Merge /pub/scm/linux/kernel/git/torvalds/linux-2.6

31 files changed, 572 insertions, 349 deletions

diff --git a/drivers/char/drm/README.drm b/drivers/char/drm/README.drm
index 6441e01..af74cd7 100644
--- a/drivers/char/drm/README.drm
+++ b/drivers/char/drm/README.drm
@@ -1,6 +1,6 @@
 ************************************************************
 * For the very latest on DRI development, please see:      *
-*     http://dri.sourceforge.net/                          *
+*     http://dri.freedesktop.org/                          *
 ************************************************************
 
 The Direct Rendering Manager (drm) is a device-independent kernel-level
@@ -26,21 +26,19 @@ ways:
 
 
 Documentation on the DRI is available from:
-    http://precisioninsight.com/piinsights.html
+    http://dri.freedesktop.org/wiki/Documentation
+    http://sourceforge.net/project/showfiles.php?group_id=387
+    http://dri.sourceforge.net/doc/
 
 For specific information about kernel-level support, see:
 
     The Direct Rendering Manager, Kernel Support for the Direct Rendering
     Infrastructure
-    http://precisioninsight.com/dr/drm.html
+    http://dri.sourceforge.net/doc/drm_low_level.html
 
     Hardware Locking for the Direct Rendering Infrastructure
-    http://precisioninsight.com/dr/locking.html
+    http://dri.sourceforge.net/doc/hardware_locking_low_level.html
 
     A Security Analysis of the Direct Rendering Infrastructure
-    http://precisioninsight.com/dr/security.html
+    http://dri.sourceforge.net/doc/security_low_level.html
 
-************************************************************
-* For the very latest on DRI development, please see:      *
-*     http://dri.sourceforge.net/                          *
-************************************************************
diff --git a/drivers/char/drm/ati_pcigart.c b/drivers/char/drm/ati_pcigart.c
index bd7be09..5b91bc0 100644
--- a/drivers/char/drm/ati_pcigart.c
+++ b/drivers/char/drm/ati_pcigart.c
@@ -33,59 +33,44 @@
 
 #include "drmP.h"
 
-#if PAGE_SIZE == 65536
-# define ATI_PCIGART_TABLE_ORDER	0
-# define ATI_PCIGART_TABLE_PAGES	(1 << 0)
-#elif PAGE_SIZE == 16384
-# define ATI_PCIGART_TABLE_ORDER	1
-# define ATI_PCIGART_TABLE_PAGES	(1 << 1)
-#elif PAGE_SIZE == 8192
-# define ATI_PCIGART_TABLE_ORDER 	2
-# define ATI_PCIGART_TABLE_PAGES 	(1 << 2)
-#elif PAGE_SIZE == 4096
-# define ATI_PCIGART_TABLE_ORDER 	3
-# define ATI_PCIGART_TABLE_PAGES 	(1 << 3)
-#else
-# error - PAGE_SIZE not 64K, 16K, 8K or 4K
-#endif
-
-# define ATI_MAX_PCIGART_PAGES		8192	/**< 32 MB aperture, 4K pages */
 # define ATI_PCIGART_PAGE_SIZE		4096	/**< PCI GART page size */
 
-static void *drm_ati_alloc_pcigart_table(void)
+static void *drm_ati_alloc_pcigart_table(int order)
 {
 	unsigned long address;
 	struct page *page;
 	int i;
-	DRM_DEBUG("%s\n", __FUNCTION__);
+
+	DRM_DEBUG("%s: alloc %d order\n", __FUNCTION__, order);
 
 	address = __get_free_pages(GFP_KERNEL | __GFP_COMP,
-				   ATI_PCIGART_TABLE_ORDER);
+				   order);
 	if (address == 0UL) {
 		return NULL;
 	}
 
 	page = virt_to_page(address);
 
-	for (i = 0; i < ATI_PCIGART_TABLE_PAGES; i++, page++)
+	for (i = 0; i < order; i++, page++)
 		SetPageReserved(page);
 
 	DRM_DEBUG("%s: returning 0x%08lx\n", __FUNCTION__, address);
 	return (void *)address;
 }
 
-static void drm_ati_free_pcigart_table(void *address)
+static void drm_ati_free_pcigart_table(void *address, int order)
 {
 	struct page *page;
 	int i;
+	int num_pages = 1 << order;
 	DRM_DEBUG("%s\n", __FUNCTION__);
 
 	page = virt_to_page((unsigned long)address);
 
-	for (i = 0; i < ATI_PCIGART_TABLE_PAGES; i++, page++)
+	for (i = 0; i < num_pages; i++, page++)
 		ClearPageReserved(page);
 
-	free_pages((unsigned long)address, ATI_PCIGART_TABLE_ORDER);
+	free_pages((unsigned long)address, order);
 }
 
 int drm_ati_pcigart_cleanup(drm_device_t *dev, drm_ati_pcigart_info *gart_info)
@@ -93,6 +78,8 @@ int drm_ati_pcigart_cleanup(drm_device_t *dev, drm_ati_pcigart_info *gart_info)
 	drm_sg_mem_t *entry = dev->sg;
 	unsigned long pages;
 	int i;
+	int order;
+	int num_pages, max_pages;
 
 	/* we need to support large memory configurations */
 	if (!entry) {
@@ -100,15 +87,19 @@ int drm_ati_pcigart_cleanup(drm_device_t *dev, drm_ati_pcigart_info *gart_info)
 		return 0;
 	}
 
+	order = drm_order((gart_info->table_size + (PAGE_SIZE-1)) / PAGE_SIZE);
+	num_pages = 1 << order;
+
 	if (gart_info->bus_addr) {
 		if (gart_info->gart_table_location == DRM_ATI_GART_MAIN) {
 			pci_unmap_single(dev->pdev, gart_info->bus_addr,
-					 ATI_PCIGART_TABLE_PAGES * PAGE_SIZE,
+					 num_pages * PAGE_SIZE,
 					 PCI_DMA_TODEVICE);
 		}
 
-		pages = (entry->pages <= ATI_MAX_PCIGART_PAGES)
-		    ? entry->pages : ATI_MAX_PCIGART_PAGES;
+		max_pages = (gart_info->table_size / sizeof(u32));
+		pages = (entry->pages <= max_pages)
+		  ? entry->pages : max_pages;
 
 		for (i = 0; i < pages; i++) {
 			if (!entry->busaddr[i])
@@ -123,13 +114,12 @@ int drm_ati_pcigart_cleanup(drm_device_t *dev, drm_ati_pcigart_info *gart_info)
 
 	if (gart_info->gart_table_location == DRM_ATI_GART_MAIN
 	    && gart_info->addr) {
-		drm_ati_free_pcigart_table(gart_info->addr);
+		drm_ati_free_pcigart_table(gart_info->addr, order);
 		gart_info->addr = NULL;
 	}
 
 	return 1;
 }
-
 EXPORT_SYMBOL(drm_ati_pcigart_cleanup);
 
 int drm_ati_pcigart_init(drm_device_t *dev, drm_ati_pcigart_info *gart_info)
@@ -139,6 +129,9 @@ int drm_ati_pcigart_init(drm_device_t *dev, drm_ati_pcigart_info *gart_info)
 	unsigned long pages;
 	u32 *pci_gart, page_base, bus_address = 0;
 	int i, j, ret = 0;
+	int order;
+	int max_pages;
+	int num_pages;
 
 	if (!entry) {
 		DRM_ERROR("no scatter/gather memory!\n");
@@ -148,7 +141,10 @@ int drm_ati_pcigart_init(drm_device_t *dev, drm_ati_pcigart_info *gart_info)
 	if (gart_info->gart_table_location == DRM_ATI_GART_MAIN) {
 		DRM_DEBUG("PCI: no table in VRAM: using normal RAM\n");
 
-		address = drm_ati_alloc_pcigart_table();
+		order = drm_order((gart_info->table_size +
+				   (PAGE_SIZE-1)) / PAGE_SIZE);
+		num_pages = 1 << order;
+		address = drm_ati_alloc_pcigart_table(order);
 		if (!address) {
 			DRM_ERROR("cannot allocate PCI GART page!\n");
 			goto done;
@@ -160,11 +156,13 @@ int drm_ati_pcigart_init(drm_device_t *dev, drm_ati_pcigart_info *gart_info)
 		}
 
 		bus_address = pci_map_single(dev->pdev, address,
-					     ATI_PCIGART_TABLE_PAGES *
-					     PAGE_SIZE, PCI_DMA_TODEVICE);
+					     num_pages * PAGE_SIZE,
+					     PCI_DMA_TODEVICE);
 		if (bus_address == 0) {
 			DRM_ERROR("unable to map PCIGART pages!\n");
-			drm_ati_free_pcigart_table(address);
+			order = drm_order((gart_info->table_size +
+					   (PAGE_SIZE-1)) / PAGE_SIZE);
+			drm_ati_free_pcigart_table(address, order);
 			address = NULL;
 			goto done;
 		}
@@ -177,10 +175,11 @@ int drm_ati_pcigart_init(drm_device_t *dev, drm_ati_pcigart_info *gart_info)
 
 	pci_gart = (u32 *) address;
 
-	pages = (entry->pages <= ATI_MAX_PCIGART_PAGES)
-	    ? entry->pages : ATI_MAX_PCIGART_PAGES;
+	max_pages = (gart_info->table_size / sizeof(u32));
+	pages = (entry->pages <= max_pages)
+	    ? entry->pages : max_pages;
 
-	memset(pci_gart, 0, ATI_MAX_PCIGART_PAGES * sizeof(u32));
+	memset(pci_gart, 0, max_pages * sizeof(u32));
 
 	for (i = 0; i < pages; i++) {
 		/* we need to support large memory configurations */
@@ -198,10 +197,18 @@ int drm_ati_pcigart_init(drm_device_t *dev, drm_ati_pcigart_info *gart_info)
 		page_base = (u32) entry->busaddr[i];
 
 		for (j = 0; j < (PAGE_SIZE / ATI_PCIGART_PAGE_SIZE); j++) {
-			if (gart_info->is_pcie)
+			switch(gart_info->gart_reg_if) {
+			case DRM_ATI_GART_IGP:
+				*pci_gart = cpu_to_le32((page_base) | 0xc);
+				break;
+			case DRM_ATI_GART_PCIE:
 				*pci_gart = cpu_to_le32((page_base >> 8) | 0xc);
-			else
+				break;
+			default:
+			case DRM_ATI_GART_PCI:
 				*pci_gart = cpu_to_le32(page_base);
+				break;
+			}
 			pci_gart++;
 			page_base += ATI_PCIGART_PAGE_SIZE;
 		}
@@ -220,5 +227,4 @@ int drm_ati_pcigart_init(drm_device_t *dev, drm_ati_pcigart_info *gart_info)
 	gart_info->bus_addr = bus_address;
 	return ret;
 }
-
 EXPORT_SYMBOL(drm_ati_pcigart_init);
diff --git a/drivers/char/drm/drm.h b/drivers/char/drm/drm.h
index 8db9041..0891984 100644
--- a/drivers/char/drm/drm.h
+++ b/drivers/char/drm/drm.h
@@ -654,11 +654,13 @@ typedef struct drm_set_version {
 
 /**
  * Device specific ioctls should only be in their respective headers
- * The device specific ioctl range is from 0x40 to 0x79.
+ * The device specific ioctl range is from 0x40 to 0x99.
+ * Generic IOCTLS restart at 0xA0.
  *
  * \sa drmCommandNone(), drmCommandRead(), drmCommandWrite(), and
  * drmCommandReadWrite().
  */
 #define DRM_COMMAND_BASE                0x40
+#define DRM_COMMAND_END			0xA0
 
 #endif
diff --git a/drivers/char/drm/drmP.h b/drivers/char/drm/drmP.h
index 85d99e2..d494315 100644
--- a/drivers/char/drm/drmP.h
+++ b/drivers/char/drm/drmP.h
@@ -414,6 +414,10 @@ typedef struct drm_lock_data {
 	struct file *filp;		/**< File descr of lock holder (0=kernel) */
 	wait_queue_head_t lock_queue;	/**< Queue of blocked processes */
 	unsigned long lock_time;	/**< Time of last lock in jiffies */
+	spinlock_t spinlock;
+	uint32_t kernel_waiters;
+	uint32_t user_waiters;
+	int idle_has_lock;
 } drm_lock_data_t;
 
 /**
@@ -515,12 +519,17 @@ typedef struct drm_vbl_sig {
 #define DRM_ATI_GART_MAIN 1
 #define DRM_ATI_GART_FB   2
 
+#define DRM_ATI_GART_PCI 1
+#define DRM_ATI_GART_PCIE 2
+#define DRM_ATI_GART_IGP 3
+
 typedef struct ati_pcigart_info {
 	int gart_table_location;
-	int is_pcie;
+	int gart_reg_if;
 	void *addr;
 	dma_addr_t bus_addr;
 	drm_local_map_t mapping;
+	int table_size;
 } drm_ati_pcigart_info;
 
 /*
@@ -590,6 +599,8 @@ struct drm_driver {
 	void (*reclaim_buffers) (struct drm_device * dev, struct file * filp);
 	void (*reclaim_buffers_locked) (struct drm_device *dev,
 					struct file *filp);
+	void (*reclaim_buffers_idlelocked) (struct drm_device *dev,
+					struct file * filp);
 	unsigned long (*get_map_ofs) (drm_map_t * map);
 	unsigned long (*get_reg_ofs) (struct drm_device * dev);
 	void (*set_version) (struct drm_device * dev, drm_set_version_t * sv);
@@ -764,7 +775,7 @@ static __inline__ int drm_core_check_feature(struct drm_device *dev,
 }
 
 #ifdef __alpha__
-#define drm_get_pci_domain(dev) dev->hose->bus->number
+#define drm_get_pci_domain(dev) dev->hose->index
 #else
 #define drm_get_pci_domain(dev) 0
 #endif
@@ -915,9 +926,18 @@ extern int drm_lock(struct inode *inode, struct file *filp,
 		    unsigned int cmd, unsigned long arg);
 extern int drm_unlock(struct inode *inode, struct file *filp,
 		      unsigned int cmd, unsigned long arg);
-extern int drm_lock_take(__volatile__ unsigned int *lock, unsigned int context);
-extern int drm_lock_free(drm_device_t * dev,
-			 __volatile__ unsigned int *lock, unsigned int context);
+extern int drm_lock_take(drm_lock_data_t *lock_data, unsigned int context);
+extern int drm_lock_free(drm_lock_data_t *lock_data, unsigned int context);
+extern void drm_idlelock_take(drm_lock_data_t *lock_data);
+extern void drm_idlelock_release(drm_lock_data_t *lock_data);
+
+/*
+ * These are exported to drivers so that they can implement fencing using
+ * DMA quiscent + idle. DMA quiescent usually requires the hardware lock.
+ */
+
+extern int drm_i_have_hw_lock(struct file *filp);
+extern int drm_kernel_take_hw_lock(struct file *filp);
 
 				/* Buffer management support (drm_bufs.h) */
 extern int drm_addbufs_agp(drm_device_t * dev, drm_buf_desc_t * request);
diff --git a/drivers/char/drm/drm_bufs.c b/drivers/char/drm/drm_bufs.c
index a6828cc..c113458 100644
--- a/drivers/char/drm/drm_bufs.c
+++ b/drivers/char/drm/drm_bufs.c
@@ -57,7 +57,8 @@ static drm_map_list_t *drm_find_matching_map(drm_device_t *dev,
 	list_for_each(list, &dev->maplist->head) {
 		drm_map_list_t *entry = list_entry(list, drm_map_list_t, head);
 		if (entry->map && map->type == entry->map->type &&
-		    entry->map->offset == map->offset) {
+		    ((entry->map->offset == map->offset) ||
+		     (map->type == _DRM_SHM && map->flags==_DRM_CONTAINS_LOCK))) {
 			return entry;
 		}
 	}
@@ -180,8 +181,20 @@ static int drm_addmap_core(drm_device_t * dev, unsigned int offset,
 		if (map->type == _DRM_REGISTERS)
 			map->handle = ioremap(map->offset, map->size);
 		break;
-
 	case _DRM_SHM:
+		list = drm_find_matching_map(dev, map);
+		if (list != NULL) {
+			if(list->map->size != map->size) {
+				DRM_DEBUG("Matching maps of type %d with "
+					  "mismatched sizes, (%ld vs %ld)\n",
+					  map->type, map->size, list->map->size);
+				list->map->size = map->size;
+			}
+
+			drm_free(map, sizeof(*map), DRM_MEM_MAPS);
+			*maplist = list;
+			return 0;
+		}
 		map->handle = vmalloc_user(map->size);
 		DRM_DEBUG("%lu %d %p\n",
 			  map->size, drm_order(map->size), map->handle);
@@ -200,15 +213,45 @@ static int drm_addmap_core(drm_device_t * dev, unsigned int offset,
 			dev->sigdata.lock = dev->lock.hw_lock = map->handle;	/* Pointer to lock */
 		}
 		break;
-	case _DRM_AGP:
-		if (drm_core_has_AGP(dev)) {
+	case _DRM_AGP: {
+		drm_agp_mem_t *entry;
+		int valid = 0;
+
+		if (!drm_core_has_AGP(dev)) {
+			drm_free(map, sizeof(*map), DRM_MEM_MAPS);
+			return -EINVAL;
+		}
 #ifdef __alpha__
-			map->offset += dev->hose->mem_space->start;
+		map->offset += dev->hose->mem_space->start;
 #endif
-			map->offset += dev->agp->base;
-			map->mtrr = dev->agp->agp_mtrr;	/* for getmap */
+		/* Note: dev->agp->base may actually be 0 when the DRM
+		 * is not in control of AGP space. But if user space is
+		 * it should already have added the AGP base itself.
+		 */
+		map->offset += dev->agp->base;
+		map->mtrr = dev->agp->agp_mtrr;	/* for getmap */
+
+		/* This assumes the DRM is in total control of AGP space.
+		 * It's not always the case as AGP can be in the control
+		 * of user space (i.e. i810 driver). So this loop will get
+		 * skipped and we double check that dev->agp->memory is
+		 * actually set as well as being invalid before EPERM'ing
+		 */
+		for (entry = dev->agp->memory; entry; entry = entry->next) {
+			if ((map->offset >= entry->bound) &&
+			    (map->offset + map->size <= entry->bound + entry->pages * PAGE_SIZE)) {
+				valid = 1;
+				break;
+			}
 		}
+		if (dev->agp->memory && !valid) {
+			drm_free(map, sizeof(*map), DRM_MEM_MAPS);
+			return -EPERM;
+		}
+		DRM_DEBUG("AGP offset = 0x%08lx, size = 0x%08lx\n", map->offset, map->size);
+
 		break;
+	}
 	case _DRM_SCATTER_GATHER:
 		if (!dev->sg) {
 			drm_free(map, sizeof(*map), DRM_MEM_MAPS);
@@ -267,7 +310,7 @@ static int drm_addmap_core(drm_device_t * dev, unsigned int offset,
 
 	*maplist = list;
 	return 0;
-}
+	}
 
 int drm_addmap(drm_device_t * dev, unsigned int offset,
 	       unsigned int size, drm_map_type_t type,
@@ -519,6 +562,7 @@ int drm_addbufs_agp(drm_device_t * dev, drm_buf_desc_t * request)
 {
 	drm_device_dma_t *dma = dev->dma;
 	drm_buf_entry_t *entry;
+	drm_agp_mem_t *agp_entry;
 	drm_buf_t *buf;
 	unsigned long offset;
 	unsigned long agp_offset;
@@ -529,7 +573,7 @@ int drm_addbufs_agp(drm_device_t * dev, drm_buf_desc_t * request)
 	int page_order;
 	int total;
 	int byte_count;
-	int i;
+	int i, valid;
 	drm_buf_t **temp_buflist;
 
 	if (!dma)
@@ -560,6 +604,19 @@ int drm_addbufs_agp(drm_device_t * dev, drm_buf_desc_t * request)
 	if (dev->queue_count)
 		return -EBUSY;	/* Not while in use */
 
+	/* Make sure buffers are located in AGP memory that we own */
+	valid = 0;
+	for (agp_entry = dev->agp->memory; agp_entry; agp_entry = agp_entry->next) {
+		if ((agp_offset >= agp_entry->bound) &&
+		    (agp_offset + total * count <= agp_entry->bound + agp_entry->pages * PAGE_SIZE)) {
+			valid = 1;
+			break;
+		}
+	}
+	if (dev->agp->memory && !valid) {
+		DRM_DEBUG("zone invalid\n");
+		return -EINVAL;
+	}
 	spin_lock(&dev->count_lock);
 	if (dev->buf_use) {
 		spin_unlock(&dev->count_lock);
diff --git a/drivers/char/drm/drm_dma.c b/drivers/char/drm/drm_dma.c
index 892db70..32ed19c 100644
--- a/drivers/char/drm/drm_dma.c
+++ b/drivers/char/drm/drm_dma.c
@@ -65,7 +65,7 @@ int drm_dma_setup(drm_device_t * dev)
  * \param dev DRM device.
  *
  * Free all pages associated with DMA buffers, the buffers and pages lists, and
- * finally the the drm_device::dma structure itself.
+ * finally the drm_device::dma structure itself.
  */
 void drm_dma_takedown(drm_device_t * dev)
 {
diff --git a/drivers/char/drm/drm_drv.c b/drivers/char/drm/drm_drv.c
index f5b9b24..8e77b7e 100644
--- a/drivers/char/drm/drm_drv.c
+++ b/drivers/char/drm/drm_drv.c
@@ -15,8 +15,6 @@
  * #define DRIVER_DESC		"Matrox G200/G400"
  * #define DRIVER_DATE		"20001127"
  *
- * #define DRIVER_IOCTL_COUNT	DRM_ARRAY_SIZE( mga_ioctls )
- *
  * #define drm_x		mga_##x
  * \endcode
  */
@@ -120,7 +118,7 @@ static drm_ioctl_desc_t drm_ioctls[] = {
 	[DRM_IOCTL_NR(DRM_IOCTL_UPDATE_DRAW)] = {drm_update_drawable_info, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY},
 };
 
-#define DRIVER_IOCTL_COUNT	ARRAY_SIZE( drm_ioctls )
+#define DRM_CORE_IOCTL_COUNT	ARRAY_SIZE( drm_ioctls )
 
 /**
  * Take down the DRM device.
@@ -496,11 +494,14 @@ int drm_ioctl(struct inode *inode, struct file *filp,
 		  (long)old_encode_dev(priv->head->device),
 		  priv->authenticated);
 
-	if (nr < DRIVER_IOCTL_COUNT)
-		ioctl = &drm_ioctls[nr];
-	else if ((nr >= DRM_COMMAND_BASE)
-		 && (nr < DRM_COMMAND_BASE + dev->driver->num_ioctls))
+	if ((nr >= DRM_CORE_IOCTL_COUNT) &&
+	    ((nr < DRM_COMMAND_BASE) || (nr >= DRM_COMMAND_END)))
+		goto err_i1;
+	if ((nr >= DRM_COMMAND_BASE) && (nr < DRM_COMMAND_END) &&
+	    (nr < DRM_COMMAND_BASE + dev->driver->num_ioctls))
 		ioctl = &dev->driver->ioctls[nr - DRM_COMMAND_BASE];
+	else if ((nr >= DRM_COMMAND_END) || (nr < DRM_COMMAND_BASE))
+		ioctl = &drm_ioctls[nr];
 	else
 		goto err_i1;
 
diff --git a/drivers/char/drm/drm_fops.c b/drivers/char/drm/drm_fops.c
index 898f47d..3b159ca 100644
--- a/drivers/char/drm/drm_fops.c
+++ b/drivers/char/drm/drm_fops.c
@@ -46,6 +46,7 @@ static int drm_setup(drm_device_t * dev)
 	drm_local_map_t *map;
 	int i;
 	int ret;
+	u32 sareapage;
 
 	if (dev->driver->firstopen) {
 		ret = dev->driver->firstopen(dev);
@@ -56,7 +57,8 @@ static int drm_setup(drm_device_t * dev)
 	dev->magicfree.next = NULL;
 
 	/* prebuild the SAREA */
-	i = drm_addmap(dev, 0, SAREA_MAX, _DRM_SHM, _DRM_CONTAINS_LOCK, &map);
+	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;
 
@@ -84,7 +86,7 @@ static int drm_setup(drm_device_t * dev)
 	INIT_LIST_HEAD(&dev->ctxlist->head);
 
 	dev->vmalist = NULL;
-	dev->sigdata.lock = dev->lock.hw_lock = NULL;
+	dev->sigdata.lock = NULL;
 	init_waitqueue_head(&dev->lock.lock_queue);
 	dev->queue_count = 0;
 	dev->queue_reserved = 0;
@@ -354,58 +356,56 @@ int drm_release(struct inode *inode, struct file *filp)
 		  current->pid, (long)old_encode_dev(priv->head->device),
 		  dev->open_count);
 
-	if (priv->lock_count && dev->lock.hw_lock &&
-	    _DRM_LOCK_IS_HELD(dev->lock.hw_lock->lock) &&
-	    dev->lock.filp == filp) {
-		DRM_DEBUG("File %p released, freeing lock for context %d\n",
-			  filp, _DRM_LOCKING_CONTEXT(dev->lock.hw_lock->lock));
-
-		if (dev->driver->reclaim_buffers_locked)
+	if (dev->driver->reclaim_buffers_locked && dev->lock.hw_lock) {
+		if (drm_i_have_hw_lock(filp)) {
 			dev->driver->reclaim_buffers_locked(dev, filp);
-
-		drm_lock_free(dev, &dev->lock.hw_lock->lock,
-			      _DRM_LOCKING_CONTEXT(dev->lock.hw_lock->lock));
-
-		/* FIXME: may require heavy-handed reset of
-		   hardware at this point, possibly
-		   processed via a callback to the X
-		   server. */
-	} else if (dev->driver->reclaim_buffers_locked && priv->lock_count
-		   && dev->lock.hw_lock) {
-		/* The lock is required to reclaim buffers */
-		DECLARE_WAITQUEUE(entry, current);
-
-		add_wait_queue(&dev->lock.lock_queue, &entry);
-		for (;;) {
-			__set_current_state(TASK_INTERRUPTIBLE);
-			if (!dev->lock.hw_lock) {
-				/* Device has been unregistered */
-				retcode = -EINTR;
-				break;
+		} else {
+			unsigned long _end=jiffies + 3*DRM_HZ;
+			int locked = 0;
+
+			drm_idlelock_take(&dev->lock);
+
+			/*
+			 * Wait for a while.
+			 */
+
+			do{
+				spin_lock(&dev->lock.spinlock);
+				locked = dev->lock.idle_has_lock;
+				spin_unlock(&dev->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");
 			}
-			if (drm_lock_take(&dev->lock.hw_lock->lock,
-					  DRM_KERNEL_CONTEXT)) {
-				dev->lock.filp = filp;
-				dev->lock.lock_time = jiffies;
-				atomic_inc(&dev->counts[_DRM_STAT_LOCKS]);
-				break;	/* Got lock */
-			}
-			/* Contention */
-			schedule();
-			if (signal_pending(current)) {
-				retcode = -ERESTARTSYS;
-				break;
-			}
-		}
-		__set_current_state(TASK_RUNNING);
-		remove_wait_queue(&dev->lock.lock_queue, &entry);
-		if (!retcode) {
+
 			dev->driver->reclaim_buffers_locked(dev, filp);
-			drm_lock_free(dev, &dev->lock.hw_lock->lock,
-				      DRM_KERNEL_CONTEXT);
+			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, filp);
+		drm_idlelock_release(&dev->lock);
+
+	}
+
+	if (drm_i_have_hw_lock(filp)) {
+		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, filp);
diff --git a/drivers/char/drm/drm_hashtab.c b/drivers/char/drm/drm_hashtab.c
index a0b2d68..31acb62 100644
--- a/drivers/char/drm/drm_hashtab.c
+++ b/drivers/char/drm/drm_hashtab.c
@@ -43,7 +43,16 @@ int drm_ht_create(drm_open_hash_t *ht, unsigned int order)
 	ht->size = 1 << order;
 	ht->order = order;
 	ht->fill = 0;
-	ht->table = vmalloc(ht->size*sizeof(*ht->table));
+	ht->table = NULL;
+	ht->use_vmalloc = ((ht->size * sizeof(*ht->table)) > PAGE_SIZE);
+	if (!ht->use_vmalloc) {
+		ht->table = drm_calloc(ht->size, sizeof(*ht->table),
+				       DRM_MEM_HASHTAB);
+	}
+	if (!ht->table) {
+		ht->use_vmalloc = 1;
+		ht->table = vmalloc(ht->size*sizeof(*ht->table));
+	}
 	if (!ht->table) {
 		DRM_ERROR("Out of memory for hash table\n");
 		return -ENOMEM;
@@ -183,7 +192,11 @@ int drm_ht_remove_item(drm_open_hash_t *ht, drm_hash_item_t *item)
 void drm_ht_remove(drm_open_hash_t *ht)
 {
 	if (ht->table) {
-		vfree(ht->table);
+		if (ht->use_vmalloc)
+			vfree(ht->table);
+		else
+			drm_free(ht->table, ht->size * sizeof(*ht->table),
+				 DRM_MEM_HASHTAB);
 		ht->table = NULL;
 	}
 }
diff --git a/drivers/char/drm/drm_hashtab.h b/drivers/char/drm/drm_hashtab.h
index 40afec0..613091c 100644
--- a/drivers/char/drm/drm_hashtab.h
+++ b/drivers/char/drm/drm_hashtab.h
@@ -47,6 +47,7 @@ typedef struct drm_open_hash{
 	unsigned int order;
 	unsigned int fill;
 	struct hlist_head *table;
+	int use_vmalloc;
 } drm_open_hash_t;
 
 
diff --git a/drivers/char/drm/drm_irq.c b/drivers/char/drm/drm_irq.c
index 9d00c51..2e75331 100644
--- a/drivers/char/drm/drm_irq.c
+++ b/drivers/char/drm/drm_irq.c
@@ -424,7 +424,7 @@ static void drm_locked_tasklet_func(unsigned long data)
 	spin_lock_irqsave(&dev->tasklet_lock, irqflags);
 
 	if (!dev->locked_tasklet_func ||
-	    !drm_lock_take(&dev->lock.hw_lock->lock,
+	    !drm_lock_take(&dev->lock,
 			   DRM_KERNEL_CONTEXT)) {
 		spin_unlock_irqrestore(&dev->tasklet_lock, irqflags);
 		return;
@@ -435,7 +435,7 @@ static void drm_locked_tasklet_func(unsigned long data)
 
 	dev->locked_tasklet_func(dev);
 
-	drm_lock_free(dev, &dev->lock.hw_lock->lock,
+	drm_lock_free(&dev->lock,
 		      DRM_KERNEL_CONTEXT);
 
 	dev->locked_tasklet_func = NULL;
diff --git a/drivers/char/drm/drm_lock.c b/drivers/char/drm/drm_lock.c
index e9993ba..befd1af 100644
--- a/drivers/char/drm/drm_lock.c
+++ b/drivers/char/drm/drm_lock.c
@@ -35,9 +35,6 @@
 
 #include "drmP.h"
 
-static int drm_lock_transfer(drm_device_t * dev,
-			     __volatile__ unsigned int *lock,
-			     unsigned int context);
 static int drm_notifier(void *priv);
 
 /**
@@ -80,6 +77,9 @@ int drm_lock(struct inode *inode, struct file *filp,
 			return -EINVAL;
 
 	add_wait_queue(&dev->lock.lock_queue, &entry);
+	spin_lock(&dev->lock.spinlock);
+	dev->lock.user_waiters++;
+	spin_unlock(&dev->lock.spinlock);
 	for (;;) {
 		__set_current_state(TASK_INTERRUPTIBLE);
 		if (!dev->lock.hw_lock) {
@@ -87,7 +87,7 @@ int drm_lock(struct inode *inode, struct file *filp,
 			ret = -EINTR;
 			break;
 		}
-		if (drm_lock_take(&dev->lock.hw_lock->lock, lock.context)) {
+		if (drm_lock_take(&dev->lock, lock.context)) {
 			dev->lock.filp = filp;
 			dev->lock.lock_time = jiffies;
 			atomic_inc(&dev->counts[_DRM_STAT_LOCKS]);
@@ -101,12 +101,14 @@ int drm_lock(struct inode *inode, struct file *filp,
 			break;
 		}
 	}
+	spin_lock(&dev->lock.spinlock);
+	dev->lock.user_waiters--;
+	spin_unlock(&dev->lock.spinlock);
 	__set_current_state(TASK_RUNNING);
 	remove_wait_queue(&dev->lock.lock_queue, &entry);
 
-	DRM_DEBUG("%d %s\n", lock.context, ret ? "interrupted" : "has lock");
-	if (ret)
-		return ret;
+	DRM_DEBUG( "%d %s\n", lock.context, ret ? "interrupted" : "has lock" );
+	if (ret) return ret;
 
 	sigemptyset(&dev->sigmask);
 	sigaddset(&dev->sigmask, SIGSTOP);
@@ -127,14 +129,12 @@ int drm_lock(struct inode *inode, struct file *filp,
 		}
 	}
 
-	/* dev->driver->kernel_context_switch isn't used by any of the x86
-	 *  drivers but is used by the Sparc driver.
-	 */
 	if (dev->driver->kernel_context_switch &&
 	    dev->last_context != lock.context) {
 		dev->driver->kernel_context_switch(dev, dev->last_context,
 						   lock.context);
 	}
+
 	return 0;
 }
 
@@ -184,12 +184,8 @@ int drm_unlock(struct inode *inode, struct file *filp,
 	if (dev->driver->kernel_context_switch_unlock)
 		dev->driver->kernel_context_switch_unlock(dev);
 	else {
-		drm_lock_transfer(dev, &dev->lock.hw_lock->lock,
-				  DRM_KERNEL_CONTEXT);
-
-		if (drm_lock_free(dev, &dev->lock.hw_lock->lock,
-				  DRM_KERNEL_CONTEXT)) {
-			DRM_ERROR("\n");
+		if (drm_lock_free(&dev->lock,lock.context)) {
+			/* FIXME: Should really bail out here. */
 		}
 	}
 
@@ -206,18 +202,26 @@ int drm_unlock(struct inode *inode, struct file *filp,
  *
  * Attempt to mark the lock as held by the given context, via the \p cmpxchg instruction.
  */
-int drm_lock_take(__volatile__ unsigned int *lock, unsigned int context)
+int drm_lock_take(drm_lock_data_t *lock_data,
+		  unsigned int context)
 {
 	unsigned int old, new, prev;
+	volatile unsigned int *lock = &lock_data->hw_lock->lock;
 
+	spin_lock(&lock_data->spinlock);
 	do {
 		old = *lock;
 		if (old & _DRM_LOCK_HELD)
 			new = old | _DRM_LOCK_CONT;
-		else
-			new = context | _DRM_LOCK_HELD;
+		else {
+			new = context | _DRM_LOCK_HELD |
+				((lock_data->user_waiters + lock_data->kernel_waiters > 1) ?
+				 _DRM_LOCK_CONT : 0);
+		}
 		prev = cmpxchg(lock, old, new);
 	} while (prev != old);
+	spin_unlock(&lock_data->spinlock);
+
 	if (_DRM_LOCKING_CONTEXT(old) == context) {
 		if (old & _DRM_LOCK_HELD) {
 			if (context != DRM_KERNEL_CONTEXT) {
@@ -227,7 +231,8 @@ int drm_lock_take(__volatile__ unsigned int *lock, unsigned int context)
 			return 0;
 		}
 	}
-	if (new == (context | _DRM_LOCK_HELD)) {
+
+	if ((_DRM_LOCKING_CONTEXT(new)) == context && (new & _DRM_LOCK_HELD)) {
 		/* Have lock */
 		return 1;
 	}
@@ -246,13 +251,13 @@ int drm_lock_take(__volatile__ unsigned int *lock, unsigned int context)
  * Resets the lock file pointer.
  * Marks the lock as held by the given context, via the \p cmpxchg instruction.
  */
-static int drm_lock_transfer(drm_device_t * dev,
-			     __volatile__ unsigned int *lock,
+static int drm_lock_transfer(drm_lock_data_t *lock_data,
 			     unsigned int context)
 {
 	unsigned int old, new, prev;
+	volatile unsigned int *lock = &lock_data->hw_lock->lock;
 
-	dev->lock.filp = NULL;
+	lock_data->filp = NULL;
 	do {
 		old = *lock;
 		new = context | _DRM_LOCK_HELD;
@@ -272,23 +277,32 @@ static int drm_lock_transfer(drm_device_t * dev,
  * Marks the lock as not held, via the \p cmpxchg instruction. Wakes any task
  * waiting on the lock queue.
  */
-int drm_lock_free(drm_device_t * dev,
-		  __volatile__ unsigned int *lock, unsigned int context)
+int drm_lock_free(drm_lock_data_t *lock_data, unsigned int context)
 {
 	unsigned int old, new, prev;
+	volatile unsigned int *lock = &lock_data->hw_lock->lock;
+
+	spin_lock(&lock_data->spinlock);
+	if (lock_data->kernel_waiters != 0) {
+		drm_lock_transfer(lock_data, 0);
+		lock_data->idle_has_lock = 1;
+		spin_unlock(&lock_data->spinlock);
+		return 1;
+	}
+	spin_unlock(&lock_data->spinlock);
 
-	dev->lock.filp = NULL;
 	do {
 		old = *lock;
-		new = 0;
+		new = _DRM_LOCKING_CONTEXT(old);
 		prev = cmpxchg(lock, old, new);
 	} while (prev != old);
+
 	if (_DRM_LOCK_IS_HELD(old) && _DRM_LOCKING_CONTEXT(old) != context) {
 		DRM_ERROR("%d freed heavyweight lock held by %d\n",
 			  context, _DRM_LOCKING_CONTEXT(old));
 		return 1;
 	}
-	wake_up_interruptible(&dev->lock.lock_queue);
+	wake_up_interruptible(&lock_data->lock_queue);
 	return 0;
 }
 
@@ -322,3 +336,67 @@ static int drm_notifier(void *priv)
 	} while (prev != old);
 	return 0;
 }
+
+/**
+ * This function returns immediately and takes the hw lock
+ * with the kernel context if it is free, otherwise it gets the highest priority when and if
+ * it is eventually released.
+ *
+ * This guarantees that the kernel will _eventually_ have the lock _unless_ it is held
+ * by a blocked process. (In the latter case an explicit wait for the hardware lock would cause
+ * a deadlock, which is why the "idlelock" was invented).
+ *
+ * This should be sufficient to wait for GPU idle without
+ * having to worry about starvation.
+ */
+
+void drm_idlelock_take(drm_lock_data_t *lock_data)
+{
+	int ret = 0;
+
+	spin_lock(&lock_data->spinlock);
+	lock_data->kernel_waiters++;
+	if (!lock_data->idle_has_lock) {
+
+		spin_unlock(&lock_data->spinlock);
+		ret = drm_lock_take(lock_data, DRM_KERNEL_CONTEXT);
+		spin_lock(&lock_data->spinlock);
+
+		if (ret == 1)
+			lock_data->idle_has_lock = 1;
+	}
+	spin_unlock(&lock_data->spinlock);
+}
+EXPORT_SYMBOL(drm_idlelock_take);
+
+void drm_idlelock_release(drm_lock_data_t *lock_data)
+{
+	unsigned int old, prev;
+	volatile unsigned int *lock = &lock_data->hw_lock->lock;
+
+	spin_lock(&lock_data->spinlock);
+	if (--lock_data->kernel_waiters == 0) {
+		if (lock_data->idle_has_lock) {
+			do {
+				old = *lock;
+				prev = cmpxchg(lock, old, DRM_KERNEL_CONTEXT);
+			} while (prev != old);
+			wake_up_interruptible(&lock_data->lock_queue);
+			lock_data->idle_has_lock = 0;
+		}
+	}
+	spin_unlock(&lock_data->spinlock);
+}
+EXPORT_SYMBOL(drm_idlelock_release);
+
+
+int drm_i_have_hw_lock(struct file *filp)
+{
+	DRM_DEVICE;
+
+	return (priv->lock_count && dev->lock.hw_lock &&
+		_DRM_LOCK_IS_HELD(dev->lock.hw_lock->lock) &&
+		dev->lock.filp == filp);
+}
+
+EXPORT_SYMBOL(drm_i_have_hw_lock);
diff --git a/drivers/char/drm/drm_mm.c b/drivers/char/drm/drm_mm.c
index 9b46b85..2ec1d9f 100644
--- a/drivers/char/drm/drm_mm.c
+++ b/drivers/char/drm/drm_mm.c
@@ -274,7 +274,6 @@ int drm_mm_init(drm_mm_t * mm, unsigned long start, unsigned long size)
 	return drm_mm_create_tail_node(mm, start, size);
 }
 
-EXPORT_SYMBOL(drm_mm_init);
 
 void drm_mm_takedown(drm_mm_t * mm)
 {
@@ -295,4 +294,3 @@ void drm_mm_takedown(drm_mm_t * mm)
 	drm_free(entry, sizeof(*entry), DRM_MEM_MM);
 }
 
-EXPORT_SYMBOL(drm_mm_takedown);
diff --git a/drivers/char/drm/drm_os_linux.h b/drivers/char/drm/drm_os_linux.h
index 2908b72..0fe7b44 100644
--- a/drivers/char/drm/drm_os_linux.h
+++ b/drivers/char/drm/drm_os_linux.h
@@ -70,9 +70,6 @@ static __inline__ int mtrr_del(int reg, unsigned long base, unsigned long size)
 
 #endif
 
-/** Task queue handler arguments */
-#define DRM_TASKQUEUE_ARGS	void *arg
-
 /** For data going into the kernel through the ioctl argument */
 #define DRM_COPY_FROM_USER_IOCTL(arg1, arg2, arg3)	\
 	if ( copy_from_user(&arg1, arg2, arg3) )	\
diff --git a/drivers/char/drm/drm_pciids.h b/drivers/char/drm/drm_pciids.h
index ad54b84..31cdde8 100644
--- a/drivers/char/drm/drm_pciids.h
+++ b/drivers/char/drm/drm_pciids.h
@@ -102,6 +102,7 @@
 	{0x1002, 0x5653, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RV410|RADEON_IS_MOBILITY|RADEON_NEW_MEMMAP}, \
 	{0x1002, 0x5834, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS300|RADEON_IS_IGP}, \
 	{0x1002, 0x5835, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS300|RADEON_IS_IGP|RADEON_IS_MOBILITY}, \
+	{0x1002, 0x5955, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RS400|RADEON_IS_IGP|RADEON_IS_MOBILITY|RADEON_IS_IGPGART}, \
 	{0x1002, 0x5960, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RV280}, \
 	{0x1002, 0x5961, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RV280}, \
 	{0x1002, 0x5962, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RV280}, \
@@ -230,10 +231,10 @@
 	{0x1106, 0x7205, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0}, \
 	{0x1106, 0x3108, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0}, \
 	{0x1106, 0x3304, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0}, \
-	{0x1106, 0x3157, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0}, \
 	{0x1106, 0x3344, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0}, \
 	{0x1106, 0x3343, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0}, \
 	{0x1106, 0x3230, PCI_ANY_ID, PCI_ANY_ID, 0, 0, VIA_DX9_0}, \
+	{0x1106, 0x3157, PCI_ANY_ID, PCI_ANY_ID, 0, 0, VIA_PRO_GROUP_A}, \
 	{0, 0, 0}
 
 #define i810_PCI_IDS \
@@ -296,5 +297,6 @@
 	{0x8086, 0x2982, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0}, \
 	{0x8086, 0x2992, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0}, \
 	{0x8086, 0x29a2, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0}, \
+	{0x8086, 0x2a02, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0}, \
 	{0, 0, 0}
 
diff --git a/drivers/char/drm/drm_proc.c b/drivers/char/drm/drm_proc.c
index 7fd0da7..b204498 100644
--- a/drivers/char/drm/drm_proc.c
+++ b/drivers/char/drm/drm_proc.c
@@ -72,7 +72,7 @@ static struct drm_proc_list {
 #endif
 };
 
-#define DRM_PROC_ENTRIES (sizeof(drm_proc_list)/sizeof(drm_proc_list[0]))
+#define DRM_PROC_ENTRIES ARRAY_SIZE(drm_proc_list)
 
 /**
  * Initialize the DRI proc filesystem for a device.
diff --git a/drivers/char/drm/drm_stub.c b/drivers/char/drm/drm_stub.c
index 120d102..19408ad 100644
--- a/drivers/char/drm/drm_stub.c
+++ b/drivers/char/drm/drm_stub.c
@@ -62,6 +62,7 @@ static int drm_fill_in_dev(drm_device_t * dev, struct pci_dev *pdev,
 	spin_lock_init(&dev->count_lock);
 	spin_lock_init(&dev->drw_lock);
 	spin_lock_init(&dev->tasklet_lock);
+	spin_lock_init(&dev->lock.spinlock);
 	init_timer(&dev->timer);
 	mutex_init(&dev->struct_mutex);
 	mutex_init(&dev->ctxlist_mutex);
diff --git a/drivers/char/drm/drm_vm.c b/drivers/char/drm/drm_vm.c
index 54a6328..b5c5b9f 100644
--- a/drivers/char/drm/drm_vm.c
+++ b/drivers/char/drm/drm_vm.c
@@ -41,6 +41,30 @@
 static void drm_vm_open(struct vm_area_struct *vma);
 static void drm_vm_close(struct vm_area_struct *vma);
 
+static pgprot_t drm_io_prot(uint32_t map_type, struct vm_area_struct *vma)
+{
+	pgprot_t tmp = vm_get_page_prot(vma->vm_flags);
+
+#if defined(__i386__) || defined(__x86_64__)
+	if (boot_cpu_data.x86 > 3 && map_type != _DRM_AGP) {
+		pgprot_val(tmp) |= _PAGE_PCD;
+		pgprot_val(tmp) &= ~_PAGE_PWT;
+	}
+#elif defined(__powerpc__)
+	pgprot_val(tmp) |= _PAGE_NO_CACHE;
+	if (map_type == _DRM_REGISTERS)
+		pgprot_val(tmp) |= _PAGE_GUARDED;
+#endif
+#if defined(__ia64__)
+	if (efi_range_is_wc(vma->vm_start, vma->vm_end -
+				    vma->vm_start))
+		tmp = pgprot_writecombine(tmp);
+	else
+		tmp = pgprot_noncached(tmp);
+#endif
+	return tmp;
+}
+
 /**
  * \c nopage method for AGP virtual memory.
  *
@@ -133,7 +157,7 @@ static __inline__ struct page *drm_do_vm_nopage(struct vm_area_struct *vma,
  * \param address access address.
  * \return pointer to the page structure.
  *
- * Get the the mapping, find the real physical page to map, get the page, and
+ * Get the mapping, find the real physical page to map, get the page, and
  * return it.
  */
 static __inline__ struct page *drm_do_vm_shm_nopage(struct vm_area_struct *vma,
@@ -151,8 +175,7 @@ static __inline__ struct page *drm_do_vm_shm_nopage(struct vm_area_struct *vma,
 
 	offset = address - vma->vm_start;
 	i = (unsigned long)map->handle + offset;
-	page = (map->type == _DRM_CONSISTENT) ?
-		virt_to_page((void *)i) : vmalloc_to_page((void *)i);
+	page = vmalloc_to_page((void *)i);
 	if (!page)
 		return NOPAGE_SIGBUS;
 	get_page(page);
@@ -389,7 +412,7 @@ static struct vm_operations_struct drm_vm_sg_ops = {
  * Create a new drm_vma_entry structure as the \p vma private data entry and
  * add it to drm_device::vmalist.
  */
-static void drm_vm_open(struct vm_area_struct *vma)
+static void drm_vm_open_locked(struct vm_area_struct *vma)
 {
 	drm_file_t *priv = vma->vm_file->private_data;
 	drm_device_t *dev = priv->head->dev;
@@ -401,15 +424,23 @@ static void drm_vm_open(struct vm_area_struct *vma)
 
 	vma_entry = drm_alloc(sizeof(*vma_entry), DRM_MEM_VMAS);
 	if (vma_entry) {
-		mutex_lock(&dev->struct_mutex);
 		vma_entry->vma = vma;
 		vma_entry->next = dev->vmalist;
 		vma_entry->pid = current->pid;
 		dev->vmalist = vma_entry;
-		mutex_unlock(&dev->struct_mutex);
 	}
 }
 
+static void drm_vm_open(struct vm_area_struct *vma)
+{
+	drm_file_t *priv = vma->vm_file->private_data;
+	drm_device_t *dev = priv->head->dev;
+
+	mutex_lock(&dev->struct_mutex);
+	drm_vm_open_locked(vma);
+	mutex_unlock(&dev->struct_mutex);
+}
+
 /**
  * \c close method for all virtual memory types.
  *
@@ -460,7 +491,6 @@ static int drm_mmap_dma(struct file *filp, struct vm_area_struct *vma)
 	drm_device_dma_t *dma;
 	unsigned long length = vma->vm_end - vma->vm_start;
 
-	lock_kernel();
 	dev = priv->head->dev;
 	dma = dev->dma;
 	DRM_DEBUG("start = 0x%lx, end = 0x%lx, page offset = 0x%lx\n",
@@ -468,10 +498,8 @@ static int drm_mmap_dma(struct file *filp, struct vm_area_struct *vma)
 
 	/* Length must match exact page count */
 	if (!dma || (length >> PAGE_SHIFT) != dma->page_count) {
-		unlock_kernel();
 		return -EINVAL;
 	}
-	unlock_kernel();
 
 	if (!capable(CAP_SYS_ADMIN) &&
 	    (dma->flags & _DRM_DMA_USE_PCI_RO)) {
@@ -494,7 +522,7 @@ static int drm_mmap_dma(struct file *filp, struct vm_area_struct *vma)
 	vma->vm_flags |= VM_RESERVED;	/* Don't swap */
 
 	vma->vm_file = filp;	/* Needed for drm_vm_open() */
-	drm_vm_open(vma);
+	drm_vm_open_locked(vma);
 	return 0;
 }
 
@@ -529,7 +557,7 @@ EXPORT_SYMBOL(drm_core_get_reg_ofs);
  * according to the mapping type and remaps the pages. Finally sets the file
  * pointer and calls vm_open().
  */
-int drm_mmap(struct file *filp, struct vm_area_struct *vma)
+static int drm_mmap_locked(struct file *filp, struct vm_area_struct *vma)
 {
 	drm_file_t *priv = filp->private_data;
 	drm_device_t *dev = priv->head->dev;
@@ -565,7 +593,7 @@ int drm_mmap(struct file *filp, struct vm_area_struct *vma)
 		return -EPERM;
 
 	/* Check for valid size. */
-	if (map->size != vma->vm_end - vma->vm_start)
+	if (map->size < vma->vm_end - vma->vm_start)
 		return -EINVAL;
 
 	if (!capable(CAP_SYS_ADMIN) && (map->flags & _DRM_READ_ONLY)) {
@@ -600,37 +628,16 @@ int drm_mmap(struct file *filp, struct vm_area_struct *vma)
 		/* fall through to _DRM_FRAME_BUFFER... */
 	case _DRM_FRAME_BUFFER:
 	case _DRM_REGISTERS:
-#if defined(__i386__) || defined(__x86_64__)
-		if (boot_cpu_data.x86 > 3 && map->type != _DRM_AGP) {
-			pgprot_val(vma->vm_page_prot) |= _PAGE_PCD;
-			pgprot_val(vma->vm_page_prot) &= ~_PAGE_PWT;
-		}
-#elif defined(__powerpc__)
-		pgprot_val(vma->vm_page_prot) |= _PAGE_NO_CACHE;
-		if (map->type == _DRM_REGISTERS)
-			pgprot_val(vma->vm_page_prot) |= _PAGE_GUARDED;
-#endif
-		vma->vm_flags |= VM_IO;	/* not in core dump */
-#if defined(__ia64__)
-		if (efi_range_is_wc(vma->vm_start, vma->vm_end - vma->vm_start))
-			vma->vm_page_prot =
-			    pgprot_writecombine(vma->vm_page_prot);
-		else
-			vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
-#endif
 		offset = dev->driver->get_reg_ofs(dev);
+		vma->vm_flags |= VM_IO;	/* not in core dump */
+		vma->vm_page_prot = drm_io_prot(map->type, vma);
 #ifdef __sparc__
 		vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
+#endif
 		if (io_remap_pfn_range(vma, vma->vm_start,
 				       (map->offset + offset) >> PAGE_SHIFT,
 				       vma->vm_end - vma->vm_start,
 				       vma->vm_page_prot))
-#else
-		if (io_remap_pfn_range(vma, vma->vm_start,
-				       (map->offset + offset) >> PAGE_SHIFT,
-				       vma->vm_end - vma->vm_start,
-				       vma->vm_page_prot))
-#endif
 			return -EAGAIN;
 		DRM_DEBUG("   Type = %d; start = 0x%lx, end = 0x%lx,"
 			  " offset = 0x%lx\n",
@@ -638,10 +645,15 @@ int drm_mmap(struct file *filp, struct vm_area_struct *vma)
 			  vma->vm_start, vma->vm_end, map->offset + offset);
 		vma->vm_ops = &drm_vm_ops;
 		break;
-	case _DRM_SHM:
 	case _DRM_CONSISTENT:
-		/* Consistent memory is really like shared memory. It's only
-		 * allocate in a different way */
+		/* Consistent memory is really like shared memory. But
+		 * it's allocated in a different way, so avoid nopage */
+		if (remap_pfn_range(vma, vma->vm_start,
+		    page_to_pfn(virt_to_page(map->handle)),
+		    vma->vm_end - vma->vm_start, vma->vm_page_prot))
+			return -EAGAIN;
+	/* fall through to _DRM_SHM */
+	case _DRM_SHM:
 		vma->vm_ops = &drm_vm_shm_ops;
 		vma->vm_private_data = (void *)map;
 		/* Don't let this area swap.  Change when
@@ -659,8 +671,20 @@ int drm_mmap(struct file *filp, struct vm_area_struct *vma)
 	vma->vm_flags |= VM_RESERVED;	/* Don't swap */
 
 	vma->vm_file = filp;	/* Needed for drm_vm_open() */
-	drm_vm_open(vma);
+	drm_vm_open_locked(vma);
 	return 0;
 }
 
+int drm_mmap(struct file *filp, struct vm_area_struct *vma)
+{
+	drm_file_t *priv = filp->private_data;
+	drm_device_t *dev = priv->head->dev;
+	int ret;
+
+	mutex_lock(&dev->struct_mutex);
+	ret = drm_mmap_locked(filp, vma);
+	mutex_unlock(&dev->struct_mutex);
+
+	return ret;
+}
 EXPORT_SYMBOL(drm_mmap);
diff --git a/drivers/char/drm/i915_dma.c b/drivers/char/drm/i915_dma.c
index 9354ce3..1ba15d9 100644
--- a/drivers/char/drm/i915_dma.c
+++ b/drivers/char/drm/i915_dma.c
@@ -34,7 +34,8 @@
 #define IS_I965G(dev) (dev->pci_device == 0x2972 || \
 		       dev->pci_device == 0x2982 || \
 		       dev->pci_device == 0x2992 || \
-		       dev->pci_device == 0x29A2)
+		       dev->pci_device == 0x29A2 || \
+		       dev->pci_device == 0x2A02)
 
 /* Really want an OS-independent resettable timer.  Would like to have
  * this loop run for (eg) 3 sec, but have the timer reset every time
diff --git a/drivers/char/drm/r128_cce.c b/drivers/char/drm/r128_cce.c
index db5a604..1014602 100644
--- a/drivers/char/drm/r128_cce.c
+++ b/drivers/char/drm/r128_cce.c
@@ -560,9 +560,10 @@ static int r128_do_init_cce(drm_device_t * dev, drm_r128_init_t * init)
 	if (dev_priv->is_pci) {
 #endif
 		dev_priv->gart_info.gart_table_location = DRM_ATI_GART_MAIN;
+		dev_priv->gart_info.table_size = R128_PCIGART_TABLE_SIZE;
 		dev_priv->gart_info.addr = NULL;
 		dev_priv->gart_info.bus_addr = 0;
-		dev_priv->gart_info.is_pcie = 0;
+		dev_priv->gart_info.gart_reg_if = DRM_ATI_GART_PCI;
 		if (!drm_ati_pcigart_init(dev, &dev_priv->gart_info)) {
 			DRM_ERROR("failed to init PCI GART!\n");
 			dev->dev_private = (void *)dev_priv;
diff --git a/drivers/char/drm/r128_drv.h b/drivers/char/drm/r128_drv.h
index f1efb49..9086835 100644
--- a/drivers/char/drm/r128_drv.h
+++ b/drivers/char/drm/r128_drv.h
@@ -383,6 +383,8 @@ extern long r128_compat_ioctl(struct file *filp, unsigned int cmd,
 
 #define R128_PERFORMANCE_BOXES		0
 
+#define R128_PCIGART_TABLE_SIZE         32768
+
 #define R128_READ(reg)		DRM_READ32(  dev_priv->mmio, (reg) )
 #define R128_WRITE(reg,val)	DRM_WRITE32( dev_priv->mmio, (reg), (val) )
 #define R128_READ8(reg)		DRM_READ8(   dev_priv->mmio, (reg) )
diff --git a/drivers/char/drm/r300_reg.h b/drivers/char/drm/r300_reg.h
index a881f96..ecda760 100644
--- a/drivers/char/drm/r300_reg.h
+++ b/drivers/char/drm/r300_reg.h
@@ -293,7 +293,7 @@ I am fairly certain that they are correct unless stated otherwise in comments.
 #       define R300_PVS_CNTL_1_PROGRAM_START_SHIFT   0
 #       define R300_PVS_CNTL_1_POS_END_SHIFT         10
 #       define R300_PVS_CNTL_1_PROGRAM_END_SHIFT     20
-/* Addresses are relative the the vertex program parameters area. */
+/* Addresses are relative to the vertex program parameters area. */
 #define R300_VAP_PVS_CNTL_2                 0x22D4
 #       define R300_PVS_CNTL_2_PARAM_OFFSET_SHIFT 0
 #       define R300_PVS_CNTL_2_PARAM_COUNT_SHIFT  16
diff --git a/drivers/char/drm/radeon_cp.c b/drivers/char/drm/radeon_cp.c
index 5ed9656..68338389 100644
--- a/drivers/char/drm/radeon_cp.c
+++ b/drivers/char/drm/radeon_cp.c
@@ -830,6 +830,15 @@ static int RADEON_READ_PCIE(drm_radeon_private_t *dev_priv, int addr)
 	return RADEON_READ(RADEON_PCIE_DATA);
 }
 
+static u32 RADEON_READ_IGPGART(drm_radeon_private_t *dev_priv, int addr)
+{
+	u32 ret;
+	RADEON_WRITE(RADEON_IGPGART_INDEX, addr & 0x7f);
+	ret = RADEON_READ(RADEON_IGPGART_DATA);
+	RADEON_WRITE(RADEON_IGPGART_INDEX, 0x7f);
+	return ret;
+}
+
 #if RADEON_FIFO_DEBUG
 static void radeon_status(drm_radeon_private_t * dev_priv)
 {
@@ -1267,7 +1276,44 @@ static void radeon_test_writeback(drm_radeon_private_t * dev_priv)
 	}
 }
 
-/* Enable or disable PCI-E GART on the chip */
+/* Enable or disable IGP GART on the chip */
+static void radeon_set_igpgart(drm_radeon_private_t * dev_priv, int on)
+{
+	u32 temp, tmp;
+
+	tmp = RADEON_READ(RADEON_AIC_CNTL);
+	if (on) {
+		DRM_DEBUG("programming igpgart %08X %08lX %08X\n",
+			 dev_priv->gart_vm_start,
+			 (long)dev_priv->gart_info.bus_addr,
+			 dev_priv->gart_size);
+
+		RADEON_WRITE_IGPGART(RADEON_IGPGART_UNK_18, 0x1000);
+		RADEON_WRITE_IGPGART(RADEON_IGPGART_ENABLE, 0x1);
+		RADEON_WRITE_IGPGART(RADEON_IGPGART_CTRL, 0x42040800);
+		RADEON_WRITE_IGPGART(RADEON_IGPGART_BASE_ADDR,
+				     dev_priv->gart_info.bus_addr);
+
+		temp = RADEON_READ_IGPGART(dev_priv, RADEON_IGPGART_UNK_39);
+		RADEON_WRITE_IGPGART(RADEON_IGPGART_UNK_39, temp);
+
+		RADEON_WRITE(RADEON_AGP_BASE, (unsigned int)dev_priv->gart_vm_start);
+		dev_priv->gart_size = 32*1024*1024;
+		RADEON_WRITE(RADEON_MC_AGP_LOCATION,
+			     (((dev_priv->gart_vm_start - 1 +
+			       dev_priv->gart_size) & 0xffff0000) |
+			     (dev_priv->gart_vm_start >> 16)));
+
+		temp = RADEON_READ_IGPGART(dev_priv, RADEON_IGPGART_ENABLE);
+		RADEON_WRITE_IGPGART(RADEON_IGPGART_ENABLE, temp);
+
+		RADEON_READ_IGPGART(dev_priv, RADEON_IGPGART_FLUSH);
+		RADEON_WRITE_IGPGART(RADEON_IGPGART_FLUSH, 0x1);
+		RADEON_READ_IGPGART(dev_priv, RADEON_IGPGART_FLUSH);
+		RADEON_WRITE_IGPGART(RADEON_IGPGART_FLUSH, 0x0);
+       }
+}
+
 static void radeon_set_pciegart(drm_radeon_private_t * dev_priv, int on)
 {
 	u32 tmp = RADEON_READ_PCIE(dev_priv, RADEON_PCIE_TX_GART_CNTL);
@@ -1302,6 +1348,11 @@ static void radeon_set_pcigart(drm_radeon_private_t * dev_priv, int on)
 {
 	u32 tmp;
 
+	if (dev_priv->flags & RADEON_IS_IGPGART) {
+		radeon_set_igpgart(dev_priv, on);
+		return;
+	}
+
 	if (dev_priv->flags & RADEON_IS_PCIE) {
 		radeon_set_pciegart(dev_priv, on);
 		return;
@@ -1560,8 +1611,8 @@ static int radeon_do_init_cp(drm_device_t * dev, drm_radeon_init_t * init)
 		if (dev_priv->flags & RADEON_IS_AGP) {
 			base = dev->agp->base;
 			/* Check if valid */
-			if ((base + dev_priv->gart_size) > dev_priv->fb_location &&
-			    base < (dev_priv->fb_location + dev_priv->fb_size)) {
+			if ((base + dev_priv->gart_size - 1) >= dev_priv->fb_location &&
+			    base < (dev_priv->fb_location + dev_priv->fb_size - 1)) {
 				DRM_INFO("Can't use AGP base @0x%08lx, won't fit\n",
 					 dev->agp->base);
 				base = 0;
@@ -1571,8 +1622,8 @@ static int radeon_do_init_cp(drm_device_t * dev, drm_radeon_init_t * init)
 		/* If not or if AGP is at 0 (Macs), try to put it elsewhere */
 		if (base == 0) {
 			base = dev_priv->fb_location + dev_priv->fb_size;
-			if (((base + dev_priv->gart_size) & 0xfffffffful)
-			    < base)
+			if (base < dev_priv->fb_location ||
+			    ((base + dev_priv->gart_size) & 0xfffffffful) < base)
 				base = dev_priv->fb_location
 					- dev_priv->gart_size;
 		}		
@@ -1620,20 +1671,22 @@ static int radeon_do_init_cp(drm_device_t * dev, drm_radeon_init_t * init)
 #endif
 	{
 		/* if we have an offset set from userspace */
-		if (dev_priv->pcigart_offset) {
+		if (dev_priv->pcigart_offset_set) {
 			dev_priv->gart_info.bus_addr =
 			    dev_priv->pcigart_offset + dev_priv->fb_location;
 			dev_priv->gart_info.mapping.offset =
 			    dev_priv->gart_info.bus_addr;
 			dev_priv->gart_info.mapping.size =
-			    RADEON_PCIGART_TABLE_SIZE;
+			    dev_priv->gart_info.table_size;
 
 			drm_core_ioremap(&dev_priv->gart_info.mapping, dev);
 			dev_priv->gart_info.addr =
 			    dev_priv->gart_info.mapping.handle;
 
-			dev_priv->gart_info.is_pcie =
-			    !!(dev_priv->flags & RADEON_IS_PCIE);
+			if (dev_priv->flags & RADEON_IS_PCIE)
+				dev_priv->gart_info.gart_reg_if = DRM_ATI_GART_PCIE;
+			else
+				dev_priv->gart_info.gart_reg_if = DRM_ATI_GART_PCI;
 			dev_priv->gart_info.gart_table_location =
 			    DRM_ATI_GART_FB;
 
@@ -1641,6 +1694,10 @@ static int radeon_do_init_cp(drm_device_t * dev, drm_radeon_init_t * init)
 				  dev_priv->gart_info.addr,
 				  dev_priv->pcigart_offset);
 		} else {
+			if (dev_priv->flags & RADEON_IS_IGPGART)
+				dev_priv->gart_info.gart_reg_if = DRM_ATI_GART_IGP;
+			else
+				dev_priv->gart_info.gart_reg_if = DRM_ATI_GART_PCI;
 			dev_priv->gart_info.gart_table_location =
 			    DRM_ATI_GART_MAIN;
 			dev_priv->gart_info.addr = NULL;
@@ -1714,7 +1771,7 @@ static int radeon_do_cleanup_cp(drm_device_t * dev)
 		if (dev_priv->gart_info.gart_table_location == DRM_ATI_GART_FB)
 		{
 			drm_core_ioremapfree(&dev_priv->gart_info.mapping, dev);
-			dev_priv->gart_info.addr = NULL;
+			dev_priv->gart_info.addr = 0;
 		}
 	}
 	/* only clear to the start of flags */
@@ -2222,6 +2279,8 @@ int radeon_driver_firstopen(struct drm_device *dev)
 	drm_local_map_t *map;
 	drm_radeon_private_t *dev_priv = dev->dev_private;
 
+	dev_priv->gart_info.table_size = RADEON_PCIGART_TABLE_SIZE;
+
 	ret = drm_addmap(dev, drm_get_resource_start(dev, 2),
 			 drm_get_resource_len(dev, 2), _DRM_REGISTERS,
 			 _DRM_READ_ONLY, &dev_priv->mmio);
diff --git a/drivers/char/drm/radeon_drm.h b/drivers/char/drm/radeon_drm.h
index 8d6350d..66c4b6f 100644
--- a/drivers/char/drm/radeon_drm.h
+++ b/drivers/char/drm/radeon_drm.h
@@ -707,6 +707,7 @@ typedef struct drm_radeon_setparam {
 #define RADEON_SETPARAM_SWITCH_TILING  2	/* enable/disable color tiling */
 #define RADEON_SETPARAM_PCIGART_LOCATION 3	/* PCI Gart Location */
 #define RADEON_SETPARAM_NEW_MEMMAP 4		/* Use new memory map */
+#define RADEON_SETPARAM_PCIGART_TABLE_SIZE 5    /* PCI GART Table Size */
 
 /* 1.14: Clients can allocate/free a surface
  */
diff --git a/drivers/char/drm/radeon_drv.h b/drivers/char/drm/radeon_drv.h
index 8b105f1..54f49ef 100644
--- a/drivers/char/drm/radeon_drv.h
+++ b/drivers/char/drm/radeon_drv.h
@@ -95,9 +95,11 @@
  * 1.24- Add general-purpose packet for manipulating scratch registers (r300)
  * 1.25- Add support for r200 vertex programs (R200_EMIT_VAP_PVS_CNTL,
  *       new packet type)
+ * 1.26- Add support for variable size PCI(E) gart aperture
+ * 1.27- Add support for IGP GART
  */
 #define DRIVER_MAJOR		1
-#define DRIVER_MINOR		25
+#define DRIVER_MINOR		27
 #define DRIVER_PATCHLEVEL	0
 
 /*
@@ -143,6 +145,7 @@ enum radeon_chip_flags {
 	RADEON_IS_PCIE = 0x00200000UL,
 	RADEON_NEW_MEMMAP = 0x00400000UL,
 	RADEON_IS_PCI = 0x00800000UL,
+	RADEON_IS_IGPGART = 0x01000000UL,
 };
 
 #define GET_RING_HEAD(dev_priv)	(dev_priv->writeback_works ? \
@@ -240,7 +243,6 @@ typedef struct drm_radeon_private {
 
 	int do_boxes;
 	int page_flipping;
-	int current_page;
 
 	u32 color_fmt;
 	unsigned int front_offset;
@@ -280,6 +282,7 @@ typedef struct drm_radeon_private {
 	struct radeon_virt_surface virt_surfaces[2 * RADEON_MAX_SURFACES];
 
 	unsigned long pcigart_offset;
+	unsigned int pcigart_offset_set;
 	drm_ati_pcigart_info gart_info;
 
 	u32 scratch_ages[5];
@@ -432,6 +435,15 @@ extern int r300_do_cp_cmdbuf(drm_device_t * dev, DRMFILE filp,
 #define RADEON_PCIE_TX_GART_END_LO	0x16
 #define RADEON_PCIE_TX_GART_END_HI	0x17
 
+#define RADEON_IGPGART_INDEX            0x168
+#define RADEON_IGPGART_DATA             0x16c
+#define RADEON_IGPGART_UNK_18           0x18
+#define RADEON_IGPGART_CTRL             0x2b
+#define RADEON_IGPGART_BASE_ADDR        0x2c
+#define RADEON_IGPGART_FLUSH            0x2e
+#define RADEON_IGPGART_ENABLE           0x38
+#define RADEON_IGPGART_UNK_39           0x39
+
 #define RADEON_MPP_TB_CONFIG		0x01c0
 #define RADEON_MEM_CNTL			0x0140
 #define RADEON_MEM_SDRAM_MODE_REG	0x0158
@@ -964,6 +976,14 @@ do {									\
 	RADEON_WRITE( RADEON_CLOCK_CNTL_DATA, (val) );			\
 } while (0)
 
+#define RADEON_WRITE_IGPGART( addr, val )				\
+do {									\
+	RADEON_WRITE( RADEON_IGPGART_INDEX,				\
+			((addr) & 0x7f) | (1 << 8));			\
+	RADEON_WRITE( RADEON_IGPGART_DATA, (val) );			\
+	RADEON_WRITE( RADEON_IGPGART_INDEX, 0x7f );			\
+} while (0)
+
 #define RADEON_WRITE_PCIE( addr, val )					\
 do {									\
 	RADEON_WRITE8( RADEON_PCIE_INDEX,				\
diff --git a/drivers/char/drm/radeon_state.c b/drivers/char/drm/radeon_state.c
index 938eccb..98c5f1d 100644
--- a/drivers/char/drm/radeon_state.c
+++ b/drivers/char/drm/radeon_state.c
@@ -773,7 +773,7 @@ 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->page_flipping && dev_priv->current_page == 1) {
+	if (dev_priv->sarea_priv->pfCurrentPage == 1) {
 		OUT_RING(dev_priv->front_pitch_offset);
 	} else {
 		OUT_RING(dev_priv->back_pitch_offset);
@@ -861,7 +861,7 @@ static void radeon_cp_dispatch_clear(drm_device_t * dev,
 
 	dev_priv->stats.clears++;
 
-	if (dev_priv->page_flipping && dev_priv->current_page == 1) {
+	if (dev_priv->sarea_priv->pfCurrentPage == 1) {
 		unsigned int tmp = flags;
 
 		flags &= ~(RADEON_FRONT | RADEON_BACK);
@@ -1382,7 +1382,7 @@ static void radeon_cp_dispatch_swap(drm_device_t * dev)
 		/* Make this work even if front & back are flipped:
 		 */
 		OUT_RING(CP_PACKET0(RADEON_SRC_PITCH_OFFSET, 1));
-		if (dev_priv->current_page == 0) {
+		if (dev_priv->sarea_priv->pfCurrentPage == 0) {
 			OUT_RING(dev_priv->back_pitch_offset);
 			OUT_RING(dev_priv->front_pitch_offset);
 		} else {
@@ -1416,12 +1416,12 @@ static void radeon_cp_dispatch_flip(drm_device_t * dev)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
 	drm_sarea_t *sarea = (drm_sarea_t *) dev_priv->sarea->handle;
-	int offset = (dev_priv->current_page == 1)
+	int offset = (dev_priv->sarea_priv->pfCurrentPage == 1)
 	    ? dev_priv->front_offset : dev_priv->back_offset;
 	RING_LOCALS;
-	DRM_DEBUG("%s: page=%d pfCurrentPage=%d\n",
+	DRM_DEBUG("%s: pfCurrentPage=%d\n",
 		  __FUNCTION__,
-		  dev_priv->current_page, dev_priv->sarea_priv->pfCurrentPage);
+		  dev_priv->sarea_priv->pfCurrentPage);
 
 	/* Do some trivial performance monitoring...
 	 */
@@ -1449,8 +1449,8 @@ static void radeon_cp_dispatch_flip(drm_device_t * dev)
 	 * performing the swapbuffer ioctl.
 	 */
 	dev_priv->sarea_priv->last_frame++;
-	dev_priv->sarea_priv->pfCurrentPage = dev_priv->current_page =
-	    1 - dev_priv->current_page;
+	dev_priv->sarea_priv->pfCurrentPage =
+		1 - dev_priv->sarea_priv->pfCurrentPage;
 
 	BEGIN_RING(2);
 
@@ -2152,24 +2152,10 @@ static int radeon_do_init_pageflip(drm_device_t * dev)
 	ADVANCE_RING();
 
 	dev_priv->page_flipping = 1;
-	dev_priv->current_page = 0;
-	dev_priv->sarea_priv->pfCurrentPage = dev_priv->current_page;
 
-	return 0;
-}
-
-/* Called whenever a client dies, from drm_release.
- * NOTE:  Lock isn't necessarily held when this is called!
- */
-static int radeon_do_cleanup_pageflip(drm_device_t * dev)
-{
-	drm_radeon_private_t *dev_priv = dev->dev_private;
-	DRM_DEBUG("\n");
-
-	if (dev_priv->current_page != 0)
-		radeon_cp_dispatch_flip(dev);
+	if (dev_priv->sarea_priv->pfCurrentPage != 1)
+		dev_priv->sarea_priv->pfCurrentPage = 0;
 
-	dev_priv->page_flipping = 0;
 	return 0;
 }
 
@@ -3145,10 +3131,16 @@ static int radeon_cp_setparam(DRM_IOCTL_ARGS)
 		break;
 	case RADEON_SETPARAM_PCIGART_LOCATION:
 		dev_priv->pcigart_offset = sp.value;
+		dev_priv->pcigart_offset_set = 1;
 		break;
 	case RADEON_SETPARAM_NEW_MEMMAP:
 		dev_priv->new_memmap = sp.value;
 		break;
+	case RADEON_SETPARAM_PCIGART_TABLE_SIZE:
+		dev_priv->gart_info.table_size = sp.value;
+		if (dev_priv->gart_info.table_size < RADEON_PCIGART_TABLE_SIZE)
+			dev_priv->gart_info.table_size = RADEON_PCIGART_TABLE_SIZE;
+		break;
 	default:
 		DRM_DEBUG("Invalid parameter %d\n", sp.param);
 		return DRM_ERR(EINVAL);
@@ -3168,9 +3160,7 @@ void radeon_driver_preclose(drm_device_t * dev, DRMFILE filp)
 {
 	if (dev->dev_private) {
 		drm_radeon_private_t *dev_priv = dev->dev_private;
-		if (dev_priv->page_flipping) {
-			radeon_do_cleanup_pageflip(dev);
-		}
+		dev_priv->page_flipping = 0;
 		radeon_mem_release(filp, dev_priv->gart_heap);
 		radeon_mem_release(filp, dev_priv->fb_heap);
 		radeon_surfaces_release(filp, dev_priv);
@@ -3179,6 +3169,14 @@ void radeon_driver_preclose(drm_device_t * dev, DRMFILE filp)
 
 void radeon_driver_lastclose(drm_device_t * 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);
 }
 
diff --git a/drivers/char/drm/sis_drv.c b/drivers/char/drm/sis_drv.c
index 3d5b321..690e0af 100644
--- a/drivers/char/drm/sis_drv.c
+++ b/drivers/char/drm/sis_drv.c
@@ -71,7 +71,7 @@ static struct drm_driver driver = {
 	.context_dtor = NULL,
 	.dma_quiescent = sis_idle,
 	.reclaim_buffers = NULL,
-	.reclaim_buffers_locked = sis_reclaim_buffers_locked,
+	.reclaim_buffers_idlelocked = sis_reclaim_buffers_locked,
 	.lastclose = sis_lastclose,
 	.get_map_ofs = drm_core_get_map_ofs,
 	.get_reg_ofs = drm_core_get_reg_ofs,
diff --git a/drivers/char/drm/via_dma.c b/drivers/char/drm/via_dma.c
index c0539c6..13a9c5c 100644
--- a/drivers/char/drm/via_dma.c
+++ b/drivers/char/drm/via_dma.c
@@ -252,7 +252,7 @@ static int via_dma_init(DRM_IOCTL_ARGS)
 		break;
 	case VIA_DMA_INITIALIZED:
 		retcode = (dev_priv->ring.virtual_start != NULL) ?
-		    0 : DRM_ERR(EFAULT);
+			0 : DRM_ERR(EFAULT);
 		break;
 	default:
 		retcode = DRM_ERR(EINVAL);
@@ -432,56 +432,34 @@ static int via_hook_segment(drm_via_private_t * dev_priv,
 {
 	int paused, count;
 	volatile uint32_t *paused_at = dev_priv->last_pause_ptr;
+	uint32_t reader,ptr;
 
+	paused = 0;
 	via_flush_write_combine();
-	while (!*(via_get_dma(dev_priv) - 1)) ;
-	*dev_priv->last_pause_ptr = pause_addr_lo;
+	(void) *(volatile uint32_t *)(via_get_dma(dev_priv) -1);
+	*paused_at = pause_addr_lo;
 	via_flush_write_combine();
-
-	/*
-	 * The below statement is inserted to really force the flush.
-	 * Not sure it is needed.
-	 */
-
-	while (!*dev_priv->last_pause_ptr) ;
+	(void) *paused_at;
+	reader = *(dev_priv->hw_addr_ptr);
+	ptr = ((volatile char *)paused_at - dev_priv->dma_ptr) +
+		dev_priv->dma_offset + (uint32_t) dev_priv->agpAddr + 4;
 	dev_priv->last_pause_ptr = via_get_dma(dev_priv) - 1;
-	while (!*dev_priv->last_pause_ptr) ;
 
-	paused = 0;
-	count = 20;
-
-	while (!(paused = (VIA_READ(0x41c) & 0x80000000)) && count--) ;
-	if ((count <= 8) && (count >= 0)) {
-		uint32_t rgtr, ptr;
-		rgtr = *(dev_priv->hw_addr_ptr);
-		ptr = ((volatile char *)dev_priv->last_pause_ptr -
-		      dev_priv->dma_ptr) + dev_priv->dma_offset +
-		      (uint32_t) dev_priv->agpAddr + 4 - CMDBUF_ALIGNMENT_SIZE;
-		if (rgtr <= ptr) {
-			DRM_ERROR
-			    ("Command regulator\npaused at count %d, address %x, "
-			     "while current pause address is %x.\n"
-			     "Please mail this message to "
-			     "<unichrome-devel@lists.sourceforge.net>\n", count,
-			     rgtr, ptr);
-		}
+	if ((ptr - reader) <= dev_priv->dma_diff ) {
+		count = 10000000;
+		while (!(paused = (VIA_READ(0x41c) & 0x80000000)) && count--);
 	}
 
 	if (paused && !no_pci_fire) {
-		uint32_t rgtr, ptr;
-		uint32_t ptr_low;
+		reader = *(dev_priv->hw_addr_ptr);
+		if ((ptr - reader) == dev_priv->dma_diff) {
 
-		count = 1000000;
-		while ((VIA_READ(VIA_REG_STATUS) & VIA_CMD_RGTR_BUSY)
-		       && count--) ;
+			/*
+			 * There is a concern that these writes may stall the PCI bus
+			 * if the GPU is not idle. However, idling the GPU first
+			 * doesn't make a difference.
+			 */
 
-		rgtr = *(dev_priv->hw_addr_ptr);
-		ptr = ((volatile char *)paused_at - dev_priv->dma_ptr) +
-		    dev_priv->dma_offset + (uint32_t) dev_priv->agpAddr + 4;
-
-		ptr_low = (ptr > 3 * CMDBUF_ALIGNMENT_SIZE) ?
-		    ptr - 3 * CMDBUF_ALIGNMENT_SIZE : 0;
-		if (rgtr <= ptr && rgtr >= ptr_low) {
 			VIA_WRITE(VIA_REG_TRANSET, (HC_ParaType_PreCR << 16));
 			VIA_WRITE(VIA_REG_TRANSPACE, pause_addr_hi);
 			VIA_WRITE(VIA_REG_TRANSPACE, pause_addr_lo);
@@ -494,6 +472,9 @@ static int via_hook_segment(drm_via_private_t * dev_priv,
 static int via_wait_idle(drm_via_private_t * dev_priv)
 {
 	int count = 10000000;
+
+	while (!(VIA_READ(VIA_REG_STATUS) & VIA_VR_QUEUE_BUSY) && count--);
+
 	while (count-- && (VIA_READ(VIA_REG_STATUS) &
 			   (VIA_CMD_RGTR_BUSY | VIA_2D_ENG_BUSY |
 			    VIA_3D_ENG_BUSY))) ;
@@ -537,6 +518,9 @@ static void via_cmdbuf_start(drm_via_private_t * dev_priv)
 	uint32_t end_addr, end_addr_lo;
 	uint32_t command;
 	uint32_t agp_base;
+	uint32_t ptr;
+	uint32_t reader;
+	int count;
 
 	dev_priv->dma_low = 0;
 
@@ -554,7 +538,7 @@ static void via_cmdbuf_start(drm_via_private_t * dev_priv)
 			  &pause_addr_hi, &pause_addr_lo, 1) - 1;
 
 	via_flush_write_combine();
-	while (!*dev_priv->last_pause_ptr) ;
+	(void) *(volatile uint32_t *)dev_priv->last_pause_ptr;
 
 	VIA_WRITE(VIA_REG_TRANSET, (HC_ParaType_PreCR << 16));
 	VIA_WRITE(VIA_REG_TRANSPACE, command);
@@ -566,6 +550,24 @@ static void via_cmdbuf_start(drm_via_private_t * dev_priv)
 	DRM_WRITEMEMORYBARRIER();
 	VIA_WRITE(VIA_REG_TRANSPACE, command | HC_HAGPCMNT_MASK);
 	VIA_READ(VIA_REG_TRANSPACE);
+
+	dev_priv->dma_diff = 0;
+
+	count = 10000000;
+	while (!(VIA_READ(0x41c) & 0x80000000) && count--);
+
+	reader = *(dev_priv->hw_addr_ptr);
+	ptr = ((volatile char *)dev_priv->last_pause_ptr - dev_priv->dma_ptr) +
+	    dev_priv->dma_offset + (uint32_t) dev_priv->agpAddr + 4;
+
+	/*
+	 * This is the difference between where we tell the
+	 * command reader to pause and where it actually pauses.
+	 * This differs between hw implementation so we need to
+	 * detect it.
+	 */
+
+	dev_priv->dma_diff = ptr - reader;
 }
 
 static void via_pad_cache(drm_via_private_t * dev_priv, int qwords)
@@ -592,7 +594,6 @@ static void via_cmdbuf_jump(drm_via_private_t * dev_priv)
 	uint32_t pause_addr_lo, pause_addr_hi;
 	uint32_t jump_addr_lo, jump_addr_hi;
 	volatile uint32_t *last_pause_ptr;
-	uint32_t dma_low_save1, dma_low_save2;
 
 	agp_base = dev_priv->dma_offset + (uint32_t) dev_priv->agpAddr;
 	via_align_cmd(dev_priv, HC_HAGPBpID_JUMP, 0, &jump_addr_hi,
@@ -619,31 +620,11 @@ static void via_cmdbuf_jump(drm_via_private_t * dev_priv)
 		      &pause_addr_lo, 0);
 
 	*last_pause_ptr = pause_addr_lo;
-	dma_low_save1 = dev_priv->dma_low;
-
-	/*
-	 * Now, set a trap that will pause the regulator if it tries to rerun the old
-	 * command buffer. (Which may happen if via_hook_segment detecs a command regulator pause
-	 * and reissues the jump command over PCI, while the regulator has already taken the jump
-	 * and actually paused at the current buffer end).
-	 * There appears to be no other way to detect this condition, since the hw_addr_pointer
-	 * does not seem to get updated immediately when a jump occurs.
-	 */
 
-	last_pause_ptr =
-	    via_align_cmd(dev_priv, HC_HAGPBpID_PAUSE, 0, &pause_addr_hi,
-			  &pause_addr_lo, 0) - 1;
-	via_align_cmd(dev_priv, HC_HAGPBpID_PAUSE, 0, &pause_addr_hi,
-		      &pause_addr_lo, 0);
-	*last_pause_ptr = pause_addr_lo;
-
-	dma_low_save2 = dev_priv->dma_low;
-	dev_priv->dma_low = dma_low_save1;
-	via_hook_segment(dev_priv, jump_addr_hi, jump_addr_lo, 0);
-	dev_priv->dma_low = dma_low_save2;
-	via_hook_segment(dev_priv, pause_addr_hi, pause_addr_lo, 0);
+	via_hook_segment( dev_priv, jump_addr_hi, jump_addr_lo, 0);
 }
 
+
 static void via_cmdbuf_rewind(drm_via_private_t * dev_priv)
 {
 	via_cmdbuf_jump(dev_priv);
diff --git a/drivers/char/drm/via_drv.c b/drivers/char/drm/via_drv.c
index bb9dde8..2d4957a 100644
--- a/drivers/char/drm/via_drv.c
+++ b/drivers/char/drm/via_drv.c
@@ -52,7 +52,8 @@ static struct drm_driver driver = {
 	.dma_quiescent = via_driver_dma_quiescent,
 	.dri_library_name = dri_library_name,
 	.reclaim_buffers = drm_core_reclaim_buffers,
-	.reclaim_buffers_locked = via_reclaim_buffers_locked,
+	.reclaim_buffers_locked = NULL,
+	.reclaim_buffers_idlelocked = via_reclaim_buffers_locked,
 	.lastclose = via_lastclose,
 	.get_map_ofs = drm_core_get_map_ofs,
 	.get_reg_ofs = drm_core_get_reg_ofs,
diff --git a/drivers/char/drm/via_drv.h b/drivers/char/drm/via_drv.h
index 8b8778d..b46ca8e6 100644
--- a/drivers/char/drm/via_drv.h
+++ b/drivers/char/drm/via_drv.h
@@ -29,11 +29,11 @@
 
 #define DRIVER_NAME		"via"
 #define DRIVER_DESC		"VIA Unichrome / Pro"
-#define DRIVER_DATE		"20061227"
+#define DRIVER_DATE		"20070202"
 
 #define DRIVER_MAJOR		2
 #define DRIVER_MINOR		11
-#define DRIVER_PATCHLEVEL	0
+#define DRIVER_PATCHLEVEL	1
 
 #include "via_verifier.h"
 
@@ -93,6 +93,7 @@ typedef struct drm_via_private {
 	unsigned long vram_offset;
 	unsigned long agp_offset;
 	drm_via_blitq_t blit_queues[VIA_NUM_BLIT_ENGINES];
+	uint32_t dma_diff;
 } drm_via_private_t;
 
 enum via_family {
diff --git a/drivers/char/drm/via_mm.h b/drivers/char/drm/via_mm.h
deleted file mode 100644
index d57efda..0000000
--- a/drivers/char/drm/via_mm.h
+++ /dev/null
@@ -1,40 +0,0 @@
-/*
- * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
- * Copyright 2001-2003 S3 Graphics, Inc. All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sub license,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
- * VIA, S3 GRAPHICS, AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
- * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
- * DEALINGS IN THE SOFTWARE.
- */
-#ifndef _via_drm_mm_h_
-#define _via_drm_mm_h_
-
-typedef struct {
-	unsigned int context;
-	unsigned int size;
-	unsigned long offset;
-	unsigned long free;
-} drm_via_mm_t;
-
-typedef struct {
-	unsigned int size;
-	unsigned long handle;
-	void *virtual;
-} drm_via_dma_t;
-
-#endif