernel_lost_context(dev);
 
@@ -572,7 +565,7 @@ static int i915_dispatch_flip(struct drm_device * dev)
 	OUT_RING(0);
 	ADVANCE_LP_RING();
 
-	dev_priv->sarea_priv->last_enqueue = dev_priv->counter++;
+	master_priv->sarea_priv->last_enqueue = dev_priv->counter++;
 
 	BEGIN_LP_RING(4);
 	OUT_RING(MI_STORE_DWORD_INDEX);
@@ -581,7 +574,7 @@ static int i915_dispatch_flip(struct drm_device * dev)
 	OUT_RING(0);
 	ADVANCE_LP_RING();
 
-	dev_priv->sarea_priv->pf_current_page = dev_priv->current_page;
+	master_priv->sarea_priv->pf_current_page = dev_priv->current_page;
 	return 0;
 }
 
@@ -611,8 +604,9 @@ static int i915_batchbuffer(struct drm_device *dev, void *data,
 			    struct drm_file *file_priv)
 {
 	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
+	struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv;
 	drm_i915_sarea_t *sarea_priv = (drm_i915_sarea_t *)
-	    dev_priv->sarea_priv;
+	    master_priv->sarea_priv;
 	drm_i915_batchbuffer_t *batch = data;
 	int ret;
 
@@ -644,8 +638,9 @@ static int i915_cmdbuffer(struct drm_device *dev, void *data,
 			  struct drm_file *file_priv)
 {
 	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
+	struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv;
 	drm_i915_sarea_t *sarea_priv = (drm_i915_sarea_t *)
-	    dev_priv->sarea_priv;
+	    master_priv->sarea_priv;
 	drm_i915_cmdbuffer_t *cmdbuf = data;
 	int ret;
 
@@ -802,6 +797,30 @@ static int i915_set_status_page(struct drm_device *dev, void *data,
 	return 0;
 }
 
+int i915_master_create(struct drm_device *dev, struct drm_master *master)
+{
+	struct drm_i915_master_private *master_priv;
+
+	master_priv = drm_calloc(1, sizeof(*master_priv), DRM_MEM_DRIVER);
+	if (!master_priv)
+		return -ENOMEM;
+
+	master->driver_priv = master_priv;
+	return 0;
+}
+
+void i915_master_destroy(struct drm_device *dev, struct drm_master *master)
+{
+	struct drm_i915_master_private *master_priv = master->driver_priv;
+
+	if (!master_priv)
+		return;
+
+	drm_free(master_priv, sizeof(*master_priv), DRM_MEM_DRIVER);
+
+	master->driver_priv = NULL;
+}
+
 int i915_driver_load(struct drm_device *dev, unsigned long flags)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index a80ead215282..c91648320a8b 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -107,6 +107,8 @@ static struct drm_driver driver = {
 	.reclaim_buffers = drm_core_reclaim_buffers,
 	.get_map_ofs = drm_core_get_map_ofs,
 	.get_reg_ofs = drm_core_get_reg_ofs,
+	.master_create = i915_master_create,
+	.master_destroy = i915_master_destroy,
 	.proc_init = i915_gem_proc_init,
 	.proc_cleanup = i915_gem_proc_cleanup,
 	.gem_init_object = i915_gem_init_object,
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index b3cc4731aa7c..ba096f9a7641 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -103,15 +103,18 @@ struct intel_opregion {
 	int enabled;
 };
 
+struct drm_i915_master_private {
+	drm_local_map_t *sarea;
+	struct _drm_i915_sarea *sarea_priv;
+};
+
 typedef struct drm_i915_private {
 	struct drm_device *dev;
 
 	int has_gem;
 
 	void __iomem *regs;
-	drm_local_map_t *sarea;
 
-	drm_i915_sarea_t *sarea_priv;
 	drm_i915_ring_buffer_t ring;
 
 	drm_dma_handle_t *status_page_dmah;
@@ -417,6 +420,9 @@ struct drm_i915_file_private {
 extern struct drm_ioctl_desc i915_ioctls[];
 extern int i915_max_ioctl;
 
+extern int i915_master_create(struct drm_device *dev, struct drm_master *master);
+extern void i915_master_destroy(struct drm_device *dev, struct drm_master *master);
+
 				/* i915_dma.c */
 extern void i915_kernel_lost_context(struct drm_device * dev);
 extern int i915_driver_load(struct drm_device *, unsigned long flags);
diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
index 69b9a42da95e..9b673d2f912b 100644
--- a/drivers/gpu/drm/i915/i915_irq.c
+++ b/drivers/gpu/drm/i915/i915_irq.c
@@ -168,6 +168,7 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS)
 {
 	struct drm_device *dev = (struct drm_device *) arg;
 	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
+	struct drm_i915_master_private *master_priv;
 	u32 iir, new_iir;
 	u32 pipea_stats, pipeb_stats;
 	u32 vblank_status;
@@ -222,9 +223,12 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS)
 		I915_WRITE(IIR, iir);
 		new_iir = I915_READ(IIR); /* Flush posted writes */
 
-		if (dev_priv->sarea_priv)
-			dev_priv->sarea_priv->last_dispatch =
-				READ_BREADCRUMB(dev_priv);
+		if (dev->primary->master) {
+			master_priv = dev->primary->master->driver_priv;
+			if (master_priv->sarea_priv)
+				master_priv->sarea_priv->last_dispatch =
+					READ_BREADCRUMB(dev_priv);
+		}
 
 		if (iir & I915_USER_INTERRUPT) {
 			dev_priv->mm.irq_gem_seqno = i915_get_gem_seqno(dev);
@@ -269,6 +273,7 @@ irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS)
 static int i915_emit_irq(struct drm_device * dev)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
+	struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv;
 	RING_LOCALS;
 
 	i915_kernel_lost_context(dev);
@@ -278,8 +283,8 @@ static int i915_emit_irq(struct drm_device * dev)
 	dev_priv->counter++;
 	if (dev_priv->counter > 0x7FFFFFFFUL)
 		dev_priv->counter = 1;
-	if (dev_priv->sarea_priv)
-		dev_priv->sarea_priv->last_enqueue = dev_priv->counter;
+	if (master_priv->sarea_priv)
+		master_priv->sarea_priv->last_enqueue = dev_priv->counter;
 
 	BEGIN_LP_RING(4);
 	OUT_RING(MI_STORE_DWORD_INDEX);
@@ -317,21 +322,20 @@ void i915_user_irq_put(struct drm_device *dev)
 static int i915_wait_irq(struct drm_device * dev, int irq_nr)
 {
 	drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private;
+	struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv;
 	int ret = 0;
 
 	DRM_DEBUG("irq_nr=%d breadcrumb=%d\n", irq_nr,
 		  READ_BREADCRUMB(dev_priv));
 
 	if (READ_BREADCRUMB(dev_priv) >= irq_nr) {
-		if (dev_priv->sarea_priv) {
-			dev_priv->sarea_priv->last_dispatch =
-				READ_BREADCRUMB(dev_priv);
-		}
+		if (master_priv->sarea_priv)
+			master_priv->sarea_priv->last_dispatch = READ_BREADCRUMB(dev_priv);
 		return 0;
 	}
 
-	if (dev_priv->sarea_priv)
-		dev_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT;
+	if (master_priv->sarea_priv)
+		master_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT;
 
 	i915_user_irq_get(dev);
 	DRM_WAIT_ON(ret, dev_priv->irq_queue, 3 * DRM_HZ,
@@ -343,10 +347,6 @@ static int i915_wait_irq(struct drm_device * dev, int irq_nr)
 			  READ_BREADCRUMB(dev_priv), (int)dev_priv->counter);
 	}
 
-	if (dev_priv->sarea_priv)
-		dev_priv->sarea_priv->last_dispatch =
-			READ_BREADCRUMB(dev_priv);
-
 	return ret;
 }
 
diff --git a/drivers/gpu/drm/i915/i915_mem.c b/drivers/gpu/drm/i915/i915_mem.c
index 6126a60dc9cb..96e271986d2a 100644
--- a/drivers/gpu/drm/i915/i915_mem.c
+++ b/drivers/gpu/drm/i915/i915_mem.c
@@ -46,7 +46,8 @@
 static void mark_block(struct drm_device * dev, struct mem_block *p, int in_use)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
-	drm_i915_sarea_t *sarea_priv = dev_priv->sarea_priv;
+	struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv;
+	drm_i915_sarea_t *sarea_priv = master_priv->sarea_priv;
 	struct drm_tex_region *list;
 	unsigned shift, nr;
 	unsigned start;
diff --git a/drivers/gpu/drm/radeon/r300_cmdbuf.c b/drivers/gpu/drm/radeon/r300_cmdbuf.c
index 4b27d9abb7bc..cace3964feeb 100644
--- a/drivers/gpu/drm/radeon/r300_cmdbuf.c
+++ b/drivers/gpu/drm/radeon/r300_cmdbuf.c
@@ -860,12 +860,12 @@ static __inline__ void r300_pacify(drm_radeon_private_t *dev_priv)
  * The actual age emit is done by r300_do_cp_cmdbuf, which is why you must
  * be careful about how this function is called.
  */
-static void r300_discard_buffer(struct drm_device * dev, struct drm_buf * buf)
+static void r300_discard_buffer(struct drm_device *dev, struct drm_master *master, struct drm_buf *buf)
 {
-	drm_radeon_private_t *dev_priv = dev->dev_private;
 	drm_radeon_buf_priv_t *buf_priv = buf->dev_private;
+	struct drm_radeon_master_private *master_priv = master->driver_priv;
 
-	buf_priv->age = ++dev_priv->sarea_priv->last_dispatch;
+	buf_priv->age = ++master_priv->sarea_priv->last_dispatch;
 	buf->pending = 1;
 	buf->used = 0;
 }
@@ -1027,6 +1027,7 @@ int r300_do_cp_cmdbuf(struct drm_device *dev,
 		      drm_radeon_kcmd_buffer_t *cmdbuf)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
+	struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv;
 	struct drm_device_dma *dma = dev->dma;
 	struct drm_buf *buf = NULL;
 	int emit_dispatch_age = 0;
@@ -1134,7 +1135,7 @@ int r300_do_cp_cmdbuf(struct drm_device *dev,
 			}
 
 			emit_dispatch_age = 1;
-			r300_discard_buffer(dev, buf);
+			r300_discard_buffer(dev, file_priv->master, buf);
 			break;
 
 		case R300_CMD_WAIT:
@@ -1189,7 +1190,7 @@ int r300_do_cp_cmdbuf(struct drm_device *dev,
 
 		/* Emit the vertex buffer age */
 		BEGIN_RING(2);
-		RADEON_DISPATCH_AGE(dev_priv->sarea_priv->last_dispatch);
+		RADEON_DISPATCH_AGE(master_priv->sarea_priv->last_dispatch);
 		ADVANCE_RING();
 	}
 
diff --git a/drivers/gpu/drm/radeon/radeon_cp.c b/drivers/gpu/drm/radeon/radeon_cp.c
index dcebb4bee7aa..7b37a4906377 100644
--- a/drivers/gpu/drm/radeon/radeon_cp.c
+++ b/drivers/gpu/drm/radeon/radeon_cp.c
@@ -31,6 +31,7 @@
 
 #include "drmP.h"
 #include "drm.h"
+#include "drm_sarea.h"
 #include "radeon_drm.h"
 #include "radeon_drv.h"
 #include "r300_reg.h"
@@ -667,15 +668,14 @@ static void radeon_cp_init_ring_buffer(struct drm_device * dev,
 		RADEON_WRITE(RADEON_BUS_CNTL, tmp);
 	} /* PCIE cards appears to not need this */
 
-	dev_priv->sarea_priv->last_frame = dev_priv->scratch[0] = 0;
-	RADEON_WRITE(RADEON_LAST_FRAME_REG, dev_priv->sarea_priv->last_frame);
+	dev_priv->scratch[0] = 0;
+	RADEON_WRITE(RADEON_LAST_FRAME_REG, 0);
 
-	dev_priv->sarea_priv->last_dispatch = dev_priv->scratch[1] = 0;
-	RADEON_WRITE(RADEON_LAST_DISPATCH_REG,
-		     dev_priv->sarea_priv->last_dispatch);
+	dev_priv->scratch[1] = 0;
+	RADEON_WRITE(RADEON_LAST_DISPATCH_REG, 0);
 
-	dev_priv->sarea_priv->last_clear = dev_priv->scratch[2] = 0;
-	RADEON_WRITE(RADEON_LAST_CLEAR_REG, dev_priv->sarea_priv->last_clear);
+	dev_priv->scratch[2] = 0;
+	RADEON_WRITE(RADEON_LAST_CLEAR_REG, 0);
 
 	radeon_do_wait_for_idle(dev_priv);
 
@@ -871,9 +871,11 @@ static void radeon_set_pcigart(drm_radeon_private_t * dev_priv, int on)
 	}
 }
 
-static int radeon_do_init_cp(struct drm_device * dev, drm_radeon_init_t * init)
+static int radeon_do_init_cp(struct drm_device *dev, drm_radeon_init_t *init,
+			     struct drm_file *file_priv)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
+	struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv;
 
 	DRM_DEBUG("\n");
 
@@ -998,8 +1000,8 @@ static int radeon_do_init_cp(struct drm_device * dev, drm_radeon_init_t * init)
 	dev_priv->buffers_offset = init->buffers_offset;
 	dev_priv->gart_textures_offset = init->gart_textures_offset;
 
-	dev_priv->sarea = drm_getsarea(dev);
-	if (!dev_priv->sarea) {
+	master_priv->sarea = drm_getsarea(dev);
+	if (!master_priv->sarea) {
 		DRM_ERROR("could not find sarea!\n");
 		radeon_do_cleanup_cp(dev);
 		return -EINVAL;
@@ -1035,10 +1037,6 @@ static int radeon_do_init_cp(struct drm_device * dev, drm_radeon_init_t * init)
 		}
 	}
 
-	dev_priv->sarea_priv =
-	    (drm_radeon_sarea_t *) ((u8 *) dev_priv->sarea->handle +
-				    init->sarea_priv_offset);
-
 #if __OS_HAS_AGP
 	if (dev_priv->flags & RADEON_IS_AGP) {
 		drm_core_ioremap(dev_priv->cp_ring, dev);
@@ -1329,7 +1327,7 @@ int radeon_cp_init(struct drm_device *dev, void *data, struct drm_file *file_pri
 	case RADEON_INIT_CP:
 	case RADEON_INIT_R200_CP:
 	case RADEON_INIT_R300_CP:
-		return radeon_do_init_cp(dev, init);
+		return radeon_do_init_cp(dev, init, file_priv);
 	case RADEON_CLEANUP_CP:
 		return radeon_do_cleanup_cp(dev);
 	}
@@ -1768,6 +1766,51 @@ int radeon_driver_load(struct drm_device *dev, unsigned long flags)
 	return ret;
 }
 
+int radeon_master_create(struct drm_device *dev, struct drm_master *master)
+{
+	struct drm_radeon_master_private *master_priv;
+	unsigned long sareapage;
+	int ret;
+
+	master_priv = drm_calloc(1, sizeof(*master_priv), DRM_MEM_DRIVER);
+	if (!master_priv)
+		return -ENOMEM;
+
+	/* prebuild the SAREA */
+	sareapage = max(SAREA_MAX, PAGE_SIZE);
+	ret = drm_addmap(dev, 0, sareapage, _DRM_SHM, _DRM_CONTAINS_LOCK|_DRM_DRIVER,
+			 &master_priv->sarea);
+	if (ret) {
+		DRM_ERROR("SAREA setup failed\n");
+		return ret;
+	}
+	master_priv->sarea_priv = master_priv->sarea->handle + sizeof(struct drm_sarea);
+	master_priv->sarea_priv->pfCurrentPage = 0;
+
+	master->driver_priv = master_priv;
+	return 0;
+}
+
+void radeon_master_destroy(struct drm_device *dev, struct drm_master *master)
+{
+	struct drm_radeon_master_private *master_priv = master->driver_priv;
+
+	if (!master_priv)
+		return;
+
+	if (master_priv->sarea_priv &&
+	    master_priv->sarea_priv->pfCurrentPage != 0)
+		radeon_cp_dispatch_flip(dev, master);
+
+	master_priv->sarea_priv = NULL;
+	if (master_priv->sarea)
+		drm_rmmap(dev, master_priv->sarea);
+
+	drm_free(master_priv, sizeof(*master_priv), DRM_MEM_DRIVER);
+
+	master->driver_priv = NULL;
+}
+
 /* Create mappings for registers and framebuffer so userland doesn't necessarily
  * have to find them.
  */
diff --git a/drivers/gpu/drm/radeon/radeon_drv.h b/drivers/gpu/drm/radeon/radeon_drv.h
index 3bbb871b25d5..490bc7ceef60 100644
--- a/drivers/gpu/drm/radeon/radeon_drv.h
+++ b/drivers/gpu/drm/radeon/radeon_drv.h
@@ -226,9 +226,13 @@ struct radeon_virt_surface {
 #define RADEON_FLUSH_EMITED	(1 < 0)
 #define RADEON_PURGE_EMITED	(1 < 1)
 
+struct drm_radeon_master_private {
+	drm_local_map_t *sarea;
+	drm_radeon_sarea_t *sarea_priv;
+};
+
 typedef struct drm_radeon_private {
 	drm_radeon_ring_buffer_t ring;
-	drm_radeon_sarea_t *sarea_priv;
 
 	u32 fb_location;
 	u32 fb_size;
@@ -409,6 +413,9 @@ extern int radeon_driver_open(struct drm_device *dev,
 extern long radeon_compat_ioctl(struct file *filp, unsigned int cmd,
 				unsigned long arg);
 
+extern int radeon_master_create(struct drm_device *dev, struct drm_master *master);
+extern void radeon_master_destroy(struct drm_device *dev, struct drm_master *master);
+extern void radeon_cp_dispatch_flip(struct drm_device *dev, struct drm_master *master);
 /* r300_cmdbuf.c */
 extern void r300_init_reg_flags(struct drm_device *dev);
 
@@ -1335,8 +1342,9 @@ do {									\
 } while (0)
 
 #define VB_AGE_TEST_WITH_RETURN( dev_priv )				\
-do {									\
-	drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv;		\
+do {								\
+	struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv;	\
+	drm_radeon_sarea_t *sarea_priv = master_priv->sarea_priv;	\
 	if ( sarea_priv->last_dispatch >= RADEON_MAX_VB_AGE ) {		\
 		int __ret = radeon_do_cp_idle( dev_priv );		\
 		if ( __ret ) return __ret;				\
diff --git a/drivers/gpu/drm/radeon/radeon_state.c b/drivers/gpu/drm/radeon/radeon_state.c
index 5d7153fcc7b0..ef940a079dcb 100644
--- a/drivers/gpu/drm/radeon/radeon_state.c
+++ b/drivers/gpu/drm/radeon/radeon_state.c
@@ -742,13 +742,14 @@ static struct {
  */
 
 static void radeon_clear_box(drm_radeon_private_t * dev_priv,
+			     struct drm_radeon_master_private *master_priv,
 			     int x, int y, int w, int h, int r, int g, int b)
 {
 	u32 color;
 	RING_LOCALS;
 
-	x += dev_priv->sarea_priv->boxes[0].x1;
-	y += dev_priv->sarea_priv->boxes[0].y1;
+	x += master_priv->sarea_priv->boxes[0].x1;
+	y += master_priv->sarea_priv->boxes[0].y1;
 
 	switch (dev_priv->color_fmt) {
 	case RADEON_COLOR_FORMAT_RGB565:
@@ -776,7 +777,7 @@ static void radeon_clear_box(drm_radeon_private_t * dev_priv,
 		 RADEON_GMC_SRC_DATATYPE_COLOR |
 		 RADEON_ROP3_P | RADEON_GMC_CLR_CMP_CNTL_DIS);
 
-	if (dev_priv->sarea_priv->pfCurrentPage == 1) {
+	if (master_priv->sarea_priv->pfCurrentPage == 1) {
 		OUT_RING(dev_priv->front_pitch_offset);
 	} else {
 		OUT_RING(dev_priv->back_pitch_offset);
@@ -790,7 +791,7 @@ static void radeon_clear_box(drm_radeon_private_t * dev_priv,
 	ADVANCE_RING();
 }
 
-static void radeon_cp_performance_boxes(drm_radeon_private_t * dev_priv)
+static void radeon_cp_performance_boxes(drm_radeon_private_t *dev_priv, struct drm_radeon_master_private *master_priv)
 {
 	/* Collapse various things into a wait flag -- trying to
 	 * guess if userspase slept -- better just to have them tell us.
@@ -807,12 +808,12 @@ static void radeon_cp_performance_boxes(drm_radeon_private_t * dev_priv)
 	/* Purple box for page flipping
 	 */
 	if (dev_priv->stats.boxes & RADEON_BOX_FLIP)
-		radeon_clear_box(dev_priv, 4, 4, 8, 8, 255, 0, 255);
+		radeon_clear_box(dev_priv, master_priv, 4, 4, 8, 8, 255, 0, 255);
 
 	/* Red box if we have to wait for idle at any point
 	 */
 	if (dev_priv->stats.boxes & RADEON_BOX_WAIT_IDLE)
-		radeon_clear_box(dev_priv, 16, 4, 8, 8, 255, 0, 0);
+		radeon_clear_box(dev_priv, master_priv, 16, 4, 8, 8, 255, 0, 0);
 
 	/* Blue box: lost context?
 	 */
@@ -820,12 +821,12 @@ static void radeon_cp_performance_boxes(drm_radeon_private_t * dev_priv)
 	/* Yellow box for texture swaps
 	 */
 	if (dev_priv->stats.boxes & RADEON_BOX_TEXTURE_LOAD)
-		radeon_clear_box(dev_priv, 40, 4, 8, 8, 255, 255, 0);
+		radeon_clear_box(dev_priv, master_priv, 40, 4, 8, 8, 255, 255, 0);
 
 	/* Green box if hardware never idles (as far as we can tell)
 	 */
 	if (!(dev_priv->stats.boxes & RADEON_BOX_DMA_IDLE))
-		radeon_clear_box(dev_priv, 64, 4, 8, 8, 0, 255, 0);
+		radeon_clear_box(dev_priv, master_priv, 64, 4, 8, 8, 0, 255, 0);
 
 	/* Draw bars indicating number of buffers allocated
 	 * (not a great measure, easily confused)
@@ -834,7 +835,7 @@ static void radeon_cp_performance_boxes(drm_radeon_private_t * dev_priv)
 		if (dev_priv->stats.requested_bufs > 100)
 			dev_priv->stats.requested_bufs = 100;
 
-		radeon_clear_box(dev_priv, 4, 16,
+		radeon_clear_box(dev_priv, master_priv, 4, 16,
 				 dev_priv->stats.requested_bufs, 4,
 				 196, 128, 128);
 	}
@@ -848,11 +849,13 @@ static void radeon_cp_performance_boxes(drm_radeon_private_t * dev_priv)
  */
 
 static void radeon_cp_dispatch_clear(struct drm_device * dev,
+				     struct drm_master *master,
 				     drm_radeon_clear_t * clear,
 				     drm_radeon_clear_rect_t * depth_boxes)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
-	drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv;
+	struct drm_radeon_master_private *master_priv = master->driver_priv;
+	drm_radeon_sarea_t *sarea_priv = master_priv->sarea_priv;
 	drm_radeon_depth_clear_t *depth_clear = &dev_priv->depth_clear;
 	int nbox = sarea_priv->nbox;
 	struct drm_clip_rect *pbox = sarea_priv->boxes;
@@ -864,7 +867,7 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev,
 
 	dev_priv->stats.clears++;
 
-	if (dev_priv->sarea_priv->pfCurrentPage == 1) {
+	if (sarea_priv->pfCurrentPage == 1) {
 		unsigned int tmp = flags;
 
 		flags &= ~(RADEON_FRONT | RADEON_BACK);
@@ -890,7 +893,7 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev,
 
 		/* Make sure we restore the 3D state next time.
 		 */
-		dev_priv->sarea_priv->ctx_owner = 0;
+		sarea_priv->ctx_owner = 0;
 
 		for (i = 0; i < nbox; i++) {
 			int x = pbox[i].x1;
@@ -967,7 +970,7 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev,
 		/* Make sure we restore the 3D state next time.
 		 * we haven't touched any "normal" state - still need this?
 		 */
-		dev_priv->sarea_priv->ctx_owner = 0;
+		sarea_priv->ctx_owner = 0;
 
 		if ((dev_priv->flags & RADEON_HAS_HIERZ)
 		    && (flags & RADEON_USE_HIERZ)) {
@@ -1214,7 +1217,7 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev,
 
 		/* Make sure we restore the 3D state next time.
 		 */
-		dev_priv->sarea_priv->ctx_owner = 0;
+		sarea_priv->ctx_owner = 0;
 
 		for (i = 0; i < nbox; i++) {
 
@@ -1285,7 +1288,7 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev,
 
 		/* Make sure we restore the 3D state next time.
 		 */
-		dev_priv->sarea_priv->ctx_owner = 0;
+		sarea_priv->ctx_owner = 0;
 
 		for (i = 0; i < nbox; i++) {
 
@@ -1328,20 +1331,21 @@ static void radeon_cp_dispatch_clear(struct drm_device * dev,
 	 * wait on this value before performing the clear ioctl.  We
 	 * need this because the card's so damned fast...
 	 */
-	dev_priv->sarea_priv->last_clear++;
+	sarea_priv->last_clear++;
 
 	BEGIN_RING(4);
 
-	RADEON_CLEAR_AGE(dev_priv->sarea_priv->last_clear);
+	RADEON_CLEAR_AGE(sarea_priv->last_clear);
 	RADEON_WAIT_UNTIL_IDLE();
 
 	ADVANCE_RING();
 }
 
-static void radeon_cp_dispatch_swap(struct drm_device * dev)
+static void radeon_cp_dispatch_swap(struct drm_device *dev, struct drm_master *master)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
-	drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv;
+	struct drm_radeon_master_private *master_priv = master->driver_priv;
+	drm_radeon_sarea_t *sarea_priv = master_priv->sarea_priv;
 	int nbox = sarea_priv->nbox;
 	struct drm_clip_rect *pbox = sarea_priv->boxes;
 	int i;
@@ -1351,7 +1355,7 @@ static void radeon_cp_dispatch_swap(struct drm_device * dev)
 	/* Do some trivial performance monitoring...
 	 */
 	if (dev_priv->do_boxes)
-		radeon_cp_performance_boxes(dev_priv);
+		radeon_cp_performance_boxes(dev_priv, master_priv);
 
 	/* Wait for the 3D stream to idle before dispatching the bitblt.
 	 * This will prevent data corruption between the two streams.
@@ -1385,7 +1389,7 @@ static void radeon_cp_dispatch_swap(struct drm_device * dev)
 		/* Make this work even if front & back are flipped:
 		 */
 		OUT_RING(CP_PACKET0(RADEON_SRC_PITCH_OFFSET, 1));
-		if (dev_priv->sarea_priv->pfCurrentPage == 0) {
+		if (sarea_priv->pfCurrentPage == 0) {
 			OUT_RING(dev_priv->back_pitch_offset);
 			OUT_RING(dev_priv->front_pitch_offset);
 		} else {
@@ -1405,31 +1409,32 @@ static void radeon_cp_dispatch_swap(struct drm_device * dev)
 	 * throttle the framerate by waiting for this value before
 	 * performing the swapbuffer ioctl.
 	 */
-	dev_priv->sarea_priv->last_frame++;
+	sarea_priv->last_frame++;
 
 	BEGIN_RING(4);
 
-	RADEON_FRAME_AGE(dev_priv->sarea_priv->last_frame);
+	RADEON_FRAME_AGE(sarea_priv->last_frame);
 	RADEON_WAIT_UNTIL_2D_IDLE();
 
 	ADVANCE_RING();
 }
 
-static void radeon_cp_dispatch_flip(struct drm_device * dev)
+void radeon_cp_dispatch_flip(struct drm_device *dev, struct drm_master *master)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
-	struct drm_sarea *sarea = (struct drm_sarea *) dev_priv->sarea->handle;
-	int offset = (dev_priv->sarea_priv->pfCurrentPage == 1)
+	struct drm_radeon_master_private *master_priv = master->driver_priv;
+	struct drm_sarea *sarea = (struct drm_sarea *)master_priv->sarea->handle;
+	int offset = (master_priv->sarea_priv->pfCurrentPage == 1)
 	    ? dev_priv->front_offset : dev_priv->back_offset;
 	RING_LOCALS;
 	DRM_DEBUG("pfCurrentPage=%d\n",
-		  dev_priv->sarea_priv->pfCurrentPage);
+		  master_priv->sarea_priv->pfCurrentPage);
 
 	/* Do some trivial performance monitoring...
 	 */
 	if (dev_priv->do_boxes) {
 		dev_priv->stats.boxes |= RADEON_BOX_FLIP;
-		radeon_cp_performance_boxes(dev_priv);
+		radeon_cp_performance_boxes(dev_priv, master_priv);
 	}
 
 	/* Update the frame offsets for both CRTCs
@@ -1441,7 +1446,7 @@ static void radeon_cp_dispatch_flip(struct drm_device * dev)
 		     ((sarea->frame.y * dev_priv->front_pitch +
 		       sarea->frame.x * (dev_priv->color_fmt - 2)) & ~7)
 		     + offset);
-	OUT_RING_REG(RADEON_CRTC2_OFFSET, dev_priv->sarea_priv->crtc2_base
+	OUT_RING_REG(RADEON_CRTC2_OFFSET, master_priv->sarea_priv->crtc2_base
 		     + offset);
 
 	ADVANCE_RING();
@@ -1450,13 +1455,13 @@ static void radeon_cp_dispatch_flip(struct drm_device * dev)
 	 * throttle the framerate by waiting for this value before
 	 * performing the swapbuffer ioctl.
 	 */
-	dev_priv->sarea_priv->last_frame++;
-	dev_priv->sarea_priv->pfCurrentPage =
-		1 - dev_priv->sarea_priv->pfCurrentPage;
+	master_priv->sarea_priv->last_frame++;
+	master_priv->sarea_priv->pfCurrentPage =
+		1 - master_priv->sarea_priv->pfCurrentPage;
 
 	BEGIN_RING(2);
 
-	RADEON_FRAME_AGE(dev_priv->sarea_priv->last_frame);
+	RADEON_FRAME_AGE(master_priv->sarea_priv->last_frame);
 
 	ADVANCE_RING();
 }
@@ -1494,11 +1499,13 @@ typedef struct {
 } drm_radeon_tcl_prim_t;
 
 static void radeon_cp_dispatch_vertex(struct drm_device * dev,
+				      struct drm_file *file_priv,
 				      struct drm_buf * buf,
 				      drm_radeon_tcl_prim_t * prim)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
-	drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv;
+	struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv;
+	drm_radeon_sarea_t *sarea_priv = master_priv->sarea_priv;
 	int offset = dev_priv->gart_buffers_offset + buf->offset + prim->start;
 	int numverts = (int)prim->numverts;
 	int nbox = sarea_priv->nbox;
@@ -1539,13 +1546,14 @@ static void radeon_cp_dispatch_vertex(struct drm_device * dev,
 	} while (i < nbox);
 }
 
-static void radeon_cp_discard_buffer(struct drm_device * dev, struct drm_buf * buf)
+static void radeon_cp_discard_buffer(struct drm_device *dev, struct drm_master *master, struct drm_buf *buf)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
+	struct drm_radeon_master_private *master_priv = master->driver_priv;
 	drm_radeon_buf_priv_t *buf_priv = buf->dev_private;
 	RING_LOCALS;
 
-	buf_priv->age = ++dev_priv->sarea_priv->last_dispatch;
+	buf_priv->age = ++master_priv->sarea_priv->last_dispatch;
 
 	/* Emit the vertex buffer age */
 	BEGIN_RING(2);
@@ -1590,12 +1598,14 @@ static void radeon_cp_dispatch_indirect(struct drm_device * dev,
 	}
 }
 
-static void radeon_cp_dispatch_indices(struct drm_device * dev,
+static void radeon_cp_dispatch_indices(struct drm_device *dev,
+				       struct drm_master *master,
 				       struct drm_buf * elt_buf,
 				       drm_radeon_tcl_prim_t * prim)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
-	drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv;
+	struct drm_radeon_master_private *master_priv = master->driver_priv;
+	drm_radeon_sarea_t *sarea_priv = master_priv->sarea_priv;
 	int offset = dev_priv->gart_buffers_offset + prim->offset;
 	u32 *data;
 	int dwords;
@@ -1870,7 +1880,7 @@ static int radeon_cp_dispatch_texture(struct drm_device * dev,
 		ADVANCE_RING();
 		COMMIT_RING();
 
-		radeon_cp_discard_buffer(dev, buf);
+		radeon_cp_discard_buffer(dev, file_priv->master, buf);
 
 		/* Update the input parameters for next time */
 		image->y += height;
@@ -2110,7 +2120,8 @@ static int radeon_surface_free(struct drm_device *dev, void *data, struct drm_fi
 static int radeon_cp_clear(struct drm_device *dev, void *data, struct drm_file *file_priv)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
-	drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv;
+	struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv;
+	drm_radeon_sarea_t *sarea_priv = master_priv->sarea_priv;
 	drm_radeon_clear_t *clear = data;
 	drm_radeon_clear_rect_t depth_boxes[RADEON_NR_SAREA_CLIPRECTS];
 	DRM_DEBUG("\n");
@@ -2126,7 +2137,7 @@ static int radeon_cp_clear(struct drm_device *dev, void *data, struct drm_file *
 			       sarea_priv->nbox * sizeof(depth_boxes[0])))
 		return -EFAULT;
 
-	radeon_cp_dispatch_clear(dev, clear, depth_boxes);
+	radeon_cp_dispatch_clear(dev, file_priv->master, clear, depth_boxes);
 
 	COMMIT_RING();
 	return 0;
@@ -2134,9 +2145,10 @@ static int radeon_cp_clear(struct drm_device *dev, void *data, struct drm_file *
 
 /* Not sure why this isn't set all the time:
  */
-static int radeon_do_init_pageflip(struct drm_device * dev)
+static int radeon_do_init_pageflip(struct drm_device *dev, struct drm_master *master)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
+	struct drm_radeon_master_private *master_priv = master->driver_priv;
 	RING_LOCALS;
 
 	DRM_DEBUG("\n");
@@ -2153,8 +2165,8 @@ static int radeon_do_init_pageflip(struct drm_device * dev)
 
 	dev_priv->page_flipping = 1;
 
-	if (dev_priv->sarea_priv->pfCurrentPage != 1)
-		dev_priv->sarea_priv->pfCurrentPage = 0;
+	if (master_priv->sarea_priv->pfCurrentPage != 1)
+		master_priv->sarea_priv->pfCurrentPage = 0;
 
 	return 0;
 }
@@ -2172,9 +2184,9 @@ static int radeon_cp_flip(struct drm_device *dev, void *data, struct drm_file *f
 	RING_SPACE_TEST_WITH_RETURN(dev_priv);
 
 	if (!dev_priv->page_flipping)
-		radeon_do_init_pageflip(dev);
+		radeon_do_init_pageflip(dev, file_priv->master);
 
-	radeon_cp_dispatch_flip(dev);
+	radeon_cp_dispatch_flip(dev, file_priv->master);
 
 	COMMIT_RING();
 	return 0;
@@ -2183,7 +2195,9 @@ static int radeon_cp_flip(struct drm_device *dev, void *data, struct drm_file *f
 static int radeon_cp_swap(struct drm_device *dev, void *data, struct drm_file *file_priv)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
-	drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv;
+	struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv;
+	drm_radeon_sarea_t *sarea_priv = master_priv->sarea_priv;
+
 	DRM_DEBUG("\n");
 
 	LOCK_TEST_WITH_RETURN(dev, file_priv);
@@ -2193,8 +2207,8 @@ static int radeon_cp_swap(struct drm_device *dev, void *data, struct drm_file *f
 	if (sarea_priv->nbox > RADEON_NR_SAREA_CLIPRECTS)
 		sarea_priv->nbox = RADEON_NR_SAREA_CLIPRECTS;
 
-	radeon_cp_dispatch_swap(dev);
-	dev_priv->sarea_priv->ctx_owner = 0;
+	radeon_cp_dispatch_swap(dev, file_priv->master);
+	sarea_priv->ctx_owner = 0;
 
 	COMMIT_RING();
 	return 0;
@@ -2203,7 +2217,8 @@ static int radeon_cp_swap(struct drm_device *dev, void *data, struct drm_file *f
 static int radeon_cp_vertex(struct drm_device *dev, void *data, struct drm_file *file_priv)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
-	drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv;
+	struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv;
+	drm_radeon_sarea_t *sarea_priv;
 	struct drm_device_dma *dma = dev->dma;
 	struct drm_buf *buf;
 	drm_radeon_vertex_t *vertex = data;
@@ -2211,6 +2226,8 @@ static int radeon_cp_vertex(struct drm_device *dev, void *data, struct drm_file
 
 	LOCK_TEST_WITH_RETURN(dev, file_priv);
 
+	sarea_priv = master_priv->sarea_priv;
+
 	DRM_DEBUG("pid=%d index=%d count=%d discard=%d\n",
 		  DRM_CURRENTPID, vertex->idx, vertex->count, vertex->discard);
 
@@ -2263,13 +2280,13 @@ static int radeon_cp_vertex(struct drm_device *dev, void *data, struct drm_file
 		prim.finish = vertex->count;	/* unused */
 		prim.prim = vertex->prim;
 		prim.numverts = vertex->count;
-		prim.vc_format = dev_priv->sarea_priv->vc_format;
+		prim.vc_format = sarea_priv->vc_format;
 
-		radeon_cp_dispatch_vertex(dev, buf, &prim);
+		radeon_cp_dispatch_vertex(dev, file_priv, buf, &prim);
 	}
 
 	if (vertex->discard) {
-		radeon_cp_discard_buffer(dev, buf);
+		radeon_cp_discard_buffer(dev, file_priv->master, buf);
 	}
 
 	COMMIT_RING();
@@ -2279,7 +2296,8 @@ static int radeon_cp_vertex(struct drm_device *dev, void *data, struct drm_file
 static int radeon_cp_indices(struct drm_device *dev, void *data, struct drm_file *file_priv)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
-	drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv;
+	struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv;
+	drm_radeon_sarea_t *sarea_priv;
 	struct drm_device_dma *dma = dev->dma;
 	struct drm_buf *buf;
 	drm_radeon_indices_t *elts = data;
@@ -2288,6 +2306,8 @@ static int radeon_cp_indices(struct drm_device *dev, void *data, struct drm_file
 
 	LOCK_TEST_WITH_RETURN(dev, file_priv);
 
+	sarea_priv = master_priv->sarea_priv;
+
 	DRM_DEBUG("pid=%d index=%d start=%d end=%d discard=%d\n",
 		  DRM_CURRENTPID, elts->idx, elts->start, elts->end,
 		  elts->discard);
@@ -2353,11 +2373,11 @@ static int radeon_cp_indices(struct drm_device *dev, void *data, struct drm_file
 	prim.prim = elts->prim;
 	prim.offset = 0;	/* offset from start of dma buffers */
 	prim.numverts = RADEON_MAX_VB_VERTS;	/* duh */
-	prim.vc_format = dev_priv->sarea_priv->vc_format;
+	prim.vc_format = sarea_priv->vc_format;
 
-	radeon_cp_dispatch_indices(dev, buf, &prim);
+	radeon_cp_dispatch_indices(dev, file_priv->master, buf, &prim);
 	if (elts->discard) {
-		radeon_cp_discard_buffer(dev, buf);
+		radeon_cp_discard_buffer(dev, file_priv->master, buf);
 	}
 
 	COMMIT_RING();
@@ -2468,7 +2488,7 @@ static int radeon_cp_indirect(struct drm_device *dev, void *data, struct drm_fil
 	 */
 	radeon_cp_dispatch_indirect(dev, buf, indirect->start, indirect->end);
 	if (indirect->discard) {
-		radeon_cp_discard_buffer(dev, buf);
+		radeon_cp_discard_buffer(dev, file_priv->master, buf);
 	}
 
 	COMMIT_RING();
@@ -2478,7 +2498,8 @@ static int radeon_cp_indirect(struct drm_device *dev, void *data, struct drm_fil
 static int radeon_cp_vertex2(struct drm_device *dev, void *data, struct drm_file *file_priv)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
-	drm_radeon_sarea_t *sarea_priv = dev_priv->sarea_priv;
+	struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv;
+	drm_radeon_sarea_t *sarea_priv;
 	struct drm_device_dma *dma = dev->dma;
 	struct drm_buf *buf;
 	drm_radeon_vertex2_t *vertex = data;
@@ -2487,6 +2508,8 @@ static int radeon_cp_vertex2(struct drm_device *dev, void *data, struct drm_file
 
 	LOCK_TEST_WITH_RETURN(dev, file_priv);
 
+	sarea_priv = master_priv->sarea_priv;
+
 	DRM_DEBUG("pid=%d index=%d discard=%d\n",
 		  DRM_CURRENTPID, vertex->idx, vertex->discard);
 
@@ -2547,12 +2570,12 @@ static int radeon_cp_vertex2(struct drm_device *dev, void *data, struct drm_file
 			tclprim.offset = prim.numverts * 64;
 			tclprim.numverts = RADEON_MAX_VB_VERTS;	/* duh */
 
-			radeon_cp_dispatch_indices(dev, buf, &tclprim);
+			radeon_cp_dispatch_indices(dev, file_priv->master, buf, &tclprim);
 		} else {
 			tclprim.numverts = prim.numverts;
 			tclprim.offset = 0;	/* not used */
 
-			radeon_cp_dispatch_vertex(dev, buf, &tclprim);
+			radeon_cp_dispatch_vertex(dev, file_priv, buf, &tclprim);
 		}
 
 		if (sarea_priv->nbox == 1)
@@ -2560,7 +2583,7 @@ static int radeon_cp_vertex2(struct drm_device *dev, void *data, struct drm_file
 	}
 
 	if (vertex->discard) {
-		radeon_cp_discard_buffer(dev, buf);
+		radeon_cp_discard_buffer(dev, file_priv->master, buf);
 	}
 
 	COMMIT_RING();
@@ -2909,7 +2932,7 @@ static int radeon_cp_cmdbuf(struct drm_device *dev, void *data, struct drm_file
 				goto err;
 			}
 
-			radeon_cp_discard_buffer(dev, buf);
+			radeon_cp_discard_buffer(dev, file_priv->master, buf);
 			break;
 
 		case RADEON_CMD_PACKET3:
@@ -3020,7 +3043,7 @@ static int radeon_cp_getparam(struct drm_device *dev, void *data, struct drm_fil
 		 */
 	case RADEON_PARAM_SAREA_HANDLE:
 		/* The lock is the first dword in the sarea. */
-		value = (long)dev->lock.hw_lock;
+		/* no users of this parameter */
 		break;
 #endif
 	case RADEON_PARAM_GART_TEX_HANDLE:
@@ -3064,6 +3087,7 @@ static int radeon_cp_getparam(struct drm_device *dev, void *data, struct drm_fil
 static int radeon_cp_setparam(struct drm_device *dev, void *data, struct drm_file *file_priv)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
+	struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv;
 	drm_radeon_setparam_t *sp = data;
 	struct drm_radeon_driver_file_fields *radeon_priv;
 
@@ -3078,12 +3102,14 @@ static int radeon_cp_setparam(struct drm_device *dev, void *data, struct drm_fil
 			DRM_DEBUG("color tiling disabled\n");
 			dev_priv->front_pitch_offset &= ~RADEON_DST_TILE_MACRO;
 			dev_priv->back_pitch_offset &= ~RADEON_DST_TILE_MACRO;
-			dev_priv->sarea_priv->tiling_enabled = 0;
+			if (master_priv->sarea_priv)
+				master_priv->sarea_priv->tiling_enabled = 0;
 		} else if (sp->value == 1) {
 			DRM_DEBUG("color tiling enabled\n");
 			dev_priv->front_pitch_offset |= RADEON_DST_TILE_MACRO;
 			dev_priv->back_pitch_offset |= RADEON_DST_TILE_MACRO;
-			dev_priv->sarea_priv->tiling_enabled = 1;
+			if (master_priv->sarea_priv)
+				master_priv->sarea_priv->tiling_enabled = 1;
 		}
 		break;
 	case RADEON_SETPARAM_PCIGART_LOCATION:
@@ -3129,14 +3155,6 @@ void radeon_driver_preclose(struct drm_device *dev, struct drm_file *file_priv)
 
 void radeon_driver_lastclose(struct drm_device *dev)
 {
-	if (dev->dev_private) {
-		drm_radeon_private_t *dev_priv = dev->dev_private;
-
-		if (dev_priv->sarea_priv &&
-		    dev_priv->sarea_priv->pfCurrentPage != 0)
-			radeon_cp_dispatch_flip(dev);
-	}
-
 	radeon_do_release(dev);
 }