Merge branch 'radeon-gem-cs' into modesetting-gem

Conflicts:

	libdrm/xf86drm.c
	linux-core/Makefile.kernel
	linux-core/drmP.h
	linux-core/drm_compat.h
	linux-core/drm_drv.c
	linux-core/drm_stub.c
	linux-core/drm_vm.c
	shared-core/i915_dma.c
	shared-core/r300_cmdbuf.c
	shared-core/radeon_drv.h

11 files changed, 678 insertions, 517 deletions

diff --git a/shared-core/drm.h b/shared-core/drm.h
index dd03bf9b..b177dfe5 100644
--- a/shared-core/drm.h
+++ b/shared-core/drm.h
@@ -675,324 +675,6 @@ struct drm_set_version {
 	int drm_dd_minor;
 };
 
-
-#define DRM_FENCE_FLAG_EMIT                0x00000001
-#define DRM_FENCE_FLAG_SHAREABLE           0x00000002
-/**
- * On hardware with no interrupt events for operation completion,
- * indicates that the kernel should sleep while waiting for any blocking
- * operation to complete rather than spinning.
- *
- * Has no effect otherwise.
- */
-#define DRM_FENCE_FLAG_WAIT_LAZY           0x00000004
-#define DRM_FENCE_FLAG_NO_USER             0x00000010
-
-/* Reserved for driver use */
-#define DRM_FENCE_MASK_DRIVER              0xFF000000
-
-#define DRM_FENCE_TYPE_EXE                 0x00000001
-
-struct drm_fence_arg {
-	unsigned int handle;
-	unsigned int fence_class;
-	unsigned int type;
-	unsigned int flags;
-	unsigned int signaled;
-	unsigned int error;
-	unsigned int sequence;
-	unsigned int pad64;
-	uint64_t expand_pad[2]; /*Future expansion */
-};
-
-/* Buffer permissions, referring to how the GPU uses the buffers.
- * these translate to fence types used for the buffers.
- * Typically a texture buffer is read, A destination buffer is write and
- *  a command (batch-) buffer is exe. Can be or-ed together.
- */
-
-#define DRM_BO_FLAG_READ        (1ULL << 0)
-#define DRM_BO_FLAG_WRITE       (1ULL << 1)
-#define DRM_BO_FLAG_EXE         (1ULL << 2)
-
-/*
- * All of the bits related to access mode
- */
-#define DRM_BO_MASK_ACCESS	(DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE | DRM_BO_FLAG_EXE)
-/*
- * Status flags. Can be read to determine the actual state of a buffer.
- * Can also be set in the buffer mask before validation.
- */
-
-/*
- * Mask: Never evict this buffer. Not even with force. This type of buffer is only
- * available to root and must be manually removed before buffer manager shutdown
- * or lock.
- * Flags: Acknowledge
- */
-#define DRM_BO_FLAG_NO_EVICT    (1ULL << 4)
-
-/*
- * Mask: Require that the buffer is placed in mappable memory when validated.
- *       If not set the buffer may or may not be in mappable memory when validated.
- * Flags: If set, the buffer is in mappable memory.
- */
-#define DRM_BO_FLAG_MAPPABLE    (1ULL << 5)
-
-/* Mask: The buffer should be shareable with other processes.
- * Flags: The buffer is shareable with other processes.
- */
-#define DRM_BO_FLAG_SHAREABLE   (1ULL << 6)
-
-/* Mask: If set, place the buffer in cache-coherent memory if available.
- *       If clear, never place the buffer in cache coherent memory if validated.
- * Flags: The buffer is currently in cache-coherent memory.
- */
-#define DRM_BO_FLAG_CACHED      (1ULL << 7)
-
-/* Mask: Make sure that every time this buffer is validated,
- *       it ends up on the same location provided that the memory mask is the same.
- *       The buffer will also not be evicted when claiming space for
- *       other buffers. Basically a pinned buffer but it may be thrown out as
- *       part of buffer manager shutdown or locking.
- * Flags: Acknowledge.
- */
-#define DRM_BO_FLAG_NO_MOVE     (1ULL << 8)
-
-/* Mask: Make sure the buffer is in cached memory when mapped.  In conjunction
- * with DRM_BO_FLAG_CACHED it also allows the buffer to be bound into the GART
- * with unsnooped PTEs instead of snooped, by using chipset-specific cache
- * flushing at bind time.  A better name might be DRM_BO_FLAG_TT_UNSNOOPED,
- * as the eviction to local memory (TTM unbind) on map is just a side effect
- * to prevent aggressive cache prefetch from the GPU disturbing the cache
- * management that the DRM is doing.
- *
- * Flags: Acknowledge.
- * Buffers allocated with this flag should not be used for suballocators
- * This type may have issues on CPUs with over-aggressive caching
- * http://marc.info/?l=linux-kernel&m=102376926732464&w=2
- */
-#define DRM_BO_FLAG_CACHED_MAPPED    (1ULL << 19)
-
-
-/* Mask: Force DRM_BO_FLAG_CACHED flag strictly also if it is set.
- * Flags: Acknowledge.
- */
-#define DRM_BO_FLAG_FORCE_CACHING  (1ULL << 13)
-
-/*
- * Mask: Force DRM_BO_FLAG_MAPPABLE flag strictly also if it is clear.
- * Flags: Acknowledge.
- */
-#define DRM_BO_FLAG_FORCE_MAPPABLE (1ULL << 14)
-#define DRM_BO_FLAG_TILE           (1ULL << 15)
-
-/*
- * Memory type flags that can be or'ed together in the mask, but only
- * one appears in flags.
- */
-
-/* System memory */
-#define DRM_BO_FLAG_MEM_LOCAL  (1ULL << 24)
-/* Translation table memory */
-#define DRM_BO_FLAG_MEM_TT     (1ULL << 25)
-/* Vram memory */
-#define DRM_BO_FLAG_MEM_VRAM   (1ULL << 26)
-/* Up to the driver to define. */
-#define DRM_BO_FLAG_MEM_PRIV0  (1ULL << 27)
-#define DRM_BO_FLAG_MEM_PRIV1  (1ULL << 28)
-#define DRM_BO_FLAG_MEM_PRIV2  (1ULL << 29)
-#define DRM_BO_FLAG_MEM_PRIV3  (1ULL << 30)
-#define DRM_BO_FLAG_MEM_PRIV4  (1ULL << 31)
-/* We can add more of these now with a 64-bit flag type */
-
-/*
- * This is a mask covering all of the memory type flags; easier to just
- * use a single constant than a bunch of | values. It covers
- * DRM_BO_FLAG_MEM_LOCAL through DRM_BO_FLAG_MEM_PRIV4
- */
-#define DRM_BO_MASK_MEM         0x00000000FF000000ULL
-/*
- * This adds all of the CPU-mapping options in with the memory
- * type to label all bits which change how the page gets mapped
- */
-#define DRM_BO_MASK_MEMTYPE     (DRM_BO_MASK_MEM | \
-				 DRM_BO_FLAG_CACHED_MAPPED | \
-				 DRM_BO_FLAG_CACHED | \
-				 DRM_BO_FLAG_MAPPABLE)
-				 
-/* Driver-private flags */
-#define DRM_BO_MASK_DRIVER      0xFFFF000000000000ULL
-
-/*
- * Don't block on validate and map. Instead, return EBUSY.
- */
-#define DRM_BO_HINT_DONT_BLOCK  0x00000002
-/*
- * Don't place this buffer on the unfenced list. This means
- * that the buffer will not end up having a fence associated
- * with it as a result of this operation
- */
-#define DRM_BO_HINT_DONT_FENCE  0x00000004
-/**
- * On hardware with no interrupt events for operation completion,
- * indicates that the kernel should sleep while waiting for any blocking
- * operation to complete rather than spinning.
- *
- * Has no effect otherwise.
- */
-#define DRM_BO_HINT_WAIT_LAZY   0x00000008
-/*
- * The client has compute relocations refering to this buffer using the
- * offset in the presumed_offset field. If that offset ends up matching
- * where this buffer lands, the kernel is free to skip executing those
- * relocations
- */
-#define DRM_BO_HINT_PRESUMED_OFFSET 0x00000010
-
-#define DRM_BO_INIT_MAGIC 0xfe769812
-#define DRM_BO_INIT_MAJOR 1
-#define DRM_BO_INIT_MINOR 0
-#define DRM_BO_INIT_PATCH 0
-
-
-struct drm_bo_info_req {
-	uint64_t mask;
-	uint64_t flags;
-	unsigned int handle;
-	unsigned int hint;
-	unsigned int fence_class;
-	unsigned int desired_tile_stride;
-	unsigned int tile_info;
-	unsigned int pad64;
-	uint64_t presumed_offset;
-};
-
-struct drm_bo_create_req {
-	uint64_t flags;
-	uint64_t size;
-	uint64_t buffer_start;
-	unsigned int hint;
-	unsigned int page_alignment;
-};
-
-
-/*
- * Reply flags
- */
-
-#define DRM_BO_REP_BUSY 0x00000001
-
-struct drm_bo_info_rep {
-	uint64_t flags;
-	uint64_t proposed_flags;
-	uint64_t size;
-	uint64_t offset;
-	uint64_t arg_handle;
-	uint64_t buffer_start;
-	unsigned int handle;
-	unsigned int fence_flags;
-	unsigned int rep_flags;
-	unsigned int page_alignment;
-	unsigned int desired_tile_stride;
-	unsigned int hw_tile_stride;
-	unsigned int tile_info;
-	unsigned int pad64;
-	uint64_t expand_pad[4]; /*Future expansion */
-};
-
-struct drm_bo_arg_rep {
-	struct drm_bo_info_rep bo_info;
-	int ret;
-	unsigned int pad64;
-};
-
-struct drm_bo_create_arg {
-	union {
-		struct drm_bo_create_req req;
-		struct drm_bo_info_rep rep;
-	} d;
-};
-
-struct drm_bo_handle_arg {
-	unsigned int handle;
-};
-
-struct drm_bo_reference_info_arg {
-	union {
-		struct drm_bo_handle_arg req;
-		struct drm_bo_info_rep rep;
-	} d;
-};
-
-struct drm_bo_map_wait_idle_arg {
-	union {
-		struct drm_bo_info_req req;
-		struct drm_bo_info_rep rep;
-	} d;
-};
-
-struct drm_bo_op_req {
-	enum {
-		drm_bo_validate,
-		drm_bo_fence,
-		drm_bo_ref_fence,
-	} op;
-	unsigned int arg_handle;
-	struct drm_bo_info_req bo_req;
-};
-
-
-struct drm_bo_op_arg {
-	uint64_t next;
-	union {
-		struct drm_bo_op_req req;
-		struct drm_bo_arg_rep rep;
-	} d;
-	int handled;
-	unsigned int pad64;
-};
-
-
-#define DRM_BO_MEM_LOCAL 0
-#define DRM_BO_MEM_TT 1
-#define DRM_BO_MEM_VRAM 2
-#define DRM_BO_MEM_PRIV0 3
-#define DRM_BO_MEM_PRIV1 4
-#define DRM_BO_MEM_PRIV2 5
-#define DRM_BO_MEM_PRIV3 6
-#define DRM_BO_MEM_PRIV4 7
-
-#define DRM_BO_MEM_TYPES 8 /* For now. */
-
-#define DRM_BO_LOCK_UNLOCK_BM       (1 << 0)
-#define DRM_BO_LOCK_IGNORE_NO_EVICT (1 << 1)
-
-struct drm_bo_version_arg {
-	uint32_t major;
-	uint32_t minor;
-	uint32_t patchlevel;
-};
-
-struct drm_mm_type_arg {
-	unsigned int mem_type;
-	unsigned int lock_flags;
-};
-
-struct drm_mm_init_arg {
-	unsigned int magic;
-	unsigned int major;
-	unsigned int minor;
-	unsigned int mem_type;
-	uint64_t p_offset;
-	uint64_t p_size;
-};
-
-struct drm_mm_info_arg {
-	unsigned int mem_type;
-	uint64_t p_size;
-};
-
 struct drm_gem_close {
 	/** Handle of the object to be closed. */
 	uint32_t handle;
@@ -1337,31 +1019,6 @@ struct drm_mode_crtc_lut {
 
 #define DRM_IOCTL_UPDATE_DRAW           DRM_IOW(0x3f, struct drm_update_draw)
 
-#define DRM_IOCTL_MM_INIT               DRM_IOWR(0xc0, struct drm_mm_init_arg)
-#define DRM_IOCTL_MM_TAKEDOWN           DRM_IOWR(0xc1, struct drm_mm_type_arg)
-#define DRM_IOCTL_MM_LOCK               DRM_IOWR(0xc2, struct drm_mm_type_arg)
-#define DRM_IOCTL_MM_UNLOCK             DRM_IOWR(0xc3, struct drm_mm_type_arg)
-
-#define DRM_IOCTL_FENCE_CREATE          DRM_IOWR(0xc4, struct drm_fence_arg)
-#define DRM_IOCTL_FENCE_REFERENCE       DRM_IOWR(0xc6, struct drm_fence_arg)
-#define DRM_IOCTL_FENCE_UNREFERENCE     DRM_IOWR(0xc7, struct drm_fence_arg)
-#define DRM_IOCTL_FENCE_SIGNALED        DRM_IOWR(0xc8, struct drm_fence_arg)
-#define DRM_IOCTL_FENCE_FLUSH           DRM_IOWR(0xc9, struct drm_fence_arg)
-#define DRM_IOCTL_FENCE_WAIT            DRM_IOWR(0xca, struct drm_fence_arg)
-#define DRM_IOCTL_FENCE_EMIT            DRM_IOWR(0xcb, struct drm_fence_arg)
-#define DRM_IOCTL_FENCE_BUFFERS         DRM_IOWR(0xcc, struct drm_fence_arg)
-
-#define DRM_IOCTL_BO_CREATE             DRM_IOWR(0xcd, struct drm_bo_create_arg)
-#define DRM_IOCTL_BO_MAP                DRM_IOWR(0xcf, struct drm_bo_map_wait_idle_arg)
-#define DRM_IOCTL_BO_UNMAP              DRM_IOWR(0xd0, struct drm_bo_handle_arg)
-#define DRM_IOCTL_BO_REFERENCE          DRM_IOWR(0xd1, struct drm_bo_reference_info_arg)
-#define DRM_IOCTL_BO_UNREFERENCE        DRM_IOWR(0xd2, struct drm_bo_handle_arg)
-#define DRM_IOCTL_BO_SETSTATUS          DRM_IOWR(0xd3, struct drm_bo_map_wait_idle_arg)
-#define DRM_IOCTL_BO_INFO               DRM_IOWR(0xd4, struct drm_bo_reference_info_arg)
-#define DRM_IOCTL_BO_WAIT_IDLE          DRM_IOWR(0xd5, struct drm_bo_map_wait_idle_arg)
-#define DRM_IOCTL_BO_VERSION          DRM_IOR(0xd6, struct drm_bo_version_arg)
-#define DRM_IOCTL_MM_INFO               DRM_IOWR(0xd7, struct drm_mm_info_arg)
-
 #define DRM_IOCTL_MODE_GETRESOURCES     DRM_IOWR(0xA0, struct drm_mode_card_res)
 #define DRM_IOCTL_MODE_GETCRTC          DRM_IOWR(0xA1, struct drm_mode_crtc)
 #define DRM_IOCTL_MODE_GETCONNECTOR        DRM_IOWR(0xA2, struct drm_mode_get_connector)
@@ -1439,11 +1096,6 @@ typedef struct drm_agp_binding drm_agp_binding_t;
 typedef struct drm_agp_info drm_agp_info_t;
 typedef struct drm_scatter_gather drm_scatter_gather_t;
 typedef struct drm_set_version drm_set_version_t;
-
-typedef struct drm_fence_arg drm_fence_arg_t;
-typedef struct drm_mm_type_arg drm_mm_type_arg_t;
-typedef struct drm_mm_init_arg drm_mm_init_arg_t;
-typedef enum drm_bo_type drm_bo_type_t;
 #endif
 
 #endif
diff --git a/shared-core/i915_dma.c b/shared-core/i915_dma.c
index 1fdc5e17..7f025e62 100644
--- a/shared-core/i915_dma.c
+++ b/shared-core/i915_dma.c
@@ -135,7 +135,7 @@ int i915_dma_cleanup(struct drm_device * dev)
 	return 0;
 }
 
-#if defined(I915_HAVE_BUFFER) && defined(DRI2)
+#if defined(DRI2)
 #define DRI2_SAREA_BLOCK_TYPE(b) ((b) >> 16)
 #define DRI2_SAREA_BLOCK_SIZE(b) ((b) & 0xffff)
 #define DRI2_SAREA_BLOCK_NEXT(p)				\
@@ -213,12 +213,7 @@ static int i915_initialize(struct drm_device * dev,
 		}
 	}
 
-#ifdef I915_HAVE_BUFFER
-	if (!drm_core_check_feature(dev, DRIVER_MODESET)) {
-		dev_priv->max_validate_buffers = I915_MAX_VALIDATE_BUFFERS;
-	}
-#endif
-
+	
 	if (init->ring_size != 0) {
 		dev_priv->ring.Size = init->ring_size;
 		dev_priv->ring.tail_mask = dev_priv->ring.Size - 1;
@@ -253,10 +248,25 @@ static int i915_initialize(struct drm_device * dev,
 	 */
 	dev_priv->vblank_pipe = DRM_I915_VBLANK_PIPE_A;
 
-#ifdef I915_HAVE_BUFFER
-	if (!drm_core_check_feature(dev, DRIVER_MODESET)) {
-		mutex_init(&dev_priv->cmdbuf_mutex);
+	/* Program Hardware Status Page */
+	if (!I915_NEED_GFX_HWS(dev)) {
+		dev_priv->status_page_dmah =
+			drm_pci_alloc(dev, PAGE_SIZE, PAGE_SIZE, 0xffffffff);
+
+		if (!dev_priv->status_page_dmah) {
+			i915_dma_cleanup(dev);
+			DRM_ERROR("Can not allocate hardware status page\n");
+			return -ENOMEM;
+		}
+		dev_priv->hws_vaddr = dev_priv->status_page_dmah->vaddr;
+		dev_priv->dma_status_page = dev_priv->status_page_dmah->busaddr;
+
+		memset(dev_priv->hws_vaddr, 0, PAGE_SIZE);
+
+		I915_WRITE(0x02080, dev_priv->dma_status_page);
 	}
+	DRM_DEBUG("Enabled hardware status page\n");
+
 #ifdef DRI2
 	if (init->func == I915_INIT_DMA2) {
 		int ret = setup_dri2_sarea(dev, file_priv, init);
@@ -267,7 +277,6 @@ static int i915_initialize(struct drm_device * dev,
 		}
 	}
 #endif /* DRI2 */
-#endif /* I915_HAVE_BUFFER */
 
 	return 0;
 }
@@ -533,9 +542,6 @@ int i915_emit_mi_flush(struct drm_device *dev, uint32_t flush)
 static int i915_dispatch_cmdbuffer(struct drm_device * dev,
 				   struct drm_i915_cmdbuffer * cmd)
 {
-#ifdef I915_HAVE_FENCE
-	struct drm_i915_private *dev_priv = dev->dev_private;
-#endif
 	int nbox = cmd->num_cliprects;
 	int i = 0, count, ret;
 
@@ -562,10 +568,6 @@ static int i915_dispatch_cmdbuffer(struct drm_device * dev,
 	}
 
 	i915_emit_breadcrumb(dev);
-#ifdef I915_HAVE_FENCE
-	if (unlikely((dev_priv->counter & 0xFF) == 0))
-		drm_fence_flush_old(dev, 0, dev_priv->counter);
-#endif
 	return 0;
 }
 
@@ -616,10 +618,6 @@ int i915_dispatch_batchbuffer(struct drm_device * dev,
 	}
 
 	i915_emit_breadcrumb(dev);
-#ifdef I915_HAVE_FENCE
-	if (unlikely((dev_priv->counter & 0xFF) == 0))
-		drm_fence_flush_old(dev, 0, dev_priv->counter);
-#endif
 	return 0;
 }
 
@@ -678,7 +676,6 @@ static void i915_do_dispatch_flip(struct drm_device * dev, int plane, int sync)
 
 void i915_dispatch_flip(struct drm_device * dev, int planes, int sync)
 {
-	struct drm_i915_private *dev_priv = dev->dev_private;
 	struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv;
 	int i;
 
@@ -692,10 +689,6 @@ void i915_dispatch_flip(struct drm_device * dev, int planes, int sync)
 			i915_do_dispatch_flip(dev, i, sync);
 
 	i915_emit_breadcrumb(dev);
-#ifdef I915_HAVE_FENCE
-	if (unlikely(!sync && ((dev_priv->counter & 0xFF) == 0)))
-		drm_fence_flush_old(dev, 0, dev_priv->counter);
-#endif
 }
 
 int i915_quiescent(struct drm_device *dev)
@@ -1049,9 +1042,6 @@ struct drm_ioctl_desc i915_ioctls[] = {
 	DRM_IOCTL_DEF(DRM_I915_VBLANK_SWAP, i915_vblank_swap, DRM_AUTH),
 	DRM_IOCTL_DEF(DRM_I915_MMIO, i915_mmio, DRM_AUTH),
 	DRM_IOCTL_DEF(DRM_I915_HWS_ADDR, i915_set_status_page, DRM_AUTH),
-#ifdef I915_HAVE_BUFFER
-	DRM_IOCTL_DEF(DRM_I915_EXECBUFFER, i915_execbuffer, DRM_AUTH),
-#endif
 	DRM_IOCTL_DEF(DRM_I915_GEM_INIT, i915_gem_init_ioctl, DRM_AUTH),
 	DRM_IOCTL_DEF(DRM_I915_GEM_EXECBUFFER, i915_gem_execbuffer, DRM_AUTH),
 	DRM_IOCTL_DEF(DRM_I915_GEM_PIN, i915_gem_pin_ioctl, DRM_AUTH|DRM_ROOT_ONLY),
diff --git a/shared-core/i915_drm.h b/shared-core/i915_drm.h
index 53087b57..ab13cd4a 100644
--- a/shared-core/i915_drm.h
+++ b/shared-core/i915_drm.h
@@ -380,58 +380,6 @@ typedef struct drm_i915_hws_addr {
 	uint64_t addr;
 } drm_i915_hws_addr_t;
 
-/*
- * Relocation header is 4 uint32_ts
- * 0 - 32 bit reloc count
- * 1 - 32-bit relocation type
- * 2-3 - 64-bit user buffer handle ptr for another list of relocs.
- */
-#define I915_RELOC_HEADER 4
-
-/*
- * type 0 relocation has 4-uint32_t stride
- * 0 - offset into buffer
- * 1 - delta to add in
- * 2 - buffer handle
- * 3 - reserved (for optimisations later).
- */
-/*
- * type 1 relocation has 4-uint32_t stride.
- * Hangs off the first item in the op list.
- * Performed after all valiations are done.
- * Try to group relocs into the same relocatee together for
- * performance reasons.
- * 0 - offset into buffer
- * 1 - delta to add in
- * 2 - buffer index in op list.
- * 3 - relocatee index in op list.
- */
-#define I915_RELOC_TYPE_0 0
-#define I915_RELOC0_STRIDE 4
-#define I915_RELOC_TYPE_1 1
-#define I915_RELOC1_STRIDE 4
-
-
-struct drm_i915_op_arg {
-	uint64_t next;
-	uint64_t reloc_ptr;
-	int handled;
-	unsigned int pad64;
-	union {
-		struct drm_bo_op_req req;
-		struct drm_bo_arg_rep rep;
-	} d;
-
-};
-
-struct drm_i915_execbuffer {
-	uint64_t ops_list;
-	uint32_t num_buffers;
-	struct drm_i915_batchbuffer batch;
-	drm_context_t context; /* for lockless use in the future */
-	struct drm_fence_arg fence_arg;
-};
-
 struct drm_i915_gem_init {
 	/**
 	 * Beginning offset in the GTT to be managed by the DRM memory
diff --git a/shared-core/i915_init.c b/shared-core/i915_init.c
index b21351c0..7b88cc8d 100644
--- a/shared-core/i915_init.c
+++ b/shared-core/i915_init.c
@@ -437,7 +437,7 @@ void i915_master_destroy(struct drm_device *dev, struct drm_master *master)
 		return;
 
 	if (master_priv->sarea)
-		drm_rmmap(dev, master_priv->sarea);
+		drm_rmmap_locked(dev, master_priv->sarea);
 		
 	drm_free(master_priv, sizeof(*master_priv), DRM_MEM_DRIVER);
 
diff --git a/shared-core/r300_cmdbuf.c b/shared-core/r300_cmdbuf.c
index 05a76ef1..53961fc7 100644
--- a/shared-core/r300_cmdbuf.c
+++ b/shared-core/r300_cmdbuf.c
@@ -166,8 +166,6 @@ void r300_init_reg_flags(struct drm_device *dev)
 		for(i=((reg)>>2);i<((reg)>>2)+(count);i++)\
 			r300_reg_flags[i]|=(mark);
 
-#define MARK_SAFE		1
-#define MARK_CHECK_OFFSET	2
 
 #define ADD_RANGE(reg, count)	ADD_RANGE_MARK(reg, count, MARK_SAFE)
 
@@ -247,6 +245,11 @@ void r300_init_reg_flags(struct drm_device *dev)
 	ADD_RANGE(R300_VAP_INPUT_ROUTE_0_0, 8);
 	ADD_RANGE(R300_VAP_INPUT_ROUTE_1_0, 8);
 
+	ADD_RANGE(R500_SU_REG_DEST, 1);
+	if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RV410) {
+		ADD_RANGE(R300_DST_PIPE_CONFIG, 1);
+	}
+
 	if ((dev_priv->flags & RADEON_FAMILY_MASK) >= CHIP_RV515) {
 		ADD_RANGE(R500_VAP_INDEX_OFFSET, 1);
 		ADD_RANGE(R500_US_CONFIG, 2);
@@ -257,6 +260,7 @@ void r300_init_reg_flags(struct drm_device *dev)
 		ADD_RANGE(R500_RB3D_COLOR_CLEAR_VALUE_AR, 2);
 		ADD_RANGE(R500_RB3D_CONSTANT_COLOR_AR, 2);
 		ADD_RANGE(R500_ZB_FIFO_SIZE, 2);
+		ADD_RANGE(R500_GA_US_VECTOR_INDEX, 2);
 	} else {
 		ADD_RANGE(R300_PFS_CNTL_0, 3);
 		ADD_RANGE(R300_PFS_NODE_0, 4);
@@ -269,9 +273,39 @@ void r300_init_reg_flags(struct drm_device *dev)
 		ADD_RANGE(R300_RS_ROUTE_0, 8);
 
 	}
+
+	/* add 2d blit engine registers for DDX */
+	ADD_RANGE(RADEON_SRC_Y_X, 3); /* 1434, 1438, 143c, 
+					 SRC_Y_X, DST_Y_X, DST_HEIGHT_WIDTH
+				       */
+	ADD_RANGE(RADEON_DP_GUI_MASTER_CNTL, 1); /* 146c */
+	ADD_RANGE(RADEON_DP_BRUSH_BKGD_CLR, 2); /* 1478, 147c */
+	ADD_RANGE(RADEON_DP_SRC_FRGD_CLR, 2); /* 15d8, 15dc */
+	ADD_RANGE(RADEON_DP_CNTL, 1); /* 16c0 */
+	ADD_RANGE(RADEON_DP_WRITE_MASK, 1); /* 16cc */
+	ADD_RANGE(RADEON_DEFAULT_SC_BOTTOM_RIGHT, 1); /* 16e8 */
+
+	ADD_RANGE(RADEON_DSTCACHE_CTLSTAT, 1);
+	ADD_RANGE(RADEON_WAIT_UNTIL, 1);
+
+	ADD_RANGE_MARK(RADEON_DST_OFFSET, 1, MARK_CHECK_OFFSET);
+	ADD_RANGE_MARK(RADEON_SRC_OFFSET, 1, MARK_CHECK_OFFSET);
+
+	ADD_RANGE_MARK(RADEON_DST_PITCH_OFFSET, 1, MARK_CHECK_OFFSET);
+	ADD_RANGE_MARK(RADEON_SRC_PITCH_OFFSET, 1, MARK_CHECK_OFFSET);
+
+	/* TODO SCISSOR */
+	ADD_RANGE_MARK(R300_SC_SCISSOR0, 2, MARK_CHECK_SCISSOR);
+
+	ADD_RANGE(R300_SC_CLIP_0_A, 2);
+	ADD_RANGE(R300_SC_CLIP_RULE, 1);
+	ADD_RANGE(R300_SC_SCREENDOOR, 1);
+
+	ADD_RANGE(R300_VAP_PVS_CODE_CNTL_0, 4);
+	ADD_RANGE(R300_VAP_PVS_VECTOR_INDX_REG, 2);
 }
 
-static __inline__ int r300_check_range(unsigned reg, int count)
+int r300_check_range(unsigned reg, int count)
 {
 	int i;
 	if (reg & ~0xffff)
@@ -282,6 +316,13 @@ static __inline__ int r300_check_range(unsigned reg, int count)
 	return 0;
 }
 
+int r300_get_reg_flags(unsigned reg)
+{
+	if (reg & ~0xffff)
+		return -1;
+	return r300_reg_flags[(reg >> 2)];
+}
+
 static __inline__ int r300_emit_carefully_checked_packet0(drm_radeon_private_t *
 							  dev_priv,
 							  drm_radeon_kcmd_buffer_t
diff --git a/shared-core/r300_reg.h b/shared-core/r300_reg.h
index d35dd39d..9e9cb526 100644
--- a/shared-core/r300_reg.h
+++ b/shared-core/r300_reg.h
@@ -129,15 +129,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 /* END: Wild guesses */
 
 #define R300_SE_VTE_CNTL                  0x20b0
-#	define     R300_VPORT_X_SCALE_ENA                0x00000001
-#	define     R300_VPORT_X_OFFSET_ENA               0x00000002
-#	define     R300_VPORT_Y_SCALE_ENA                0x00000004
-#	define     R300_VPORT_Y_OFFSET_ENA               0x00000008
-#	define     R300_VPORT_Z_SCALE_ENA                0x00000010
-#	define     R300_VPORT_Z_OFFSET_ENA               0x00000020
-#	define     R300_VTX_XY_FMT                       0x00000100
-#	define     R300_VTX_Z_FMT                        0x00000200
-#	define     R300_VTX_W0_FMT                       0x00000400
 #	define     R300_VTX_W0_NORMALIZE                 0x00000800
 #	define     R300_VTX_ST_DENORMALIZED              0x00001000
 
@@ -493,7 +484,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #	define R300_OFIFO_HIGHWATER_SHIFT	22	/* two bits only */
 #	define R300_CUBE_FIFO_HIGHWATER_COL_SHIFT	24
 
-#define R300_GB_SELECT	0x401C
+
 #	define R300_GB_FOG_SELECT_C0A		0
 #	define R300_GB_FOG_SELECT_C1A		1
 #	define R300_GB_FOG_SELECT_C2A		2
@@ -955,7 +946,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 /* 32 bit chroma key */
 #define R300_TX_CHROMA_KEY_0                      0x4580
 /* ff00ff00 == { 0, 1.0, 0, 1.0 } */
-#define R300_TX_BORDER_COLOR_0              0x45C0
 
 /* END: Texture specification */
 
@@ -1340,7 +1330,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 /* gap */
 
-#define R300_RB3D_COLOROFFSET0              0x4E28
 #       define R300_COLOROFFSET_MASK             0xFFFFFFF0 /* GUESS */
 #define R300_RB3D_COLOROFFSET1              0x4E2C /* GUESS */
 #define R300_RB3D_COLOROFFSET2              0x4E30 /* GUESS */
@@ -1352,7 +1341,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
  * Bit 17: 4x2 tiles
  * Bit 18: Extremely weird tile like, but some pixels duplicated?
  */
-#define R300_RB3D_COLORPITCH0               0x4E38
 #       define R300_COLORPITCH_MASK              0x00001FF8 /* GUESS */
 #       define R300_COLOR_TILE_ENABLE            (1 << 16) /* GUESS */
 #       define R300_COLOR_MICROTILE_ENABLE       (1 << 17) /* GUESS */
diff --git a/shared-core/radeon_cp.c b/shared-core/radeon_cp.c
index 6fbcfeab..04f4b1f8 100644
--- a/shared-core/radeon_cp.c
+++ b/shared-core/radeon_cp.c
@@ -190,7 +190,7 @@ void radeon_pll_errata_after_data(struct drm_radeon_private *dev_priv)
 	}
 }
 
-int RADEON_READ_PLL(struct drm_radeon_private *dev_priv, int addr)
+u32 RADEON_READ_PLL(struct drm_radeon_private *dev_priv, int addr)
 {
 	uint32_t data;
 
@@ -661,8 +661,8 @@ static void radeon_cp_init_ring_buffer(struct drm_device * dev,
 					 ((dev_priv->gart_vm_start - 1) & 0xffff0000)
 					 | (dev_priv->fb_location >> 16));
 	
-	if (dev_priv->mm.ring) {
-		ring_start = dev_priv->mm.ring->offset +
+	if (dev_priv->mm.ring.bo) {
+		ring_start = dev_priv->mm.ring.bo->offset +
 			dev_priv->gart_vm_start;
 	} else
 #if __OS_HAS_AGP
@@ -695,9 +695,9 @@ static void radeon_cp_init_ring_buffer(struct drm_device * dev,
 	dev_priv->ring.tail = cur_read_ptr;
 
 
-	if (dev_priv->mm.ring_read_ptr) {
+	if (dev_priv->mm.ring_read.bo) {
 		RADEON_WRITE(RADEON_CP_RB_RPTR_ADDR,
-			     dev_priv->mm.ring_read_ptr->offset +
+			     dev_priv->mm.ring_read.bo->offset +
 			     dev_priv->gart_vm_start);
 	} else
 #if __OS_HAS_AGP
@@ -748,9 +748,9 @@ static void radeon_cp_init_ring_buffer(struct drm_device * dev,
 	RADEON_WRITE(RADEON_SCRATCH_ADDR, RADEON_READ(RADEON_CP_RB_RPTR_ADDR)
 		     + RADEON_SCRATCH_REG_OFFSET);
 
-	if (dev_priv->mm.ring_read_ptr)
+	if (dev_priv->mm.ring_read.bo)
 		dev_priv->scratch = ((__volatile__ u32 *)
-				     dev_priv->mm.ring_read_ptr_map.virtual +
+				     dev_priv->mm.ring_read.kmap.virtual +
 				     (RADEON_SCRATCH_REG_OFFSET / sizeof(u32)));
 	else
 		dev_priv->scratch = ((__volatile__ u32 *)
@@ -775,12 +775,18 @@ static void radeon_cp_init_ring_buffer(struct drm_device * dev,
 	radeon_do_wait_for_idle(dev_priv);
 
 	/* Sync everything up */
+	if (dev_priv->chip_family > CHIP_RV280) {
 	RADEON_WRITE(RADEON_ISYNC_CNTL,
 		     (RADEON_ISYNC_ANY2D_IDLE3D |
 		      RADEON_ISYNC_ANY3D_IDLE2D |
 		      RADEON_ISYNC_WAIT_IDLEGUI |
 		      RADEON_ISYNC_CPSCRATCH_IDLEGUI));
-
+	} else {
+	RADEON_WRITE(RADEON_ISYNC_CNTL,
+		     (RADEON_ISYNC_ANY2D_IDLE3D |
+		      RADEON_ISYNC_ANY3D_IDLE2D |
+		      RADEON_ISYNC_WAIT_IDLEGUI));
+	}
 }
 
 static void radeon_test_writeback(drm_radeon_private_t * dev_priv)
@@ -788,8 +794,8 @@ static void radeon_test_writeback(drm_radeon_private_t * dev_priv)
 	u32 tmp;
 	void *ring_read_ptr;
 
-	if (dev_priv->mm.ring_read_ptr)
-		ring_read_ptr = dev_priv->mm.ring_read_ptr_map.virtual;
+	if (dev_priv->mm.ring_read.bo)
+		ring_read_ptr = dev_priv->mm.ring_read.kmap.virtual;
 	else
 		ring_read_ptr = dev_priv->ring_rptr->handle;
 
@@ -1353,8 +1359,7 @@ static int radeon_do_cleanup_cp(struct drm_device * dev)
 		if (dev_priv->gart_info.bus_addr) {
 			/* Turn off PCI GART */
 			radeon_set_pcigart(dev_priv, 0);
-			if (!drm_ati_pcigart_cleanup(dev, &dev_priv->gart_info))
-				DRM_ERROR("failed to cleanup PCI GART!\n");
+			drm_ati_pcigart_cleanup(dev, &dev_priv->gart_info);
 		}
 
 		if (dev_priv->gart_info.gart_table_location == DRM_ATI_GART_FB)
@@ -1362,6 +1367,7 @@ static int radeon_do_cleanup_cp(struct drm_device * dev)
 			if (dev_priv->pcigart_offset_set == 1) {
 				drm_core_ioremapfree(&dev_priv->gart_info.mapping, dev);
 				dev_priv->gart_info.addr = NULL;
+				dev_priv->pcigart_offset_set = 0;
 			}
 		}
 	}
@@ -1553,8 +1559,10 @@ void radeon_do_release(struct drm_device * dev)
 		radeon_mem_takedown(&(dev_priv->gart_heap));
 		radeon_mem_takedown(&(dev_priv->fb_heap));
 
-
-		radeon_gem_mm_fini(dev);
+		if (dev_priv->user_mm_enable) {
+			radeon_gem_mm_fini(dev);
+			dev_priv->user_mm_enable = false;
+		}
 
 		/* deallocate kernel resources */
 		radeon_do_cleanup_cp(dev);
@@ -2270,6 +2278,7 @@ static void radeon_set_dynamic_clock(struct drm_device *dev, int mode)
 int radeon_modeset_cp_init(struct drm_device *dev)
 {
 	drm_radeon_private_t *dev_priv = dev->dev_private;
+	uint32_t tmp;
 
 	/* allocate a ring and ring rptr bits from GART space */
 	/* these are allocated in GEM files */
@@ -2278,8 +2287,8 @@ int radeon_modeset_cp_init(struct drm_device *dev)
 	dev_priv->ring.size = RADEON_DEFAULT_RING_SIZE;
 	dev_priv->cp_mode = RADEON_CSQ_PRIBM_INDBM;
 
-	dev_priv->ring.start = (u32 *)(void *)(unsigned long)dev_priv->mm.ring_map.virtual;
-	dev_priv->ring.end = (u32 *)(void *)(unsigned long)dev_priv->mm.ring_map.virtual +
+	dev_priv->ring.start = (u32 *)(void *)(unsigned long)dev_priv->mm.ring.kmap.virtual;
+	dev_priv->ring.end = (u32 *)(void *)(unsigned long)dev_priv->mm.ring.kmap.virtual +
 		dev_priv->ring.size / sizeof(u32);
 	dev_priv->ring.size_l2qw = drm_order(dev_priv->ring.size / 8);
 	dev_priv->ring.rptr_update = 4096;
@@ -2289,14 +2298,21 @@ int radeon_modeset_cp_init(struct drm_device *dev)
 	dev_priv->ring.tail_mask = (dev_priv->ring.size / sizeof(u32)) - 1;
 	dev_priv->ring.high_mark = RADEON_RING_HIGH_MARK;
 
-	dev_priv->new_memmap = 1;
+	dev_priv->new_memmap = true;
 
+	r300_init_reg_flags(dev);
+		
 	radeon_cp_load_microcode(dev_priv);
 	
-	DRM_DEBUG("ring offset is %x %x\n", dev_priv->mm.ring->offset, dev_priv->mm.ring_read_ptr->offset);
+	DRM_DEBUG("ring offset is %x %x\n", dev_priv->mm.ring.bo->offset, dev_priv->mm.ring_read.bo->offset);
 
 	radeon_cp_init_ring_buffer(dev, dev_priv);
 
+	/* need to enable BUS mastering in Buscntl */
+	tmp = RADEON_READ(RADEON_BUS_CNTL);
+	tmp &= ~RADEON_BUS_MASTER_DIS;
+	RADEON_WRITE(RADEON_BUS_CNTL, tmp);
+
 	radeon_do_engine_reset(dev);
 	radeon_test_writeback(dev_priv);
 
@@ -2367,8 +2383,8 @@ int radeon_modeset_preinit(struct drm_device *dev)
 
 	if (dev_priv->is_atom_bios) {
 		dev_priv->mode_info.atom_context = atom_parse(&card, dev_priv->bios);
-		radeon_get_clock_info(dev);
 	}
+	radeon_get_clock_info(dev);
 	return 0;
 }
 
@@ -2523,7 +2539,7 @@ void radeon_master_destroy(struct drm_device *dev, struct drm_master *master)
 
 	master_priv->sarea_priv = NULL;
 	if (master_priv->sarea)
-		drm_rmmap(dev, master_priv->sarea);
+		drm_rmmap_locked(dev, master_priv->sarea);
 		
 	drm_free(master_priv, sizeof(*master_priv), DRM_MEM_DRIVER);
 
diff --git a/shared-core/radeon_cs.c b/shared-core/radeon_cs.c
new file mode 100644
index 00000000..b0c4abe8
--- /dev/null
+++ b/shared-core/radeon_cs.c
@@ -0,0 +1,411 @@
+/*
+ * Copyright 2008 Jerome Glisse.
+ * 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, sublicense,
+ * 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 NONINFRINGEMENT.  IN NO EVENT SHALL
+ * PRECISION INSIGHT 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.
+ *
+ * Authors:
+ *    Jerome Glisse <glisse@freedesktop.org>
+ */
+#include "drmP.h"
+#include "radeon_drm.h"
+#include "radeon_drv.h"
+#include "r300_reg.h"
+
+int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *fpriv)
+{
+	struct drm_radeon_private *dev_priv = dev->dev_private;
+	struct drm_radeon_cs *cs = data;
+	uint32_t *packets = NULL;
+	uint32_t cs_id;
+	uint32_t card_offset;
+	void *ib = NULL;
+	long size;
+	int r;
+	RING_LOCALS;
+
+	/* set command stream id to 0 which is fake id */
+	cs_id = 0;
+	DRM_COPY_TO_USER(&cs->cs_id, &cs_id, sizeof(uint32_t));
+
+	if (dev_priv == NULL) {
+		DRM_ERROR("called with no initialization\n");
+		return -EINVAL;
+	}
+	if (!cs->dwords) {
+		return 0;
+	}
+	/* limit cs to 64K ib */
+	if (cs->dwords > (16 * 1024)) {
+		return -EINVAL;
+	}
+	/* copy cs from userspace maybe we should copy into ib to save
+	 * one copy but ib will be mapped wc so not good for cmd checking
+	 * somethings worth testing i guess (Jerome)
+	 */
+	size = cs->dwords * sizeof(uint32_t);
+	packets = drm_alloc(size, DRM_MEM_DRIVER);
+	if (packets == NULL) {
+		return -ENOMEM;
+	}
+	if (DRM_COPY_FROM_USER(packets, (void __user *)(unsigned long)cs->packets, size)) {
+		r = -EFAULT;
+		goto out;
+	}
+	/* get ib */
+	r = dev_priv->cs.ib_get(dev, &ib, cs->dwords, &card_offset);
+	if (r) {
+		goto out;
+	}
+
+	/* now parse command stream */
+	r = dev_priv->cs.parse(dev, fpriv, ib, packets, cs->dwords);
+	if (r) {
+		goto out;
+	}
+
+	BEGIN_RING(4);
+	OUT_RING(CP_PACKET0(RADEON_CP_IB_BASE, 1));
+	OUT_RING(card_offset);
+	OUT_RING(cs->dwords);
+	OUT_RING(CP_PACKET2());
+	ADVANCE_RING();
+
+	/* emit cs id sequence */
+	dev_priv->cs.id_emit(dev, &cs_id);
+	COMMIT_RING();
+
+	DRM_COPY_TO_USER(&cs->cs_id, &cs_id, sizeof(uint32_t));
+out:
+	dev_priv->cs.ib_free(dev, ib, cs->dwords);
+	drm_free(packets, size, DRM_MEM_DRIVER);
+	return r;
+}
+
+/* for non-mm */
+static int radeon_nomm_relocate(struct drm_device *dev, struct drm_file *file_priv, uint32_t *reloc, uint32_t *offset)
+{
+	*offset = reloc[1];
+	return 0;
+}
+#define RELOC_SIZE 2
+#define RADEON_2D_OFFSET_MASK 0x3fffff
+
+static __inline__ int radeon_cs_relocate_packet0(struct drm_device *dev, struct drm_file *file_priv,
+						 uint32_t *packets, uint32_t offset_dw)
+{
+	drm_radeon_private_t *dev_priv = dev->dev_private;
+	uint32_t hdr = packets[offset_dw];
+	uint32_t reg = (hdr & R300_CP_PACKET0_REG_MASK) << 2;
+	uint32_t val = packets[offset_dw + 1];
+	uint32_t packet3_hdr = packets[offset_dw+2];
+	uint32_t tmp, offset;
+	int ret;
+
+	/* this is too strict we may want to expand the length in the future and have
+	 old kernels ignore it. */ 
+	if (packet3_hdr != (RADEON_CP_PACKET3 | RADEON_CP_NOP | (RELOC_SIZE << 16))) {
+		DRM_ERROR("Packet 3 was %x should have been %x\n", packet3_hdr, RADEON_CP_PACKET3 | RADEON_CP_NOP | (RELOC_SIZE << 16));
+		return -EINVAL;
+	}
+	
+	switch(reg) {
+	case RADEON_DST_PITCH_OFFSET:
+	case RADEON_SRC_PITCH_OFFSET:
+		/* pass in the start of the reloc */
+		ret = dev_priv->cs.relocate(dev, file_priv, packets + offset_dw + 2, &offset);
+		if (ret)
+			return ret;
+		tmp = (val & RADEON_2D_OFFSET_MASK) << 10;
+		val &= ~RADEON_2D_OFFSET_MASK;
+		offset += tmp;
+		offset >>= 10;
+		val |= offset;
+		break;
+	case R300_RB3D_COLOROFFSET0:
+	case R300_ZB_DEPTHOFFSET:
+	case R300_TX_OFFSET_0:
+	case R300_TX_OFFSET_0+4:
+	        ret = dev_priv->cs.relocate(dev, file_priv, packets + offset_dw + 2, &offset);
+		if (ret)
+			return ret;
+
+		offset &= 0xffffffe0;
+		val += offset;
+		break;
+	default:
+		break;
+	}
+
+	packets[offset_dw + 1] = val;
+	return 0;
+}
+
+static int radeon_cs_relocate_packet3(struct drm_device *dev, struct drm_file *file_priv,
+				      uint32_t *packets, uint32_t offset_dw)
+{
+	drm_radeon_private_t *dev_priv = dev->dev_private;
+	uint32_t hdr = packets[offset_dw];
+	int num_dw = (hdr & RADEON_CP_PACKET_COUNT_MASK) >> 16;
+	uint32_t reg = hdr & 0xff00;
+	uint32_t offset, val, tmp;
+	int ret;
+
+	switch(reg) {
+	case RADEON_CNTL_HOSTDATA_BLT:
+	{
+		val = packets[offset_dw + 2];
+		ret = dev_priv->cs.relocate(dev, file_priv, packets + offset_dw + num_dw + 2, &offset);
+		if (ret)
+			return ret;
+
+		tmp = (val & RADEON_2D_OFFSET_MASK) << 10;
+		val &= ~RADEON_2D_OFFSET_MASK;
+		offset += tmp;
+		offset >>= 10;
+		val |= offset;
+
+		packets[offset_dw + 2] = val;
+	}
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+static __inline__ int radeon_cs_check_offset(struct drm_device *dev,
+					     uint32_t reg, uint32_t val)
+{
+	uint32_t offset;
+
+	switch(reg) {
+	case RADEON_DST_PITCH_OFFSET:
+	case RADEON_SRC_PITCH_OFFSET:
+		offset = val & ((1 << 22) - 1);
+		offset <<= 10;
+		break;
+	case R300_RB3D_COLOROFFSET0:
+	case R300_ZB_DEPTHOFFSET:
+		offset = val;
+		break;
+	case R300_TX_OFFSET_0:
+	case R300_TX_OFFSET_0+4:
+		offset = val & 0xffffffe0;
+		break;
+	}
+	
+	return 0;
+}
+
+int radeon_cs_packet0(struct drm_device *dev, struct drm_file *file_priv,
+		      uint32_t *packets, uint32_t offset_dw)
+{
+	drm_radeon_private_t *dev_priv = dev->dev_private;
+	uint32_t hdr = packets[offset_dw];
+	int num_dw = ((hdr & RADEON_CP_PACKET_COUNT_MASK) >> 16) + 2;
+	int need_reloc = 0;
+	int reg = (hdr & R300_CP_PACKET0_REG_MASK) << 2;
+	int count_dw = 1;
+	int ret;
+
+	while (count_dw < num_dw) {
+		/* need to have something like the r300 validation here - 
+		   list of allowed registers */
+		int flags;
+
+		ret = r300_check_range(reg, 1);
+		switch(ret) {
+		case -1:
+			DRM_ERROR("Illegal register %x\n", reg);
+			break;
+		case 0:
+			break;
+		case 1:
+			flags = r300_get_reg_flags(reg);
+			if (flags == MARK_CHECK_OFFSET) {
+				if (num_dw > 2) {
+					DRM_ERROR("Cannot relocate inside type stream of reg0 packets\n");
+					return -EINVAL;
+				}
+
+				ret = radeon_cs_relocate_packet0(dev, file_priv, packets, offset_dw);
+				if (ret)
+					return ret;
+				DRM_DEBUG("need to relocate %x %d\n", reg, flags);
+				/* okay it should be followed by a NOP */
+			} else if (flags == MARK_CHECK_SCISSOR) {
+				DRM_DEBUG("need to validate scissor %x %d\n", reg, flags);
+			} else {
+				DRM_DEBUG("illegal register %x %d\n", reg, flags);
+				return -EINVAL;
+			}
+			break;
+		}
+		count_dw++;
+		reg += 4;
+	}
+	return 0;
+}
+
+int radeon_cs_parse(struct drm_device *dev, struct drm_file *file_priv,
+		    void *ib, uint32_t *packets, uint32_t dwords)
+{
+	drm_radeon_private_t *dev_priv = dev->dev_private;
+	volatile int rb;
+	int size_dw = dwords;
+	/* scan the packet for various things */
+	int count_dw = 0;
+	int ret = 0;
+
+	while (count_dw < size_dw && ret == 0) {
+		int hdr = packets[count_dw];
+		int num_dw = (hdr & RADEON_CP_PACKET_COUNT_MASK) >> 16;
+		int reg;
+
+		switch (hdr & RADEON_CP_PACKET_MASK) {
+		case RADEON_CP_PACKET0:
+			ret = radeon_cs_packet0(dev, file_priv, packets, count_dw);
+			break;
+		case RADEON_CP_PACKET1:
+		case RADEON_CP_PACKET2:
+			reg = hdr & RADEON_CP_PACKET0_REG_MASK;
+			DRM_DEBUG("Packet 1/2: %d  %x\n", num_dw, reg);
+			break;
+
+		case RADEON_CP_PACKET3:
+			reg = hdr & 0xff00;
+			
+			switch(reg) {
+			case RADEON_CNTL_HOSTDATA_BLT:
+				radeon_cs_relocate_packet3(dev, file_priv, packets, count_dw);
+				break;
+
+			case RADEON_CNTL_BITBLT_MULTI:
+			case RADEON_3D_LOAD_VBPNTR:	/* load vertex array pointers */
+			case RADEON_CP_INDX_BUFFER:
+				DRM_ERROR("need relocate packet 3 for %x\n", reg);
+				break;
+
+			case RADEON_CP_3D_DRAW_IMMD_2:	/* triggers drawing using in-packet vertex data */
+			case RADEON_CP_3D_DRAW_VBUF_2:	/* triggers drawing of vertex buffers setup elsewhere */
+			case RADEON_CP_3D_DRAW_INDX_2:	/* triggers drawing using indices to vertex buffer */
+			case RADEON_WAIT_FOR_IDLE:
+			case RADEON_CP_NOP:
+				break;
+			default:
+				DRM_ERROR("unknown packet 3 %x\n", reg);
+				ret = -EINVAL;
+			}
+			break;
+		}
+
+		count_dw += num_dw+2;
+	}
+
+	if (ret)
+		return ret;
+	     
+
+	/* copy the packet into the IB */
+	memcpy(ib, packets, dwords * sizeof(uint32_t));
+
+	/* read back last byte to flush WC buffers */
+	rb = readl((ib + (dwords-1) * sizeof(uint32_t)));
+
+	return 0;
+}
+
+uint32_t radeon_cs_id_get(struct drm_radeon_private *radeon)
+{
+	/* FIXME: protect with a spinlock */
+	/* FIXME: check if wrap affect last reported wrap & sequence */
+	radeon->cs.id_scnt = (radeon->cs.id_scnt + 1) & 0x00FFFFFF;
+	if (!radeon->cs.id_scnt) {
+		/* increment wrap counter */
+		radeon->cs.id_wcnt += 0x01000000;
+		/* valid sequence counter start at 1 */
+		radeon->cs.id_scnt = 1;
+	}
+	return (radeon->cs.id_scnt | radeon->cs.id_wcnt);
+}
+
+void r100_cs_id_emit(struct drm_device *dev, uint32_t *id)
+{
+	drm_radeon_private_t *dev_priv = dev->dev_private;
+	RING_LOCALS;
+
+	/* ISYNC_CNTL should have CPSCRACTH bit set */
+	*id = radeon_cs_id_get(dev_priv);
+	/* emit id in SCRATCH4 (not used yet in old drm) */
+	BEGIN_RING(2);
+	OUT_RING(CP_PACKET0(RADEON_SCRATCH_REG4, 0));
+	OUT_RING(*id);
+	ADVANCE_RING();	
+}
+
+void r300_cs_id_emit(struct drm_device *dev, uint32_t *id)
+{
+	drm_radeon_private_t *dev_priv = dev->dev_private;
+	RING_LOCALS;
+
+	/* ISYNC_CNTL should not have CPSCRACTH bit set */
+	*id = radeon_cs_id_get(dev_priv);
+	/* emit id in SCRATCH6 */
+	BEGIN_RING(6);
+	OUT_RING(CP_PACKET0(R300_CP_RESYNC_ADDR, 0));
+	OUT_RING(6);
+	OUT_RING(CP_PACKET0(R300_CP_RESYNC_DATA, 0));
+	OUT_RING(*id);
+	OUT_RING(CP_PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));
+	OUT_RING(R300_RB3D_DC_FINISH);
+	ADVANCE_RING();	
+}
+
+uint32_t r100_cs_id_last_get(struct drm_device *dev)
+{
+	drm_radeon_private_t *dev_priv = dev->dev_private;
+
+	return RADEON_READ(RADEON_SCRATCH_REG4);
+}
+
+uint32_t r300_cs_id_last_get(struct drm_device *dev)
+{
+	drm_radeon_private_t *dev_priv = dev->dev_private;
+
+	return RADEON_READ(RADEON_SCRATCH_REG6);
+}
+
+int radeon_cs_init(struct drm_device *dev)
+{
+	drm_radeon_private_t *dev_priv = dev->dev_private;
+
+	if (dev_priv->chip_family < CHIP_RV280) {
+		dev_priv->cs.id_emit = r100_cs_id_emit;
+		dev_priv->cs.id_last_get = r100_cs_id_last_get;
+	} else if (dev_priv->chip_family < CHIP_R600) {
+		dev_priv->cs.id_emit = r300_cs_id_emit;
+		dev_priv->cs.id_last_get = r300_cs_id_last_get;
+	}
+
+	dev_priv->cs.parse = radeon_cs_parse;
+	/* ib get depends on memory manager or not so memory manager */
+	dev_priv->cs.relocate = radeon_nomm_relocate;
+	return 0;
+}
diff --git a/shared-core/radeon_drm.h b/shared-core/radeon_drm.h
index 7fcf9305..8bb5d87c 100644
--- a/shared-core/radeon_drm.h
+++ b/shared-core/radeon_drm.h
@@ -455,6 +455,13 @@ typedef struct {
 	int tiling_enabled;	/* set by drm, read by 2d + 3d clients */
 
 	unsigned int last_fence;
+
+	uint32_t front_handle;
+	uint32_t back_handle;
+	uint32_t depth_handle;
+	uint32_t front_pitch;
+	uint32_t back_pitch;
+	uint32_t depth_pitch;
 } drm_radeon_sarea_t;
 
 
@@ -506,6 +513,7 @@ typedef struct {
 #define DRM_RADEON_GEM_SET_DOMAIN 0x23
 #define DRM_RADEON_GEM_INDIRECT 0x24 // temporary for X server
 
+#define DRM_RADEON_CS           0x25
 
 #define DRM_IOCTL_RADEON_CP_INIT    DRM_IOW( DRM_COMMAND_BASE + DRM_RADEON_CP_INIT, drm_radeon_init_t)
 #define DRM_IOCTL_RADEON_CP_START   DRM_IO(  DRM_COMMAND_BASE + DRM_RADEON_CP_START)
@@ -545,6 +553,7 @@ typedef struct {
 #define DRM_IOCTL_RADEON_GEM_SET_DOMAIN  DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_SET_DOMAIN, struct drm_radeon_gem_set_domain)
 #define DRM_IOCTL_RADEON_GEM_INDIRECT DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_GEM_INDIRECT, struct drm_radeon_gem_indirect)
 
+#define DRM_IOCTL_RADEON_CS DRM_IOWR(DRM_COMMAND_BASE + DRM_RADEON_CS, struct drm_radeon_cs)
 
 
 typedef struct drm_radeon_init {
@@ -703,6 +712,7 @@ typedef struct drm_radeon_indirect {
 #define RADEON_PARAM_VBLANK_CRTC           13   /* VBLANK CRTC */
 #define RADEON_PARAM_FB_LOCATION           14   /* FB location */
 #define RADEON_PARAM_NUM_GB_PIPES          15   /* num GB pipes */
+#define RADEON_PARAM_KERNEL_MM             16
 
 typedef struct drm_radeon_getparam {
 	int param;
@@ -758,6 +768,7 @@ typedef struct drm_radeon_setparam {
 #define RADEON_SETPARAM_NEW_MEMMAP 4		/* Use new memory map */
 #define RADEON_SETPARAM_PCIGART_TABLE_SIZE 5    /* PCI GART Table Size */
 #define RADEON_SETPARAM_VBLANK_CRTC 6           /* VBLANK CRTC */
+#define RADEON_SETPARAM_MM_INIT 7		/* Initialise the mm */
 /* 1.14: Clients can allocate/free a surface
  */
 typedef struct drm_radeon_surface_alloc {
@@ -861,4 +872,17 @@ struct drm_radeon_gem_indirect {
 	uint32_t used;
 };
 
+/* New interface which obsolete all previous interface.
+ */
+
+
+struct drm_radeon_cs {
+//	uint32_t __user     *packets;
+	uint32_t            dwords;
+	uint32_t            cs_id;
+	uint64_t            packets;
+
+};
+
+
 #endif
diff --git a/shared-core/radeon_drv.h b/shared-core/radeon_drv.h
index 5ce97b64..a40ff6dd 100644
--- a/shared-core/radeon_drv.h
+++ b/shared-core/radeon_drv.h
@@ -195,11 +195,11 @@ enum radeon_mac_model {
 
 
 #define GET_RING_HEAD(dev_priv)	(dev_priv->writeback_works ? \
-				 (dev_priv->mm.ring_read_ptr ? readl(dev_priv->mm.ring_read_ptr_map.virtual + 0) : DRM_READ32((dev_priv)->ring_rptr, 0 )) : \
+				 (dev_priv->mm.ring_read.bo ? readl(dev_priv->mm.ring_read.kmap.virtual + 0) : DRM_READ32((dev_priv)->ring_rptr, 0 )) : \
 				 RADEON_READ(RADEON_CP_RB_RPTR))
 
-#define SET_RING_HEAD(dev_priv,val) (dev_priv->mm.ring_read_ptr ? \
-				     writel((val), dev_priv->mm.ring_read_ptr_map.virtual) : \
+#define SET_RING_HEAD(dev_priv,val) (dev_priv->mm.ring_read.bo ? \
+				     writel((val), dev_priv->mm.ring_read.kmap.virtual) : \
 				     DRM_WRITE32((dev_priv)->ring_rptr, 0, (val)))
 
 typedef struct drm_radeon_freelist {
@@ -261,6 +261,11 @@ struct radeon_virt_surface {
 	struct drm_file *file_priv;
 };
 
+struct radeon_mm_obj {
+	struct drm_buffer_object *bo;
+	struct drm_bo_kmap_obj kmap;
+};
+
 struct radeon_mm_info {
 	uint64_t vram_offset; // Offset into GPU space
 	uint64_t vram_size;
@@ -268,15 +273,13 @@ struct radeon_mm_info {
 	
 	uint64_t gart_start;
 	uint64_t gart_size;
+	
+	struct radeon_mm_obj pcie_table;
+	struct radeon_mm_obj ring;
+	struct radeon_mm_obj ring_read;
 
-	struct drm_buffer_object *pcie_table;
-	struct drm_bo_kmap_obj pcie_table_map;
-
-	struct drm_buffer_object *ring;
-	struct drm_bo_kmap_obj ring_map;
-
-	struct drm_buffer_object *ring_read_ptr;
-	struct drm_bo_kmap_obj ring_read_ptr_map;
+	struct radeon_mm_obj dma_bufs;
+	struct drm_map fake_agp_map;
 };
 
 #include "radeon_mode.h"
@@ -289,11 +292,35 @@ struct drm_radeon_master_private {
 #define RADEON_FLUSH_EMITED	(1 < 0)
 #define RADEON_PURGE_EMITED	(1 < 1)
 
+/* command submission struct */
+struct drm_radeon_cs_priv {
+	uint32_t id_wcnt;
+	uint32_t id_scnt;
+	uint32_t id_last_wcnt;
+	uint32_t id_last_scnt;
+
+	int (*parse)(struct drm_device *dev, struct drm_file *file_priv,
+		     void *ib, uint32_t *packets, uint32_t dwords);
+	void (*id_emit)(struct drm_device *dev, uint32_t *id);
+	uint32_t (*id_last_get)(struct drm_device *dev);
+	/* this ib handling callback are for hidding memory manager drm
+	 * from memory manager less drm, free have to emit ib discard
+	 * sequence into the ring */
+	int (*ib_get)(struct drm_device *dev, void **ib, uint32_t dwords, uint32_t *card_offset);
+	uint32_t (*ib_get_ptr)(struct drm_device *dev, void *ib);
+	void (*ib_free)(struct drm_device *dev, void *ib, uint32_t dwords);
+	/* do a relocation either MM or non-MM */
+	int (*relocate)(struct drm_device *dev, struct drm_file *file_priv,
+			 uint32_t *reloc, uint32_t *offset);
+};
+
 typedef struct drm_radeon_private {
 
 	drm_radeon_ring_buffer_t ring;
 
-	int new_memmap;
+	bool new_memmap;
+
+	bool user_mm_enable;
 
 	int gart_size;
 	u32 gart_vm_start;
@@ -372,6 +399,7 @@ typedef struct drm_radeon_private {
 	uint32_t flags;		/* see radeon_chip_flags */
 	unsigned long fb_aper_offset;
 
+	bool mm_enabled;
 	struct radeon_mm_info mm;
 	drm_local_map_t *mmio;
 
@@ -395,6 +423,11 @@ typedef struct drm_radeon_private {
 	int num_gb_pipes;
 	int track_flush;
 	uint32_t chip_family; /* extract from flags */
+
+	struct radeon_mm_obj **ib_objs;
+	/* ib bitmap */
+	uint64_t ib_alloc_bitmap; // TO DO replace with a real bitmap
+	struct drm_radeon_cs_priv cs;
 } drm_radeon_private_t;
 
 typedef struct drm_radeon_buf_priv {
@@ -672,14 +705,15 @@ extern int r300_do_cp_cmdbuf(struct drm_device *dev,
 #define RADEON_SCRATCH_REG3		0x15ec
 #define RADEON_SCRATCH_REG4		0x15f0
 #define RADEON_SCRATCH_REG5		0x15f4
+#define RADEON_SCRATCH_REG6		0x15f8
 #define RADEON_SCRATCH_UMSK		0x0770
 #define RADEON_SCRATCH_ADDR		0x0774
 
 #define RADEON_SCRATCHOFF( x )		(RADEON_SCRATCH_REG_OFFSET + 4*(x))
 
 #define GET_SCRATCH( x )	(dev_priv->writeback_works ?			\
-				 (dev_priv->mm.ring_read_ptr ? \
-				  readl(dev_priv->mm.ring_read_ptr_map.virtual + RADEON_SCRATCHOFF(0)) : \
+				 (dev_priv->mm.ring_read.bo ? \
+				  readl(dev_priv->mm.ring_read.kmap.virtual + RADEON_SCRATCHOFF(0)) : \
 				  DRM_READ32(dev_priv->ring_rptr, RADEON_SCRATCHOFF(x))) : \
 				 RADEON_READ( RADEON_SCRATCH_REG0 + 4*(x)))
 
@@ -1243,44 +1277,62 @@ extern int r300_do_cp_cmdbuf(struct drm_device *dev,
 #define RADEON_READ8(reg)	DRM_READ8(  dev_priv->mmio, (reg) )
 #define RADEON_WRITE8(reg,val)	DRM_WRITE8( dev_priv->mmio, (reg), (val) )
 
-extern int RADEON_READ_PLL(struct drm_radeon_private *dev_priv, int addr);
+extern u32 RADEON_READ_PLL(struct drm_radeon_private *dev_priv, int addr);
 extern void RADEON_WRITE_PLL(struct drm_radeon_private *dev_priv, int addr, uint32_t data);
 
-#define RADEON_WRITE_PCIE( addr, val )					\
+#define RADEON_WRITE_P(reg, val, mask)		\
+do {						\
+	uint32_t tmp = RADEON_READ(reg);	\
+	tmp &= (mask);				\
+	tmp |= ((val) & ~(mask));		\
+	RADEON_WRITE(reg, tmp);			\
+} while(0)
+
+#define RADEON_WRITE_PLL_P(dev_priv, addr, val, mask)		\
+do {								\
+	uint32_t tmp_ = RADEON_READ_PLL(dev_priv, addr);	\
+	tmp_ &= (mask);						\
+	tmp_ |= ((val) & ~(mask));				\
+	RADEON_WRITE_PLL(dev_priv, addr, tmp_);			\
+} while (0)
+
+
+
+#define RADEON_WRITE_PCIE(addr, val)					\
 do {									\
-	RADEON_WRITE8( RADEON_PCIE_INDEX,				\
+	RADEON_WRITE8(RADEON_PCIE_INDEX,				\
 			((addr) & 0xff));				\
-	RADEON_WRITE( RADEON_PCIE_DATA, (val) );			\
+	RADEON_WRITE(RADEON_PCIE_DATA, (val));			\
 } while (0)
 
-#define R500_WRITE_MCIND( addr, val )					\
+#define R500_WRITE_MCIND(addr, val)					\
 do {								\
 	RADEON_WRITE(R520_MC_IND_INDEX, 0xff0000 | ((addr) & 0xff));	\
 	RADEON_WRITE(R520_MC_IND_DATA, (val));			\
 	RADEON_WRITE(R520_MC_IND_INDEX, 0);	\
 } while (0)
 
-#define RS480_WRITE_MCIND( addr, val )				\
+#define RS480_WRITE_MCIND(addr, val)				\
 do {									\
-	RADEON_WRITE( RS480_NB_MC_INDEX,				\
+	RADEON_WRITE(RS480_NB_MC_INDEX,				\
 			((addr) & 0xff) | RS480_NB_MC_IND_WR_EN);	\
-	RADEON_WRITE( RS480_NB_MC_DATA, (val) );			\
-	RADEON_WRITE( RS480_NB_MC_INDEX, 0xff );			\
+	RADEON_WRITE(RS480_NB_MC_DATA, (val));			\
+	RADEON_WRITE(RS480_NB_MC_INDEX, 0xff);			\
 } while (0)
 
-#define RS690_WRITE_MCIND( addr, val )					\
+#define RS690_WRITE_MCIND(addr, val)					\
 do {								\
 	RADEON_WRITE(RS690_MC_INDEX, RS690_MC_INDEX_WR_EN | ((addr) & RS690_MC_INDEX_MASK));	\
 	RADEON_WRITE(RS690_MC_DATA, val);			\
 	RADEON_WRITE(RS690_MC_INDEX, RS690_MC_INDEX_WR_ACK);	\
 } while (0)
 
-#define IGP_WRITE_MCIND( addr, val )				\
+#define IGP_WRITE_MCIND(addr, val)				\
 do {									\
-        if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690)       \
-	        RS690_WRITE_MCIND( addr, val );                         \
-	else                                                            \
-	        RS480_WRITE_MCIND( addr, val );                         \
+	if ((dev_priv->flags & RADEON_FAMILY_MASK) == CHIP_RS690)       \
+		RS690_WRITE_MCIND(addr, val);				\
+	else								\
+		RS480_WRITE_MCIND(addr, val);				\
 } while (0)
 
 #define CP_PACKET0( reg, n )						\
@@ -1324,42 +1376,42 @@ do {									\
 
 #define RADEON_FLUSH_CACHE() do {					\
 	if ((dev_priv->flags & RADEON_FAMILY_MASK) <= CHIP_RV280) {	\
-	        OUT_RING(CP_PACKET0(RADEON_RB3D_DSTCACHE_CTLSTAT, 0));	\
-	        OUT_RING(RADEON_RB3D_DC_FLUSH);				\
+		OUT_RING(CP_PACKET0(RADEON_RB3D_DSTCACHE_CTLSTAT, 0));	\
+		OUT_RING(RADEON_RB3D_DC_FLUSH);				\
 	} else {                                                        \
-	        OUT_RING(CP_PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));	\
-	        OUT_RING(R300_RB3D_DC_FLUSH);				\
-        }                                                               \
+		OUT_RING(CP_PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));	\
+		OUT_RING(RADEON_RB3D_DC_FLUSH);				\
+	}                                                               \
 } while (0)
 
 #define RADEON_PURGE_CACHE() do {					\
 	if ((dev_priv->flags & RADEON_FAMILY_MASK) <= CHIP_RV280) {	\
-	        OUT_RING(CP_PACKET0( RADEON_RB3D_DSTCACHE_CTLSTAT, 0));	\
-	        OUT_RING(RADEON_RB3D_DC_FLUSH | RADEON_RB3D_DC_FREE);	\
+		OUT_RING(CP_PACKET0(RADEON_RB3D_DSTCACHE_CTLSTAT, 0));	\
+		OUT_RING(RADEON_RB3D_DC_FLUSH | RADEON_RB3D_DC_FREE);	\
 	} else {                                                        \
-	        OUT_RING(CP_PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));	\
-	        OUT_RING(R300_RB3D_DC_FLUSH | R300_RB3D_DC_FREE );	\
-        }                                                               \
+		OUT_RING(CP_PACKET0(R300_RB3D_DSTCACHE_CTLSTAT, 0));	\
+		OUT_RING(R300_RB3D_DC_FLUSH | R300_RB3D_DC_FREE );	\
+	}                                                               \
 } while (0)
 
 #define RADEON_FLUSH_ZCACHE() do {					\
 	if ((dev_priv->flags & RADEON_FAMILY_MASK) <= CHIP_RV280) {	\
-	        OUT_RING( CP_PACKET0( RADEON_RB3D_ZCACHE_CTLSTAT, 0 ) ); \
-	        OUT_RING( RADEON_RB3D_ZC_FLUSH );			\
+		OUT_RING(CP_PACKET0(RADEON_RB3D_ZCACHE_CTLSTAT, 0));	\
+		OUT_RING(RADEON_RB3D_ZC_FLUSH);				\
 	} else {                                                        \
-	        OUT_RING( CP_PACKET0( R300_ZB_ZCACHE_CTLSTAT, 0 ) );	\
-	        OUT_RING( R300_ZC_FLUSH );				\
-        }                                                               \
+		OUT_RING(CP_PACKET0(R300_ZB_ZCACHE_CTLSTAT, 0));	\
+		OUT_RING(R300_ZC_FLUSH);				\
+	}                                                               \
 } while (0)
 
 #define RADEON_PURGE_ZCACHE() do {					\
 	if ((dev_priv->flags & RADEON_FAMILY_MASK) <= CHIP_RV280) {	\
-	        OUT_RING(CP_PACKET0(RADEON_RB3D_ZCACHE_CTLSTAT, 0));	\
-	        OUT_RING(RADEON_RB3D_ZC_FLUSH | RADEON_RB3D_ZC_FREE);	\
+		OUT_RING(CP_PACKET0(RADEON_RB3D_ZCACHE_CTLSTAT, 0));	\
+		OUT_RING(RADEON_RB3D_ZC_FLUSH | RADEON_RB3D_ZC_FREE);	\
 	} else {                                                        \
-	        OUT_RING(CP_PACKET0(R300_ZB_ZCACHE_CTLSTAT, 0));	\
-	        OUT_RING(R300_ZC_FLUSH | R300_ZC_FREE);			\
-        }                                                               \
+		OUT_RING(CP_PACKET0(R300_ZB_ZCACHE_CTLSTAT, 0));	\
+		OUT_RING(R300_ZC_FLUSH | R300_ZC_FREE);			\
+	}                                                               \
 } while (0)
 
 /* ================================================================
@@ -1380,7 +1432,7 @@ do {									\
 #define VB_AGE_TEST_WITH_RETURN( dev_priv )				\
 do {									\
 	struct drm_radeon_master_private *master_priv = file_priv->master->driver_priv;		\
-	drm_radeon_sarea_t *sarea_priv = master_priv->sarea_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;				\
@@ -1556,6 +1608,23 @@ static inline int radeon_update_breadcrumb(struct drm_device *dev)
 
 #define radeon_is_dce3(dev_priv) ((dev_priv->chip_family >= CHIP_RV620))
 
+#define radeon_is_rv100(dev_priv) ((dev_priv->chip_family == CHIP_RV100) || \
+				   (dev_priv->chip_family == CHIP_RV200) || \
+				   (dev_priv->chip_family == CHIP_RS100) || \
+				   (dev_priv->chip_family == CHIP_RS200) || \
+				   (dev_priv->chip_family == CHIP_RV250) || \
+				   (dev_priv->chip_family == CHIP_RV280) || \
+				   (dev_priv->chip_family == CHIP_RS300))
+
+#define radeon_is_r300(dev_priv) ((dev_priv->chip_family == CHIP_R300)  || \
+				  (dev_priv->chip_family == CHIP_RV350) || \
+				  (dev_priv->chip_family == CHIP_R350)  || \
+				  (dev_priv->chip_family == CHIP_RV380) || \
+				  (dev_priv->chip_family == CHIP_R420)  || \
+				  (dev_priv->chip_family == CHIP_RV410) || \
+				  (dev_priv->chip_family == CHIP_RS400) || \
+				  (dev_priv->chip_family == CHIP_RS480))
+
 #define radeon_bios8(dev_priv, v) (dev_priv->bios[v])
 #define radeon_bios16(dev_priv, v) (dev_priv->bios[v] | (dev_priv->bios[(v) + 1] << 8))
 #define radeon_bios32(dev_priv, v) ((dev_priv->bios[v]) | \
@@ -1563,6 +1632,7 @@ static inline int radeon_update_breadcrumb(struct drm_device *dev)
 				    (dev_priv->bios[(v) + 2] << 16) | \
 				    (dev_priv->bios[(v) + 3] << 24))
 
+extern void radeon_pll_errata_after_index(struct drm_radeon_private *dev_priv);
 extern int radeon_emit_irq(struct drm_device * dev);
 
 extern void radeon_gem_free_object(struct drm_gem_object *obj);
@@ -1592,4 +1662,14 @@ extern void radeon_set_pcigart(drm_radeon_private_t * dev_priv, int on);
 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);
+extern int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *fpriv);
+extern int radeon_cs_init(struct drm_device *dev);
+void radeon_gem_update_offsets(struct drm_device *dev, struct drm_master *master);
+
+#define MARK_SAFE		1
+#define MARK_CHECK_OFFSET	2
+#define MARK_CHECK_SCISSOR	3
+
+extern int r300_check_range(unsigned reg, int count);
+extern int r300_get_reg_flags(unsigned reg);
 #endif				/* __RADEON_DRV_H__ */
diff --git a/shared-core/radeon_state.c b/shared-core/radeon_state.c
index 6de4b135..7262b2aa 100644
--- a/shared-core/radeon_state.c
+++ b/shared-core/radeon_state.c
@@ -2223,6 +2223,9 @@ 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;
 
+	if (dev_priv->mm.vram_offset)
+		radeon_gem_update_offsets(dev, file_priv->master);
+
 	radeon_cp_dispatch_swap(dev, file_priv->master);
 	sarea_priv->ctx_owner = 0;
 
@@ -3117,6 +3120,9 @@ static int radeon_cp_getparam(struct drm_device *dev, void *data, struct drm_fil
 	case RADEON_PARAM_NUM_GB_PIPES:
 		value = dev_priv->num_gb_pipes;
 		break;
+	case RADEON_PARAM_KERNEL_MM:
+		value = dev_priv->mm_enabled;
+		break;
 	default:
 		DRM_DEBUG( "Invalid parameter %d\n", param->param );
 		return -EINVAL;
@@ -3178,6 +3184,10 @@ static int radeon_cp_setparam(struct drm_device *dev, void *data, struct drm_fil
 	case RADEON_SETPARAM_VBLANK_CRTC:
 		return radeon_vblank_crtc_set(dev, sp->value);
 		break;
+	case RADEON_SETPARAM_MM_INIT:
+		dev_priv->user_mm_enable = true;
+		dev_priv->new_memmap = true;
+		return radeon_gem_mm_init(dev);
 	default:
 		DRM_DEBUG("Invalid parameter %d\n", sp->param);
 		return -EINVAL;
@@ -3279,6 +3289,7 @@ struct drm_ioctl_desc radeon_ioctls[] = {
 	DRM_IOCTL_DEF(DRM_RADEON_GEM_PWRITE, radeon_gem_pwrite_ioctl, DRM_AUTH),
 	DRM_IOCTL_DEF(DRM_RADEON_GEM_SET_DOMAIN, radeon_gem_set_domain_ioctl, DRM_AUTH),
 	DRM_IOCTL_DEF(DRM_RADEON_GEM_INDIRECT, radeon_gem_indirect_ioctl, DRM_AUTH),
+	DRM_IOCTL_DEF(DRM_RADEON_CS, radeon_cs_ioctl, DRM_AUTH),
 };
 
 int radeon_max_ioctl = DRM_ARRAY_SIZE(radeon_ioctls);