Merge branch 'radeon-fbo-hacking' into radeon-rewrite

26 files changed, 1311 insertions, 799 deletions

diff --git a/src/mesa/drivers/dri/r200/r200_context.c b/src/mesa/drivers/dri/r200/r200_context.c
index 10a6362afd..6fd0575898 100644
--- a/src/mesa/drivers/dri/r200/r200_context.c
+++ b/src/mesa/drivers/dri/r200/r200_context.c
@@ -72,6 +72,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define need_GL_EXT_blend_func_separate
 #define need_GL_NV_vertex_program
 #define need_GL_ARB_point_parameters
+#define need_GL_EXT_framebuffer_object
 #include "extension_helper.h"
 
 #define DRIVER_DATE	"20060602"
@@ -124,6 +125,7 @@ const struct dri_extension card_extensions[] =
     { "GL_EXT_blend_minmax",               GL_EXT_blend_minmax_functions },
     { "GL_EXT_blend_subtract",             NULL },
     { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
+    { "GL_EXT_packed_depth_stencil",	   NULL},
     { "GL_EXT_secondary_color",            GL_EXT_secondary_color_functions },
     { "GL_EXT_stencil_wrap",               NULL },
     { "GL_EXT_texture_edge_clamp",         NULL },
@@ -165,6 +167,11 @@ const struct dri_extension point_extensions[] = {
     { NULL,                                NULL }
 };
 
+const struct dri_extension mm_extensions[] = {
+  { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions },
+  { NULL, NULL }
+};
+
 extern const struct tnl_pipeline_stage _r200_render_stage;
 extern const struct tnl_pipeline_stage _r200_tcl_stage;
 
@@ -418,6 +425,9 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
    _math_matrix_set_identity( &rmesa->tmpmat );
 
    driInitExtensions( ctx, card_extensions, GL_TRUE );
+
+   if (rmesa->radeon.radeonScreen->kernel_mm)
+     driInitExtensions(ctx, mm_extensions, GL_FALSE);
    if (!(rmesa->radeon.radeonScreen->chip_flags & R200_CHIPSET_YCBCR_BROKEN)) {
      /* yuv textures don't work with some chips - R200 / rv280 okay so far
 	others get the bit ordering right but don't actually do YUV-RGB conversion */
@@ -453,6 +463,7 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
 #endif
    /* plug in a few more device driver functions */
    /* XXX these should really go right after _mesa_init_driver_functions() */
+   radeon_fbo_init(&rmesa->radeon);
    radeonInitSpanFuncs( ctx );
    r200InitPixelFuncs( ctx );
    r200InitTnlFuncs( ctx );
diff --git a/src/mesa/drivers/dri/r200/r200_context.h b/src/mesa/drivers/dri/r200/r200_context.h
index f7bad2a241..fcbe725d6f 100644
--- a/src/mesa/drivers/dri/r200/r200_context.h
+++ b/src/mesa/drivers/dri/r200/r200_context.h
@@ -621,21 +621,6 @@ struct r200_context {
    GLboolean texmicrotile;
 
   struct ati_fragment_shader *afs_loaded;
-
-  struct {
-      struct gl_fragment_program *bitmap_fp;
-      struct gl_vertex_program *passthrough_vp;
-
-      struct gl_fragment_program *saved_fp;
-      GLboolean saved_fp_enable;
-      struct gl_vertex_program *saved_vp;
-      GLboolean saved_vp_enable;
-
-      GLint saved_vp_x, saved_vp_y;
-      GLsizei saved_vp_width, saved_vp_height;
-      GLenum saved_matrix_mode;
-   } meta;
-
 };
 
 #define R200_CONTEXT(ctx)		((r200ContextPtr)(ctx->DriverCtx))
diff --git a/src/mesa/drivers/dri/r200/r200_ioctl.c b/src/mesa/drivers/dri/r200/r200_ioctl.c
index 96ed49665b..ccb56202f6 100644
--- a/src/mesa/drivers/dri/r200/r200_ioctl.c
+++ b/src/mesa/drivers/dri/r200/r200_ioctl.c
@@ -41,19 +41,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "main/context.h"
 #include "swrast/swrast.h"
 
-#include "main/blend.h"
-#include "main/bufferobj.h"
-#include "main/buffers.h"
-#include "main/depth.h"
-#include "main/shaders.h"
-#include "main/texstate.h"
-#include "main/varray.h"
-#include "glapi/dispatch.h"
-#include "swrast/swrast.h"
-#include "main/stencil.h"
-#include "main/matrix.h"
-#include "main/attrib.h"
-#include "main/enable.h"
+
 
 #include "radeon_common.h"
 #include "radeon_lock.h"
@@ -70,217 +58,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define R200_TIMEOUT             512
 #define R200_IDLE_RETRY           16
 
-static void
-r200_meta_set_passthrough_transform(r200ContextPtr r200)
-{
-   GLcontext *ctx = r200->radeon.glCtx;
-
-   r200->meta.saved_vp_x = ctx->Viewport.X;
-   r200->meta.saved_vp_y = ctx->Viewport.Y;
-   r200->meta.saved_vp_width = ctx->Viewport.Width;
-   r200->meta.saved_vp_height = ctx->Viewport.Height;
-   r200->meta.saved_matrix_mode = ctx->Transform.MatrixMode;
-
-   _mesa_Viewport(0, 0, ctx->DrawBuffer->Width, ctx->DrawBuffer->Height);
-
-   _mesa_MatrixMode(GL_PROJECTION);
-   _mesa_PushMatrix();
-   _mesa_LoadIdentity();
-   _mesa_Ortho(0, ctx->DrawBuffer->Width, 0, ctx->DrawBuffer->Height, 1, -1);
-
-   _mesa_MatrixMode(GL_MODELVIEW);
-   _mesa_PushMatrix();
-   _mesa_LoadIdentity();
-}
-
-static void
-r200_meta_restore_transform(r200ContextPtr r200)
-{
-   _mesa_MatrixMode(GL_PROJECTION);
-   _mesa_PopMatrix();
-   _mesa_MatrixMode(GL_MODELVIEW);
-   _mesa_PopMatrix();
-
-   _mesa_MatrixMode(r200->meta.saved_matrix_mode);
-
-   _mesa_Viewport(r200->meta.saved_vp_x, r200->meta.saved_vp_y,
-		  r200->meta.saved_vp_width, r200->meta.saved_vp_height);
-}
-
-/**
- * Perform glClear where mask contains only color, depth, and/or stencil.
- *
- * The implementation is based on calling into Mesa to set GL state and
- * performing normal triangle rendering.  The intent of this path is to
- * have as generic a path as possible, so that any driver could make use of
- * it.
- */
-static void radeon_clear_tris(GLcontext *ctx, GLbitfield mask)
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat vertices[4][3];
-   GLfloat color[4][4];
-   GLfloat dst_z;
-   struct gl_framebuffer *fb = ctx->DrawBuffer;
-   int i;
-   GLboolean saved_fp_enable = GL_FALSE, saved_vp_enable = GL_FALSE;
-   GLboolean saved_shader_program = 0;
-   unsigned int saved_active_texture;
-   
-   assert((mask & ~(BUFFER_BIT_BACK_LEFT | BUFFER_BIT_FRONT_LEFT |
-		    BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL)) == 0);
-
-   _mesa_PushAttrib(GL_COLOR_BUFFER_BIT |
-		    GL_CURRENT_BIT |
-		    GL_DEPTH_BUFFER_BIT |
-		    GL_ENABLE_BIT |
-		    GL_STENCIL_BUFFER_BIT |
-		    GL_TRANSFORM_BIT |
-		    GL_CURRENT_BIT);
-   _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT);
-   saved_active_texture = ctx->Texture.CurrentUnit;
-  
-  /* Disable existing GL state we don't want to apply to a clear. */
-   _mesa_Disable(GL_ALPHA_TEST);
-   _mesa_Disable(GL_BLEND);
-   _mesa_Disable(GL_CULL_FACE);
-   _mesa_Disable(GL_FOG);
-   _mesa_Disable(GL_POLYGON_SMOOTH);
-   _mesa_Disable(GL_POLYGON_STIPPLE);
-   _mesa_Disable(GL_POLYGON_OFFSET_FILL);
-   _mesa_Disable(GL_LIGHTING);
-   _mesa_Disable(GL_CLIP_PLANE0);
-   _mesa_Disable(GL_CLIP_PLANE1);
-   _mesa_Disable(GL_CLIP_PLANE2);
-   _mesa_Disable(GL_CLIP_PLANE3);
-   _mesa_Disable(GL_CLIP_PLANE4);
-   _mesa_Disable(GL_CLIP_PLANE5);
-   if (ctx->Extensions.ARB_fragment_program && ctx->FragmentProgram.Enabled) {
-      saved_fp_enable = GL_TRUE;
-      _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
-   }
-   if (ctx->Extensions.ARB_vertex_program && ctx->VertexProgram.Enabled) {
-      saved_vp_enable = GL_TRUE;
-      _mesa_Disable(GL_VERTEX_PROGRAM_ARB);
-   }
-   if (ctx->Extensions.ARB_shader_objects && ctx->Shader.CurrentProgram) {
-      saved_shader_program = ctx->Shader.CurrentProgram->Name;
-      _mesa_UseProgramObjectARB(0);
-   }
-   
-   if (ctx->Texture._EnabledUnits != 0) {
-      int i;
-      
-      for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-	 _mesa_ActiveTextureARB(GL_TEXTURE0 + i);
-	 _mesa_Disable(GL_TEXTURE_1D);
-	 _mesa_Disable(GL_TEXTURE_2D);
-	 _mesa_Disable(GL_TEXTURE_3D);
-	 if (ctx->Extensions.ARB_texture_cube_map)
-	    _mesa_Disable(GL_TEXTURE_CUBE_MAP_ARB);
-	 if (ctx->Extensions.NV_texture_rectangle)
-	    _mesa_Disable(GL_TEXTURE_RECTANGLE_NV);
-	 if (ctx->Extensions.MESA_texture_array) {
-	    _mesa_Disable(GL_TEXTURE_1D_ARRAY_EXT);
-	    _mesa_Disable(GL_TEXTURE_2D_ARRAY_EXT);
-	 }
-      }
-   }
-  
-   r200_meta_set_passthrough_transform(rmesa);
-   
-   for (i = 0; i < 4; i++) {
-      color[i][0] = ctx->Color.ClearColor[0];
-      color[i][1] = ctx->Color.ClearColor[1];
-      color[i][2] = ctx->Color.ClearColor[2];
-      color[i][3] = ctx->Color.ClearColor[3];
-   }
-
-   /* convert clear Z from [0,1] to NDC coord in [-1,1] */
-   dst_z = -1.0 + 2.0 * ctx->Depth.Clear;
-   
-   /* Prepare the vertices, which are the same regardless of which buffer we're
-    * drawing to.
-    */
-   vertices[0][0] = fb->_Xmin;
-   vertices[0][1] = fb->_Ymin;
-   vertices[0][2] = dst_z;
-   vertices[1][0] = fb->_Xmax;
-   vertices[1][1] = fb->_Ymin;
-   vertices[1][2] = dst_z;
-   vertices[2][0] = fb->_Xmax;
-   vertices[2][1] = fb->_Ymax;
-   vertices[2][2] = dst_z;
-   vertices[3][0] = fb->_Xmin;
-   vertices[3][1] = fb->_Ymax;
-   vertices[3][2] = dst_z;
-
-   _mesa_ColorPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &color);
-   _mesa_VertexPointer(3, GL_FLOAT, 3 * sizeof(GLfloat), &vertices);
-   _mesa_Enable(GL_COLOR_ARRAY);
-   _mesa_Enable(GL_VERTEX_ARRAY);
-
-   while (mask != 0) {
-      GLuint this_mask = 0;
-
-      if (mask & BUFFER_BIT_BACK_LEFT)
-	 this_mask = BUFFER_BIT_BACK_LEFT;
-      else if (mask & BUFFER_BIT_FRONT_LEFT)
-	 this_mask = BUFFER_BIT_FRONT_LEFT;
-
-      /* Clear depth/stencil in the same pass as color. */
-      this_mask |= (mask & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL));
-
-      /* Select the current color buffer and use the color write mask if
-       * we have one, otherwise don't write any color channels.
-       */
-      if (this_mask & BUFFER_BIT_FRONT_LEFT)
-	 _mesa_DrawBuffer(GL_FRONT_LEFT);
-      else if (this_mask & BUFFER_BIT_BACK_LEFT)
-	 _mesa_DrawBuffer(GL_BACK_LEFT);
-      else
-	 _mesa_ColorMask(GL_FALSE, GL_FALSE, GL_FALSE, GL_FALSE);
-
-      /* Control writing of the depth clear value to depth. */
-      if (this_mask & BUFFER_BIT_DEPTH) {
-	 _mesa_DepthFunc(GL_ALWAYS);
-	 _mesa_Enable(GL_DEPTH_TEST);
-      } else {
-	 _mesa_Disable(GL_DEPTH_TEST);
-	 _mesa_DepthMask(GL_FALSE);
-      }
-
-      /* Control writing of the stencil clear value to stencil. */
-      if (this_mask & BUFFER_BIT_STENCIL) {
-	 _mesa_Enable(GL_STENCIL_TEST);
-	 _mesa_StencilOp(GL_REPLACE, GL_REPLACE, GL_REPLACE);
-	 _mesa_StencilFuncSeparate(GL_FRONT, GL_ALWAYS, ctx->Stencil.Clear,
-				   ctx->Stencil.WriteMask[0]);
-      } else {
-	 _mesa_Disable(GL_STENCIL_TEST);
-      }
-
-      CALL_DrawArrays(ctx->Exec, (GL_TRIANGLE_FAN, 0, 4));
-
-      mask &= ~this_mask;
-   }
-
-   r200_meta_restore_transform(rmesa);
-
-   _mesa_ActiveTextureARB(GL_TEXTURE0 + saved_active_texture);
-   if (saved_fp_enable)
-      _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB);
-   if (saved_vp_enable)
-      _mesa_Enable(GL_VERTEX_PROGRAM_ARB);
-
-   if (saved_shader_program)
-      _mesa_UseProgramObjectARB(saved_shader_program);
-
-   _mesa_PopClientAttrib();
-   _mesa_PopAttrib();
-}
-
-
 static void r200UserClear(GLcontext *ctx, GLuint mask)
 {
    radeon_clear_tris(ctx, mask);
@@ -449,7 +226,7 @@ static void r200Clear( GLcontext *ctx, GLbitfield mask )
       mask &= ~BUFFER_BIT_DEPTH;
    }
 
-   if ( (mask & BUFFER_BIT_STENCIL) && rmesa->radeon.state.stencil.hwBuffer ) {
+   if ( (mask & BUFFER_BIT_STENCIL) ) {
       flags |= RADEON_STENCIL;
       mask &= ~BUFFER_BIT_STENCIL;
    }
@@ -467,8 +244,7 @@ static void r200Clear( GLcontext *ctx, GLbitfield mask )
       flags |= RADEON_USE_COMP_ZBUF;
 /*      if (rmesa->radeon.radeonScreen->chip_family == CHIP_FAMILY_R200)
 	 flags |= RADEON_USE_HIERZ; */
-      if (!(rmesa->radeon.state.stencil.hwBuffer) ||
-	 ((flags & RADEON_DEPTH) && (flags & RADEON_STENCIL) &&
+      if (!((flags & RADEON_DEPTH) && (flags & RADEON_STENCIL) &&
 	    ((rmesa->radeon.state.stencil.clear & R200_STENCIL_WRITE_MASK) == R200_STENCIL_WRITE_MASK))) {
 	  flags |= RADEON_CLEAR_FASTZ;
       }
diff --git a/src/mesa/drivers/dri/r200/r200_state.c b/src/mesa/drivers/dri/r200/r200_state.c
index 5eb61c8804..ca4dee8a5b 100644
--- a/src/mesa/drivers/dri/r200/r200_state.c
+++ b/src/mesa/drivers/dri/r200/r200_state.c
@@ -740,7 +740,8 @@ static void r200PolygonOffset( GLcontext *ctx,
 			       GLfloat factor, GLfloat units )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   float_ui32_type constant =  { units * rmesa->radeon.state.depth.scale };
+   const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+   float_ui32_type constant =  { units * depthScale };
    float_ui32_type factoru = { factor };
 
 /*    factor *= 2; */
@@ -1610,13 +1611,24 @@ void r200UpdateWindow( GLcontext *ctx )
    GLfloat xoffset = dPriv ? (GLfloat) dPriv->x : 0;
    GLfloat yoffset = dPriv ? (GLfloat) dPriv->y + dPriv->h : 0;
    const GLfloat *v = ctx->Viewport._WindowMap.m;
+   const GLboolean render_to_fbo = (ctx->DrawBuffer ? (ctx->DrawBuffer->Name != 0) : 0);
+   const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+   GLfloat y_scale, y_bias;
+
+   if (render_to_fbo) {
+      y_scale = 1.0;
+      y_bias = 0;
+   } else {
+      y_scale = -1.0;
+      y_bias = yoffset;
+   }
 
    float_ui32_type sx = { v[MAT_SX] };
    float_ui32_type tx = { v[MAT_TX] + xoffset + SUBPIXEL_X };
-   float_ui32_type sy = { - v[MAT_SY] };
-   float_ui32_type ty = { (- v[MAT_TY]) + yoffset + SUBPIXEL_Y };
-   float_ui32_type sz = { v[MAT_SZ] * rmesa->radeon.state.depth.scale };
-   float_ui32_type tz = { v[MAT_TZ] * rmesa->radeon.state.depth.scale };
+   float_ui32_type sy = { v[MAT_SY] * y_scale };
+   float_ui32_type ty = { (v[MAT_TY] * y_scale) + y_bias + SUBPIXEL_Y };
+   float_ui32_type sz = { v[MAT_SZ] * depthScale };
+   float_ui32_type tz = { v[MAT_TZ] * depthScale };
 
    R200_STATECHANGE( rmesa, vpt );
 
@@ -2004,15 +2016,24 @@ static void r200Enable( GLcontext *ctx, GLenum cap, GLboolean state )
       break;
 
    case GL_STENCIL_TEST:
-      if ( rmesa->radeon.state.stencil.hwBuffer ) {
-	 R200_STATECHANGE( rmesa, ctx );
-	 if ( state ) {
-	    rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  R200_STENCIL_ENABLE;
+      {
+	 GLboolean hw_stencil = GL_FALSE;
+	 if (ctx->DrawBuffer) {
+	    struct radeon_renderbuffer *rrbStencil
+	       = radeon_get_renderbuffer(ctx->DrawBuffer, BUFFER_STENCIL);
+	    hw_stencil = (rrbStencil && rrbStencil->bo);
+	 }
+
+	 if (hw_stencil) {
+	    R200_STATECHANGE( rmesa, ctx );
+	    if ( state ) {
+	       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  R200_STENCIL_ENABLE;
+	    } else {
+	       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~R200_STENCIL_ENABLE;
+	    }
 	 } else {
-	    rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~R200_STENCIL_ENABLE;
+	    FALLBACK( rmesa, R200_FALLBACK_STENCIL, state );
 	 }
-      } else {
-	 FALLBACK( rmesa, R200_FALLBACK_STENCIL, state );
       }
       break;
 
diff --git a/src/mesa/drivers/dri/r200/r200_state_init.c b/src/mesa/drivers/dri/r200/r200_state_init.c
index a71f33ca3b..30326c2960 100644
--- a/src/mesa/drivers/dri/r200/r200_state_init.c
+++ b/src/mesa/drivers/dri/r200/r200_state_init.c
@@ -674,21 +674,15 @@ void r200InitState( r200ContextPtr rmesa )
    switch ( ctx->Visual.depthBits ) {
    case 16:
       rmesa->radeon.state.depth.clear = 0x0000ffff;
-      rmesa->radeon.state.depth.scale = 1.0 / (GLfloat)0xffff;
       rmesa->radeon.state.stencil.clear = 0x00000000;
       break;
    case 24:
    default:
       rmesa->radeon.state.depth.clear = 0x00ffffff;
-      rmesa->radeon.state.depth.scale = 1.0 / (GLfloat)0xffffff;
       rmesa->radeon.state.stencil.clear = 0xffff0000;
       break;
    }
 
-   /* Only have hw stencil when depth buffer is 24 bits deep */
-   rmesa->radeon.state.stencil.hwBuffer = ( ctx->Visual.stencilBits > 0 &&
-				     ctx->Visual.depthBits == 24 );
-
    rmesa->radeon.Fallback = 0;
 
    rmesa->radeon.hw.max_state_size = 0;
diff --git a/src/mesa/drivers/dri/r300/Makefile b/src/mesa/drivers/dri/r300/Makefile
index 497b1ec528..0dff9a1273 100644
--- a/src/mesa/drivers/dri/r300/Makefile
+++ b/src/mesa/drivers/dri/r300/Makefile
@@ -31,7 +31,8 @@ RADEON_COMMON_SOURCES = \
 	radeon_bo_legacy.c \
 	radeon_cs_legacy.c \
 	radeon_mipmap_tree.c \
-	radeon_span.c
+	radeon_span.c \
+	radeon_fbo.c
 
 DRIVER_SOURCES = \
 		 radeon_screen.c \
@@ -96,7 +97,8 @@ COMMON_SYMLINKS = \
 	radeon_mipmap_tree.c \
 	radeon_mipmap_tree.h \
 	radeon_texture.c \
-	radeon_texture.h
+	radeon_texture.h \
+	radeon_fbo.c
 
 DRI_LIB_DEPS += $(RADEON_LDFLAGS)
 
diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
index 2c6618fc60..d85f106c11 100644
--- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c
+++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
@@ -213,6 +213,7 @@ static void emit_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
 	BATCH_LOCALS(&r300->radeon);
 	struct radeon_renderbuffer *rrb;
 	uint32_t cbpitch;
+	uint32_t offset = r300->radeon.state.color.draw_offset;
 
 	rrb = radeon_get_colorbuffer(&r300->radeon);
 	if (!rrb || !rrb->bo) {
@@ -231,7 +232,7 @@ static void emit_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
 
 	BEGIN_BATCH_NO_AUTOSTATE(6);
 	OUT_BATCH_REGSEQ(R300_RB3D_COLOROFFSET0, 1);
-	OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+	OUT_BATCH_RELOC(offset, rrb->bo, offset, 0, RADEON_GEM_DOMAIN_VRAM, 0);
 	OUT_BATCH_REGSEQ(R300_RB3D_COLORPITCH0, 1);
 	OUT_BATCH(cbpitch);
 	END_BATCH();
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index 5bae37e1b1..c6bd69ed14 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -82,14 +82,17 @@ int hw_tcl_on = 1;
 #define need_GL_EXT_blend_equation_separate
 #define need_GL_EXT_blend_func_separate
 #define need_GL_EXT_blend_minmax
+#define need_GL_EXT_framebuffer_object
 #define need_GL_EXT_fog_coord
 #define need_GL_EXT_gpu_program_parameters
 #define need_GL_EXT_secondary_color
 #define need_GL_EXT_stencil_two_side
 #define need_GL_ATI_separate_stencil
 #define need_GL_NV_vertex_program
+
 #include "extension_helper.h"
 
+
 const struct dri_extension card_extensions[] = {
   /* *INDENT-OFF* */
   {"GL_ARB_depth_texture",		NULL},
@@ -110,6 +113,7 @@ const struct dri_extension card_extensions[] = {
   {"GL_EXT_blend_func_separate",	GL_EXT_blend_func_separate_functions},
   {"GL_EXT_blend_minmax",		GL_EXT_blend_minmax_functions},
   {"GL_EXT_blend_subtract",		NULL},
+  {"GL_EXT_packed_depth_stencil",	NULL},
   {"GL_EXT_fog_coord",			GL_EXT_fog_coord_functions },
   {"GL_EXT_gpu_program_parameters",     GL_EXT_gpu_program_parameters_functions},
   {"GL_EXT_secondary_color", 		GL_EXT_secondary_color_functions},
@@ -137,6 +141,11 @@ const struct dri_extension card_extensions[] = {
 };
 
 
+const struct dri_extension mm_extensions[] = {
+  { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions },
+  { NULL, NULL }
+};
+
 /**
  * The GL 2.0 functions are needed to make display lists work with
  * functions added by GL_ATI_separate_stencil.
@@ -417,6 +426,8 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
 
 	driInitExtensions(ctx, card_extensions, GL_TRUE);
+	if (r300->radeon.radeonScreen->kernel_mm)
+	  driInitExtensions(ctx, mm_extensions, GL_FALSE);
 
 	if (driQueryOptionb
 	    (&r300->radeon.optionCache, "disable_stencil_two_side"))
@@ -435,7 +446,7 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	r300->disable_lowimpact_fallback =
 	    driQueryOptionb(&r300->radeon.optionCache,
 			    "disable_lowimpact_fallback");
-
+	radeon_fbo_init(&r300->radeon);
    	radeonInitSpanFuncs( ctx );
 	r300InitCmdBuf(r300);
 	r300InitState(r300);
diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.c b/src/mesa/drivers/dri/r300/r300_ioctl.c
index d1cf57959f..71661eef19 100644
--- a/src/mesa/drivers/dri/r300/r300_ioctl.c
+++ b/src/mesa/drivers/dri/r300/r300_ioctl.c
@@ -66,6 +66,13 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define CLEARBUFFER_DEPTH	0x2
 #define CLEARBUFFER_STENCIL	0x4
 
+static void r300EmitClearState(GLcontext * ctx);
+
+static void r300UserClear(GLcontext *ctx, GLuint mask)
+{
+	radeon_clear_tris(ctx, mask);
+}
+
 static void r300ClearBuffer(r300ContextPtr r300, int flags,
 			    struct radeon_renderbuffer *rrb,
 			    struct radeon_renderbuffer *rrbd)
@@ -534,6 +541,47 @@ static void r300EmitClearState(GLcontext * ctx)
 	}
 }
 
+static void r300KernelClear(GLcontext *ctx, GLuint flags)
+{	  	
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	__DRIdrawablePrivate *dPriv = r300->radeon.dri.drawable;
+	struct radeon_framebuffer *rfb = dPriv->driverPrivate;
+	struct radeon_renderbuffer *rrb;
+	struct radeon_renderbuffer *rrbd;
+	int bits = 0;
+
+	/* Make sure it fits there. */
+	rcommonEnsureCmdBufSpace(&r300->radeon, 421 * 3, __FUNCTION__);
+	if (flags || bits)
+		r300EmitClearState(ctx);
+	rrbd = radeon_get_renderbuffer(&rfb->base, BUFFER_DEPTH);
+	if (rrbd && (flags & BUFFER_BIT_DEPTH))
+		bits |= CLEARBUFFER_DEPTH;
+
+	if (flags & BUFFER_BIT_COLOR0) {
+		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_COLOR0);
+		r300ClearBuffer(r300, CLEARBUFFER_COLOR, rrb, NULL);
+		bits = 0;
+	}
+		
+	if (flags & BUFFER_BIT_FRONT_LEFT) {
+		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_FRONT_LEFT);
+		r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, rrb, rrbd);
+		bits = 0;
+	}
+
+	if (flags & BUFFER_BIT_BACK_LEFT) {
+		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_BACK_LEFT);
+		r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, rrb, rrbd);
+		bits = 0;
+	}
+
+	if (bits)
+		r300ClearBuffer(r300, bits, NULL, rrbd);
+
+	COMMIT_BATCH();
+}
+
 /**
  * Buffer clear
  */
@@ -541,16 +589,15 @@ static void r300Clear(GLcontext * ctx, GLbitfield mask)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	__DRIdrawablePrivate *dPriv = r300->radeon.dri.drawable;
-	struct radeon_framebuffer *rfb = dPriv->driverPrivate;
-	struct radeon_renderbuffer *rrb;
-	struct radeon_renderbuffer *rrbd;
-	int flags = 0;
-	int bits = 0;
+	const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask);
+	GLbitfield swrast_mask = 0, tri_mask = 0;
+	int i;
+	struct gl_framebuffer *fb = ctx->DrawBuffer;
 
 	if (RADEON_DEBUG & DEBUG_IOCTL)
 		fprintf(stderr, "r300Clear\n");
 
-	{
+	if (!r300->radeon.radeonScreen->driScreen->dri2.enabled) {
 		LOCK_HARDWARE(&r300->radeon);
 		UNLOCK_HARDWARE(&r300->radeon);
 		if (dPriv->numClipRects == 0)
@@ -563,57 +610,52 @@ static void r300Clear(GLcontext * ctx, GLbitfield mask)
 	 */
 	R300_NEWPRIM(r300);
 
-	if (mask & BUFFER_BIT_FRONT_LEFT) {
-		flags |= BUFFER_BIT_FRONT_LEFT;
-		mask &= ~BUFFER_BIT_FRONT_LEFT;
-	}
+	if (colorMask == ~0)
+	  tri_mask |= (mask & BUFFER_BITS_COLOR);
+
 
-	if (mask & BUFFER_BIT_BACK_LEFT) {
-		flags |= BUFFER_BIT_BACK_LEFT;
-		mask &= ~BUFFER_BIT_BACK_LEFT;
+	/* HW stencil */
+	if (mask & BUFFER_BIT_STENCIL) {
+		tri_mask |= BUFFER_BIT_STENCIL;
 	}
 
+	/* HW depth */
 	if (mask & BUFFER_BIT_DEPTH) {
-		bits |= CLEARBUFFER_DEPTH;
-		mask &= ~BUFFER_BIT_DEPTH;
+    	  tri_mask |= BUFFER_BIT_DEPTH;
 	}
 
-	if ((mask & BUFFER_BIT_STENCIL) && r300->radeon.state.stencil.hwBuffer) {
-		bits |= CLEARBUFFER_STENCIL;
-		mask &= ~BUFFER_BIT_STENCIL;
-	}
+	/* If we're doing a tri pass for depth/stencil, include a likely color
+	 * buffer with it.
+	 */
 
-	if (mask) {
-		if (RADEON_DEBUG & DEBUG_FALLBACKS)
-			fprintf(stderr, "%s: swrast clear, mask: %x\n",
-				__FUNCTION__, mask);
-		_swrast_Clear(ctx, mask);
+	for (i = 0; i < BUFFER_COUNT; i++) {
+	  GLuint bufBit = 1 << i;
+	  if ((tri_mask) & bufBit) {
+	    if (!fb->Attachment[i].Renderbuffer->ClassID) {
+	      tri_mask &= ~bufBit;
+	      swrast_mask |= bufBit;
+	    }
+	  }
 	}
 
-	/* Make sure it fits there. */
-	rcommonEnsureCmdBufSpace(&r300->radeon, 421 * 3, __FUNCTION__);
-	if (flags || bits)
-		r300EmitClearState(ctx);
-	rrbd = (void *)rfb->base.Attachment[BUFFER_DEPTH].Renderbuffer;
+	/* SW fallback clearing */
+	swrast_mask = mask & ~tri_mask;
 
-	if (flags & BUFFER_BIT_FRONT_LEFT) {
-		rrb = (void *)rfb->base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer;
-		r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, rrb, rrbd);
-		bits = 0;
+	if (tri_mask) {
+		if (r300->radeon.radeonScreen->kernel_mm)
+			r300UserClear(ctx, tri_mask);
+		else
+			r300KernelClear(ctx, tri_mask);
 	}
-
-	if (flags & BUFFER_BIT_BACK_LEFT) {
-		rrb = (void *)rfb->base.Attachment[BUFFER_BACK_LEFT].Renderbuffer;
-		r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, rrb, rrbd);
-		bits = 0;
+	if (swrast_mask) {
+		if (RADEON_DEBUG & DEBUG_FALLBACKS)
+			fprintf(stderr, "%s: swrast clear, mask: %x\n",
+				__FUNCTION__, swrast_mask);
+		_swrast_Clear(ctx, swrast_mask);
 	}
-
-	if (bits)
-		r300ClearBuffer(r300, bits, NULL, rrbd);
-
-	COMMIT_BATCH();
 }
 
+
 void r300InitIoctlFuncs(struct dd_function_table *functions)
 {
 	functions->Clear = r300Clear;
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index f423029ee6..f49b43c207 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -587,8 +587,14 @@ static void r300SetDepthState(GLcontext * ctx)
 static void r300SetStencilState(GLcontext * ctx, GLboolean state)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	GLboolean hw_stencil = GL_FALSE;
+	if (ctx->DrawBuffer) {
+		struct radeon_renderbuffer *rrbStencil
+			= radeon_get_renderbuffer(ctx->DrawBuffer, BUFFER_STENCIL);
+		hw_stencil = (rrbStencil && rrbStencil->bo);
+	}
 
-	if (r300->radeon.state.stencil.hwBuffer) {
+	if (hw_stencil) {
 		R300_STATECHANGE(r300, zs);
 		if (state) {
 			r300->hw.zs.cmd[R300_ZS_CNTL_0] |=
@@ -933,13 +939,24 @@ static void r300UpdateWindow(GLcontext * ctx)
 	GLfloat xoffset = dPriv ? (GLfloat) dPriv->x : 0;
 	GLfloat yoffset = dPriv ? (GLfloat) dPriv->y + dPriv->h : 0;
 	const GLfloat *v = ctx->Viewport._WindowMap.m;
+	const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+	const GLboolean render_to_fbo = (ctx->DrawBuffer->Name != 0);
+	GLfloat y_scale, y_bias;
+
+	if (render_to_fbo) {
+		y_scale = 1.0;
+		y_bias = 0;
+	} else {
+		y_scale = -1.0;
+		y_bias = yoffset;
+	}
 
 	GLfloat sx = v[MAT_SX];
 	GLfloat tx = v[MAT_TX] + xoffset + SUBPIXEL_X;
-	GLfloat sy = -v[MAT_SY];
-	GLfloat ty = (-v[MAT_TY]) + yoffset + SUBPIXEL_Y;
-	GLfloat sz = v[MAT_SZ] * rmesa->radeon.state.depth.scale;
-	GLfloat tz = v[MAT_TZ] * rmesa->radeon.state.depth.scale;
+	GLfloat sy = v[MAT_SY] * y_scale;
+	GLfloat ty = (v[MAT_TY] * y_scale) + y_bias + SUBPIXEL_Y;
+	GLfloat sz = v[MAT_SZ] * depthScale;
+	GLfloat tz = v[MAT_TZ] * depthScale;
 
 	R300_STATECHANGE(rmesa, vpt);
 
@@ -2022,7 +2039,7 @@ static void r300ResetHwState(r300ContextPtr r300)
 		fprintf(stderr, "%s\n", __FUNCTION__);
 
 	radeon_firevertices(&r300->radeon);
-	r300UpdateWindow(ctx);
+	//r300UpdateWindow(ctx);
 
 	r300ColorMask(ctx,
 		      ctx->Color.ColorMask[RCOMP],
@@ -2197,16 +2214,6 @@ static void r300ResetHwState(r300ContextPtr r300)
 	r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[1] = 0x00000000;
 	r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[2] = 0xffffffff;
 
-	rrb = r300->radeon.state.depth.rrb;
-	if (rrb && rrb->bo && (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE)) {
-		/* XXX: Turn off when clearing buffers ? */
-		r300->hw.zb.cmd[R300_ZB_PITCH] |= R300_DEPTHMACROTILE_ENABLE;
-
-		if (ctx->Visual.depthBits == 24)
-			r300->hw.zb.cmd[R300_ZB_PITCH] |=
-			    R300_DEPTHMICROTILE_TILED;
-	}
-
 	r300->hw.zb_depthclearvalue.cmd[1] = 0;
 
 	r300->hw.zstencil_format.cmd[2] = R300_ZTOP_DISABLE;
@@ -2520,10 +2527,6 @@ void r300InitState(r300ContextPtr r300)
 	GLcontext *ctx = r300->radeon.glCtx;
 	GLuint depth_fmt;
 
-	/* Only have hw stencil when depth buffer is 24 bits deep */
-	r300->radeon.state.stencil.hwBuffer = (ctx->Visual.stencilBits > 0 &&
-					       ctx->Visual.depthBits == 24);
-
 	memset(&(r300->state.texture), 0, sizeof(r300->state.texture));
 
 	r300ResetHwState(r300);
diff --git a/src/mesa/drivers/dri/radeon/radeon_bocs_wrapper.h b/src/mesa/drivers/dri/radeon/radeon_bocs_wrapper.h
index f80f0f7b22..42607df967 100644
--- a/src/mesa/drivers/dri/radeon/radeon_bocs_wrapper.h
+++ b/src/mesa/drivers/dri/radeon/radeon_bocs_wrapper.h
@@ -36,7 +36,10 @@ struct drm_radeon_gem_info {
 #endif
 
 
-
+uint32_t radeon_gem_bo_name(struct radeon_bo *dummy)
+{
+  return 0;
+}
 
 static inline void *radeon_bo_manager_gem_ctor(int fd)
 {
diff --git a/src/mesa/drivers/dri/radeon/radeon_common.c b/src/mesa/drivers/dri/radeon/radeon_common.c
index 840233ff89..9f646c4386 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common.c
+++ b/src/mesa/drivers/dri/radeon/radeon_common.c
@@ -58,6 +58,21 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "tnl/t_pipeline.h"
 #include "swrast_setup/swrast_setup.h"
 
+#include "main/blend.h"
+#include "main/bufferobj.h"
+#include "main/buffers.h"
+#include "main/depth.h"
+#include "main/shaders.h"
+#include "main/texstate.h"
+#include "main/varray.h"
+#include "glapi/dispatch.h"
+#include "swrast/swrast.h"
+#include "main/stencil.h"
+#include "main/matrix.h"
+#include "main/attrib.h"
+#include "main/enable.h"
+#include "main/viewport.h"
+
 #include "dri_util.h"
 #include "vblank.h"
 
@@ -132,10 +147,10 @@ void radeonRecalcScissorRects(radeonContextPtr radeon)
 	}
 }
 
-static void radeon_get_cliprects(radeonContextPtr radeon,
-				 struct drm_clip_rect **cliprects,
-				 unsigned int *num_cliprects,
-				 int *x_off, int *y_off)
+void radeon_get_cliprects(radeonContextPtr radeon,
+			  struct drm_clip_rect **cliprects,
+			  unsigned int *num_cliprects,
+			  int *x_off, int *y_off)
 {
 	__DRIdrawablePrivate *dPriv = radeon->dri.drawable;
 	struct radeon_framebuffer *rfb = dPriv->driverPrivate;
@@ -619,6 +634,7 @@ void radeon_draw_buffer(GLcontext *ctx, struct gl_framebuffer *fb)
 	radeonContextPtr radeon = RADEON_CONTEXT(ctx);
 	struct radeon_renderbuffer *rrbDepth = NULL, *rrbStencil = NULL,
 		*rrbColor = NULL;
+	uint32_t offset = 0;
        
 
 	if (!fb) {
@@ -657,17 +673,20 @@ void radeon_draw_buffer(GLcontext *ctx, struct gl_framebuffer *fb)
 		/* none */
 	if (fb->Name == 0) {
 		if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) {
-			rrbColor = (void *)fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer;
+			rrbColor = radeon_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
 			radeon->front_cliprects = GL_TRUE;
 		} else {
-			rrbColor = (void *)fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer;
+			rrbColor = radeon_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer);
 			radeon->front_cliprects = GL_FALSE;
 		}
 	} else {
 		/* user FBO in theory */
 		struct radeon_renderbuffer *rrb;
-		rrb = (void *)fb->_ColorDrawBuffers[0];
-		rrbColor = rrb;
+		rrb = radeon_renderbuffer(fb->_ColorDrawBuffers[0]);
+		if (rrb) {
+			offset = rrb->draw_offset;
+			rrbColor = rrb;
+		}
 		radeon->constant_cliprect = GL_TRUE;
 	}
 
@@ -677,9 +696,8 @@ void radeon_draw_buffer(GLcontext *ctx, struct gl_framebuffer *fb)
 		radeon->vtbl.fallback(ctx, RADEON_FALLBACK_DRAW_BUFFER, GL_FALSE);
 
 
-
 	if (fb->_DepthBuffer && fb->_DepthBuffer->Wrapped) {
-		rrbDepth = (struct radeon_renderbuffer *)fb->_DepthBuffer->Wrapped;
+		rrbDepth = radeon_renderbuffer(fb->_DepthBuffer->Wrapped);
 		if (rrbDepth && rrbDepth->bo) {
 			radeon->vtbl.fallback(ctx, RADEON_FALLBACK_DEPTH_BUFFER, GL_FALSE);
 		} else {
@@ -690,16 +708,11 @@ void radeon_draw_buffer(GLcontext *ctx, struct gl_framebuffer *fb)
 		rrbDepth = NULL;
 	}
 
-	/* TODO stencil things */
 	if (fb->_StencilBuffer && fb->_StencilBuffer->Wrapped) {
-		rrbStencil = (struct radeon_renderbuffer *)fb->_DepthBuffer->Wrapped;
+		rrbStencil = radeon_renderbuffer(fb->_DepthBuffer->Wrapped);
 		if (rrbStencil && rrbStencil->bo) {
 			radeon->vtbl.fallback(ctx, RADEON_FALLBACK_STENCIL_BUFFER, GL_FALSE);
 			/* need to re-compute stencil hw state */
-			if (ctx->Driver.Enable != NULL)
-				ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled);
-			else
-				ctx->NewState |= _NEW_STENCIL;
 			if (!rrbDepth)
 				rrbDepth = rrbStencil;
 		} else {
@@ -725,26 +738,28 @@ void radeon_draw_buffer(GLcontext *ctx, struct gl_framebuffer *fb)
 	 * Update depth test state
 	 */
 	if (ctx->Driver.Enable) {
-		if (ctx->Depth.Test && fb->Visual.depthBits > 0) {
-			ctx->Driver.Enable(ctx, GL_DEPTH_TEST, GL_TRUE);
-		} else {
-			ctx->Driver.Enable(ctx, GL_DEPTH_TEST, GL_FALSE);
-		}
+		ctx->Driver.Enable(ctx, GL_DEPTH_TEST,
+				   (ctx->Depth.Test && fb->Visual.depthBits > 0));
+		ctx->Driver.Enable(ctx, GL_STENCIL_TEST,
+				   (ctx->Stencil._Enabled && fb->Visual.stencilBits > 0));
 	} else {
-		ctx->NewState |= _NEW_DEPTH;
+		ctx->NewState |= (_NEW_DEPTH | _NEW_STENCIL);
 	}
 	
 	radeon->state.depth.rrb = rrbDepth;
-
 	radeon->state.color.rrb = rrbColor;
+	radeon->state.color.draw_offset = offset;
 
+#if 0
 	/* update viewport since it depends on window size */
 	if (ctx->Driver.Viewport) {
 		ctx->Driver.Viewport(ctx, ctx->Viewport.X, ctx->Viewport.Y,
 				     ctx->Viewport.Width, ctx->Viewport.Height);
 	} else {
-		ctx->NewState |= _NEW_VIEWPORT;
+	
 	}
+#endif
+	ctx->NewState |= _NEW_VIEWPORT;
 
 	/* Set state we know depends on drawable parameters:
 	 */
@@ -752,6 +767,19 @@ void radeon_draw_buffer(GLcontext *ctx, struct gl_framebuffer *fb)
 		ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y,
 				    ctx->Scissor.Width, ctx->Scissor.Height);
 	radeon->NewGLState |= _NEW_SCISSOR;
+
+	if (ctx->Driver.DepthRange)
+		ctx->Driver.DepthRange(ctx,
+				       ctx->Viewport.Near,
+				       ctx->Viewport.Far);
+
+	/* Update culling direction which changes depending on the
+	 * orientation of the buffer:
+	 */
+	if (ctx->Driver.FrontFace)
+		ctx->Driver.FrontFace(ctx, ctx->Polygon.FrontFace);
+	else
+		ctx->NewState |= _NEW_POLYGON;
 }
 
 /**
@@ -799,10 +827,6 @@ void radeonUpdatePageFlipping(radeonContextPtr radeon)
 
 void radeon_window_moved(radeonContextPtr radeon)
 {
-	GLcontext *ctx = radeon->glCtx;
-	__DRIdrawablePrivate *dPriv = radeon->dri.drawable;
-	struct radeon_framebuffer *rfb = dPriv->driverPrivate;
-
 	if (!radeon->radeonScreen->driScreen->dri2.enabled) {
 		radeonUpdatePageFlipping(radeon);
 	}
@@ -946,8 +970,14 @@ void radeonFinish(GLcontext * ctx)
 	if (radeon->radeonScreen->kernel_mm) {
 		for (i = 0; i < fb->_NumColorDrawBuffers; i++) {
 			struct radeon_renderbuffer *rrb;
-			rrb = (struct radeon_renderbuffer *)fb->_ColorDrawBuffers[i];
-			if (rrb->bo)
+			rrb = radeon_renderbuffer(fb->_ColorDrawBuffers[i]);
+			if (rrb && rrb->bo)
+				radeon_bo_wait(rrb->bo);
+		}
+		{
+			struct radeon_renderbuffer *rrb;
+			rrb = radeon_get_depthbuffer(radeon);
+			if (rrb && rrb->bo)
 				radeon_bo_wait(rrb->bo);
 		}
 	} else if (radeon->do_irqs) {
@@ -1105,3 +1135,219 @@ void rcommonBeginBatch(radeonContextPtr rmesa, int n,
 
 
 
+static void
+radeon_meta_set_passthrough_transform(radeonContextPtr radeon)
+{
+   GLcontext *ctx = radeon->glCtx;
+
+   radeon->meta.saved_vp_x = ctx->Viewport.X;
+   radeon->meta.saved_vp_y = ctx->Viewport.Y;
+   radeon->meta.saved_vp_width = ctx->Viewport.Width;
+   radeon->meta.saved_vp_height = ctx->Viewport.Height;
+   radeon->meta.saved_matrix_mode = ctx->Transform.MatrixMode;
+
+   _mesa_Viewport(0, 0, ctx->DrawBuffer->Width, ctx->DrawBuffer->Height);
+
+   _mesa_MatrixMode(GL_PROJECTION);
+   _mesa_PushMatrix();
+   _mesa_LoadIdentity();
+   _mesa_Ortho(0, ctx->DrawBuffer->Width, 0, ctx->DrawBuffer->Height, 1, -1);
+
+   _mesa_MatrixMode(GL_MODELVIEW);
+   _mesa_PushMatrix();
+   _mesa_LoadIdentity();
+}
+
+static void
+radeon_meta_restore_transform(radeonContextPtr radeon)
+{
+   _mesa_MatrixMode(GL_PROJECTION);
+   _mesa_PopMatrix();
+   _mesa_MatrixMode(GL_MODELVIEW);
+   _mesa_PopMatrix();
+
+   _mesa_MatrixMode(radeon->meta.saved_matrix_mode);
+
+   _mesa_Viewport(radeon->meta.saved_vp_x, radeon->meta.saved_vp_y,
+		  radeon->meta.saved_vp_width, radeon->meta.saved_vp_height);
+}
+
+
+/**
+ * Perform glClear where mask contains only color, depth, and/or stencil.
+ *
+ * The implementation is based on calling into Mesa to set GL state and
+ * performing normal triangle rendering.  The intent of this path is to
+ * have as generic a path as possible, so that any driver could make use of
+ * it.
+ */
+
+
+void radeon_clear_tris(GLcontext *ctx, GLbitfield mask)
+{
+   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+   GLfloat vertices[4][3];
+   GLfloat color[4][4];
+   GLfloat dst_z;
+   struct gl_framebuffer *fb = ctx->DrawBuffer;
+   int i;
+   GLboolean saved_fp_enable = GL_FALSE, saved_vp_enable = GL_FALSE;
+   GLboolean saved_shader_program = 0;
+   unsigned int saved_active_texture;
+
+   assert((mask & ~(TRI_CLEAR_COLOR_BITS | BUFFER_BIT_DEPTH |
+		    BUFFER_BIT_STENCIL)) == 0);   
+
+   _mesa_PushAttrib(GL_COLOR_BUFFER_BIT |
+		    GL_CURRENT_BIT |
+		    GL_DEPTH_BUFFER_BIT |
+		    GL_ENABLE_BIT |
+		    GL_STENCIL_BUFFER_BIT |
+		    GL_TRANSFORM_BIT |
+		    GL_CURRENT_BIT);
+   _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT);
+   saved_active_texture = ctx->Texture.CurrentUnit;
+  
+  /* Disable existing GL state we don't want to apply to a clear. */
+   _mesa_Disable(GL_ALPHA_TEST);
+   _mesa_Disable(GL_BLEND);
+   _mesa_Disable(GL_CULL_FACE);
+   _mesa_Disable(GL_FOG);
+   _mesa_Disable(GL_POLYGON_SMOOTH);
+   _mesa_Disable(GL_POLYGON_STIPPLE);
+   _mesa_Disable(GL_POLYGON_OFFSET_FILL);
+   _mesa_Disable(GL_LIGHTING);
+   _mesa_Disable(GL_CLIP_PLANE0);
+   _mesa_Disable(GL_CLIP_PLANE1);
+   _mesa_Disable(GL_CLIP_PLANE2);
+   _mesa_Disable(GL_CLIP_PLANE3);
+   _mesa_Disable(GL_CLIP_PLANE4);
+   _mesa_Disable(GL_CLIP_PLANE5);
+   if (ctx->Extensions.ARB_fragment_program && ctx->FragmentProgram.Enabled) {
+      saved_fp_enable = GL_TRUE;
+      _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
+   }
+   if (ctx->Extensions.ARB_vertex_program && ctx->VertexProgram.Enabled) {
+      saved_vp_enable = GL_TRUE;
+      _mesa_Disable(GL_VERTEX_PROGRAM_ARB);
+   }
+   if (ctx->Extensions.ARB_shader_objects && ctx->Shader.CurrentProgram) {
+      saved_shader_program = ctx->Shader.CurrentProgram->Name;
+      _mesa_UseProgramObjectARB(0);
+   }
+   
+   if (ctx->Texture._EnabledUnits != 0) {
+      int i;
+      
+      for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
+	 _mesa_ActiveTextureARB(GL_TEXTURE0 + i);
+	 _mesa_Disable(GL_TEXTURE_1D);
+	 _mesa_Disable(GL_TEXTURE_2D);
+	 _mesa_Disable(GL_TEXTURE_3D);
+	 if (ctx->Extensions.ARB_texture_cube_map)
+	    _mesa_Disable(GL_TEXTURE_CUBE_MAP_ARB);
+	 if (ctx->Extensions.NV_texture_rectangle)
+	    _mesa_Disable(GL_TEXTURE_RECTANGLE_NV);
+	 if (ctx->Extensions.MESA_texture_array) {
+	    _mesa_Disable(GL_TEXTURE_1D_ARRAY_EXT);
+	    _mesa_Disable(GL_TEXTURE_2D_ARRAY_EXT);
+	 }
+      }
+   }
+  
+   radeon_meta_set_passthrough_transform(rmesa);
+   
+   for (i = 0; i < 4; i++) {
+      color[i][0] = ctx->Color.ClearColor[0];
+      color[i][1] = ctx->Color.ClearColor[1];
+      color[i][2] = ctx->Color.ClearColor[2];
+      color[i][3] = ctx->Color.ClearColor[3];
+   }
+
+   /* convert clear Z from [0,1] to NDC coord in [-1,1] */
+
+   dst_z = -1.0 + 2.0 * ctx->Depth.Clear;
+   /* Prepare the vertices, which are the same regardless of which buffer we're
+    * drawing to.
+    */
+   vertices[0][0] = fb->_Xmin;
+   vertices[0][1] = fb->_Ymin;
+   vertices[0][2] = dst_z;
+   vertices[1][0] = fb->_Xmax;
+   vertices[1][1] = fb->_Ymin;
+   vertices[1][2] = dst_z;
+   vertices[2][0] = fb->_Xmax;
+   vertices[2][1] = fb->_Ymax;
+   vertices[2][2] = dst_z;
+   vertices[3][0] = fb->_Xmin;
+   vertices[3][1] = fb->_Ymax;
+   vertices[3][2] = dst_z;
+
+   _mesa_ColorPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &color);
+   _mesa_VertexPointer(3, GL_FLOAT, 3 * sizeof(GLfloat), &vertices);
+   _mesa_Enable(GL_COLOR_ARRAY);
+   _mesa_Enable(GL_VERTEX_ARRAY);
+
+   while (mask != 0) {
+      GLuint this_mask = 0;
+      GLuint color_bit;
+
+      color_bit = _mesa_ffs(mask & TRI_CLEAR_COLOR_BITS);
+      if (color_bit != 0)
+	 this_mask |= (1 << (color_bit - 1));
+
+      /* Clear depth/stencil in the same pass as color. */
+      this_mask |= (mask & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL));
+
+      /* Select the current color buffer and use the color write mask if
+       * we have one, otherwise don't write any color channels.
+       */
+      if (this_mask & BUFFER_BIT_FRONT_LEFT)
+	 _mesa_DrawBuffer(GL_FRONT_LEFT);
+      else if (this_mask & BUFFER_BIT_BACK_LEFT)
+	 _mesa_DrawBuffer(GL_BACK_LEFT);
+      else if (color_bit != 0)
+	 _mesa_DrawBuffer(GL_COLOR_ATTACHMENT0 +
+			  (color_bit - BUFFER_COLOR0 - 1));
+      else
+	 _mesa_ColorMask(GL_FALSE, GL_FALSE, GL_FALSE, GL_FALSE);
+
+      /* Control writing of the depth clear value to depth. */
+      if (this_mask & BUFFER_BIT_DEPTH) {
+	 _mesa_DepthFunc(GL_ALWAYS);
+	 _mesa_DepthMask(GL_TRUE);
+	 _mesa_Enable(GL_DEPTH_TEST);
+      } else {
+	 _mesa_Disable(GL_DEPTH_TEST);
+	 _mesa_DepthMask(GL_FALSE);
+      }
+
+      /* Control writing of the stencil clear value to stencil. */
+      if (this_mask & BUFFER_BIT_STENCIL) {
+	 _mesa_Enable(GL_STENCIL_TEST);
+	 _mesa_StencilOp(GL_REPLACE, GL_REPLACE, GL_REPLACE);
+	 _mesa_StencilFuncSeparate(GL_FRONT, GL_ALWAYS, ctx->Stencil.Clear,
+				   ctx->Stencil.WriteMask[0]);
+      } else {
+	 _mesa_Disable(GL_STENCIL_TEST);
+      }
+
+      CALL_DrawArrays(ctx->Exec, (GL_TRIANGLE_FAN, 0, 4));
+
+      mask &= ~this_mask;
+   }
+
+   radeon_meta_restore_transform(rmesa);
+
+   _mesa_ActiveTextureARB(GL_TEXTURE0 + saved_active_texture);
+   if (saved_fp_enable)
+      _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB);
+   if (saved_vp_enable)
+      _mesa_Enable(GL_VERTEX_PROGRAM_ARB);
+
+   if (saved_shader_program)
+      _mesa_UseProgramObjectARB(saved_shader_program);
+
+   _mesa_PopClientAttrib();
+   _mesa_PopAttrib();
+}
diff --git a/src/mesa/drivers/dri/radeon/radeon_common.h b/src/mesa/drivers/dri/radeon/radeon_common.h
index 814da1b4f5..f3e2290cab 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common.h
+++ b/src/mesa/drivers/dri/radeon/radeon_common.h
@@ -5,6 +5,18 @@
 #include "radeon_dma.h"
 #include "radeon_texture.h"
 
+
+#define TRI_CLEAR_COLOR_BITS (BUFFER_BIT_BACK_LEFT |			\
+			      BUFFER_BIT_FRONT_LEFT |			\
+			      BUFFER_BIT_COLOR0 |			\
+			      BUFFER_BIT_COLOR1 |			\
+			      BUFFER_BIT_COLOR2 |			\
+			      BUFFER_BIT_COLOR3 |			\
+			      BUFFER_BIT_COLOR4 |			\
+			      BUFFER_BIT_COLOR5 |			\
+			      BUFFER_BIT_COLOR6 |			\
+			      BUFFER_BIT_COLOR7)
+
 void radeonRecalcScissorRects(radeonContextPtr radeon);
 void radeonSetCliprects(radeonContextPtr radeon);
 void radeonUpdateScissor( GLcontext *ctx );
@@ -24,12 +36,24 @@ void radeonFlush(GLcontext *ctx);
 void radeonFinish(GLcontext * ctx);
 void radeonEmitState(radeonContextPtr radeon);
 
+void radeon_clear_tris(GLcontext *ctx, GLbitfield mask);
+
 void radeon_window_moved(radeonContextPtr radeon);
 void radeon_draw_buffer(GLcontext *ctx, struct gl_framebuffer *fb);
 void radeonDrawBuffer( GLcontext *ctx, GLenum mode );
 void radeonReadBuffer( GLcontext *ctx, GLenum mode );
 void radeon_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height);
+void radeon_get_cliprects(radeonContextPtr radeon,
+			  struct drm_clip_rect **cliprects,
+			  unsigned int *num_cliprects,
+			  int *x_off, int *y_off);
 
+void radeon_fbo_init(struct radeon_context *radeon);
+void
+radeon_renderbuffer_set_bo(struct radeon_renderbuffer *rb,
+			   struct radeon_bo *bo);
+struct radeon_renderbuffer *
+radeon_create_renderbuffer(GLenum format, __DRIdrawablePrivate *driDrawPriv);
 static inline struct radeon_renderbuffer *radeon_renderbuffer(struct gl_renderbuffer *rb)
 {
 	struct radeon_renderbuffer *rrb = (struct radeon_renderbuffer *)rb;
@@ -60,12 +84,8 @@ static inline struct radeon_renderbuffer *radeon_get_depthbuffer(radeonContextPt
 static inline struct radeon_renderbuffer *radeon_get_colorbuffer(radeonContextPtr rmesa)
 {
 	struct radeon_renderbuffer *rrb;
-	struct radeon_framebuffer *rfb = rmesa->dri.drawable->driverPrivate;
 
 	rrb = rmesa->state.color.rrb;
-	if (rmesa->radeonScreen->driScreen->dri2.enabled) {
-		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_BACK_LEFT);
-	}
 	if (!rrb)
 		return NULL;
 	return rrb;
diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.c b/src/mesa/drivers/dri/radeon/radeon_common_context.c
index a818440faf..f335eb0313 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common_context.c
+++ b/src/mesa/drivers/dri/radeon/radeon_common_context.c
@@ -392,6 +392,7 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
 	struct radeon_framebuffer *draw;
 	radeonContextPtr radeon;
 	char *regname;
+	struct radeon_bo *depth_bo, *bo;
 
 	if (RADEON_DEBUG & DEBUG_DRI)
 	    fprintf(stderr, "enter %s, drawable %p\n", __func__, drawable);
@@ -448,7 +449,7 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
 			regname = "dri2 depth buffer";
 			break;
 		case __DRI_BUFFER_STENCIL:
-			rb = radeon_get_renderbuffer(&draw->base, BUFFER_DEPTH);
+			rb = radeon_get_renderbuffer(&draw->base, BUFFER_STENCIL);
 			regname = "dri2 stencil buffer";
 			break;
 		case __DRI_BUFFER_ACCUM:
@@ -463,25 +464,49 @@ radeon_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
 			continue;
 
 		if (rb->bo) {
-			radeon_bo_unref(rb->bo);
-			rb->bo = NULL;
+			uint32_t name = radeon_gem_name_bo(rb->bo);
+			if (name == buffers[i].name)
+				continue;
 		}
+
+		if (RADEON_DEBUG & DEBUG_DRI)
+			fprintf(stderr,
+				"attaching buffer %s, %d, at %d, cpp %d, pitch %d\n",
+				regname, buffers[i].name, buffers[i].attachment,
+				buffers[i].cpp, buffers[i].pitch);
+
 		rb->cpp = buffers[i].cpp;
 		rb->pitch = buffers[i].pitch;
 		rb->width = drawable->w;
 		rb->height = drawable->h;
 		rb->has_surface = 0;
-		rb->bo = radeon_bo_open(radeon->radeonScreen->bom,
-					buffers[i].name,
-					0,
-					0,
-					RADEON_GEM_DOMAIN_VRAM,
-					buffers[i].flags);
-		if (rb->bo == NULL) {
-			fprintf(stderr, "failed to attach %s %d\n",
-				regname, buffers[i].name);
 
+		if (buffers[i].attachment == __DRI_BUFFER_STENCIL && depth_bo) {
+			if (RADEON_DEBUG & DEBUG_DRI)
+				fprintf(stderr, "(reusing depth buffer as stencil)\n");
+			bo = depth_bo;
+			radeon_bo_ref(bo);
+		} else {
+			bo = radeon_bo_open(radeon->radeonScreen->bom,
+						buffers[i].name,
+						0,
+						0,
+						RADEON_GEM_DOMAIN_VRAM,
+						buffers[i].flags);
+			if (bo == NULL) {
+
+				fprintf(stderr, "failed to attach %s %d\n",
+					regname, buffers[i].name);
+				
+			}
 		}
+
+		if (buffers[i].attachment == __DRI_BUFFER_DEPTH)
+			depth_bo = bo;
+
+		radeon_renderbuffer_set_bo(rb, bo);
+		radeon_bo_unref(bo);
+		    
 	}
 
 	driUpdateFramebufferSize(radeon->glCtx, drawable);
diff --git a/src/mesa/drivers/dri/radeon/radeon_common_context.h b/src/mesa/drivers/dri/radeon/radeon_common_context.h
index e5ada622b0..0ce72c9198 100644
--- a/src/mesa/drivers/dri/radeon/radeon_common_context.h
+++ b/src/mesa/drivers/dri/radeon/radeon_common_context.h
@@ -82,6 +82,7 @@ struct radeon_renderbuffer
 	unsigned int width;
 	unsigned int height;
 
+	uint32_t draw_offset; /* FBO */
 	/* boo Xorg 6.8.2 compat */
 	int has_surface;
 
@@ -117,11 +118,11 @@ struct radeon_colorbuffer_state {
 	GLuint clear;
 	int roundEnable;
 	struct radeon_renderbuffer *rrb;
+	uint32_t draw_offset; /* offset into color renderbuffer - FBOs */
 };
 
 struct radeon_depthbuffer_state {
 	GLuint clear;
-	GLfloat scale;
 	struct radeon_renderbuffer *rrb;
 };
 
@@ -135,7 +136,6 @@ struct radeon_scissor_state {
 };
 
 struct radeon_stencilbuffer_state {
-	GLboolean hwBuffer;
 	GLuint clear;		/* rb3d_stencilrefmask value */
 };
 
@@ -442,9 +442,23 @@ struct radeon_context {
 
    struct radeon_cmdbuf cmdbuf;
 	
-	drm_clip_rect_t fboRect;
-	GLboolean constant_cliprect; /* use for FBO or DRI2 rendering */
-	GLboolean front_cliprects;
+  drm_clip_rect_t fboRect;
+  GLboolean constant_cliprect; /* use for FBO or DRI2 rendering */
+  GLboolean front_cliprects;
+
+  struct {
+      struct gl_fragment_program *bitmap_fp;
+      struct gl_vertex_program *passthrough_vp;
+
+      struct gl_fragment_program *saved_fp;
+      GLboolean saved_fp_enable;
+      struct gl_vertex_program *saved_vp;
+      GLboolean saved_vp_enable;
+
+      GLint saved_vp_x, saved_vp_y;
+      GLsizei saved_vp_width, saved_vp_height;
+      GLenum saved_matrix_mode;
+   } meta;
 
    struct {
 	   void (*get_lock)(radeonContextPtr radeon);
diff --git a/src/mesa/drivers/dri/radeon/radeon_context.c b/src/mesa/drivers/dri/radeon/radeon_context.c
index 47ebe41c6d..ac945ecc4d 100644
--- a/src/mesa/drivers/dri/radeon/radeon_context.c
+++ b/src/mesa/drivers/dri/radeon/radeon_context.c
@@ -66,6 +66,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define need_GL_EXT_blend_minmax
 #define need_GL_EXT_fog_coord
 #define need_GL_EXT_secondary_color
+#define need_GL_EXT_framebuffer_object
 #include "extension_helper.h"
 
 #define DRIVER_DATE	"20061018"
@@ -88,6 +89,7 @@ const struct dri_extension card_extensions[] =
     { "GL_EXT_blend_logic_op",             NULL },
     { "GL_EXT_blend_subtract",             GL_EXT_blend_minmax_functions },
     { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
+    { "GL_EXT_packed_depth_stencil",	   NULL},
     { "GL_EXT_secondary_color",            GL_EXT_secondary_color_functions },
     { "GL_EXT_stencil_wrap",               NULL },
     { "GL_EXT_texture_edge_clamp",         NULL },
@@ -104,6 +106,11 @@ const struct dri_extension card_extensions[] =
     { NULL,                                NULL }
 };
 
+const struct dri_extension mm_extensions[] = {
+  { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions },
+  { NULL, NULL }
+};
+
 extern const struct tnl_pipeline_stage _radeon_render_stage;
 extern const struct tnl_pipeline_stage _radeon_tcl_stage;
 
@@ -338,6 +345,8 @@ radeonCreateContext( const __GLcontextModes *glVisual,
    }
 
    driInitExtensions( ctx, card_extensions, GL_TRUE );
+   if (rmesa->radeon.radeonScreen->kernel_mm)
+     driInitExtensions(ctx, mm_extensions, GL_FALSE);
    if (rmesa->radeon.radeonScreen->drmSupportsCubeMapsR100)
       _mesa_enable_extension( ctx, "GL_ARB_texture_cube_map" );
    if (rmesa->radeon.glCtx->Mesa_DXTn) {
@@ -352,6 +361,7 @@ radeonCreateContext( const __GLcontextModes *glVisual,
       _mesa_enable_extension( ctx, "GL_NV_texture_rectangle");
 
    /* XXX these should really go right after _mesa_init_driver_functions() */
+   radeon_fbo_init(&rmesa->radeon);
    radeonInitSpanFuncs( ctx );
    radeonInitIoctlFuncs( ctx );
    radeonInitStateFuncs( ctx );
diff --git a/src/mesa/drivers/dri/radeon/radeon_context.h b/src/mesa/drivers/dri/radeon/radeon_context.h
index 2015e96a74..5235a6e374 100644
--- a/src/mesa/drivers/dri/radeon/radeon_context.h
+++ b/src/mesa/drivers/dri/radeon/radeon_context.h
@@ -435,20 +435,6 @@ struct r100_context {
 	GLuint c_textureBytes;
 	GLuint c_vertexBuffers;
 
-  struct {
-      struct gl_fragment_program *bitmap_fp;
-      struct gl_vertex_program *passthrough_vp;
-
-      struct gl_fragment_program *saved_fp;
-      GLboolean saved_fp_enable;
-      struct gl_vertex_program *saved_vp;
-      GLboolean saved_vp_enable;
-
-      GLint saved_vp_x, saved_vp_y;
-      GLsizei saved_vp_width, saved_vp_height;
-      GLenum saved_matrix_mode;
-   } meta;
-
 };
 
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_fbo.c b/src/mesa/drivers/dri/radeon/radeon_fbo.c
new file mode 100644
index 0000000000..f914c8c8d0
--- /dev/null
+++ b/src/mesa/drivers/dri/radeon/radeon_fbo.c
@@ -0,0 +1,574 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Red Hat Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "main/imports.h"
+#include "main/macros.h"
+#include "main/mtypes.h"
+#include "main/fbobject.h"
+#include "main/framebuffer.h"
+#include "main/renderbuffer.h"
+#include "main/context.h"
+#include "main/texformat.h"
+#include "main/texrender.h"
+
+#include "radeon_common.h"
+#include "radeon_mipmap_tree.h"
+
+#define FILE_DEBUG_FLAG DEBUG_TEXTURE
+#define DBG(...) do {                                           \
+        if (RADEON_DEBUG & FILE_DEBUG_FLAG)                      \
+                _mesa_printf(__VA_ARGS__);                      \
+} while(0)
+
+static struct gl_framebuffer *
+radeon_new_framebuffer(GLcontext *ctx, GLuint name)
+{
+  return _mesa_new_framebuffer(ctx, name);
+}
+
+static void
+radeon_delete_renderbuffer(struct gl_renderbuffer *rb)
+{
+  struct radeon_renderbuffer *rrb = radeon_renderbuffer(rb);
+
+  ASSERT(rrb);
+
+  if (rrb && rrb->bo) {
+    radeon_bo_unref(rrb->bo);
+  }
+  _mesa_free(rrb);
+}
+
+static void *
+radeon_get_pointer(GLcontext *ctx, struct gl_renderbuffer *rb,
+		   GLint x, GLint y)
+{
+  return NULL;
+}
+
+/**
+ * Called via glRenderbufferStorageEXT() to set the format and allocate
+ * storage for a user-created renderbuffer.
+ */
+static GLboolean
+radeon_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
+                                 GLenum internalFormat,
+                                 GLuint width, GLuint height)
+{
+  struct radeon_context *radeon = RADEON_CONTEXT(ctx);
+  struct radeon_renderbuffer *rrb = radeon_renderbuffer(rb);
+  GLboolean software_buffer = GL_FALSE;
+  int cpp;
+
+   ASSERT(rb->Name != 0);
+  switch (internalFormat) {
+   case GL_R3_G3_B2:
+   case GL_RGB4:
+   case GL_RGB5:
+      rb->_ActualFormat = GL_RGB5;
+      rb->DataType = GL_UNSIGNED_BYTE;
+      rb->RedBits = 5;
+      rb->GreenBits = 6;
+      rb->BlueBits = 5;
+      cpp = 2;
+      break;
+   case GL_RGB:
+   case GL_RGB8:
+   case GL_RGB10:
+   case GL_RGB12:
+   case GL_RGB16:
+      rb->_ActualFormat = GL_RGB8;
+      rb->DataType = GL_UNSIGNED_BYTE;
+      rb->RedBits = 8;
+      rb->GreenBits = 8;
+      rb->BlueBits = 8;
+      rb->AlphaBits = 0;
+      cpp = 4;
+      break;
+   case GL_RGBA:
+   case GL_RGBA2:
+   case GL_RGBA4:
+   case GL_RGB5_A1:
+   case GL_RGBA8:
+   case GL_RGB10_A2:
+   case GL_RGBA12:
+   case GL_RGBA16:
+      rb->_ActualFormat = GL_RGBA8;
+      rb->DataType = GL_UNSIGNED_BYTE;
+      rb->RedBits = 8;
+      rb->GreenBits = 8;
+      rb->BlueBits = 8;
+      rb->AlphaBits = 8;
+      cpp = 4;
+      break;
+   case GL_STENCIL_INDEX:
+   case GL_STENCIL_INDEX1_EXT:
+   case GL_STENCIL_INDEX4_EXT:
+   case GL_STENCIL_INDEX8_EXT:
+   case GL_STENCIL_INDEX16_EXT:
+      /* alloc a depth+stencil buffer */
+      rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT;
+      rb->DataType = GL_UNSIGNED_INT_24_8_EXT;
+      rb->StencilBits = 8;
+      cpp = 4;
+      break;
+   case GL_DEPTH_COMPONENT16:
+      rb->_ActualFormat = GL_DEPTH_COMPONENT16;
+      rb->DataType = GL_UNSIGNED_SHORT;
+      rb->DepthBits = 16;
+      cpp = 2;
+      break;
+   case GL_DEPTH_COMPONENT:
+   case GL_DEPTH_COMPONENT24:
+   case GL_DEPTH_COMPONENT32:
+      rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT;
+      rb->DataType = GL_UNSIGNED_INT_24_8_EXT;
+      rb->DepthBits = 24;
+      cpp = 4;
+      break;
+   case GL_DEPTH_STENCIL_EXT:
+   case GL_DEPTH24_STENCIL8_EXT:
+      rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT;
+      rb->DataType = GL_UNSIGNED_INT_24_8_EXT;
+      rb->DepthBits = 24;
+      rb->StencilBits = 8;
+      cpp = 4;
+      break;
+   default:
+      _mesa_problem(ctx,
+                    "Unexpected format in intel_alloc_renderbuffer_storage");
+      return GL_FALSE;
+   }
+
+  radeonFlush(ctx);
+
+  if (rrb->bo)
+    radeon_bo_unref(rrb->bo);
+  
+    
+   if (software_buffer) {
+      return _mesa_soft_renderbuffer_storage(ctx, rb, internalFormat,
+                                             width, height);
+   }
+   else {
+     uint32_t size = width * height * cpp;
+     uint32_t pitch = ((cpp * width + 63) & ~63) / cpp;
+
+     fprintf(stderr,"Allocating %d x %d radeon RBO (pitch %d)\n", width,
+	  height, pitch);
+
+     rrb->pitch = pitch * cpp;
+     rrb->cpp = cpp;
+     rrb->bo = radeon_bo_open(radeon->radeonScreen->bom,
+			      0,
+			      size,
+			      0,
+			      RADEON_GEM_DOMAIN_VRAM,
+			      0);
+     rb->Width = width;
+     rb->Height = height;
+       return GL_TRUE;
+   }    
+   
+}
+
+
+/**
+ * Called for each hardware renderbuffer when a _window_ is resized.
+ * Just update fields.
+ * Not used for user-created renderbuffers!
+ */
+static GLboolean
+radeon_alloc_window_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
+                           GLenum internalFormat, GLuint width, GLuint height)
+{
+   ASSERT(rb->Name == 0);
+   rb->Width = width;
+   rb->Height = height;
+   rb->_ActualFormat = internalFormat;
+
+   return GL_TRUE;
+}
+
+
+static void
+radeon_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb,
+		     GLuint width, GLuint height)
+{
+     struct radeon_framebuffer *radeon_fb = (struct radeon_framebuffer*)fb;
+   int i;
+
+   _mesa_resize_framebuffer(ctx, fb, width, height);
+
+   fb->Initialized = GL_TRUE; /* XXX remove someday */
+
+   if (fb->Name != 0) {
+      return;
+   }
+
+   /* Make sure all window system renderbuffers are up to date */
+   for (i = 0; i < 2; i++) {
+      struct gl_renderbuffer *rb = &radeon_fb->color_rb[i]->base;
+
+      /* only resize if size is changing */
+      if (rb && (rb->Width != width || rb->Height != height)) {
+	 rb->AllocStorage(ctx, rb, rb->InternalFormat, width, height);
+      }
+   }
+}
+
+
+/** Dummy function for gl_renderbuffer::AllocStorage() */
+static GLboolean
+radeon_nop_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
+			 GLenum internalFormat, GLuint width, GLuint height)
+{
+   _mesa_problem(ctx, "radeon_op_alloc_storage should never be called.");
+   return GL_FALSE;
+}
+
+struct radeon_renderbuffer *
+radeon_create_renderbuffer(GLenum format, __DRIdrawablePrivate *driDrawPriv)
+{
+    struct radeon_renderbuffer *rrb;
+
+    rrb = CALLOC_STRUCT(radeon_renderbuffer);
+    if (!rrb)
+	return NULL;
+
+    _mesa_init_renderbuffer(&rrb->base, 0);
+    rrb->base.ClassID = RADEON_RB_CLASS;
+
+    /* XXX format junk */
+    switch (format) {
+	case GL_RGB5:
+	    rrb->base._ActualFormat = GL_RGB5;
+	    rrb->base._BaseFormat = GL_RGBA;
+	    rrb->base.RedBits = 5;
+	    rrb->base.GreenBits = 6;
+	    rrb->base.BlueBits = 5;
+	    rrb->base.DataType = GL_UNSIGNED_BYTE;
+	    break;
+	case GL_RGBA8:
+	    rrb->base._ActualFormat = GL_RGBA8;
+	    rrb->base._BaseFormat = GL_RGBA;
+	    rrb->base.RedBits = 8;
+	    rrb->base.GreenBits = 8;
+	    rrb->base.BlueBits = 8;
+	    rrb->base.AlphaBits = 8;
+	    rrb->base.DataType = GL_UNSIGNED_BYTE;
+	    break;
+	case GL_STENCIL_INDEX8_EXT:
+	    rrb->base._ActualFormat = GL_STENCIL_INDEX8_EXT;
+	    rrb->base._BaseFormat = GL_STENCIL_INDEX;
+	    rrb->base.StencilBits = 8;
+	    rrb->base.DataType = GL_UNSIGNED_BYTE;
+	    break;
+	case GL_DEPTH_COMPONENT16:
+	    rrb->base._ActualFormat = GL_DEPTH_COMPONENT16;
+	    rrb->base._BaseFormat = GL_DEPTH_COMPONENT;
+	    rrb->base.DepthBits = 16;
+	    rrb->base.DataType = GL_UNSIGNED_SHORT;
+	    break;
+	case GL_DEPTH_COMPONENT24:
+	    rrb->base._ActualFormat = GL_DEPTH24_STENCIL8_EXT;
+	    rrb->base._BaseFormat = GL_DEPTH_COMPONENT;
+	    rrb->base.DepthBits = 24;
+	    rrb->base.DataType = GL_UNSIGNED_INT;
+	    break;
+	case GL_DEPTH24_STENCIL8_EXT:
+	    rrb->base._ActualFormat = GL_DEPTH24_STENCIL8_EXT;
+	    rrb->base._BaseFormat = GL_DEPTH_STENCIL_EXT;
+	    rrb->base.DepthBits = 24;
+	    rrb->base.StencilBits = 8;
+	    rrb->base.DataType = GL_UNSIGNED_INT_24_8_EXT;
+	    break;
+	default:
+	    fprintf(stderr, "%s: Unknown format 0x%04x\n", __FUNCTION__, format);
+	    _mesa_delete_renderbuffer(&rrb->base);
+	    return NULL;
+    }
+
+    rrb->dPriv = driDrawPriv;
+    rrb->base.InternalFormat = format;
+
+    rrb->base.Delete = radeon_delete_renderbuffer;
+    rrb->base.AllocStorage = radeon_alloc_window_storage;
+    rrb->base.GetPointer = radeon_get_pointer;
+
+    rrb->bo = NULL;
+    return rrb;
+}
+
+static struct gl_renderbuffer *
+radeon_new_renderbuffer(GLcontext * ctx, GLuint name)
+{
+  struct radeon_renderbuffer *rrb;
+
+  rrb = CALLOC_STRUCT(radeon_renderbuffer);
+  if (!rrb)
+    return NULL;
+
+  _mesa_init_renderbuffer(&rrb->base, name);
+  rrb->base.ClassID = RADEON_RB_CLASS;
+
+  rrb->base.Delete = radeon_delete_renderbuffer;
+  rrb->base.AllocStorage = radeon_alloc_renderbuffer_storage;
+  rrb->base.GetPointer = radeon_get_pointer;
+
+  return &rrb->base;
+}
+
+static void
+radeon_bind_framebuffer(GLcontext * ctx, GLenum target,
+                       struct gl_framebuffer *fb, struct gl_framebuffer *fbread)
+{
+   if (target == GL_FRAMEBUFFER_EXT || target == GL_DRAW_FRAMEBUFFER_EXT) {
+      radeon_draw_buffer(ctx, fb);
+   }
+   else {
+      /* don't need to do anything if target == GL_READ_FRAMEBUFFER_EXT */
+   }
+}
+
+static void
+radeon_framebuffer_renderbuffer(GLcontext * ctx,
+                               struct gl_framebuffer *fb,
+                               GLenum attachment, struct gl_renderbuffer *rb)
+{
+
+   radeonFlush(ctx);
+
+   _mesa_framebuffer_renderbuffer(ctx, fb, attachment, rb);
+   radeon_draw_buffer(ctx, fb);
+}
+
+
+static GLboolean
+radeon_update_wrapper(GLcontext *ctx, struct radeon_renderbuffer *rrb, 
+		     struct gl_texture_image *texImage)
+{
+   if (texImage->TexFormat == &_mesa_texformat_argb8888) {
+      rrb->cpp = 4;
+      rrb->base._ActualFormat = GL_RGBA8;
+      rrb->base._BaseFormat = GL_RGBA;
+      rrb->base.DataType = GL_UNSIGNED_BYTE;
+      DBG("Render to RGBA8 texture OK\n");
+   }
+   else if (texImage->TexFormat == &_mesa_texformat_argb4444) {
+      rrb->cpp = 2;
+      rrb->base._ActualFormat = GL_RGBA4;
+      rrb->base._BaseFormat = GL_RGBA;
+      rrb->base.DataType = GL_UNSIGNED_BYTE;
+      DBG("Render to RGBA4 texture OK\n");
+   }
+   else if (texImage->TexFormat == &_mesa_texformat_rgb565) {
+      rrb->cpp = 2;
+      rrb->base._ActualFormat = GL_RGB5;
+      rrb->base._BaseFormat = GL_RGB;
+      rrb->base.DataType = GL_UNSIGNED_SHORT;
+      DBG("Render to RGB5 texture OK\n");
+   }
+   else if (texImage->TexFormat == &_mesa_texformat_z16) {
+      rrb->cpp = 2;
+      rrb->base._ActualFormat = GL_DEPTH_COMPONENT16;
+      rrb->base._BaseFormat = GL_DEPTH_COMPONENT;
+      rrb->base.DataType = GL_UNSIGNED_SHORT;
+      DBG("Render to DEPTH16 texture OK\n");
+   }
+   else if (texImage->TexFormat == &_mesa_texformat_s8_z24) {
+      rrb->cpp = 4;
+      rrb->base._ActualFormat = GL_DEPTH24_STENCIL8_EXT;
+      rrb->base._BaseFormat = GL_DEPTH_STENCIL_EXT;
+      rrb->base.DataType = GL_UNSIGNED_INT_24_8_EXT;
+      DBG("Render to DEPTH_STENCIL texture OK\n");
+   }
+   else {
+      DBG("Render to texture BAD FORMAT %d\n",
+	  texImage->TexFormat->MesaFormat);
+      return GL_FALSE;
+   }
+
+   rrb->pitch = texImage->Width * rrb->cpp;
+   rrb->base.InternalFormat = rrb->base._ActualFormat;
+   rrb->base.Width = texImage->Width;
+   rrb->base.Height = texImage->Height;
+   rrb->base.RedBits = texImage->TexFormat->RedBits;
+   rrb->base.GreenBits = texImage->TexFormat->GreenBits;
+   rrb->base.BlueBits = texImage->TexFormat->BlueBits;
+   rrb->base.AlphaBits = texImage->TexFormat->AlphaBits;
+   rrb->base.DepthBits = texImage->TexFormat->DepthBits;
+
+   rrb->base.Delete = radeon_delete_renderbuffer;
+   rrb->base.AllocStorage = radeon_nop_alloc_storage;
+
+   return GL_TRUE;
+}
+
+
+static struct radeon_renderbuffer *
+radeon_wrap_texture(GLcontext * ctx, struct gl_texture_image *texImage)
+{
+  const GLuint name = ~0;   /* not significant, but distinct for debugging */
+  struct radeon_renderbuffer *rrb;
+
+   /* make an radeon_renderbuffer to wrap the texture image */
+   rrb = CALLOC_STRUCT(radeon_renderbuffer);
+   if (!rrb) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glFramebufferTexture");
+      return NULL;
+   }
+
+   _mesa_init_renderbuffer(&rrb->base, name);
+   rrb->base.ClassID = RADEON_RB_CLASS;
+
+   if (!radeon_update_wrapper(ctx, rrb, texImage)) {
+      _mesa_free(rrb);
+      return NULL;
+   }
+
+   return rrb;
+  
+}
+static void
+radeon_render_texture(GLcontext * ctx,
+                     struct gl_framebuffer *fb,
+                     struct gl_renderbuffer_attachment *att)
+{
+   struct gl_texture_image *newImage
+      = att->Texture->Image[att->CubeMapFace][att->TextureLevel];
+   struct radeon_renderbuffer *rrb = radeon_renderbuffer(att->Renderbuffer);
+   radeon_texture_image *radeon_image;
+   GLuint imageOffset;
+
+   (void) fb;
+
+   ASSERT(newImage);
+
+   if (newImage->Border != 0) {
+      /* Fallback on drawing to a texture with a border, which won't have a
+       * miptree.
+       */
+      _mesa_reference_renderbuffer(&att->Renderbuffer, NULL);
+      _mesa_render_texture(ctx, fb, att);
+      return;
+   }
+   else if (!rrb) {
+      rrb = radeon_wrap_texture(ctx, newImage);
+      if (rrb) {
+         /* bind the wrapper to the attachment point */
+         _mesa_reference_renderbuffer(&att->Renderbuffer, &rrb->base);
+      }
+      else {
+         /* fallback to software rendering */
+         _mesa_render_texture(ctx, fb, att);
+         return;
+      }
+   }
+
+   if (!radeon_update_wrapper(ctx, rrb, newImage)) {
+       _mesa_reference_renderbuffer(&att->Renderbuffer, NULL);
+       _mesa_render_texture(ctx, fb, att);
+       return;
+   }
+
+   DBG("Begin render texture tid %x tex=%u w=%d h=%d refcount=%d\n",
+       _glthread_GetID(),
+       att->Texture->Name, newImage->Width, newImage->Height,
+       rrb->base.RefCount);
+
+   /* point the renderbufer's region to the texture image region */
+   radeon_image = (radeon_texture_image *)newImage;
+   if (rrb->bo != radeon_image->mt->bo) {
+      if (rrb->bo)
+  	radeon_bo_unref(rrb->bo);
+      rrb->bo = radeon_image->mt->bo;
+      radeon_bo_ref(rrb->bo);
+   }
+
+   /* compute offset of the particular 2D image within the texture region */
+   imageOffset = radeon_miptree_image_offset(radeon_image->mt,
+                                            att->CubeMapFace,
+                                            att->TextureLevel);
+
+   if (att->Texture->Target == GL_TEXTURE_3D) {
+      const GLuint *offsets = radeon_miptree_depth_offsets(radeon_image->mt,
+                                                          att->TextureLevel);
+      imageOffset += offsets[att->Zoffset];
+   }
+
+   /* store that offset in the region */
+   rrb->draw_offset = imageOffset;
+
+   /* update drawing region, etc */
+   radeon_draw_buffer(ctx, fb);
+}
+
+static void
+radeon_finish_render_texture(GLcontext * ctx,
+                            struct gl_renderbuffer_attachment *att)
+{
+
+}
+static void
+radeon_validate_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb)
+{
+}
+
+static void
+radeon_blit_framebuffer(GLcontext *ctx,
+                       GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1,
+                       GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1,
+                       GLbitfield mask, GLenum filter)
+{
+}
+
+void radeon_fbo_init(struct radeon_context *radeon)
+{
+  radeon->glCtx->Driver.NewFramebuffer = radeon_new_framebuffer;
+  radeon->glCtx->Driver.NewRenderbuffer = radeon_new_renderbuffer;
+  radeon->glCtx->Driver.BindFramebuffer = radeon_bind_framebuffer;
+  radeon->glCtx->Driver.FramebufferRenderbuffer = radeon_framebuffer_renderbuffer;
+  radeon->glCtx->Driver.RenderTexture = radeon_render_texture;
+  radeon->glCtx->Driver.FinishRenderTexture = radeon_finish_render_texture;
+  radeon->glCtx->Driver.ResizeBuffers = radeon_resize_buffers;
+  radeon->glCtx->Driver.ValidateFramebuffer = radeon_validate_framebuffer;
+  radeon->glCtx->Driver.BlitFramebuffer = radeon_blit_framebuffer;
+}
+
+  
+void radeon_renderbuffer_set_bo(struct radeon_renderbuffer *rb,
+				struct radeon_bo *bo)
+{
+  struct radeon_bo *old;
+  old = rb->bo;
+  rb->bo = bo;
+  radeon_bo_ref(bo);
+  if (old)
+    radeon_bo_unref(old);
+}
diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.c b/src/mesa/drivers/dri/radeon/radeon_ioctl.c
index 22584f4817..f18aa1a4da 100644
--- a/src/mesa/drivers/dri/radeon/radeon_ioctl.c
+++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.c
@@ -419,218 +419,6 @@ void radeonEmitAOS( r100ContextPtr rmesa,
  */
 #define RADEON_MAX_CLEARS	256
 
-
-
-static void
-r100_meta_set_passthrough_transform(r100ContextPtr r100)
-{
-   GLcontext *ctx = r100->radeon.glCtx;
-
-   r100->meta.saved_vp_x = ctx->Viewport.X;
-   r100->meta.saved_vp_y = ctx->Viewport.Y;
-   r100->meta.saved_vp_width = ctx->Viewport.Width;
-   r100->meta.saved_vp_height = ctx->Viewport.Height;
-   r100->meta.saved_matrix_mode = ctx->Transform.MatrixMode;
-
-   _mesa_Viewport(0, 0, ctx->DrawBuffer->Width, ctx->DrawBuffer->Height);
-
-   _mesa_MatrixMode(GL_PROJECTION);
-   _mesa_PushMatrix();
-   _mesa_LoadIdentity();
-   _mesa_Ortho(0, ctx->DrawBuffer->Width, 0, ctx->DrawBuffer->Height, 1, -1);
-
-   _mesa_MatrixMode(GL_MODELVIEW);
-   _mesa_PushMatrix();
-   _mesa_LoadIdentity();
-}
-
-static void
-r100_meta_restore_transform(r100ContextPtr r100)
-{
-   _mesa_MatrixMode(GL_PROJECTION);
-   _mesa_PopMatrix();
-   _mesa_MatrixMode(GL_MODELVIEW);
-   _mesa_PopMatrix();
-
-   _mesa_MatrixMode(r100->meta.saved_matrix_mode);
-
-   _mesa_Viewport(r100->meta.saved_vp_x, r100->meta.saved_vp_y,
-		  r100->meta.saved_vp_width, r100->meta.saved_vp_height);
-}
-
-/**
- * Perform glClear where mask contains only color, depth, and/or stencil.
- *
- * The implementation is based on calling into Mesa to set GL state and
- * performing normal triangle rendering.  The intent of this path is to
- * have as generic a path as possible, so that any driver could make use of
- * it.
- */
-static void radeon_clear_tris(GLcontext *ctx, GLbitfield mask)
-{
-   r100ContextPtr rmesa = R100_CONTEXT(ctx);
-   GLfloat vertices[4][3];
-   GLfloat color[4][4];
-   GLfloat dst_z;
-   struct gl_framebuffer *fb = ctx->DrawBuffer;
-   int i;
-   GLboolean saved_fp_enable = GL_FALSE, saved_vp_enable = GL_FALSE;
-   GLboolean saved_shader_program = 0;
-   unsigned int saved_active_texture;
-   
-   assert((mask & ~(BUFFER_BIT_BACK_LEFT | BUFFER_BIT_FRONT_LEFT |
-		    BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL)) == 0);
-
-   _mesa_PushAttrib(GL_COLOR_BUFFER_BIT |
-		    GL_CURRENT_BIT |
-		    GL_DEPTH_BUFFER_BIT |
-		    GL_ENABLE_BIT |
-		    GL_STENCIL_BUFFER_BIT |
-		    GL_TRANSFORM_BIT |
-		    GL_CURRENT_BIT);
-   _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT);
-   saved_active_texture = ctx->Texture.CurrentUnit;
-  
-  /* Disable existing GL state we don't want to apply to a clear. */
-   _mesa_Disable(GL_ALPHA_TEST);
-   _mesa_Disable(GL_BLEND);
-   _mesa_Disable(GL_CULL_FACE);
-   _mesa_Disable(GL_FOG);
-   _mesa_Disable(GL_POLYGON_SMOOTH);
-   _mesa_Disable(GL_POLYGON_STIPPLE);
-   _mesa_Disable(GL_POLYGON_OFFSET_FILL);
-   _mesa_Disable(GL_LIGHTING);
-   _mesa_Disable(GL_CLIP_PLANE0);
-   _mesa_Disable(GL_CLIP_PLANE1);
-   _mesa_Disable(GL_CLIP_PLANE2);
-   _mesa_Disable(GL_CLIP_PLANE3);
-   _mesa_Disable(GL_CLIP_PLANE4);
-   _mesa_Disable(GL_CLIP_PLANE5);
-   if (ctx->Extensions.ARB_fragment_program && ctx->FragmentProgram.Enabled) {
-      saved_fp_enable = GL_TRUE;
-      _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
-   }
-   if (ctx->Extensions.ARB_vertex_program && ctx->VertexProgram.Enabled) {
-      saved_vp_enable = GL_TRUE;
-      _mesa_Disable(GL_VERTEX_PROGRAM_ARB);
-   }
-   if (ctx->Extensions.ARB_shader_objects && ctx->Shader.CurrentProgram) {
-      saved_shader_program = ctx->Shader.CurrentProgram->Name;
-      _mesa_UseProgramObjectARB(0);
-   }
-   
-   if (ctx->Texture._EnabledUnits != 0) {
-      int i;
-      
-      for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-	 _mesa_ActiveTextureARB(GL_TEXTURE0 + i);
-	 _mesa_Disable(GL_TEXTURE_1D);
-	 _mesa_Disable(GL_TEXTURE_2D);
-	 _mesa_Disable(GL_TEXTURE_3D);
-	 if (ctx->Extensions.ARB_texture_cube_map)
-	    _mesa_Disable(GL_TEXTURE_CUBE_MAP_ARB);
-	 if (ctx->Extensions.NV_texture_rectangle)
-	    _mesa_Disable(GL_TEXTURE_RECTANGLE_NV);
-	 if (ctx->Extensions.MESA_texture_array) {
-	    _mesa_Disable(GL_TEXTURE_1D_ARRAY_EXT);
-	    _mesa_Disable(GL_TEXTURE_2D_ARRAY_EXT);
-	 }
-      }
-   }
-  
-   r100_meta_set_passthrough_transform(rmesa);
-   
-   for (i = 0; i < 4; i++) {
-      color[i][0] = ctx->Color.ClearColor[0];
-      color[i][1] = ctx->Color.ClearColor[1];
-      color[i][2] = ctx->Color.ClearColor[2];
-      color[i][3] = ctx->Color.ClearColor[3];
-   }
-
-   /* convert clear Z from [0,1] to NDC coord in [-1,1] */
-   dst_z = -1.0 + 2.0 * ctx->Depth.Clear;
-   
-   /* Prepare the vertices, which are the same regardless of which buffer we're
-    * drawing to.
-    */
-   vertices[0][0] = fb->_Xmin;
-   vertices[0][1] = fb->_Ymin;
-   vertices[0][2] = dst_z;
-   vertices[1][0] = fb->_Xmax;
-   vertices[1][1] = fb->_Ymin;
-   vertices[1][2] = dst_z;
-   vertices[2][0] = fb->_Xmax;
-   vertices[2][1] = fb->_Ymax;
-   vertices[2][2] = dst_z;
-   vertices[3][0] = fb->_Xmin;
-   vertices[3][1] = fb->_Ymax;
-   vertices[3][2] = dst_z;
-
-   _mesa_ColorPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &color);
-   _mesa_VertexPointer(3, GL_FLOAT, 3 * sizeof(GLfloat), &vertices);
-   _mesa_Enable(GL_COLOR_ARRAY);
-   _mesa_Enable(GL_VERTEX_ARRAY);
-
-   while (mask != 0) {
-      GLuint this_mask = 0;
-
-      if (mask & BUFFER_BIT_BACK_LEFT)
-	 this_mask = BUFFER_BIT_BACK_LEFT;
-      else if (mask & BUFFER_BIT_FRONT_LEFT)
-	 this_mask = BUFFER_BIT_FRONT_LEFT;
-
-      /* Clear depth/stencil in the same pass as color. */
-      this_mask |= (mask & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL));
-
-      /* Select the current color buffer and use the color write mask if
-       * we have one, otherwise don't write any color channels.
-       */
-      if (this_mask & BUFFER_BIT_FRONT_LEFT)
-	 _mesa_DrawBuffer(GL_FRONT_LEFT);
-      else if (this_mask & BUFFER_BIT_BACK_LEFT)
-	 _mesa_DrawBuffer(GL_BACK_LEFT);
-      else
-	 _mesa_ColorMask(GL_FALSE, GL_FALSE, GL_FALSE, GL_FALSE);
-
-      /* Control writing of the depth clear value to depth. */
-      if (this_mask & BUFFER_BIT_DEPTH) {
-	 _mesa_DepthFunc(GL_ALWAYS);
-	 _mesa_Enable(GL_DEPTH_TEST);
-      } else {
-	 _mesa_Disable(GL_DEPTH_TEST);
-	 _mesa_DepthMask(GL_FALSE);
-      }
-
-      /* Control writing of the stencil clear value to stencil. */
-      if (this_mask & BUFFER_BIT_STENCIL) {
-	 _mesa_Enable(GL_STENCIL_TEST);
-	 _mesa_StencilOp(GL_REPLACE, GL_REPLACE, GL_REPLACE);
-	 _mesa_StencilFuncSeparate(GL_FRONT, GL_ALWAYS, ctx->Stencil.Clear,
-				   ctx->Stencil.WriteMask[0]);
-      } else {
-	 _mesa_Disable(GL_STENCIL_TEST);
-      }
-
-      CALL_DrawArrays(ctx->Exec, (GL_TRIANGLE_FAN, 0, 4));
-
-      mask &= ~this_mask;
-   }
-
-   r100_meta_restore_transform(rmesa);
-
-   _mesa_ActiveTextureARB(GL_TEXTURE0 + saved_active_texture);
-   if (saved_fp_enable)
-      _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB);
-   if (saved_vp_enable)
-      _mesa_Enable(GL_VERTEX_PROGRAM_ARB);
-
-   if (saved_shader_program)
-      _mesa_UseProgramObjectARB(saved_shader_program);
-
-   _mesa_PopClientAttrib();
-   _mesa_PopAttrib();
-}
-
 static void radeonUserClear(GLcontext *ctx, GLuint mask)
 {
    radeon_clear_tris(ctx, mask);
@@ -795,7 +583,7 @@ static void radeonClear( GLcontext *ctx, GLbitfield mask )
       mask &= ~BUFFER_BIT_DEPTH;
    }
 
-   if ( (mask & BUFFER_BIT_STENCIL) && rmesa->radeon.state.stencil.hwBuffer ) {
+   if ( (mask & BUFFER_BIT_STENCIL) ) {
       flags |= RADEON_STENCIL;
       mask &= ~BUFFER_BIT_STENCIL;
    }
@@ -813,8 +601,7 @@ static void radeonClear( GLcontext *ctx, GLbitfield mask )
       flags |= RADEON_USE_COMP_ZBUF;
 /*      if (rmesa->radeon.radeonScreen->chipset & RADEON_CHIPSET_TCL) 
          flags |= RADEON_USE_HIERZ; */
-      if (!(rmesa->radeon.state.stencil.hwBuffer) ||
-	 ((flags & RADEON_DEPTH) && (flags & RADEON_STENCIL) &&
+      if (((flags & RADEON_DEPTH) && (flags & RADEON_STENCIL) &&
 	    ((rmesa->radeon.state.stencil.clear & RADEON_STENCIL_WRITE_MASK) == RADEON_STENCIL_WRITE_MASK))) {
 	  flags |= RADEON_CLEAR_FASTZ;
       }
diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c
index 3203ee1cba..228629e3c4 100644
--- a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c
+++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c
@@ -358,3 +358,29 @@ void radeon_try_alloc_miptree(radeonContextPtr rmesa, radeonTexObj *t,
 		texImage->Width, texImage->Height, texImage->Depth,
 		texImage->TexFormat->TexelBytes, t->tile_bits, compressed);
 }
+
+/* Although we use the image_offset[] array to store relative offsets
+ * to cube faces, Mesa doesn't know anything about this and expects
+ * each cube face to be treated as a separate image.
+ *
+ * These functions present that view to mesa:
+ */
+const GLuint *
+radeon_miptree_depth_offsets(radeon_mipmap_tree *mt, GLuint level)
+{
+     static const GLuint zero = 0;
+     if (mt->target != GL_TEXTURE_3D || mt->faces == 1)
+       return &zero;
+     else
+       return mt->levels[level].faces[0].offset;
+}
+
+GLuint
+radeon_miptree_image_offset(radeon_mipmap_tree *mt,
+			    GLuint face, GLuint level)
+{
+   if (mt->target == GL_TEXTURE_CUBE_MAP_ARB)
+      return (mt->levels[level].faces[face].offset);
+   else
+      return mt->levels[level].faces[0].offset;
+}
diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h
index 43dfa48aa7..d9ad5ad39a 100644
--- a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h
+++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h
@@ -92,6 +92,8 @@ GLboolean radeon_miptree_matches_image(radeon_mipmap_tree *mt,
 GLboolean radeon_miptree_matches_texture(radeon_mipmap_tree *mt, struct gl_texture_object *texObj);
 void radeon_try_alloc_miptree(radeonContextPtr rmesa, radeonTexObj *t,
 			      struct gl_texture_image *texImage, GLuint face, GLuint level);
-
-
+GLuint radeon_miptree_image_offset(radeon_mipmap_tree *mt,
+				   GLuint face, GLuint level);
+const GLuint *
+radeon_miptree_depth_offsets(radeon_mipmap_tree *mt, GLuint level);
 #endif /* __RADEON_MIPMAP_TREE_H_ */
diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c
index bbcf19ccee..2d3bb619cd 100644
--- a/src/mesa/drivers/dri/radeon/radeon_screen.c
+++ b/src/mesa/drivers/dri/radeon/radeon_screen.c
@@ -1089,6 +1089,9 @@ radeonCreateScreen2(__DRIscreenPrivate *sPriv)
    else
       screen->chip_flags |= RADEON_CLASS_R300;
 
+   if (getenv("R300_NO_TCL"))
+     screen->chip_flags &= ~RADEON_CHIPSET_TCL;
+
    i = 0;
    screen->extensions[i++] = &driCopySubBufferExtension.base;
    screen->extensions[i++] = &driFrameTrackingExtension.base;
@@ -1178,90 +1181,7 @@ radeonInitDriver( __DRIscreenPrivate *sPriv )
     return GL_TRUE;
 }
 
-static GLboolean
-radeon_alloc_window_storage(GLcontext *ctx, struct gl_renderbuffer *rb,
-			    GLenum intFormat, GLuint w, GLuint h)
-{
-    rb->Width = w;
-    rb->Height = h;
-    rb->_ActualFormat = intFormat;
 
-    return GL_TRUE;
-}
-
-
-static struct radeon_renderbuffer *
-radeon_create_renderbuffer(GLenum format, __DRIdrawablePrivate *driDrawPriv)
-{
-    struct radeon_renderbuffer *ret;
-
-    ret = CALLOC_STRUCT(radeon_renderbuffer);
-    if (!ret)
-	return NULL;
-
-    _mesa_init_renderbuffer(&ret->base, 0);
-    ret->base.ClassID = RADEON_RB_CLASS;
-
-    /* XXX format junk */
-    switch (format) {
-	case GL_RGB5:
-	    ret->base._ActualFormat = GL_RGB5;
-	    ret->base._BaseFormat = GL_RGBA;
-	    ret->base.RedBits = 5;
-	    ret->base.GreenBits = 6;
-	    ret->base.BlueBits = 5;
-	    ret->base.DataType = GL_UNSIGNED_BYTE;
-	    break;
-	case GL_RGBA8:
-	    ret->base._ActualFormat = GL_RGBA8;
-	    ret->base._BaseFormat = GL_RGBA;
-	    ret->base.RedBits = 8;
-	    ret->base.GreenBits = 8;
-	    ret->base.BlueBits = 8;
-	    ret->base.AlphaBits = 8;
-	    ret->base.DataType = GL_UNSIGNED_BYTE;
-	    break;
-	case GL_STENCIL_INDEX8_EXT:
-	    ret->base._ActualFormat = GL_STENCIL_INDEX8_EXT;
-	    ret->base._BaseFormat = GL_STENCIL_INDEX;
-	    ret->base.StencilBits = 8;
-	    ret->base.DataType = GL_UNSIGNED_BYTE;
-	    break;
-	case GL_DEPTH_COMPONENT16:
-	    ret->base._ActualFormat = GL_DEPTH_COMPONENT16;
-	    ret->base._BaseFormat = GL_DEPTH_COMPONENT;
-	    ret->base.DepthBits = 16;
-	    ret->base.DataType = GL_UNSIGNED_SHORT;
-	    break;
-	case GL_DEPTH_COMPONENT24:
-	    ret->base._ActualFormat = GL_DEPTH24_STENCIL8_EXT;
-	    ret->base._BaseFormat = GL_DEPTH_COMPONENT;
-	    ret->base.DepthBits = 24;
-	    ret->base.DataType = GL_UNSIGNED_INT;
-	    break;
-	case GL_DEPTH24_STENCIL8_EXT:
-	    ret->base._ActualFormat = GL_DEPTH24_STENCIL8_EXT;
-	    ret->base._BaseFormat = GL_DEPTH_STENCIL_EXT;
-	    ret->base.DepthBits = 24;
-	    ret->base.StencilBits = 8;
-	    ret->base.DataType = GL_UNSIGNED_INT_24_8_EXT;
-	    break;
-	default:
-	    fprintf(stderr, "%s: Unknown format 0x%04x\n", __FUNCTION__, format);
-	    _mesa_delete_renderbuffer(&ret->base);
-	    return NULL;
-    }
-
-    ret->dPriv = driDrawPriv;
-    ret->base.InternalFormat = format;
-
-    ret->base.AllocStorage = radeon_alloc_window_storage;
-
-    radeonSetSpanFunctions(ret);
-
-    ret->bo = NULL;
-    return ret;
-}
 
 /**
  * Create the Mesa framebuffer and renderbuffers for a given window/drawable.
@@ -1283,7 +1203,6 @@ radeonCreateBuffer( __DRIscreenPrivate *driScrnPriv,
     const GLboolean swStencil = mesaVis->stencilBits > 0 &&
 	mesaVis->depthBits != 24;
     GLenum rgbFormat = (mesaVis->redBits == 5 ? GL_RGB5 : GL_RGBA8);
-    GLenum depthFormat = GL_NONE;
     struct radeon_framebuffer *rfb;
 
     if (isPixmap)
@@ -1295,11 +1214,6 @@ radeonCreateBuffer( __DRIscreenPrivate *driScrnPriv,
 
     _mesa_initialize_framebuffer(&rfb->base, mesaVis);
 
-    if (mesaVis->depthBits == 16)
-	depthFormat = GL_DEPTH_COMPONENT16;
-    else if (mesaVis->depthBits == 24)
-	depthFormat = GL_DEPTH_COMPONENT24;
-
     /* front color renderbuffer */
     rfb->color_rb[0] = radeon_create_renderbuffer(rgbFormat, driDrawPriv);
     _mesa_add_renderbuffer(&rfb->base, BUFFER_FRONT_LEFT, &rfb->color_rb[0]->base);
@@ -1307,25 +1221,28 @@ radeonCreateBuffer( __DRIscreenPrivate *driScrnPriv,
 
     /* back color renderbuffer */
     if (mesaVis->doubleBufferMode) {
-	rfb->color_rb[1] = radeon_create_renderbuffer(rgbFormat, driDrawPriv);
+      rfb->color_rb[1] = radeon_create_renderbuffer(rgbFormat, driDrawPriv);
 	_mesa_add_renderbuffer(&rfb->base, BUFFER_BACK_LEFT, &rfb->color_rb[1]->base);
 	rfb->color_rb[1]->has_surface = 1;
     }
 
-    /* depth renderbuffer */
-    if (depthFormat != GL_NONE) {
-	struct radeon_renderbuffer *depth =
-	    radeon_create_renderbuffer(depthFormat, driDrawPriv);
+    if (mesaVis->depthBits == 24) {
+      if (mesaVis->stencilBits == 8) {
+	struct radeon_renderbuffer *depthStencilRb = radeon_create_renderbuffer(GL_DEPTH24_STENCIL8_EXT, driDrawPriv);
+	_mesa_add_renderbuffer(&rfb->base, BUFFER_DEPTH, &depthStencilRb->base);
+	_mesa_add_renderbuffer(&rfb->base, BUFFER_STENCIL, &depthStencilRb->base);
+	depthStencilRb->has_surface = screen->depthHasSurface;
+      } else {
+	/* depth renderbuffer */
+	struct radeon_renderbuffer *depth = radeon_create_renderbuffer(GL_DEPTH_COMPONENT24, driDrawPriv);
+	_mesa_add_renderbuffer(&rfb->base, BUFFER_DEPTH, &depth->base);
+	depth->has_surface = screen->depthHasSurface;
+      }
+    } else if (mesaVis->depthBits == 16) {
+      /* just 16-bit depth buffer, no hw stencil */
+	struct radeon_renderbuffer *depth = radeon_create_renderbuffer(GL_DEPTH_COMPONENT16, driDrawPriv);
 	_mesa_add_renderbuffer(&rfb->base, BUFFER_DEPTH, &depth->base);
 	depth->has_surface = screen->depthHasSurface;
-    }
-
-    /* stencil renderbuffer */
-    if (mesaVis->stencilBits > 0 && !swStencil) {
-	struct radeon_renderbuffer *stencil =
-	    radeon_create_renderbuffer(GL_STENCIL_INDEX8_EXT, driDrawPriv);
-	_mesa_add_renderbuffer(&rfb->base, BUFFER_STENCIL, &stencil->base);
-	stencil->has_surface = screen->depthHasSurface;
     }
 
     _mesa_add_soft_renderbuffers(&rfb->base,
diff --git a/src/mesa/drivers/dri/radeon/radeon_span.c b/src/mesa/drivers/dri/radeon/radeon_span.c
index 49ec2c378e..3d2c5da4c0 100644
--- a/src/mesa/drivers/dri/radeon/radeon_span.c
+++ b/src/mesa/drivers/dri/radeon/radeon_span.c
@@ -49,19 +49,17 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #define DBG 0
 
+static void radeonSetSpanFunctions(struct radeon_renderbuffer *rrb);
+
 static GLubyte *radeon_ptr32(const struct radeon_renderbuffer * rrb,
 			     GLint x, GLint y)
 {
     GLubyte *ptr = rrb->bo->ptr;
-    const __DRIdrawablePrivate *dPriv = rrb->dPriv;
     uint32_t mask = RADEON_BO_FLAGS_MACRO_TILE | RADEON_BO_FLAGS_MICRO_TILE;
     GLint offset;
     GLint nmacroblkpl;
     GLint nmicroblkpl;
 
-    x += dPriv->x;
-    y += dPriv->y;
-
     if (rrb->has_surface || !(rrb->bo->flags & mask)) {
         offset = x * rrb->cpp + y * rrb->pitch;
     } else {
@@ -97,15 +95,11 @@ static GLubyte *radeon_ptr16(const struct radeon_renderbuffer * rrb,
 			     GLint x, GLint y)
 {
     GLubyte *ptr = rrb->bo->ptr;
-    const __DRIdrawablePrivate *dPriv = rrb->dPriv;
     uint32_t mask = RADEON_BO_FLAGS_MACRO_TILE | RADEON_BO_FLAGS_MICRO_TILE;
     GLint offset;
     GLint nmacroblkpl;
     GLint nmicroblkpl;
 
-    x += dPriv->x;
-    y += dPriv->y;
-
     if (rrb->has_surface || !(rrb->bo->flags & mask)) {
         offset = x * rrb->cpp + y * rrb->pitch;
     } else {
@@ -141,7 +135,6 @@ static GLubyte *radeon_ptr(const struct radeon_renderbuffer * rrb,
 			   GLint x, GLint y)
 {
     GLubyte *ptr = rrb->bo->ptr;
-    const __DRIdrawablePrivate *dPriv = rrb->dPriv;
     uint32_t mask = RADEON_BO_FLAGS_MACRO_TILE | RADEON_BO_FLAGS_MICRO_TILE;
     GLint offset;
     GLint microblkxs;
@@ -149,9 +142,6 @@ static GLubyte *radeon_ptr(const struct radeon_renderbuffer * rrb,
     GLint nmacroblkpl;
     GLint nmicroblkpl;
 
-    x += dPriv->x;
-    y += dPriv->y;
-
     if (rrb->has_surface || !(rrb->bo->flags & mask)) {
         offset = x * rrb->cpp + y * rrb->pitch;
     } else {
@@ -195,25 +185,48 @@ static GLubyte *radeon_ptr(const struct radeon_renderbuffer * rrb,
  * information.
  */
 #define LOCAL_VARS						\
+   struct radeon_context *radeon = RADEON_CONTEXT(ctx);			\
    struct radeon_renderbuffer *rrb = (void *) rb;		\
-   const __DRIdrawablePrivate *dPriv = rrb->dPriv;		\
-   const GLuint bottom = dPriv->h - 1;				\
+   const GLint yScale = ctx->DrawBuffer->Name ? 1 : -1;			\
+   const GLint yBias = ctx->DrawBuffer->Name ? 0 : rrb->base.Height - 1;\
+   unsigned int num_cliprects;						\
+   struct drm_clip_rect *cliprects;					\
+   int x_off, y_off;							\
    GLuint p;						\
-   (void)p;
+   (void)p;						\
+   radeon_get_cliprects(radeon, &cliprects, &num_cliprects, &x_off, &y_off);
 
 #define LOCAL_DEPTH_VARS				\
+   struct radeon_context *radeon = RADEON_CONTEXT(ctx);			\
    struct radeon_renderbuffer *rrb = (void *) rb;	\
-   const __DRIdrawablePrivate *dPriv = rrb->dPriv;	\
-   const GLuint bottom = dPriv->h - 1;
+   const GLint yScale = ctx->DrawBuffer->Name ? 1 : -1;			\
+   const GLint yBias = ctx->DrawBuffer->Name ? 0 : rrb->base.Height - 1;\
+   unsigned int num_cliprects;						\
+   struct drm_clip_rect *cliprects;					\
+   int x_off, y_off;							\
+  radeon_get_cliprects(radeon, &cliprects, &num_cliprects, &x_off, &y_off);
 
 #define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS
 
-#define Y_FLIP(Y) (bottom - (Y))
+#define Y_FLIP(_y) ((_y) * yScale + yBias)
 
 #define HW_LOCK()
 
 #define HW_UNLOCK()
 
+/* XXX FBO: this is identical to the macro in spantmp2.h except we get
+ * the cliprect info from the context, not the driDrawable.
+ * Move this into spantmp2.h someday.
+ */
+#define HW_CLIPLOOP()							\
+   do {									\
+      int _nc = num_cliprects;						\
+      while ( _nc-- ) {							\
+	 int minx = cliprects[_nc].x1 - x_off;				\
+	 int miny = cliprects[_nc].y1 - y_off;				\
+	 int maxx = cliprects[_nc].x2 - x_off;				\
+	 int maxy = cliprects[_nc].y2 - y_off;
+	
 /* ================================================================
  * Color buffer
  */
@@ -225,7 +238,17 @@ static GLubyte *radeon_ptr(const struct radeon_renderbuffer * rrb,
 
 #define TAG(x)    radeon##x##_RGB565
 #define TAG2(x,y) radeon##x##_RGB565##y
-#define GET_PTR(X,Y) radeon_ptr16(rrb, (X), (Y))
+#define GET_PTR(X,Y) radeon_ptr16(rrb, (X) + x_off, (Y) + y_off)
+#include "spantmp2.h"
+
+/* 32 bit, xRGB8888 color spanline and pixel functions
+ */
+#define SPANTMP_PIXEL_FMT GL_BGRA
+#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
+
+#define TAG(x)    radeon##x##_xRGB8888
+#define TAG2(x,y) radeon##x##_xRGB8888##y
+#define GET_PTR(X,Y) radeon_ptr32(rrb, (X) + x_off, (Y) + y_off)
 #include "spantmp2.h"
 
 /* 32 bit, ARGB8888 color spanline and pixel functions
@@ -235,7 +258,7 @@ static GLubyte *radeon_ptr(const struct radeon_renderbuffer * rrb,
 
 #define TAG(x)    radeon##x##_ARGB8888
 #define TAG2(x,y) radeon##x##_ARGB8888##y
-#define GET_PTR(X,Y) radeon_ptr32(rrb, (X), (Y))
+#define GET_PTR(X,Y) radeon_ptr32(rrb, (X) + x_off, (Y) + y_off)
 #include "spantmp2.h"
 
 /* ================================================================
@@ -257,10 +280,10 @@ static GLubyte *radeon_ptr(const struct radeon_renderbuffer * rrb,
 #define VALUE_TYPE GLushort
 
 #define WRITE_DEPTH( _x, _y, d )					\
-   *(GLushort *)radeon_ptr(rrb, _x, _y) = d
+   *(GLushort *)radeon_ptr(rrb, _x + x_off, _y + y_off) = d
 
 #define READ_DEPTH( d, _x, _y )						\
-   d = *(GLushort *)radeon_ptr(rrb, _x, _y)
+   d = *(GLushort *)radeon_ptr(rrb, _x + x_off, _y + y_off)
 
 #define TAG(x) radeon##x##_z16
 #include "depthtmp.h"
@@ -275,7 +298,7 @@ static GLubyte *radeon_ptr(const struct radeon_renderbuffer * rrb,
 #ifdef COMPILE_R300
 #define WRITE_DEPTH( _x, _y, d )					\
 do {									\
-   GLuint *_ptr = (GLuint*)radeon_ptr32( rrb, _x, _y );		\
+   GLuint *_ptr = (GLuint*)radeon_ptr32( rrb, _x + x_off, _y + y_off );		\
    GLuint tmp = *_ptr;				\
    tmp &= 0x000000ff;							\
    tmp |= ((d << 8) & 0xffffff00);					\
@@ -284,8 +307,8 @@ do {									\
 #else
 #define WRITE_DEPTH( _x, _y, d )					\
 do {									\
-   GLuint *_ptr = (GLuint*)radeon_ptr32( rrb, _x, _y );		\
-   GLuint tmp = *_ptr;				\
+   GLuint *_ptr = (GLuint*)radeon_ptr32( rrb, _x + x_off, _y + y_off );	\
+   GLuint tmp = *_ptr;							\
    tmp &= 0xff000000;							\
    tmp |= ((d) & 0x00ffffff);						\
    *_ptr = tmp;					\
@@ -295,11 +318,11 @@ do {									\
 #ifdef COMPILE_R300
 #define READ_DEPTH( d, _x, _y )						\
   do { \
-    d = (*(GLuint*)(radeon_ptr32(rrb, _x, _y)) & 0xffffff00) >> 8; \
+    d = (*(GLuint*)(radeon_ptr32(rrb, _x + x_off, _y + y_off)) & 0xffffff00) >> 8; \
   }while(0)
 #else
 #define READ_DEPTH( d, _x, _y )						\
-   d = *(GLuint*)(radeon_ptr32(rrb, _x,	_y )) & 0x00ffffff;
+   d = *(GLuint*)(radeon_ptr32(rrb, _x + x_off,	_y + y_off )) & 0x00ffffff;
 #endif
 /*
     fprintf(stderr, "dval(%d, %d, %d, %d)=0x%08X\n", _x, xo, _y, yo, d);\
@@ -317,7 +340,7 @@ do {									\
 #ifdef COMPILE_R300
 #define WRITE_STENCIL( _x, _y, d )					\
 do {									\
-   GLuint *_ptr = (GLuint*)radeon_ptr32(rrb, _x, _y);		\
+   GLuint *_ptr = (GLuint*)radeon_ptr32(rrb, _x + x_off, _y + y_off);		\
    GLuint tmp = *_ptr;				\
    tmp &= 0xffffff00;							\
    tmp |= (d) & 0xff;							\
@@ -326,7 +349,7 @@ do {									\
 #else
 #define WRITE_STENCIL( _x, _y, d )					\
 do {									\
-   GLuint *_ptr = (GLuint*)radeon_ptr32(rrb, _x, _y);		\
+   GLuint *_ptr = (GLuint*)radeon_ptr32(rrb, _x + x_off, _y + y_off);		\
    GLuint tmp = *_ptr;				\
    tmp &= 0x00ffffff;							\
    tmp |= (((d) & 0xff) << 24);						\
@@ -337,14 +360,14 @@ do {									\
 #ifdef COMPILE_R300
 #define READ_STENCIL( d, _x, _y )					\
 do {									\
-   GLuint *_ptr = (GLuint*)radeon_ptr32( rrb, _x, _y );		\
+   GLuint *_ptr = (GLuint*)radeon_ptr32( rrb, _x + x_off, _y + y_off );		\
    GLuint tmp = *_ptr;				\
    d = tmp & 0x000000ff;						\
 } while (0)
 #else
 #define READ_STENCIL( d, _x, _y )					\
 do {									\
-   GLuint *_ptr = (GLuint*)radeon_ptr32( rrb, _x, _y );		\
+   GLuint *_ptr = (GLuint*)radeon_ptr32( rrb, _x + x_off, _y + y_off );		\
    GLuint tmp = *_ptr;				\
    d = (tmp & 0xff000000) >> 24;					\
 } while (0)
@@ -354,29 +377,64 @@ do {									\
 #include "stenciltmp.h"
 
 
-static void map_buffer(struct gl_renderbuffer *rb, GLboolean write)
+void map_unmap_rb(struct gl_renderbuffer *rb, int flag)
 {
-	struct radeon_renderbuffer *rrb = (void*)rb;
+	struct radeon_renderbuffer *rrb = radeon_renderbuffer(rb);
 	int r;
 	
-	if (rrb->bo) {
-		r = radeon_bo_map(rrb->bo, write);
+	if (rrb == NULL || !rrb->bo)
+		return;
+
+	if (flag) {
+		r = radeon_bo_map(rrb->bo, 1);
 		if (r) {
 			fprintf(stderr, "(%s) error(%d) mapping buffer.\n",
 				__FUNCTION__, r);
 		}
+
+		radeonSetSpanFunctions(rrb);
+	} else {
+		radeon_bo_unmap(rrb->bo);
+		rb->GetRow = NULL;
+		rb->PutRow = NULL;
 	}
 }
 
-static void unmap_buffer(struct gl_renderbuffer *rb)
+static void
+radeon_map_unmap_buffers(GLcontext *ctx, GLboolean map)
 {
-	struct radeon_renderbuffer *rrb = (void*)rb;
+	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
+	GLuint i, j;
 
-	if (rrb->bo) {
-		radeon_bo_unmap(rrb->bo);
+	/* color draw buffers */
+	for (j = 0; j < ctx->DrawBuffer->_NumColorDrawBuffers; j++)
+		map_unmap_rb(ctx->DrawBuffer->_ColorDrawBuffers[j], map);
+
+	/* check for render to textures */
+	for (i = 0; i < BUFFER_COUNT; i++) {
+		struct gl_renderbuffer_attachment *att =
+			ctx->DrawBuffer->Attachment + i;
+		struct gl_texture_object *tex = att->Texture;
+		if (tex) {
+			/* render to texture */
+			ASSERT(att->Renderbuffer);
+			if (map)
+				ctx->Driver.MapTexture(ctx, tex);
+			else
+				ctx->Driver.UnmapTexture(ctx, tex);
+		}
 	}
-}
+	
+	map_unmap_rb(ctx->ReadBuffer->_ColorReadBuffer, map);
+
+	/* depth buffer (Note wrapper!) */
+	if (ctx->DrawBuffer->_DepthBuffer)
+		map_unmap_rb(ctx->DrawBuffer->_DepthBuffer->Wrapped, map);
+	
+	if (ctx->DrawBuffer->_StencilBuffer)
+		map_unmap_rb(ctx->DrawBuffer->_StencilBuffer->Wrapped, map);
 
+}
 static void radeonSpanRenderStart(GLcontext * ctx)
 {
 	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
@@ -389,18 +447,7 @@ static void radeonSpanRenderStart(GLcontext * ctx)
 			ctx->Driver.MapTexture(ctx, ctx->Texture.Unit[i]._Current);
 	}
 
-	/* color draw buffers */
-	for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++) {
-		map_buffer(ctx->DrawBuffer->_ColorDrawBuffers[i], GL_TRUE);
-	}
-
-	map_buffer(ctx->ReadBuffer->_ColorReadBuffer, GL_FALSE);
-
-	if (ctx->DrawBuffer->_DepthBuffer) {
-		map_buffer(ctx->DrawBuffer->_DepthBuffer->Wrapped, GL_TRUE);
-	}
-	if (ctx->DrawBuffer->_StencilBuffer)
-		map_buffer(ctx->DrawBuffer->_StencilBuffer->Wrapped, GL_TRUE);
+	radeon_map_unmap_buffers(ctx, 1);
 
 	/* The locking and wait for idle should really only be needed in classic mode.
 	 * In a future memory manager based implementation, this should become
@@ -423,16 +470,7 @@ static void radeonSpanRenderFinish(GLcontext * ctx)
 			ctx->Driver.UnmapTexture(ctx, ctx->Texture.Unit[i]._Current);
 	}
 
-	/* color draw buffers */
-	for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++)
-		unmap_buffer(ctx->DrawBuffer->_ColorDrawBuffers[i]);
-
-	unmap_buffer(ctx->ReadBuffer->_ColorReadBuffer);
-
-	if (ctx->DrawBuffer->_DepthBuffer)
-		unmap_buffer(ctx->DrawBuffer->_DepthBuffer->Wrapped);
-	if (ctx->DrawBuffer->_StencilBuffer)
-		unmap_buffer(ctx->DrawBuffer->_StencilBuffer->Wrapped);
+	radeon_map_unmap_buffers(ctx, 0);
 }
 
 void radeonInitSpanFuncs(GLcontext * ctx)
@@ -446,17 +484,21 @@ void radeonInitSpanFuncs(GLcontext * ctx)
 /**
  * Plug in the Get/Put routines for the given driRenderbuffer.
  */
-void radeonSetSpanFunctions(struct radeon_renderbuffer *rrb)
+static void radeonSetSpanFunctions(struct radeon_renderbuffer *rrb)
 {
-	if (rrb->base.InternalFormat == GL_RGB5) {
+	if (rrb->base._ActualFormat == GL_RGB5) {
 		radeonInitPointers_RGB565(&rrb->base);
-	} else if (rrb->base.InternalFormat == GL_RGBA8) {
+	} else if (rrb->base._ActualFormat == GL_RGB8) {
+		radeonInitPointers_xRGB8888(&rrb->base);
+	} else if (rrb->base._ActualFormat == GL_RGBA8) {
 		radeonInitPointers_ARGB8888(&rrb->base);
-	} else if (rrb->base.InternalFormat == GL_DEPTH_COMPONENT16) {
+	} else if (rrb->base._ActualFormat == GL_DEPTH_COMPONENT16) {
 		radeonInitDepthPointers_z16(&rrb->base);
-	} else if (rrb->base.InternalFormat == GL_DEPTH_COMPONENT24) {
+	} else if (rrb->base._ActualFormat == GL_DEPTH_COMPONENT24) {
 		radeonInitDepthPointers_z24_s8(&rrb->base);
-	} else if (rrb->base.InternalFormat == GL_STENCIL_INDEX8_EXT) {
+	} else if (rrb->base._ActualFormat == GL_DEPTH24_STENCIL8_EXT) {
+		radeonInitStencilPointers_z24_s8(&rrb->base);
+	} else if (rrb->base._ActualFormat == GL_STENCIL_INDEX8_EXT) {
 		radeonInitStencilPointers_z24_s8(&rrb->base);
 	}
 }
diff --git a/src/mesa/drivers/dri/radeon/radeon_span.h b/src/mesa/drivers/dri/radeon/radeon_span.h
index dd44ab517a..ea6a2e7fb4 100644
--- a/src/mesa/drivers/dri/radeon/radeon_span.h
+++ b/src/mesa/drivers/dri/radeon/radeon_span.h
@@ -44,5 +44,4 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 extern void radeonInitSpanFuncs(GLcontext * ctx);
 
-extern void radeonSetSpanFunctions(struct radeon_renderbuffer *rrb);
 #endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_state.c b/src/mesa/drivers/dri/radeon/radeon_state.c
index 42b65cbee6..19ff2688e6 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state.c
+++ b/src/mesa/drivers/dri/radeon/radeon_state.c
@@ -528,7 +528,8 @@ static void radeonPolygonOffset( GLcontext *ctx,
 				 GLfloat factor, GLfloat units )
 {
    r100ContextPtr rmesa = R100_CONTEXT(ctx);
-   float_ui32_type constant =  { units * rmesa->radeon.state.depth.scale };
+   const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+   float_ui32_type constant =  { units * depthScale };
    float_ui32_type factoru = { factor };
 
    RADEON_STATECHANGE( rmesa, zbs );
@@ -1390,13 +1391,24 @@ void radeonUpdateWindow( GLcontext *ctx )
    GLfloat xoffset = dPriv ? (GLfloat) dPriv->x : 0;
    GLfloat yoffset = dPriv ? (GLfloat) dPriv->y + dPriv->h : 0;
    const GLfloat *v = ctx->Viewport._WindowMap.m;
+   const GLboolean render_to_fbo = (ctx->DrawBuffer ? (ctx->DrawBuffer->Name != 0) : 0);
+   const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF;
+   GLfloat y_scale, y_bias;
+
+   if (render_to_fbo) {
+      y_scale = 1.0;
+      y_bias = 0;
+   } else {
+      y_scale = -1.0;
+      y_bias = yoffset;
+   }
 
    float_ui32_type sx = { v[MAT_SX] };
    float_ui32_type tx = { v[MAT_TX] + xoffset + SUBPIXEL_X };
-   float_ui32_type sy = { - v[MAT_SY] };
-   float_ui32_type ty = { (- v[MAT_TY]) + yoffset + SUBPIXEL_Y };
-   float_ui32_type sz = { v[MAT_SZ] * rmesa->radeon.state.depth.scale };
-   float_ui32_type tz = { v[MAT_TZ] * rmesa->radeon.state.depth.scale };
+   float_ui32_type sy = { v[MAT_SY] * y_scale };
+   float_ui32_type ty = { (v[MAT_TY] * y_scale) + y_bias + SUBPIXEL_Y };
+   float_ui32_type sz = { v[MAT_SZ] * depthScale };
+   float_ui32_type tz = { v[MAT_TZ] * depthScale };
 
    RADEON_STATECHANGE( rmesa, vpt );
 
@@ -1790,15 +1802,24 @@ static void radeonEnable( GLcontext *ctx, GLenum cap, GLboolean state )
       break;
 
    case GL_STENCIL_TEST:
-      if ( rmesa->radeon.state.stencil.hwBuffer ) {
-	 RADEON_STATECHANGE( rmesa, ctx );
-	 if ( state ) {
-	    rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  RADEON_STENCIL_ENABLE;
+      {
+	 GLboolean hw_stencil = GL_FALSE;
+	 if (ctx->DrawBuffer) {
+	    struct radeon_renderbuffer *rrbStencil
+	       = radeon_get_renderbuffer(ctx->DrawBuffer, BUFFER_STENCIL);
+	    hw_stencil = (rrbStencil && rrbStencil->bo);
+	 }
+
+	 if (hw_stencil) {
+	    RADEON_STATECHANGE( rmesa, ctx );
+	    if ( state ) {
+	       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] |=  RADEON_STENCIL_ENABLE;
+	    } else {
+	       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~RADEON_STENCIL_ENABLE;
+	    }
 	 } else {
-	    rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &= ~RADEON_STENCIL_ENABLE;
+	    FALLBACK( rmesa, RADEON_FALLBACK_STENCIL, state );
 	 }
-      } else {
-	 FALLBACK( rmesa, RADEON_FALLBACK_STENCIL, state );
       }
       break;
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c
index 8b6caf19d3..3d0cd8d3f8 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state_init.c
+++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c
@@ -592,22 +592,16 @@ void radeonInitState( r100ContextPtr rmesa )
    switch ( ctx->Visual.depthBits ) {
    case 16:
       rmesa->radeon.state.depth.clear = 0x0000ffff;
-      rmesa->radeon.state.depth.scale = 1.0 / (GLfloat)0xffff;
       rmesa->radeon.state.stencil.clear = 0x00000000;
       break;
    case 24:
       rmesa->radeon.state.depth.clear = 0x00ffffff;
-      rmesa->radeon.state.depth.scale = 1.0 / (GLfloat)0xffffff;
       rmesa->radeon.state.stencil.clear = 0xffff0000;
       break;
    default:
       break;
    }
 
-   /* Only have hw stencil when depth buffer is 24 bits deep */
-   rmesa->radeon.state.stencil.hwBuffer = ( ctx->Visual.stencilBits > 0 &&
-				     ctx->Visual.depthBits == 24 );
-
    rmesa->radeon.Fallback = 0;