Merge branch 'vbo_clean'

Conflicts:
	src/mesa/drivers/dri/r300/r300_draw.c

13 files changed, 621 insertions, 334 deletions

diff --git a/src/mesa/drivers/dri/r300/Makefile b/src/mesa/drivers/dri/r300/Makefile
index 95c6765bc4..2390d1896a 100644
--- a/src/mesa/drivers/dri/r300/Makefile
+++ b/src/mesa/drivers/dri/r300/Makefile
@@ -36,7 +36,8 @@ RADEON_COMMON_SOURCES = \
 	radeon_cs_legacy.c \
 	radeon_mipmap_tree.c \
 	radeon_span.c \
-	radeon_fbo.c
+	radeon_fbo.c \
+	radeon_buffer_objects.c
 
 DRIVER_SOURCES = \
 		 radeon_screen.c \
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index db404b3847..1baae8fc76 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -64,9 +64,9 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r300_ioctl.h"
 #include "r300_tex.h"
 #include "r300_emit.h"
-#include "r300_render.h"
 #include "r300_swtcl.h"
 #include "radeon_bocs_wrapper.h"
+#include "radeon_buffer_objects.h"
 
 
 #include "vblank.h"
@@ -154,7 +154,6 @@ const struct dri_extension gl_20_extension[] = {
 };
 
 static const struct tnl_pipeline_stage *r300_pipeline[] = {
-
 	/* Catch any t&l fallbacks
 	 */
 	&_tnl_vertex_transform_stage,
@@ -165,21 +164,7 @@ static const struct tnl_pipeline_stage *r300_pipeline[] = {
 	&_tnl_texture_transform_stage,
 	&_tnl_point_attenuation_stage,
 	&_tnl_vertex_program_stage,
-
-	/* Try again to go to tcl?
-	 *     - no good for asymmetric-twoside (do with multipass)
-	 *     - no good for asymmetric-unfilled (do with multipass)
-	 *     - good for material
-	 *     - good for texgen
-	 *     - need to manipulate a bit of state
-	 *
-	 * - worth it/not worth it?
-	 */
-
-	/* Else do them here.
-	 */
-	&_r300_render_stage,
-	&_tnl_render_stage,	/* FALLBACK  */
+	&_tnl_render_stage,
 	0,
 };
 
@@ -398,6 +383,7 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	r300InitStateFuncs(&functions);
 	r300InitTextureFuncs(&functions);
 	r300InitShaderFuncs(&functions);
+	radeonInitBufferObjectFuncs(&functions);
 
 	if (!radeonInitContext(&r300->radeon, &functions,
 			       glVisual, driContextPriv,
diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index 24dc6bc6a3..d620417422 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -478,11 +478,12 @@ struct r300_vertex_buffer {
 	struct vertex_attribute {
 		/* generic */
 		GLubyte element;
-		GLvoid *data;
-		GLboolean free_needed;
 		GLuint stride;
 		GLuint dwords;
 		GLubyte size; /* number of components */
+		GLboolean is_named_bo;
+		struct radeon_bo *bo;
+		GLint bo_offset;
 
 		/* hw specific */
 		uint32_t data_type:4;
@@ -497,9 +498,10 @@ struct r300_vertex_buffer {
 };
 
 struct r300_index_buffer {
-	GLvoid *ptr;
+	struct radeon_bo *bo;
+	int bo_offset;
+
 	GLboolean is_32bit;
-	GLboolean free_needed;
 	GLuint count;
 };
 
diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c
index aedc6cfb2a..ab2287a5e2 100644
--- a/src/mesa/drivers/dri/r300/r300_draw.c
+++ b/src/mesa/drivers/dri/r300/r300_draw.c
@@ -39,38 +39,63 @@
 #include "r300_state.h"
 #include "r300_tex.h"
 
+#include "radeon_buffer_objects.h"
+
 #include "tnl/tnl.h"
 #include "tnl/t_vp_build.h"
 #include "vbo/vbo_context.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 
-static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer *mesa_ind_buf, struct gl_buffer_object **bo, GLuint *nr_bo)
+
+static int getTypeSize(GLenum type)
+{
+	switch (type) {
+		case GL_DOUBLE:
+			return sizeof(GLdouble);
+		case GL_FLOAT:
+			return sizeof(GLfloat);
+		case GL_INT:
+			return sizeof(GLint);
+		case GL_UNSIGNED_INT:
+			return sizeof(GLuint);
+		case GL_SHORT:
+			return sizeof(GLshort);
+		case GL_UNSIGNED_SHORT:
+			return sizeof(GLushort);
+		case GL_BYTE:
+			return sizeof(GLbyte);
+		case GL_UNSIGNED_BYTE:
+			return sizeof(GLubyte);
+		default:
+			assert(0);
+			return 0;
+	}
+}
+
+static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer *mesa_ind_buf)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-	struct r300_index_buffer *ind_buf = &r300->ind_buf;
 	GLvoid *src_ptr;
+	GLuint *out;
+	int i;
+	GLboolean mapped_named_bo = GL_FALSE;
 
-	if (!mesa_ind_buf) {
-		ind_buf->ptr = NULL;
-		return;
-	}
-
-	ind_buf->count = mesa_ind_buf->count;
 	if (mesa_ind_buf->obj->Name && !mesa_ind_buf->obj->Pointer) {
-		bo[*nr_bo] = mesa_ind_buf->obj;
-		(*nr_bo)++;
 		ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, GL_READ_ONLY_ARB, mesa_ind_buf->obj);
+		mapped_named_bo = GL_TRUE;
 		assert(mesa_ind_buf->obj->Pointer != NULL);
 	}
 	src_ptr = ADD_POINTERS(mesa_ind_buf->obj->Pointer, mesa_ind_buf->ptr);
 
 	if (mesa_ind_buf->type == GL_UNSIGNED_BYTE) {
+		GLuint size = sizeof(GLushort) * ((mesa_ind_buf->count + 1) & ~1);
 		GLubyte *in = (GLubyte *)src_ptr;
-		GLuint *out = _mesa_malloc(sizeof(GLushort) * ((mesa_ind_buf->count + 1) & ~1));
-		int i;
 
-		ind_buf->ptr = out;
+		radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, &r300->ind_buf.bo_offset, size, 4);
+
+		assert(r300->ind_buf.bo->ptr != NULL);
+		out = (GLuint *)ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset);
 
 		for (i = 0; i + 1 < mesa_ind_buf->count; i += 2) {
 			*out++ = in[i] | in[i + 1] << 16;
@@ -80,16 +105,15 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer
 			*out++ = in[i];
 		}
 
-		ind_buf->free_needed = GL_TRUE;
-		ind_buf->is_32bit = GL_FALSE;
-	} else if (mesa_ind_buf->type == GL_UNSIGNED_SHORT) {
 #if MESA_BIG_ENDIAN
+	} else { /* if (mesa_ind_buf->type == GL_UNSIGNED_SHORT) */
 		GLushort *in = (GLushort *)src_ptr;
-		GLuint *out = _mesa_malloc(sizeof(GLushort) *
-					   ((mesa_ind_buf->count + 1) & ~1));
-		int i;
+		size = sizeof(GLushort) * ((mesa_ind_buf->count + 1) & ~1);
+
+		radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, &r300->ind_buf.bo_offet, size, 4);
 
-		ind_buf->ptr = out;
+		assert(r300->ind_buf.bo->ptr != NULL)
+		out = (GLuint *)ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset);
 
 		for (i = 0; i + 1 < mesa_ind_buf->count; i += 2) {
 			*out++ = in[i] | in[i + 1] << 16;
@@ -98,42 +122,60 @@ static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer
 		if (i < mesa_ind_buf->count) {
 			*out++ = in[i];
 		}
-
-		ind_buf->free_needed = GL_TRUE;
-#else
-		ind_buf->ptr = src_ptr;
-		ind_buf->free_needed = GL_FALSE;
 #endif
-		ind_buf->is_32bit = GL_FALSE;
-	} else {
-		ind_buf->ptr = src_ptr;
-		ind_buf->free_needed = GL_FALSE;
-		ind_buf->is_32bit = GL_TRUE;
+	}
+
+	r300->ind_buf.is_32bit = GL_FALSE;
+	r300->ind_buf.count = mesa_ind_buf->count;
+
+	if (mapped_named_bo) {
+		ctx->Driver.UnmapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, mesa_ind_buf->obj);
 	}
 }
 
-static int getTypeSize(GLenum type)
+
+static void r300SetupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer *mesa_ind_buf)
 {
-	switch (type) {
-		case GL_DOUBLE:
-			return sizeof(GLdouble);
-		case GL_FLOAT:
-			return sizeof(GLfloat);
-		case GL_INT:
-			return sizeof(GLint);
-		case GL_UNSIGNED_INT:
-			return sizeof(GLuint);
-		case GL_SHORT:
-			return sizeof(GLshort);
-		case GL_UNSIGNED_SHORT:
-			return sizeof(GLushort);
-		case GL_BYTE:
-			return sizeof(GLbyte);
-		case GL_UNSIGNED_BYTE:
-			return sizeof(GLubyte);
-		default:
-			assert(0);
-			return 0;
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+
+	if (!mesa_ind_buf) {
+		r300->ind_buf.bo = NULL;
+		return;
+	}
+
+#if MESA_BIG_ENDIAN
+	if (mesa_ind_buf->type == GL_UNSIGNED_INT) {
+#else
+	if (mesa_ind_buf->type != GL_UNSIGNED_BYTE) {
+#endif
+		const GLvoid *src_ptr;
+		GLvoid *dst_ptr;
+		GLboolean mapped_named_bo = GL_FALSE;
+
+		if (mesa_ind_buf->obj->Name && !mesa_ind_buf->obj->Pointer) {
+			ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, GL_READ_ONLY_ARB, mesa_ind_buf->obj);
+			assert(mesa_ind_buf->obj->Pointer != NULL);
+			mapped_named_bo = GL_TRUE;
+		}
+
+		src_ptr = ADD_POINTERS(mesa_ind_buf->obj->Pointer, mesa_ind_buf->ptr);
+
+		const GLuint size = mesa_ind_buf->count * getTypeSize(mesa_ind_buf->type);
+
+		radeonAllocDmaRegion(&r300->radeon, &r300->ind_buf.bo, &r300->ind_buf.bo_offset, size, 4);
+
+		assert(r300->ind_buf.bo->ptr != NULL);
+		dst_ptr = ADD_POINTERS(r300->ind_buf.bo->ptr, r300->ind_buf.bo_offset);
+		_mesa_memcpy(dst_ptr, src_ptr, size);
+
+		r300->ind_buf.is_32bit = (mesa_ind_buf->type == GL_UNSIGNED_INT);
+		r300->ind_buf.count = mesa_ind_buf->count;
+
+		if (mapped_named_bo) {
+			ctx->Driver.UnmapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, mesa_ind_buf->obj);
+		}
+	} else {
+		r300FixupIndexBuffer(ctx, mesa_ind_buf);
 	}
 }
 
@@ -161,26 +203,118 @@ static int getTypeSize(GLenum type)
 	}						\
 } while (0)
 
-static void r300TranslateAttrib(GLcontext *ctx, GLuint attr, int count, const struct gl_client_array *input, struct gl_buffer_object **bo, GLuint *nr_bo)
+/**
+ * Convert attribute data type to float
+ * If the attribute uses named buffer object replace the bo with newly allocated bo
+ */
+static void r300ConvertAttrib(GLcontext *ctx, int count, const struct gl_client_array *input, struct vertex_attribute *attr)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-	struct r300_vertex_buffer *vbuf = &r300->vbuf;
-	struct vertex_attribute r300_attr;
-	const void *src_ptr;
-	GLenum type;
+	const GLvoid *src_ptr;
+	GLboolean mapped_named_bo = GL_FALSE;
+	GLfloat *dst_ptr;
 	GLuint stride;
 
+	stride = (input->StrideB == 0) ? getTypeSize(input->Type) * input->Size : input->StrideB;
+
+	/* Convert value for first element only */
+	if (input->StrideB == 0)
+		count = 1;
+
 	if (input->BufferObj->Name) {
 		if (!input->BufferObj->Pointer) {
-			bo[*nr_bo] = input->BufferObj;
-			(*nr_bo)++;
 			ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY_ARB, input->BufferObj);
-			assert(input->BufferObj->Pointer != NULL);
+			mapped_named_bo = GL_TRUE;
 		}
 
 		src_ptr = ADD_POINTERS(input->BufferObj->Pointer, input->Ptr);
-	} else
+	} else {
 		src_ptr = input->Ptr;
+	}
+
+	radeonAllocDmaRegion(&r300->radeon, &attr->bo, &attr->bo_offset, sizeof(GLfloat) * input->Size * count, 32);
+	dst_ptr = (GLfloat *)ADD_POINTERS(attr->bo->ptr, attr->bo_offset);
+
+	if (RADEON_DEBUG & DEBUG_FALLBACKS) {
+		fprintf(stderr, "%s: Converting vertex attributes, attribute data format %x,", __FUNCTION__, input->Type);
+		fprintf(stderr, "stride %d, components %d\n", stride, input->Size);
+	}
+
+	assert(src_ptr != NULL);
+
+	switch (input->Type) {
+		case GL_DOUBLE:
+			CONVERT(GLdouble, (GLfloat));
+			break;
+		case GL_UNSIGNED_INT:
+			CONVERT(GLuint, UINT_TO_FLOAT);
+			break;
+		case GL_INT:
+			CONVERT(GLint, INT_TO_FLOAT);
+			break;
+		case GL_UNSIGNED_SHORT:
+			CONVERT(GLushort, USHORT_TO_FLOAT);
+			break;
+		case GL_SHORT:
+			CONVERT(GLshort, SHORT_TO_FLOAT);
+			break;
+		case GL_UNSIGNED_BYTE:
+			assert(input->Format != GL_BGRA);
+			CONVERT(GLubyte, UBYTE_TO_FLOAT);
+			break;
+		case GL_BYTE:
+			CONVERT(GLbyte, BYTE_TO_FLOAT);
+			break;
+		default:
+			assert(0);
+			break;
+	}
+
+	if (mapped_named_bo) {
+		ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj);
+	}
+}
+
+static void r300AlignDataToDword(GLcontext *ctx, const struct gl_client_array *input, int count, struct vertex_attribute *attr)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	const int dst_stride = (input->StrideB + 3) & ~3;
+	const int size = getTypeSize(input->Type) * input->Size * count;
+	GLboolean mapped_named_bo = GL_FALSE;
+
+	radeonAllocDmaRegion(&r300->radeon, &attr->bo, &attr->bo_offset, size, 32);
+
+	if (!input->BufferObj->Pointer) {
+		ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY_ARB, input->BufferObj);
+		mapped_named_bo = GL_TRUE;
+	}
+
+	{
+		GLvoid *src_ptr = ADD_POINTERS(input->BufferObj->Pointer, input->Ptr);
+		GLvoid *dst_ptr = ADD_POINTERS(attr->bo->ptr, attr->bo_offset);
+		int i;
+
+		for (i = 0; i < count; ++i) {
+			_mesa_memcpy(dst_ptr, src_ptr, input->StrideB);
+			src_ptr += input->StrideB;
+			dst_ptr += dst_stride;
+		}
+	}
+
+	if (mapped_named_bo) {
+		ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER, input->BufferObj);
+	}
+
+	attr->stride = dst_stride;
+}
+
+static void r300TranslateAttrib(GLcontext *ctx, GLuint attr, int count, const struct gl_client_array *input)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_vertex_buffer *vbuf = &r300->vbuf;
+	struct vertex_attribute r300_attr;
+	GLenum type;
+	GLuint stride;
 
 	stride = (input->StrideB == 0) ? getTypeSize(input->Type) * input->Size : input->StrideB;
 
@@ -189,62 +323,57 @@ static void r300TranslateAttrib(GLcontext *ctx, GLuint attr, int count, const st
 	    getTypeSize(input->Type) != 4 ||
 #endif
 	    stride < 4) {
-		if (RADEON_DEBUG & DEBUG_FALLBACKS) {
-			fprintf(stderr, "%s: Converting vertex attributes, attribute data format %x,", __FUNCTION__, input->Type);
-			fprintf(stderr, "stride %d, components %d\n", stride, input->Size);
-		}
-
-		GLfloat *dst_ptr, *tmp;
-
-		/* Convert value for first element only */
-		if (input->StrideB == 0)
-			count = 1;
-
-		tmp = dst_ptr = _mesa_malloc(sizeof(GLfloat) * input->Size * count);
-
-		switch (input->Type) {
-			case GL_DOUBLE:
-				CONVERT(GLdouble, (GLfloat));
-				break;
-			case GL_UNSIGNED_INT:
-				CONVERT(GLuint, UINT_TO_FLOAT);
-				break;
-			case GL_INT:
-				CONVERT(GLint, INT_TO_FLOAT);
-				break;
-			case GL_UNSIGNED_SHORT:
-				CONVERT(GLushort, USHORT_TO_FLOAT);
-				break;
-			case GL_SHORT:
-				CONVERT(GLshort, SHORT_TO_FLOAT);
-				break;
-			case GL_UNSIGNED_BYTE:
-				assert(input->Format != GL_BGRA);
-				CONVERT(GLubyte, UBYTE_TO_FLOAT);
-				break;
-			case GL_BYTE:
-				CONVERT(GLbyte, BYTE_TO_FLOAT);
-				break;
-			default:
-				assert(0);
-				break;
-		}
 
 		type = GL_FLOAT;
-		r300_attr.free_needed = GL_TRUE;
-		r300_attr.data = tmp;
+
+		r300ConvertAttrib(ctx, count, input, &r300_attr);
 		if (input->StrideB == 0) {
 			r300_attr.stride = 0;
 		} else {
 			r300_attr.stride = sizeof(GLfloat) * input->Size;
 		}
 		r300_attr.dwords = input->Size;
+		r300_attr.is_named_bo = GL_FALSE;
 	} else {
 		type = input->Type;
-		r300_attr.free_needed = GL_FALSE;
-		r300_attr.data = (GLvoid *)src_ptr;
-		r300_attr.stride = input->StrideB;
-		r300_attr.dwords = (getTypeSize(type) * input->Size  + 3)/ 4;
+		r300_attr.dwords = (getTypeSize(type) * input->Size + 3)/ 4;
+		if (input->BufferObj->Name) {
+			if (stride % 4 != 0) {
+				assert(((int) input->Ptr) % input->StrideB == 0);
+				r300AlignDataToDword(ctx, input, count, &r300_attr);
+				r300_attr.is_named_bo = GL_FALSE;
+			} else {
+				r300_attr.stride = input->StrideB;
+				r300_attr.bo_offset = (GLuint) input->Ptr;
+				r300_attr.bo = get_radeon_buffer_object(input->BufferObj)->bo;
+				r300_attr.is_named_bo = GL_TRUE;
+			}
+		} else {
+			int size;
+			uint32_t *dst;
+
+			if (input->StrideB == 0) {
+				size = getTypeSize(input->Type) * input->Size;
+				count = 1;
+				r300_attr.stride = 0;
+			} else {
+				size = getTypeSize(input->Type) * input->Size * count;
+				r300_attr.stride = (getTypeSize(type) * input->Size + 3) & ~3;
+			}
+
+			radeonAllocDmaRegion(&r300->radeon, &r300_attr.bo, &r300_attr.bo_offset, size, 32);
+			assert(r300_attr.bo->ptr != NULL);
+			dst = (uint32_t *)ADD_POINTERS(r300_attr.bo->ptr, r300_attr.bo_offset);
+			switch (r300_attr.dwords) {
+				case 1: radeonEmitVec4(dst, input->Ptr, input->StrideB, count); break;
+				case 2: radeonEmitVec8(dst, input->Ptr, input->StrideB, count); break;
+				case 3: radeonEmitVec12(dst, input->Ptr, input->StrideB, count); break;
+				case 4: radeonEmitVec16(dst, input->Ptr, input->StrideB, count); break;
+				default: assert(0); break;
+			}
+
+			r300_attr.is_named_bo = GL_FALSE;
+		}
 	}
 
 	r300_attr.size = input->Size;
@@ -333,7 +462,7 @@ static void r300TranslateAttrib(GLcontext *ctx, GLuint attr, int count, const st
 	++vbuf->num_attribs;
 }
 
-static void r300SetVertexFormat(GLcontext *ctx, const struct gl_client_array *arrays[], int count, struct gl_buffer_object **bo, GLuint *nr_bo)
+static void r300SetVertexFormat(GLcontext *ctx, const struct gl_client_array *arrays[], int count)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	struct r300_vertex_buffer *vbuf = &r300->vbuf;
@@ -351,7 +480,7 @@ static void r300SetVertexFormat(GLcontext *ctx, const struct gl_client_array *ar
 				++i;
 			}
 
-			r300TranslateAttrib(ctx, i, count, arrays[i], bo, nr_bo);
+			r300TranslateAttrib(ctx, i, count, arrays[i]);
 
 			tmp >>= 1;
 			++i;
@@ -366,38 +495,54 @@ static void r300SetVertexFormat(GLcontext *ctx, const struct gl_client_array *ar
 		int i;
 
 		for (i = 0; i < vbuf->num_attribs; i++) {
-			rcommon_emit_vector(ctx, &r300->radeon.tcl.aos[i],
-						vbuf->attribs[i].data, vbuf->attribs[i].dwords,
-						vbuf->attribs[i].stride, count);
+			struct radeon_aos *aos = &r300->radeon.tcl.aos[i];
+
+			aos->count = vbuf->attribs[i].stride == 0 ? 1 : count;
+			aos->stride = vbuf->attribs[i].stride / sizeof(float);
+			aos->offset = vbuf->attribs[i].bo_offset;
+			aos->components = vbuf->attribs[i].dwords;
+			aos->bo = vbuf->attribs[i].bo;
+
+			radeon_cs_space_check_with_bo(r300->radeon.cmdbuf.cs,
+										  r300->vbuf.attribs[i].bo,
+										  RADEON_GEM_DOMAIN_GTT, 0);
+			if (vbuf->attribs[i].is_named_bo) {
+				radeon_cs_space_add_persistent_bo(r300->radeon.cmdbuf.cs,
+												  r300->vbuf.attribs[i].bo,
+												  RADEON_GEM_DOMAIN_GTT, 0);
+			}
 		}
-
 		r300->radeon.tcl.aos_count = vbuf->num_attribs;
+
+		if (r300->ind_buf.bo) {
+			radeon_cs_space_check_with_bo(r300->radeon.cmdbuf.cs,
+										  r300->ind_buf.bo,
+										  RADEON_GEM_DOMAIN_GTT, 0);
+		}
 	}
 }
 
-static void r300FreeData(GLcontext *ctx, struct gl_buffer_object **bo, GLuint nr_bo)
+static void r300FreeData(GLcontext *ctx)
 {
+	/* Need to zero tcl.aos[n].bo and tcl.elt_dma_bo
+	 * to prevent double unref in radeonReleaseArrays
+	 * called during context destroy
+	 */
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	{
-		struct r300_vertex_buffer *vbuf = &R300_CONTEXT(ctx)->vbuf;
 		int i;
 
-		for (i = 0; i < vbuf->num_attribs; i++) {
-			if (vbuf->attribs[i].free_needed)
-				_mesa_free(vbuf->attribs[i].data);
+		for (i = 0; i < r300->vbuf.num_attribs; i++) {
+			if (!r300->vbuf.attribs[i].is_named_bo) {
+				radeon_bo_unref(r300->vbuf.attribs[i].bo);
+			}
+			r300->radeon.tcl.aos[i].bo = NULL;
 		}
 	}
 
 	{
-		struct r300_index_buffer *ind_buf = &R300_CONTEXT(ctx)->ind_buf;
-		if (ind_buf->free_needed)
-			_mesa_free(ind_buf->ptr);
-	}
-
-	{
-		int i;
-
-		for (i = 0; i < nr_bo; ++i) {
-			ctx->Driver.UnmapBuffer(ctx, 0, bo[i]);
+		if (r300->ind_buf.bo != NULL) {
+			radeon_bo_unref(r300->ind_buf.bo);
 		}
 	}
 }
@@ -411,8 +556,7 @@ static GLboolean r300TryDrawPrims(GLcontext *ctx,
 					 GLuint max_index )
 {
 	struct r300_context *r300 = R300_CONTEXT(ctx);
-	struct gl_buffer_object *bo[VERT_ATTRIB_MAX+1];
-	GLuint i, nr_bo = 0;
+	GLuint i;
 
 	if (ctx->NewState)
 		_mesa_update_state( ctx );
@@ -424,7 +568,7 @@ static GLboolean r300TryDrawPrims(GLcontext *ctx,
 
 	r300SwitchFallback(ctx, R300_FALLBACK_INVALID_BUFFERS, !r300ValidateBuffers(ctx));
 
-	r300FixupIndexBuffer(ctx, ib, bo, &nr_bo);
+	r300SetupIndexBuffer(ctx, ib);
 
 	/* ensure we have the cmd buf space in advance to cover
  	 * the state + DMA AOS pointers */
@@ -432,7 +576,7 @@ static GLboolean r300TryDrawPrims(GLcontext *ctx,
                            r300->radeon.hw.max_state_size + (50*sizeof(int)),
                            __FUNCTION__);
 
-	r300SetVertexFormat(ctx, arrays, max_index + 1, bo, &nr_bo);
+	r300SetVertexFormat(ctx, arrays, max_index + 1);
 
 	if (r300->fallback)
 		return GL_FALSE;
@@ -450,9 +594,7 @@ static GLboolean r300TryDrawPrims(GLcontext *ctx,
 
 	r300EmitCacheFlush(r300);
 
-	radeonReleaseArrays(ctx, ~0);
-
-	r300FreeData(ctx, bo, nr_bo);
+	r300FreeData(ctx);
 
 	return GL_TRUE;
 }
@@ -466,31 +608,19 @@ static void r300DrawPrims(GLcontext *ctx,
 			 GLuint min_index,
 			 GLuint max_index)
 {
-	struct split_limits limits;
 	GLboolean retval;
 
-	if (ib)
-		limits.max_verts = 0xffffffff;
-	else
-		limits.max_verts = 65535;
-
-	limits.max_indices = 65535;
-	limits.max_vb_size = 1024*1024;
-
 	/* This check should get folded into just the places that
 	 * min/max index are really needed.
 	 */
-	if (!index_bounds_valid)
-	   vbo_get_minmax_index(ctx, prim, ib, &min_index, &max_index);
+	if (!index_bounds_valid) {
+		vbo_get_minmax_index(ctx, prim, ib, &min_index, &max_index);
+	}
 
 	if (min_index) {
 		vbo_rebase_prims( ctx, arrays, prim, nr_prims, ib, min_index, max_index, r300DrawPrims );
 		return;
 	}
-	if ((ib && ib->count > 65535)) {
-		vbo_split_prims (ctx, arrays, prim, nr_prims, ib, min_index, max_index, r300DrawPrims, &limits);
-		return;
-	}
 
 	/* Make an attempt at drawing */
 	retval = r300TryDrawPrims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
diff --git a/src/mesa/drivers/dri/r300/r300_emit.c b/src/mesa/drivers/dri/r300/r300_emit.c
index feb3370f37..07e6223087 100644
--- a/src/mesa/drivers/dri/r300/r300_emit.c
+++ b/src/mesa/drivers/dri/r300/r300_emit.c
@@ -124,41 +124,6 @@ GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint vp_writes)
 	return ret;
 }
 
-GLboolean r300EmitArrays(GLcontext * ctx)
-{
-	r300ContextPtr r300 = R300_CONTEXT(ctx);
-	struct r300_vertex_buffer *vbuf = &r300->vbuf;
-	GLuint InputsRead, OutputsWritten;
-
-	r300ChooseSwtclVertexFormat(ctx, &InputsRead, &OutputsWritten);
-
-	r300SwitchFallback(ctx, R300_FALLBACK_AOS_LIMIT, vbuf->num_attribs > R300_MAX_AOS_ARRAYS);
-	if (r300->fallback & R300_RASTER_FALLBACK_MASK)
-		return GL_FALSE;
-
-	{
-		struct vertex_buffer *mesa_vb = &TNL_CONTEXT(ctx)->vb;
-		GLuint attr, i;
-
-		for (i = 0; i < vbuf->num_attribs; i++) {
-			attr = vbuf->attribs[i].element;
-			rcommon_emit_vector(ctx, &r300->radeon.tcl.aos[i], mesa_vb->AttribPtr[attr]->data,
-					mesa_vb->AttribPtr[attr]->size, mesa_vb->AttribPtr[attr]->stride, mesa_vb->Count);
-		}
-
-		r300->radeon.tcl.aos_count = vbuf->num_attribs;
-
-		/* Fill index buffer info */
-		r300->ind_buf.ptr = mesa_vb->Elts;
-		r300->ind_buf.is_32bit = GL_TRUE;
-		r300->ind_buf.free_needed = GL_FALSE;
-	}
-
-	r300SetupVAP(ctx, InputsRead, OutputsWritten);
-
-	return GL_TRUE;
-}
-
 void r300EmitCacheFlush(r300ContextPtr rmesa)
 {
 	BATCH_LOCALS(&rmesa->radeon);
diff --git a/src/mesa/drivers/dri/r300/r300_emit.h b/src/mesa/drivers/dri/r300/r300_emit.h
index 3f8c60ffae..8e57e354d1 100644
--- a/src/mesa/drivers/dri/r300/r300_emit.h
+++ b/src/mesa/drivers/dri/r300/r300_emit.h
@@ -104,7 +104,7 @@ static INLINE uint32_t cmdpacket3(struct radeon_screen *rscrn, int packet)
 	return cmd.u;
 }
 
-static INLINE uint32_t cmdcpdelay(struct radeon_screen *rscrn,  
+static INLINE uint32_t cmdcpdelay(struct radeon_screen *rscrn,
                                   unsigned short count)
 {
 	drm_r300_cmd_header_t cmd;
@@ -216,8 +216,6 @@ void static INLINE cp_wait(radeonContextPtr radeon, unsigned char flags)
 	}
 }
 
-extern GLboolean r300EmitArrays(GLcontext * ctx);
-
 extern int r300PrimitiveType(r300ContextPtr rmesa, int prim);
 extern int r300NumVerts(r300ContextPtr rmesa, int num_verts, int prim);
 
diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c
index 08d67b73ed..26953cd9d1 100644
--- a/src/mesa/drivers/dri/r300/r300_render.c
+++ b/src/mesa/drivers/dri/r300/r300_render.c
@@ -172,64 +172,42 @@ int r300NumVerts(r300ContextPtr rmesa, int num_verts, int prim)
 	return num_verts - verts_off;
 }
 
-static void r300EmitElts(GLcontext * ctx, unsigned long n_elts)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	void *out;
-	GLuint size;
-
-	size = ((rmesa->ind_buf.is_32bit ? 4 : 2) * n_elts + 3) & ~3;
-
-	radeonAllocDmaRegion(&rmesa->radeon, &rmesa->radeon.tcl.elt_dma_bo,
-			     &rmesa->radeon.tcl.elt_dma_offset, size, 4);
-	radeon_bo_map(rmesa->radeon.tcl.elt_dma_bo, 1);
-	out = rmesa->radeon.tcl.elt_dma_bo->ptr + rmesa->radeon.tcl.elt_dma_offset;
-	memcpy(out, rmesa->ind_buf.ptr, size);
-	radeon_bo_unmap(rmesa->radeon.tcl.elt_dma_bo);
-}
-
 static void r300FireEB(r300ContextPtr rmesa, int vertex_count, int type)
 {
 	BATCH_LOCALS(&rmesa->radeon);
+	int size;
 
-    r300_emit_scissor(rmesa->radeon.glCtx);
-	if (vertex_count > 0) {
-		int size;
-
-		BEGIN_BATCH(10);
-		OUT_BATCH_PACKET3(R300_PACKET3_3D_DRAW_INDX_2, 0);
-		if (rmesa->ind_buf.is_32bit) {
-			size = vertex_count;
-			OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_INDICES |
-			  ((vertex_count + 0) << 16) | type |
-			  R300_VAP_VF_CNTL__INDEX_SIZE_32bit);
-		} else {
-			size = (vertex_count + 1) >> 1;
-			OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_INDICES |
-			   ((vertex_count + 0) << 16) | type);
-		}
+	r300_emit_scissor(rmesa->radeon.glCtx);
 
-		if (!rmesa->radeon.radeonScreen->kernel_mm) {
-			OUT_BATCH_PACKET3(R300_PACKET3_INDX_BUFFER, 2);
-			OUT_BATCH(R300_INDX_BUFFER_ONE_REG_WR | (0 << R300_INDX_BUFFER_SKIP_SHIFT) |
-	    			 (R300_VAP_PORT_IDX0 >> 2));
-			OUT_BATCH_RELOC(rmesa->radeon.tcl.elt_dma_offset,
-					rmesa->radeon.tcl.elt_dma_bo,
-					rmesa->radeon.tcl.elt_dma_offset,
-					RADEON_GEM_DOMAIN_GTT, 0, 0);
-			OUT_BATCH(size);
-		} else {
-			OUT_BATCH_PACKET3(R300_PACKET3_INDX_BUFFER, 2);
-			OUT_BATCH(R300_INDX_BUFFER_ONE_REG_WR | (0 << R300_INDX_BUFFER_SKIP_SHIFT) |
-	    			 (R300_VAP_PORT_IDX0 >> 2));
-			OUT_BATCH(rmesa->radeon.tcl.elt_dma_offset);
-			OUT_BATCH(size);
-			radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
-					      rmesa->radeon.tcl.elt_dma_bo,
-					      RADEON_GEM_DOMAIN_GTT, 0, 0);
-		}
-		END_BATCH();
+	BEGIN_BATCH(10);
+	OUT_BATCH_PACKET3(R300_PACKET3_3D_DRAW_INDX_2, 0);
+	if (rmesa->ind_buf.is_32bit) {
+		size = vertex_count;
+		OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_INDICES |
+		  (vertex_count << 16) | type |
+		  R300_VAP_VF_CNTL__INDEX_SIZE_32bit);
+	} else {
+		size = (vertex_count + 1) >> 1;
+		OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_INDICES |
+		   (vertex_count << 16) | type);
 	}
+
+	if (!rmesa->radeon.radeonScreen->kernel_mm) {
+		OUT_BATCH_PACKET3(R300_PACKET3_INDX_BUFFER, 2);
+		OUT_BATCH(R300_INDX_BUFFER_ONE_REG_WR | (0 << R300_INDX_BUFFER_SKIP_SHIFT) |
+				 (R300_VAP_PORT_IDX0 >> 2));
+		OUT_BATCH_RELOC(0, rmesa->ind_buf.bo, rmesa->ind_buf.bo_offset, RADEON_GEM_DOMAIN_GTT, 0, 0);
+		OUT_BATCH(size);
+	} else {
+		OUT_BATCH_PACKET3(R300_PACKET3_INDX_BUFFER, 2);
+		OUT_BATCH(R300_INDX_BUFFER_ONE_REG_WR | (0 << R300_INDX_BUFFER_SKIP_SHIFT) |
+				 (R300_VAP_PORT_IDX0 >> 2));
+		OUT_BATCH(rmesa->ind_buf.bo_offset);
+		OUT_BATCH(size);
+		radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
+				      rmesa->ind_buf.bo, RADEON_GEM_DOMAIN_GTT, 0, 0);
+	}
+	END_BATCH();
 }
 
 static void r300EmitAOS(r300ContextPtr rmesa, GLuint nr, GLuint offset)
@@ -359,14 +337,18 @@ void r300RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim)
 	if (type < 0 || num_verts <= 0)
 		return;
 
+	if (num_verts > 65535) {
+		WARN_ONCE("Can't handle more then 65535 vertices at once\n");
+		return;
+	}
+
 	/* Make space for at least 128 dwords.
 	 * This is supposed to ensure that we can get all rendering
 	 * commands into a single command buffer.
 	 */
 	rcommonEnsureCmdBufSpace(&rmesa->radeon, 128, __FUNCTION__);
 
-	if (rmesa->ind_buf.ptr) {
-		r300EmitElts(ctx, num_verts);
+	if (rmesa->ind_buf.bo) {
 		r300EmitAOS(rmesa, rmesa->radeon.tcl.aos_count, 0);
 		if (rmesa->radeon.radeonScreen->kernel_mm) {
 			BEGIN_BATCH_NO_AUTOSTATE(2);
@@ -382,37 +364,6 @@ void r300RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim)
 	COMMIT_BATCH();
 }
 
-static void r300RunRender(GLcontext * ctx, struct tnl_pipeline_stage *stage)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	int i;
-	TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *vb = &tnl->vb;
-
-	if (RADEON_DEBUG & DEBUG_PRIMS)
-		fprintf(stderr, "%s\n", __FUNCTION__);
-
-	r300UpdateShaders(rmesa);
-	r300EmitArrays(ctx);
-
-	r300UpdateShaderStates(rmesa);
-
-	r300EmitCacheFlush(rmesa);
-	radeonEmitState(&rmesa->radeon);
-
-	for (i = 0; i < vb->PrimitiveCount; i++) {
-		GLuint prim = _tnl_translate_prim(&vb->Primitive[i]);
-		GLuint start = vb->Primitive[i].start;
-		GLuint end = vb->Primitive[i].start + vb->Primitive[i].count;
-		r300RunRenderPrimitive(ctx, start, end, prim);
-	}
-
-	r300EmitCacheFlush(rmesa);
-
-	radeonReleaseArrays(ctx, ~0);
-}
-
-
 static const char *getFallbackString(uint32_t bit)
 {
 	switch (bit) {
@@ -449,7 +400,7 @@ void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode)
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	uint32_t old_fallback = rmesa->fallback;
 	static uint32_t fallback_warn = 0;
-	
+
 	if (mode) {
 		if ((fallback_warn & bit) == 0) {
 			if (RADEON_DEBUG & DEBUG_FALLBACKS)
@@ -470,7 +421,7 @@ void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode)
 		/* update only if we change from no raster fallbacks to some raster fallbacks */
 		if (((old_fallback & R300_RASTER_FALLBACK_MASK) == 0) &&
 			((bit & R300_RASTER_FALLBACK_MASK) > 0)) {
-			
+
 			radeon_firevertices(&rmesa->radeon);
 			rmesa->radeon.swtcl.RenderIndex = ~0;
 			_swsetup_Wakeup( ctx );
@@ -489,7 +440,7 @@ void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode)
 		/* update only if we have disabled all raster fallbacks */
 		if ((old_fallback & R300_RASTER_FALLBACK_MASK) == bit) {
 			_swrast_flush( ctx );
-			
+
 			tnl->Driver.Render.Start = r300RenderStart;
 			tnl->Driver.Render.Finish = r300RenderFinish;
 			tnl->Driver.Render.PrimitiveNotify = r300RenderPrimitive;
@@ -497,38 +448,10 @@ void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode)
 			tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
 			tnl->Driver.Render.CopyPV = _tnl_copy_pv;
 			tnl->Driver.Render.Interp = _tnl_interp;
-			
+
 			_tnl_invalidate_vertex_state( ctx, ~0 );
 			_tnl_invalidate_vertices( ctx, ~0 );
 		}
 	}
-	
-}
-
-static GLboolean r300RunNonTCLRender(GLcontext * ctx,
-				     struct tnl_pipeline_stage *stage)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-
-	if (RADEON_DEBUG & DEBUG_PRIMS)
-		fprintf(stderr, "%s\n", __FUNCTION__);
-
-	if (rmesa->fallback & R300_RASTER_FALLBACK_MASK)
-		return GL_TRUE;
 
-	if (rmesa->options.hw_tcl_enabled == GL_FALSE)
-		return GL_TRUE;
-
-	r300RunRender(ctx, stage);
-
-	return GL_FALSE;
 }
-
-const struct tnl_pipeline_stage _r300_render_stage = {
-	"r300 Hardware Rasterization",
-	NULL,
-	NULL,
-	NULL,
-	NULL,
-	r300RunNonTCLRender
-};
diff --git a/src/mesa/drivers/dri/r300/radeon_buffer_objects.c b/src/mesa/drivers/dri/r300/radeon_buffer_objects.c
new file mode 120000
index 0000000000..f6a5f66470
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_buffer_objects.c
@@ -0,0 +1 @@
+../radeon/radeon_buffer_objects.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/r300/radeon_buffer_objects.h b/src/mesa/drivers/dri/r300/radeon_buffer_objects.h
new file mode 120000
index 0000000000..2f134fd17b
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/radeon_buffer_objects.h
@@ -0,0 +1 @@
+../radeon/radeon_buffer_objects.h
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/radeon/radeon_buffer_objects.c b/src/mesa/drivers/dri/radeon/radeon_buffer_objects.c
new file mode 100644
index 0000000000..69556923b6
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_buffer_objects.c
@@ -0,0 +1,226 @@
+/*
+ * Copyright 2009 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * 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 THE COPYRIGHT OWNER(S) 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.
+ *
+ */
+
+#include "radeon_buffer_objects.h"
+
+#include "main/imports.h"
+#include "main/mtypes.h"
+#include "main/bufferobj.h"
+
+#include "radeon_common.h"
+
+struct radeon_buffer_object *
+get_radeon_buffer_object(struct gl_buffer_object *obj)
+{
+    return (struct radeon_buffer_object *) obj;
+}
+
+static struct gl_buffer_object *
+radeonNewBufferObject(GLcontext * ctx,
+                      GLuint name,
+                      GLenum target)
+{
+    struct radeon_buffer_object *obj = CALLOC_STRUCT(radeon_buffer_object);
+
+    _mesa_initialize_buffer_object(&obj->Base, name, target);
+
+    obj->bo = NULL;
+
+    return &obj->Base;
+}
+
+/**
+ * Called via glDeleteBuffersARB().
+ */
+static void
+radeonDeleteBufferObject(GLcontext * ctx,
+                         struct gl_buffer_object *obj)
+{
+    struct radeon_buffer_object *radeon_obj = get_radeon_buffer_object(obj);
+
+    if (obj->Pointer) {
+        radeon_bo_unmap(radeon_obj->bo);
+    }
+
+    if (radeon_obj->bo) {
+        radeon_bo_unref(radeon_obj->bo);
+    }
+
+    _mesa_free(radeon_obj);
+}
+
+
+/**
+ * Allocate space for and store data in a buffer object.  Any data that was
+ * previously stored in the buffer object is lost.  If data is NULL,
+ * memory will be allocated, but no copy will occur.
+ * Called via glBufferDataARB().
+ */
+static void
+radeonBufferData(GLcontext * ctx,
+                 GLenum target,
+                 GLsizeiptrARB size,
+                 const GLvoid * data,
+                 GLenum usage,
+                 struct gl_buffer_object *obj)
+{
+    radeonContextPtr radeon = RADEON_CONTEXT(ctx);
+    struct radeon_buffer_object *radeon_obj = get_radeon_buffer_object(obj);
+
+    radeon_obj->Base.Size = size;
+    radeon_obj->Base.Usage = usage;
+
+    if (radeon_obj->bo != NULL) {
+        radeon_bo_unref(radeon_obj->bo);
+        radeon_obj->bo = NULL;
+    }
+
+    if (size != 0) {
+#ifdef RADEON_DEBUG_BO
+        radeon_obj->bo = radeon_bo_open(radeon->radeonScreen->bom,
+                                        0,
+                                        size,
+                                        32,
+                                        RADEON_GEM_DOMAIN_GTT,
+                                        0,
+                                        "Radeon Named object");
+#else
+        radeon_obj->bo = radeon_bo_open(radeon->radeonScreen->bom,
+                                        0,
+                                        size,
+                                        32,
+                                        RADEON_GEM_DOMAIN_GTT,
+                                        0);
+#endif
+        if (data != NULL) {
+            radeon_bo_map(radeon_obj->bo, GL_TRUE);
+
+            _mesa_memcpy(radeon_obj->bo->ptr, data, size);
+
+            radeon_bo_unmap(radeon_obj->bo);
+        }
+    }
+}
+
+/**
+ * Replace data in a subrange of buffer object.  If the data range
+ * specified by size + offset extends beyond the end of the buffer or
+ * if data is NULL, no copy is performed.
+ * Called via glBufferSubDataARB().
+ */
+static void
+radeonBufferSubData(GLcontext * ctx,
+                    GLenum target,
+                    GLintptrARB offset,
+                    GLsizeiptrARB size,
+                    const GLvoid * data,
+                    struct gl_buffer_object *obj)
+{
+    struct radeon_buffer_object *radeon_obj = get_radeon_buffer_object(obj);
+
+    radeon_bo_map(radeon_obj->bo, GL_TRUE);
+
+    _mesa_memcpy(radeon_obj->bo->ptr + offset, data, size);
+
+    radeon_bo_unmap(radeon_obj->bo);
+}
+
+/**
+ * Called via glGetBufferSubDataARB()
+ */
+static void
+radeonGetBufferSubData(GLcontext * ctx,
+                       GLenum target,
+                       GLintptrARB offset,
+                       GLsizeiptrARB size,
+                       GLvoid * data,
+                       struct gl_buffer_object *obj)
+{
+    struct radeon_buffer_object *radeon_obj = get_radeon_buffer_object(obj);
+
+    radeon_bo_map(radeon_obj->bo, GL_FALSE);
+
+    _mesa_memcpy(data, radeon_obj->bo->ptr + offset, size);
+
+    radeon_bo_unmap(radeon_obj->bo);
+}
+
+/**
+ * Called via glMapBufferARB()
+ */
+static void *
+radeonMapBuffer(GLcontext * ctx,
+                GLenum target,
+                GLenum access,
+                struct gl_buffer_object *obj)
+{
+    struct radeon_buffer_object *radeon_obj = get_radeon_buffer_object(obj);
+
+    if (access == GL_WRITE_ONLY_ARB) {
+        radeonFlush(ctx);
+    }
+
+    if (radeon_obj->bo == NULL) {
+        obj->Pointer = NULL;
+        return NULL;
+    }
+
+    radeon_bo_map(radeon_obj->bo, access == GL_WRITE_ONLY_ARB);
+
+    return obj->Pointer = radeon_obj->bo->ptr;
+}
+
+
+/**
+ * Called via glUnmapBufferARB()
+ */
+static GLboolean
+radeonUnmapBuffer(GLcontext * ctx,
+                  GLenum target,
+                  struct gl_buffer_object *obj)
+{
+    struct radeon_buffer_object *radeon_obj = get_radeon_buffer_object(obj);
+
+    if (radeon_obj->bo != NULL) {
+        radeon_bo_unmap(radeon_obj->bo);
+        obj->Pointer = NULL;
+    }
+
+    return GL_TRUE;
+}
+
+void
+radeonInitBufferObjectFuncs(struct dd_function_table *functions)
+{
+    functions->NewBufferObject = radeonNewBufferObject;
+    functions->DeleteBuffer = radeonDeleteBufferObject;
+    functions->BufferData = radeonBufferData;
+    functions->BufferSubData = radeonBufferSubData;
+    functions->GetBufferSubData = radeonGetBufferSubData;
+    functions->MapBuffer = radeonMapBuffer;
+    functions->UnmapBuffer = radeonUnmapBuffer;
+}
diff --git a/src/mesa/drivers/dri/radeon/radeon_buffer_objects.h b/src/mesa/drivers/dri/radeon/radeon_buffer_objects.h
new file mode 100644
index 0000000000..d681960825
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_buffer_objects.h
@@ -0,0 +1,52 @@
+/*
+ * Copyright 2009 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * 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 THE COPYRIGHT OWNER(S) 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 RADEON_BUFFER_OBJECTS_H
+#define RADEON_BUFFER_OBJECTS_H
+
+#include "main/mtypes.h"
+
+struct radeon_bo;
+
+/**
+ * Radeon vertex/pixel buffer object, derived from Mesa's gl_buffer_object.
+ */
+struct radeon_buffer_object
+{
+   struct gl_buffer_object Base;
+   struct radeon_bo *bo;
+};
+
+struct radeon_buffer_object *
+get_radeon_buffer_object(struct gl_buffer_object *obj);
+
+/**
+ * Hook the bufferobject implementation into mesa:
+ */
+void radeonInitBufferObjectFuncs(struct dd_function_table *functions);
+
+#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c
index 48114a0012..7144abbe00 100644
--- a/src/mesa/drivers/dri/radeon/radeon_dma.c
+++ b/src/mesa/drivers/dri/radeon/radeon_dma.c
@@ -52,7 +52,7 @@ do {						\
 } while (0)
 #endif
 
-static void radeonEmitVec4(uint32_t *out, GLvoid * data, int stride, int count)
+void radeonEmitVec4(uint32_t *out, const GLvoid * data, int stride, int count)
 {
 	int i;
 
@@ -70,7 +70,7 @@ static void radeonEmitVec4(uint32_t *out, GLvoid * data, int stride, int count)
 		}
 }
 
-void radeonEmitVec8(uint32_t *out, GLvoid * data, int stride, int count)
+void radeonEmitVec8(uint32_t *out, const GLvoid * data, int stride, int count)
 {
 	int i;
 
@@ -89,7 +89,7 @@ void radeonEmitVec8(uint32_t *out, GLvoid * data, int stride, int count)
 		}
 }
 
-void radeonEmitVec12(uint32_t *out, GLvoid * data, int stride, int count)
+void radeonEmitVec12(uint32_t *out, const GLvoid * data, int stride, int count)
 {
 	int i;
 
@@ -110,7 +110,7 @@ void radeonEmitVec12(uint32_t *out, GLvoid * data, int stride, int count)
 		}
 }
 
-static void radeonEmitVec16(uint32_t *out, GLvoid * data, int stride, int count)
+void radeonEmitVec16(uint32_t *out, const GLvoid * data, int stride, int count)
 {
 	int i;
 
@@ -132,7 +132,7 @@ static void radeonEmitVec16(uint32_t *out, GLvoid * data, int stride, int count)
 }
 
 void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos,
-			 GLvoid * data, int size, int stride, int count)
+			 const GLvoid * data, int size, int stride, int count)
 {
 	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
 	uint32_t *out;
diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.h b/src/mesa/drivers/dri/radeon/radeon_dma.h
index 06e388fc1d..55509ed00c 100644
--- a/src/mesa/drivers/dri/radeon/radeon_dma.h
+++ b/src/mesa/drivers/dri/radeon/radeon_dma.h
@@ -33,11 +33,13 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #ifndef RADEON_DMA_H
 #define RADEON_DMA_H
 
-void radeonEmitVec8(uint32_t *out, GLvoid * data, int stride, int count);
-void radeonEmitVec12(uint32_t *out, GLvoid * data, int stride, int count);
+void radeonEmitVec4(uint32_t *out, const GLvoid * data, int stride, int count);
+void radeonEmitVec8(uint32_t *out, const GLvoid * data, int stride, int count);
+void radeonEmitVec12(uint32_t *out, const GLvoid * data, int stride, int count);
+void radeonEmitVec16(uint32_t *out, const GLvoid * data, int stride, int count);
 
 void rcommon_emit_vector(GLcontext * ctx, struct radeon_aos *aos,
-			 GLvoid * data, int size, int stride, int count);
+			 const GLvoid * data, int size, int stride, int count);
 
 void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size);
 void radeonAllocDmaRegion(radeonContextPtr rmesa,