Merge commit 'origin/master' into gallium-sw-api-2

Conflicts:
	src/gallium/drivers/llvmpipe/lp_setup.c
	src/gallium/drivers/softpipe/sp_texture.c
	src/gallium/drivers/softpipe/sp_winsys.h
	src/gallium/state_trackers/egl/common/egl_g3d.c
	src/gallium/state_trackers/egl/x11/native_x11.c
	src/gallium/state_trackers/egl/x11/native_x11.h
	src/gallium/state_trackers/egl/x11/native_ximage.c

114 files changed, 3900 insertions, 2092 deletions

diff --git a/src/gallium/auxiliary/Makefile b/src/gallium/auxiliary/Makefile
index 1d0930e024b..68b1a3e8546 100644
--- a/src/gallium/auxiliary/Makefile
+++ b/src/gallium/auxiliary/Makefile
@@ -107,6 +107,7 @@ C_SOURCES = \
 	util/u_draw_quad.c \
 	util/u_format_access.c \
 	util/u_format_table.c \
+	util/u_format_tests.c \
 	util/u_gen_mipmap.c \
 	util/u_handle_table.c \
 	util/u_hash_table.c \
diff --git a/src/gallium/auxiliary/SConscript b/src/gallium/auxiliary/SConscript
index f365c4bbdd6..d6185bcb5e3 100644
--- a/src/gallium/auxiliary/SConscript
+++ b/src/gallium/auxiliary/SConscript
@@ -7,6 +7,8 @@ env.Append(CPPPATH = [
     'util',
 ])
 
+env.Tool('udis86')
+
 env.CodeGenerate(
     target = 'indices/u_indices_gen.c', 
     script = 'indices/u_indices_gen.py', 
@@ -149,6 +151,7 @@ source = [
     'util/u_draw_quad.c',
     'util/u_format_access.c',
     'util/u_format_table.c',
+    'util/u_format_tests.c',
     'util/u_gen_mipmap.c',
     'util/u_handle_table.c',
     'util/u_hash.c',
diff --git a/src/gallium/auxiliary/draw/draw_pt_decompose.h b/src/gallium/auxiliary/draw/draw_pt_decompose.h
index 4ca5b520204..3c44f7c11ee 100644
--- a/src/gallium/auxiliary/draw/draw_pt_decompose.h
+++ b/src/gallium/auxiliary/draw/draw_pt_decompose.h
@@ -105,40 +105,20 @@ static void FUNC( ARGS,
 
 
    case PIPE_PRIM_QUADS:
-      if (flatfirst) {
-         for (i = 0; i+3 < count; i += 4) {
-            QUAD( (i + 1),
-                  (i + 2),
-                  (i + 3),
-                  (i + 0) );
-         }
-      }
-      else {
-         for (i = 0; i+3 < count; i += 4) {
-            QUAD( (i + 0),
-                  (i + 1),
-                  (i + 2),
-                  (i + 3));
-         }
+      for (i = 0; i+3 < count; i += 4) {
+         QUAD( (i + 0),
+               (i + 1),
+               (i + 2),
+               (i + 3));
       }
       break;
 
    case PIPE_PRIM_QUAD_STRIP:
-      if (flatfirst) {
-         for (i = 0; i+3 < count; i += 2) {
-            QUAD( (i + 1),
-                  (i + 3),
-                  (i + 2),
-                  (i + 0) );
-         }
-      }
-      else {
-         for (i = 0; i+3 < count; i += 2) {
-            QUAD( (i + 2),
-                  (i + 0),
-                  (i + 1),
-                  (i + 3));
-         }
+      for (i = 0; i+3 < count; i += 2) {
+         QUAD( (i + 2),
+               (i + 0),
+               (i + 1),
+               (i + 3));
       }
       break;
 
diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h
index 62822a3d562..7cba8547f15 100644
--- a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h
+++ b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h
@@ -118,39 +118,21 @@ static void FUNC( struct draw_pt_front_end *frontend,
 
    case PIPE_PRIM_QUADS:
       for (i = 0; i+3 < count; i += 4) {
-         if (flatfirst) {
-            QUAD( vcache,
-                  get_elt(elts, i + 0),
-                  get_elt(elts, i + 1),
-                  get_elt(elts, i + 2),
-                  get_elt(elts, i + 3) );
-         }
-         else {
-            QUAD( vcache,
-                  get_elt(elts, i + 0),
-                  get_elt(elts, i + 1),
-                  get_elt(elts, i + 2),
-                  get_elt(elts, i + 3) );
-         }
+         QUAD( vcache,
+               get_elt(elts, i + 0),
+               get_elt(elts, i + 1),
+               get_elt(elts, i + 2),
+               get_elt(elts, i + 3) );
       }
       break;
 
    case PIPE_PRIM_QUAD_STRIP:
       for (i = 0; i+3 < count; i += 2) {
-         if (flatfirst) {
-            QUAD( vcache,
-                  get_elt(elts, i + 0),
-                  get_elt(elts, i + 1),
-                  get_elt(elts, i + 3),
-                  get_elt(elts, i + 2) );
-         }
-         else {
-            QUAD( vcache,
-                  get_elt(elts, i + 2),
-                  get_elt(elts, i + 0),
-                  get_elt(elts, i + 1),
-                  get_elt(elts, i + 3) );
-         }
+         QUAD( vcache,
+               get_elt(elts, i + 2),
+               get_elt(elts, i + 0),
+               get_elt(elts, i + 1),
+               get_elt(elts, i + 3) );
       }
       break;
 
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_arit.c b/src/gallium/auxiliary/gallivm/lp_bld_arit.c
index bbce31f9eb2..32f9e5201c5 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_arit.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_arit.c
@@ -669,6 +669,14 @@ lp_build_abs(struct lp_build_context *bld,
 
 
 LLVMValueRef
+lp_build_negate(struct lp_build_context *bld,
+                LLVMValueRef a)
+{
+   return LLVMBuildNeg(bld->builder, a, "");
+}
+
+
+LLVMValueRef
 lp_build_sgn(struct lp_build_context *bld,
              LLVMValueRef a)
 {
@@ -710,6 +718,41 @@ lp_build_sgn(struct lp_build_context *bld,
 
 
 /**
+ * Set the sign of float vector 'a' according to 'sign'.
+ * If sign==0, return abs(a).
+ * If sign==1, return -abs(a);
+ * Other values for sign produce undefined results.
+ */
+LLVMValueRef
+lp_build_set_sign(struct lp_build_context *bld,
+                  LLVMValueRef a, LLVMValueRef sign)
+{
+   const struct lp_type type = bld->type;
+   LLVMTypeRef int_vec_type = lp_build_int_vec_type(type);
+   LLVMTypeRef vec_type = lp_build_vec_type(type);
+   LLVMValueRef shift = lp_build_int_const_scalar(type, type.width - 1);
+   LLVMValueRef mask = lp_build_int_const_scalar(type,
+                             ~((unsigned long long) 1 << (type.width - 1)));
+   LLVMValueRef val, res;
+
+   assert(type.floating);
+
+   /* val = reinterpret_cast<int>(a) */
+   val = LLVMBuildBitCast(bld->builder, a, int_vec_type, "");
+   /* val = val & mask */
+   val = LLVMBuildAnd(bld->builder, val, mask, "");
+   /* sign = sign << shift */
+   sign = LLVMBuildShl(bld->builder, sign, shift, "");
+   /* res = val | sign */
+   res = LLVMBuildOr(bld->builder, val, sign, "");
+   /* res = reinterpret_cast<float>(res) */
+   res = LLVMBuildBitCast(bld->builder, res, vec_type, "");
+
+   return res;
+}
+
+
+/**
  * Convert vector of int to vector of float.
  */
 LLVMValueRef
@@ -857,6 +900,19 @@ lp_build_ceil(struct lp_build_context *bld,
 
 
 /**
+ * Return fractional part of 'a' computed as a - floor(f)
+ * Typically used in texture coord arithmetic.
+ */
+LLVMValueRef
+lp_build_fract(struct lp_build_context *bld,
+               LLVMValueRef a)
+{
+   assert(bld->type.floating);
+   return lp_build_sub(bld, a, lp_build_floor(bld, a));
+}
+
+
+/**
  * Convert to integer, through whichever rounding method that's fastest,
  * typically truncating toward zero.
  */
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_arit.h b/src/gallium/auxiliary/gallivm/lp_bld_arit.h
index da84b7ca02d..55385e3a66a 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_arit.h
+++ b/src/gallium/auxiliary/gallivm/lp_bld_arit.h
@@ -117,10 +117,18 @@ lp_build_abs(struct lp_build_context *bld,
              LLVMValueRef a);
 
 LLVMValueRef
+lp_build_negate(struct lp_build_context *bld,
+                LLVMValueRef a);
+
+LLVMValueRef
 lp_build_sgn(struct lp_build_context *bld,
              LLVMValueRef a);
 
 LLVMValueRef
+lp_build_set_sign(struct lp_build_context *bld,
+                  LLVMValueRef a, LLVMValueRef sign);
+
+LLVMValueRef
 lp_build_int_to_float(struct lp_build_context *bld,
                       LLVMValueRef a);
 
@@ -141,6 +149,10 @@ lp_build_trunc(struct lp_build_context *bld,
                LLVMValueRef a);
 
 LLVMValueRef
+lp_build_fract(struct lp_build_context *bld,
+               LLVMValueRef a);
+
+LLVMValueRef
 lp_build_ifloor(struct lp_build_context *bld,
                 LLVMValueRef a);
 LLVMValueRef
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_logic.c b/src/gallium/auxiliary/gallivm/lp_bld_logic.c
index 41ac81b7441..2726747eaea 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_logic.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_logic.c
@@ -45,6 +45,7 @@
 /**
  * Build code to compare two values 'a' and 'b' of 'type' using the given func.
  * \param func  one of PIPE_FUNC_x
+ * The result values will be 0 for false or ~0 for true.
  */
 LLVMValueRef
 lp_build_compare(LLVMBuilderRef builder,
@@ -311,6 +312,7 @@ lp_build_compare(LLVMBuilderRef builder,
 /**
  * Build code to compare two values 'a' and 'b' using the given func.
  * \param func  one of PIPE_FUNC_x
+ * The result values will be 0 for false or ~0 for true.
  */
 LLVMValueRef
 lp_build_cmp(struct lp_build_context *bld,
@@ -322,6 +324,9 @@ lp_build_cmp(struct lp_build_context *bld,
 }
 
 
+/**
+ * Return mask ? a : b;
+ */
 LLVMValueRef
 lp_build_select(struct lp_build_context *bld,
                 LLVMValueRef mask,
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_pack.c b/src/gallium/auxiliary/gallivm/lp_bld_pack.c
index bc360ad77ad..4c61d107494 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_pack.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_pack.c
@@ -256,7 +256,9 @@ lp_build_pack2(LLVMBuilderRef builder,
                LLVMValueRef lo,
                LLVMValueRef hi)
 {
+#if !(HAVE_LLVM >= 0x0207)
    LLVMTypeRef src_vec_type = lp_build_vec_type(src_type);
+#endif
    LLVMTypeRef dst_vec_type = lp_build_vec_type(dst_type);
    LLVMValueRef shuffle;
    LLVMValueRef res;
@@ -272,11 +274,14 @@ lp_build_pack2(LLVMBuilderRef builder,
       switch(src_type.width) {
       case 32:
          if(dst_type.sign) {
+#if HAVE_LLVM >= 0x0207
+            res = lp_build_intrinsic_binary(builder, "llvm.x86.sse2.packssdw.128", dst_vec_type, lo, hi);
+#else
             res = lp_build_intrinsic_binary(builder, "llvm.x86.sse2.packssdw.128", src_vec_type, lo, hi);
+#endif
          }
          else {
             if (util_cpu_caps.has_sse4_1) {
-               /* PACKUSDW is the only instrinsic with a consistent signature */
                return lp_build_intrinsic_binary(builder, "llvm.x86.sse41.packusdw", dst_vec_type, lo, hi);
             }
             else {
@@ -288,9 +293,17 @@ lp_build_pack2(LLVMBuilderRef builder,
 
       case 16:
          if(dst_type.sign)
+#if HAVE_LLVM >= 0x0207
+            res = lp_build_intrinsic_binary(builder, "llvm.x86.sse2.packsswb.128", dst_vec_type, lo, hi);
+#else
             res = lp_build_intrinsic_binary(builder, "llvm.x86.sse2.packsswb.128", src_vec_type, lo, hi);
+#endif
          else
+#if HAVE_LLVM >= 0x0207
+            res = lp_build_intrinsic_binary(builder, "llvm.x86.sse2.packuswb.128", dst_vec_type, lo, hi);
+#else
             res = lp_build_intrinsic_binary(builder, "llvm.x86.sse2.packuswb.128", src_vec_type, lo, hi);
+#endif
          break;
 
       default:
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_sample.c b/src/gallium/auxiliary/gallivm/lp_bld_sample.c
index a133b56ac55..311c9f1b9e4 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_sample.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_sample.c
@@ -44,6 +44,11 @@
 #include "lp_bld_sample.h"
 
 
+/**
+ * Initialize lp_sampler_static_state object with the gallium sampler
+ * and texture state.
+ * The former is considered to be static and the later dynamic.
+ */
 void
 lp_sampler_static_state(struct lp_sampler_static_state *state,
                         const struct pipe_texture *texture,
@@ -57,6 +62,18 @@ lp_sampler_static_state(struct lp_sampler_static_state *state,
    if(!sampler)
       return;
 
+   /*
+    * We don't copy sampler state over unless it is actually enabled, to avoid
+    * spurious recompiles, as the sampler static state is part of the shader
+    * key.
+    *
+    * Ideally the state tracker or cso_cache module would make all state
+    * canonical, but until that happens it's better to be safe than sorry here.
+    *
+    * XXX: Actually there's much more than can be done here, especially
+    * regarding 1D/2D/3D/CUBE textures, wrap modes, etc.
+    */
+
    state->format            = texture->format;
    state->target            = texture->target;
    state->pot_width         = util_is_pot(texture->width0);
@@ -69,11 +86,20 @@ lp_sampler_static_state(struct lp_sampler_static_state *state,
    state->min_img_filter    = sampler->min_img_filter;
    state->min_mip_filter    = sampler->min_mip_filter;
    state->mag_img_filter    = sampler->mag_img_filter;
+
    state->compare_mode      = sampler->compare_mode;
-   if(sampler->compare_mode != PIPE_TEX_COMPARE_NONE) {
-      state->compare_func      = sampler->compare_func;
+   if (sampler->compare_mode != PIPE_TEX_COMPARE_NONE) {
+      state->compare_func   = sampler->compare_func;
    }
+
    state->normalized_coords = sampler->normalized_coords;
+   state->lod_bias          = sampler->lod_bias;
+   state->min_lod           = sampler->min_lod;
+   state->max_lod           = sampler->max_lod;
+   state->border_color[0]   = sampler->border_color[0];
+   state->border_color[1]   = sampler->border_color[1];
+   state->border_color[2]   = sampler->border_color[2];
+   state->border_color[3]   = sampler->border_color[3];
 }
 
 
@@ -136,8 +162,7 @@ lp_build_sample_offset(struct lp_build_context *bld,
                        const struct util_format_description *format_desc,
                        LLVMValueRef x,
                        LLVMValueRef y,
-                       LLVMValueRef y_stride,
-                       LLVMValueRef data_ptr)
+                       LLVMValueRef y_stride)
 {
    LLVMValueRef x_stride;
    LLVMValueRef offset;
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_sample.h b/src/gallium/auxiliary/gallivm/lp_bld_sample.h
index 39edcf13d14..68db91d6fd6 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_sample.h
+++ b/src/gallium/auxiliary/gallivm/lp_bld_sample.h
@@ -70,6 +70,8 @@ struct lp_sampler_static_state
    unsigned compare_mode:1;
    unsigned compare_func:3;
    unsigned normalized_coords:1;
+   float lod_bias, min_lod, max_lod;
+   float border_color[4];
 };
 
 
@@ -98,6 +100,18 @@ struct lp_sampler_dynamic_state
               LLVMBuilderRef builder,
               unsigned unit);
 
+   /** Obtain the base texture depth. */
+   LLVMValueRef
+   (*depth)( struct lp_sampler_dynamic_state *state,
+             LLVMBuilderRef builder,
+             unsigned unit);
+
+   /** Obtain the number of mipmap levels (minus one). */
+   LLVMValueRef
+   (*last_level)( struct lp_sampler_dynamic_state *state,
+                  LLVMBuilderRef builder,
+                  unsigned unit);
+
    LLVMValueRef
    (*stride)( struct lp_sampler_dynamic_state *state,
               LLVMBuilderRef builder,
@@ -134,8 +148,7 @@ lp_build_sample_offset(struct lp_build_context *bld,
                        const struct util_format_description *format_desc,
                        LLVMValueRef x,
                        LLVMValueRef y,
-                       LLVMValueRef y_stride,
-                       LLVMValueRef data_ptr);
+                       LLVMValueRef y_stride);
 
 
 void
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c b/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c
index e2688622829..1dca29cdd58 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c
@@ -69,7 +69,11 @@ struct lp_build_sample_context
    struct lp_type coord_type;
    struct lp_build_context coord_bld;
 
-   /** Integer coordinates */
+   /** Unsigned integer coordinates */
+   struct lp_type uint_coord_type;
+   struct lp_build_context uint_coord_bld;
+
+   /** Signed integer coordinates */
    struct lp_type int_coord_type;
    struct lp_build_context int_coord_bld;
 
@@ -79,36 +83,135 @@ struct lp_build_sample_context
 };
 
 
+/**
+ * Does the given texture wrap mode allow sampling the texture border color?
+ * XXX maybe move this into gallium util code.
+ */
+static boolean
+wrap_mode_uses_border_color(unsigned mode)
+{
+   switch (mode) {
+   case PIPE_TEX_WRAP_REPEAT:
+   case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+   case PIPE_TEX_WRAP_MIRROR_REPEAT:
+   case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+      return FALSE;
+   case PIPE_TEX_WRAP_CLAMP:
+   case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+   case PIPE_TEX_WRAP_MIRROR_CLAMP:
+   case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+      return TRUE;
+   default:
+      assert(0 && "unexpected wrap mode");
+      return FALSE;
+   }
+}
+
+
+
+/**
+ * Gen code to fetch a texel from a texture at int coords (x, y).
+ * The result, texel, will be:
+ *   texel[0] = red values
+ *   texel[1] = green values
+ *   texel[2] = blue values
+ *   texel[3] = alpha values
+ */
 static void
 lp_build_sample_texel_soa(struct lp_build_sample_context *bld,
+                          LLVMValueRef width,
+                          LLVMValueRef height,
                           LLVMValueRef x,
                           LLVMValueRef y,
                           LLVMValueRef y_stride,
-                          LLVMValueRef data_ptr,
+                          LLVMValueRef data_array,
                           LLVMValueRef *texel)
 {
+   struct lp_build_context *int_coord_bld = &bld->int_coord_bld;
    LLVMValueRef offset;
    LLVMValueRef packed;
+   LLVMValueRef use_border = NULL;
+   LLVMValueRef data_ptr;
+
+   /* use_border = x < 0 || x >= width || y < 0 || y >= height */
+   if (wrap_mode_uses_border_color(bld->static_state->wrap_s)) {
+      LLVMValueRef b1, b2;
+      b1 = lp_build_cmp(int_coord_bld, PIPE_FUNC_LESS, x, int_coord_bld->zero);
+      b2 = lp_build_cmp(int_coord_bld, PIPE_FUNC_GEQUAL, x, width);
+      use_border = LLVMBuildOr(bld->builder, b1, b2, "b1_or_b2");
+   }
 
-   offset = lp_build_sample_offset(&bld->int_coord_bld,
+   if (wrap_mode_uses_border_color(bld->static_state->wrap_t)) {
+      LLVMValueRef b1, b2;
+      b1 = lp_build_cmp(int_coord_bld, PIPE_FUNC_LESS, y, int_coord_bld->zero);
+      b2 = lp_build_cmp(int_coord_bld, PIPE_FUNC_GEQUAL, y, height);
+      if (use_border) {
+         use_border = LLVMBuildOr(bld->builder, use_border, b1, "ub_or_b1");
+         use_border = LLVMBuildOr(bld->builder, use_border, b2, "ub_or_b2");
+      }
+      else {
+         use_border = LLVMBuildOr(bld->builder, b1, b2, "b1_or_b2");
+      }
+   }
+
+   /* XXX always use mipmap level 0 for now */
+   {
+      const int level = 0;
+      LLVMValueRef indexes[2];
+      indexes[0] = LLVMConstInt(LLVMInt32Type(), 0, 0);
+      indexes[1] = LLVMConstInt(LLVMInt32Type(), level, 0);
+      data_ptr = LLVMBuildGEP(bld->builder, data_array, indexes, 2, "");
+      data_ptr = LLVMBuildLoad(bld->builder, data_ptr, "");
+   }
+
+   /*
+    * Note: if we find an app which frequently samples the texture border
+    * we might want to implement a true conditional here to avoid sampling
+    * the texture whenever possible (since that's quite a bit of code).
+    * Ex:
+    *   if (use_border) {
+    *      texel = border_color;
+    *   }
+    *   else {
+    *      texel = sample_texture(coord);
+    *   }
+    * As it is now, we always sample the texture, then selectively replace
+    * the texel color results with the border color.
+    */
+
+   /* convert x,y coords to linear offset from start of texture, in bytes */
+   offset = lp_build_sample_offset(&bld->uint_coord_bld,
                                    bld->format_desc,
-                                   x, y, y_stride,
-                                   data_ptr);
+                                   x, y, y_stride);
 
    assert(bld->format_desc->block.width == 1);
    assert(bld->format_desc->block.height == 1);
    assert(bld->format_desc->block.bits <= bld->texel_type.width);
 
+   /* gather the texels from the texture */
    packed = lp_build_gather(bld->builder,
                             bld->texel_type.length,
                             bld->format_desc->block.bits,
                             bld->texel_type.width,
                             data_ptr, offset);
 
+   /* convert texels to float rgba */
    lp_build_unpack_rgba_soa(bld->builder,
                             bld->format_desc,
                             bld->texel_type,
                             packed, texel);
+
+   if (use_border) {
+      /* select texel color or border color depending on use_border */
+      int chan;
+      for (chan = 0; chan < 4; chan++) {
+         LLVMValueRef border_chan =
+            lp_build_const_scalar(bld->texel_type,
+                                  bld->static_state->border_color[chan]);
+         texel[chan] = lp_build_select(&bld->texel_bld, use_border,
+                                       border_chan, texel[chan]);
+      }
+   }
 }
 
 
@@ -117,19 +220,31 @@ lp_build_sample_packed(struct lp_build_sample_context *bld,
                        LLVMValueRef x,
                        LLVMValueRef y,
                        LLVMValueRef y_stride,
-                       LLVMValueRef data_ptr)
+                       LLVMValueRef data_array)
 {
    LLVMValueRef offset;
+   LLVMValueRef data_ptr;
 
-   offset = lp_build_sample_offset(&bld->int_coord_bld,
+   offset = lp_build_sample_offset(&bld->uint_coord_bld,
                                    bld->format_desc,
-                                   x, y, y_stride,
-                                   data_ptr);
+                                   x, y, y_stride);
 
    assert(bld->format_desc->block.width == 1);
    assert(bld->format_desc->block.height == 1);
    assert(bld->format_desc->block.bits <= bld->texel_type.width);
 
+   /* XXX always use mipmap level 0 for now */
+   {
+      const int level = 0;
+      LLVMValueRef indexes[2];
+      /* get data_ptr[level] */
+      indexes[0] = LLVMConstInt(LLVMInt32Type(), 0, 0);
+      indexes[1] = LLVMConstInt(LLVMInt32Type(), level, 0);
+      data_ptr = LLVMBuildGEP(bld->builder, data_array, indexes, 2, "");
+      /* load texture base address */
+      data_ptr = LLVMBuildLoad(bld->builder, data_ptr, "");
+   }
+
    return lp_build_gather(bld->builder,
                           bld->texel_type.length,
                           bld->format_desc->block.bits,
@@ -138,17 +253,77 @@ lp_build_sample_packed(struct lp_build_sample_context *bld,
 }
 
 
+/**
+ * Helper to compute the mirror function for the PIPE_WRAP_MIRROR modes.
+ */
 static LLVMValueRef
-lp_build_sample_wrap(struct lp_build_sample_context *bld,
-                     LLVMValueRef coord,
-                     LLVMValueRef length,
-                     boolean is_pot,
-                     unsigned wrap_mode)
+lp_build_coord_mirror(struct lp_build_sample_context *bld,
+                      LLVMValueRef coord)
 {
+   struct lp_build_context *coord_bld = &bld->coord_bld;
+   struct lp_build_context *int_coord_bld = &bld->int_coord_bld;
+   LLVMValueRef fract, flr, isOdd;
+
+   /* fract = coord - floor(coord) */
+   fract = lp_build_sub(coord_bld, coord, lp_build_floor(coord_bld, coord));
+
+   /* flr = ifloor(coord); */
+   flr = lp_build_ifloor(coord_bld, coord);
+
+   /* isOdd = flr & 1 */
+   isOdd = LLVMBuildAnd(bld->builder, flr, int_coord_bld->one, "");
+
+   /* make coord positive or negative depending on isOdd */
+   coord = lp_build_set_sign(coord_bld, fract, isOdd);
+
+   /* convert isOdd to float */
+   isOdd = lp_build_int_to_float(coord_bld, isOdd);
+
+   /* add isOdd to coord */
+   coord = lp_build_add(coord_bld, coord, isOdd);
+
+   return coord;
+}
+
+
+/**
+ * We only support a few wrap modes in lp_build_sample_wrap_int() at this time.
+ * Return whether the given mode is supported by that function.
+ */
+static boolean
+is_simple_wrap_mode(unsigned mode)
+{
+   switch (mode) {
+   case PIPE_TEX_WRAP_REPEAT:
+   case PIPE_TEX_WRAP_CLAMP:
+   case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+      return TRUE;
+   case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+   default:
+      return FALSE;
+   }
+}
+
+
+/**
+ * Build LLVM code for texture wrap mode, for scaled integer texcoords.
+ * \param coord  the incoming texcoord (s,t,r or q) scaled to the texture size
+ * \param length  the texture size along one dimension
+ * \param is_pot  if TRUE, length is a power of two
+ * \param wrap_mode  one of PIPE_TEX_WRAP_x
+ */
+static LLVMValueRef
+lp_build_sample_wrap_int(struct lp_build_sample_context *bld,
+                         LLVMValueRef coord,
+                         LLVMValueRef length,
+                         boolean is_pot,
+                         unsigned wrap_mode)
+{
+   struct lp_build_context *uint_coord_bld = &bld->uint_coord_bld;
    struct lp_build_context *int_coord_bld = &bld->int_coord_bld;
    LLVMValueRef length_minus_one;
 
-   length_minus_one = lp_build_sub(int_coord_bld, length, int_coord_bld->one);
+   length_minus_one = lp_build_sub(uint_coord_bld, length, uint_coord_bld->one);
 
    switch(wrap_mode) {
    case PIPE_TEX_WRAP_REPEAT:
@@ -161,12 +336,12 @@ lp_build_sample_wrap(struct lp_build_sample_context *bld,
       break;
 
    case PIPE_TEX_WRAP_CLAMP:
+   case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+   case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
       coord = lp_build_max(int_coord_bld, coord, int_coord_bld->zero);
       coord = lp_build_min(int_coord_bld, coord, length_minus_one);
       break;
 
-   case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
-   case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
    case PIPE_TEX_WRAP_MIRROR_REPEAT:
    case PIPE_TEX_WRAP_MIRROR_CLAMP:
    case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
@@ -174,8 +349,8 @@ lp_build_sample_wrap(struct lp_build_sample_context *bld,
       /* FIXME */
       _debug_printf("llvmpipe: failed to translate texture wrap mode %s\n",
                     util_dump_tex_wrap(wrap_mode, TRUE));
-      coord = lp_build_max(int_coord_bld, coord, int_coord_bld->zero);
-      coord = lp_build_min(int_coord_bld, coord, length_minus_one);
+      coord = lp_build_max(uint_coord_bld, coord, uint_coord_bld->zero);
+      coord = lp_build_min(uint_coord_bld, coord, length_minus_one);
       break;
 
    default:
@@ -186,6 +361,380 @@ lp_build_sample_wrap(struct lp_build_sample_context *bld,
 }
 
 
+/**
+ * Build LLVM code for texture wrap mode for linear filtering.
+ * \param x0_out  returns first integer texcoord
+ * \param x1_out  returns second integer texcoord
+ * \param weight_out  returns linear interpolation weight
+ */
+static void
+lp_build_sample_wrap_linear(struct lp_build_sample_context *bld,
+                            LLVMValueRef coord,
+                            LLVMValueRef length,
+                            boolean is_pot,
+                            unsigned wrap_mode,
+                            LLVMValueRef *x0_out,
+                            LLVMValueRef *x1_out,
+                            LLVMValueRef *weight_out)
+{
+   struct lp_build_context *coord_bld = &bld->coord_bld;
+   struct lp_build_context *int_coord_bld = &bld->int_coord_bld;
+   struct lp_build_context *uint_coord_bld = &bld->uint_coord_bld;
+   LLVMValueRef two = lp_build_const_scalar(coord_bld->type, 2.0);
+   LLVMValueRef half = lp_build_const_scalar(coord_bld->type, 0.5);
+   LLVMValueRef length_f = lp_build_int_to_float(coord_bld, length);
+   LLVMValueRef length_minus_one = lp_build_sub(uint_coord_bld, length, uint_coord_bld->one);
+   LLVMValueRef length_f_minus_one = lp_build_sub(coord_bld, length_f, coord_bld->one);
+   LLVMValueRef coord0, coord1, weight;
+
+   switch(wrap_mode) {
+   case PIPE_TEX_WRAP_REPEAT:
+      /* mul by size and subtract 0.5 */
+      coord = lp_build_mul(coord_bld, coord, length_f);
+      coord = lp_build_sub(coord_bld, coord, half);
+      /* convert to int */
+      coord0 = lp_build_ifloor(coord_bld, coord);
+      coord1 = lp_build_add(uint_coord_bld, coord0, uint_coord_bld->one);
+      /* compute lerp weight */
+      weight = lp_build_fract(coord_bld, coord);
+      /* repeat wrap */
+      if (is_pot) {
+         coord0 = LLVMBuildAnd(bld->builder, coord0, length_minus_one, "");
+         coord1 = LLVMBuildAnd(bld->builder, coord1, length_minus_one, "");
+      }
+      else {
+         /* Signed remainder won't give the right results for negative
+          * dividends but unsigned remainder does.*/
+         coord0 = LLVMBuildURem(bld->builder, coord0, length, "");
+         coord1 = LLVMBuildURem(bld->builder, coord1, length, "");
+      }
+      break;
+
+   case PIPE_TEX_WRAP_CLAMP:
+      if (bld->static_state->normalized_coords) {
+         coord = lp_build_mul(coord_bld, coord, length_f);
+      }
+      weight = lp_build_fract(coord_bld, coord);
+      coord0 = lp_build_clamp(coord_bld, coord, coord_bld->zero,
+                              length_f_minus_one);
+      coord1 = lp_build_add(coord_bld, coord, coord_bld->one);
+      coord1 = lp_build_clamp(coord_bld, coord1, coord_bld->zero,
+                              length_f_minus_one);
+      coord0 = lp_build_ifloor(coord_bld, coord0);
+      coord1 = lp_build_ifloor(coord_bld, coord1);
+      break;
+
+   case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+      if (bld->static_state->normalized_coords) {
+         /* clamp to [0,1] */
+         coord = lp_build_clamp(coord_bld, coord, coord_bld->zero, coord_bld->one);
+         /* mul by tex size and subtract 0.5 */
+         coord = lp_build_mul(coord_bld, coord, length_f);
+         coord = lp_build_sub(coord_bld, coord, half);
+      }
+      else {
+         LLVMValueRef min, max;
+         /* clamp to [0.5, length - 0.5] */
+         min = lp_build_const_scalar(coord_bld->type, 0.5F);
+         max = lp_build_sub(coord_bld, length_f, min);
+         coord = lp_build_clamp(coord_bld, coord, min, max);
+      }
+      /* compute lerp weight */
+      weight = lp_build_fract(coord_bld, coord);
+      /* coord0 = floor(coord); */
+      coord0 = lp_build_ifloor(coord_bld, coord);
+      coord1 = lp_build_add(int_coord_bld, coord0, int_coord_bld->one);
+      /* coord0 = max(coord0, 0) */
+      coord0 = lp_build_max(int_coord_bld, coord0, int_coord_bld->zero);
+      /* coord1 = min(coord1, length-1) */
+      coord1 = lp_build_min(int_coord_bld, coord1, length_minus_one);
+      break;
+
+   case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+      {
+         LLVMValueRef min, max;
+         if (bld->static_state->normalized_coords) {
+            /* min = -1.0 / (2 * length) = -0.5 / length */
+            min = lp_build_mul(coord_bld,
+                               lp_build_const_scalar(coord_bld->type, -0.5F),
+                               lp_build_rcp(coord_bld, length_f));
+            /* max = 1.0 - min */
+            max = lp_build_sub(coord_bld, coord_bld->one, min);
+            /* coord = clamp(coord, min, max) */
+            coord = lp_build_clamp(coord_bld, coord, min, max);
+            /* scale coord to length (and sub 0.5?) */
+            coord = lp_build_mul(coord_bld, coord, length_f);
+            coord = lp_build_sub(coord_bld, coord, half);
+         }
+         else {
+            /* clamp to [-0.5, length + 0.5] */
+            min = lp_build_const_scalar(coord_bld->type, -0.5F);
+            max = lp_build_sub(coord_bld, length_f, min);
+            coord = lp_build_clamp(coord_bld, coord, min, max);
+            coord = lp_build_sub(coord_bld, coord, half);
+         }
+         /* compute lerp weight */
+         weight = lp_build_fract(coord_bld, coord);
+         /* convert to int */
+         coord0 = lp_build_ifloor(coord_bld, coord);
+         coord1 = lp_build_add(int_coord_bld, coord0, int_coord_bld->one);
+      }
+      break;
+
+   case PIPE_TEX_WRAP_MIRROR_REPEAT:
+      /* compute mirror function */
+      coord = lp_build_coord_mirror(bld, coord);
+
+      /* scale coord to length */
+      coord = lp_build_mul(coord_bld, coord, length_f);
+      coord = lp_build_sub(coord_bld, coord, half);
+
+      /* compute lerp weight */
+      weight = lp_build_fract(coord_bld, coord);
+
+      /* convert to int coords */
+      coord0 = lp_build_ifloor(coord_bld, coord);
+      coord1 = lp_build_add(int_coord_bld, coord0, int_coord_bld->one);
+
+      /* coord0 = max(coord0, 0) */
+      coord0 = lp_build_max(int_coord_bld, coord0, int_coord_bld->zero);
+      /* coord1 = min(coord1, length-1) */
+      coord1 = lp_build_min(int_coord_bld, coord1, length_minus_one);
+      break;
+
+   case PIPE_TEX_WRAP_MIRROR_CLAMP:
+      {
+         LLVMValueRef min, max;
+         /* min = 1.0 / (2 * length) */
+         min = lp_build_rcp(coord_bld, lp_build_mul(coord_bld, two, length_f));
+         /* max = 1.0 - min */
+         max = lp_build_sub(coord_bld, coord_bld->one, min);
+
+         coord = lp_build_abs(coord_bld, coord);
+         coord = lp_build_clamp(coord_bld, coord, min, max);
+         coord = lp_build_mul(coord_bld, coord, length_f);
+         if(0)coord = lp_build_sub(coord_bld, coord, half);
+         weight = lp_build_fract(coord_bld, coord);
+         coord0 = lp_build_ifloor(coord_bld, coord);
+         coord1 = lp_build_add(int_coord_bld, coord0, int_coord_bld->one);
+      }
+      break;
+
+   case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+      {
+         LLVMValueRef min, max;
+         /* min = 1.0 / (2 * length) */
+         min = lp_build_rcp(coord_bld, lp_build_mul(coord_bld, two, length_f));
+         /* max = 1.0 - min */
+         max = lp_build_sub(coord_bld, coord_bld->one, min);
+
+         coord = lp_build_abs(coord_bld, coord);
+         coord = lp_build_clamp(coord_bld, coord, min, max);
+         coord = lp_build_mul(coord_bld, coord, length_f);
+         coord = lp_build_sub(coord_bld, coord, half);
+         weight = lp_build_fract(coord_bld, coord);
+         coord0 = lp_build_ifloor(coord_bld, coord);
+         coord1 = lp_build_add(int_coord_bld, coord0, int_coord_bld->one);
+      }
+      break;
+
+   case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+      {
+         LLVMValueRef min, max;
+         /* min = -1.0 / (2 * length) = -0.5 / length */
+         min = lp_build_mul(coord_bld,
+                            lp_build_const_scalar(coord_bld->type, -0.5F),
+                            lp_build_rcp(coord_bld, length_f));
+         /* max = 1.0 - min */
+         max = lp_build_sub(coord_bld, coord_bld->one, min);
+
+         coord = lp_build_abs(coord_bld, coord);
+         coord = lp_build_clamp(coord_bld, coord, min, max);
+         coord = lp_build_mul(coord_bld, coord, length_f);
+         coord = lp_build_sub(coord_bld, coord, half);
+         weight = lp_build_fract(coord_bld, coord);
+         coord0 = lp_build_ifloor(coord_bld, coord);
+         coord1 = lp_build_add(int_coord_bld, coord0, int_coord_bld->one);
+      }
+      break;
+
+   default:
+      assert(0);
+      coord0 = NULL;
+      coord1 = NULL;
+      weight = NULL;
+   }
+
+   *x0_out = coord0;
+   *x1_out = coord1;
+   *weight_out = weight;
+}
+
+
+/**
+ * Build LLVM code for texture wrap mode for nearest filtering.
+ * \param coord  the incoming texcoord (nominally in [0,1])
+ * \param length  the texture size along one dimension, as int
+ * \param is_pot  if TRUE, length is a power of two
+ * \param wrap_mode  one of PIPE_TEX_WRAP_x
+ */
+static LLVMValueRef
+lp_build_sample_wrap_nearest(struct lp_build_sample_context *bld,
+                             LLVMValueRef coord,
+                             LLVMValueRef length,
+                             boolean is_pot,
+                             unsigned wrap_mode)
+{
+   struct lp_build_context *coord_bld = &bld->coord_bld;
+   struct lp_build_context *int_coord_bld = &bld->int_coord_bld;
+   struct lp_build_context *uint_coord_bld = &bld->uint_coord_bld;
+   LLVMValueRef two = lp_build_const_scalar(coord_bld->type, 2.0);
+   LLVMValueRef length_f = lp_build_int_to_float(coord_bld, length);
+   LLVMValueRef length_minus_one = lp_build_sub(uint_coord_bld, length, uint_coord_bld->one);
+   LLVMValueRef length_f_minus_one = lp_build_sub(coord_bld, length_f, coord_bld->one);
+   LLVMValueRef icoord;
+   
+   switch(wrap_mode) {
+   case PIPE_TEX_WRAP_REPEAT:
+      coord = lp_build_mul(coord_bld, coord, length_f);
+      icoord = lp_build_ifloor(coord_bld, coord);
+      if (is_pot)
+         icoord = LLVMBuildAnd(bld->builder, icoord, length_minus_one, "");
+      else
+         /* Signed remainder won't give the right results for negative
+          * dividends but unsigned remainder does.*/
+         icoord = LLVMBuildURem(bld->builder, icoord, length, "");
+      break;
+
+   case PIPE_TEX_WRAP_CLAMP:
+      /* mul by size */
+      if (bld->static_state->normalized_coords) {
+         coord = lp_build_mul(coord_bld, coord, length_f);
+      }
+      /* floor */
+      icoord = lp_build_ifloor(coord_bld, coord);
+      /* clamp to [0, size-1].  Note: int coord builder type */
+      icoord = lp_build_clamp(int_coord_bld, icoord, int_coord_bld->zero,
+                              length_minus_one);
+      break;
+
+   case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+      {
+         LLVMValueRef min, max;
+         if (bld->static_state->normalized_coords) {
+            /* min = 1.0 / (2 * length) */
+            min = lp_build_rcp(coord_bld, lp_build_mul(coord_bld, two, length_f));
+            /* max = length - min */
+            max = lp_build_sub(coord_bld, length_f, min);
+            /* scale coord to length */
+            coord = lp_build_mul(coord_bld, coord, length_f);
+         }
+         else {
+            /* clamp to [0.5, length - 0.5] */
+            min = lp_build_const_scalar(coord_bld->type, 0.5F);
+            max = lp_build_sub(coord_bld, length_f, min);
+         }
+         /* coord = clamp(coord, min, max) */
+         coord = lp_build_clamp(coord_bld, coord, min, max);
+         icoord = lp_build_ifloor(coord_bld, coord);
+      }
+      break;
+
+   case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+      /* Note: this is the same as CLAMP_TO_EDGE, except min = -min */
+      {
+         LLVMValueRef min, max;
+         if (bld->static_state->normalized_coords) {
+            /* min = -1.0 / (2 * length) = -0.5 / length */
+            min = lp_build_mul(coord_bld,
+                               lp_build_const_scalar(coord_bld->type, -0.5F),
+                               lp_build_rcp(coord_bld, length_f));
+            /* max = length - min */
+            max = lp_build_sub(coord_bld, length_f, min);
+            /* scale coord to length */
+            coord = lp_build_mul(coord_bld, coord, length_f);
+         }
+         else {
+            /* clamp to [-0.5, length + 0.5] */
+            min = lp_build_const_scalar(coord_bld->type, -0.5F);
+            max = lp_build_sub(coord_bld, length_f, min);
+         }
+         /* coord = clamp(coord, min, max) */
+         coord = lp_build_clamp(coord_bld, coord, min, max);
+         icoord = lp_build_ifloor(coord_bld, coord);
+      }
+      break;
+
+   case PIPE_TEX_WRAP_MIRROR_REPEAT:
+      {
+         LLVMValueRef min, max;
+         /* min = 1.0 / (2 * length) */
+         min = lp_build_rcp(coord_bld, lp_build_mul(coord_bld, two, length_f));
+         /* max = length - min */
+         max = lp_build_sub(coord_bld, length_f, min);
+
+         /* compute mirror function */
+         coord = lp_build_coord_mirror(bld, coord);
+
+         /* scale coord to length */
+         coord = lp_build_mul(coord_bld, coord, length_f);
+
+         /* coord = clamp(coord, min, max) */
+         coord = lp_build_clamp(coord_bld, coord, min, max);
+         icoord = lp_build_ifloor(coord_bld, coord);
+      }
+      break;
+
+   case PIPE_TEX_WRAP_MIRROR_CLAMP:
+      coord = lp_build_abs(coord_bld, coord);
+      coord = lp_build_mul(coord_bld, coord, length_f);
+      coord = lp_build_clamp(coord_bld, coord, coord_bld->zero, length_f_minus_one);
+      icoord = lp_build_ifloor(coord_bld, coord);
+      break;
+
+   case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+      {
+         LLVMValueRef min, max;
+         /* min = 1.0 / (2 * length) */
+         min = lp_build_rcp(coord_bld, lp_build_mul(coord_bld, two, length_f));
+         /* max = length - min */
+         max = lp_build_sub(coord_bld, length_f, min);
+
+         coord = lp_build_abs(coord_bld, coord);
+         coord = lp_build_mul(coord_bld, coord, length_f);
+         coord = lp_build_clamp(coord_bld, coord, min, max);
+         icoord = lp_build_ifloor(coord_bld, coord);
+      }
+      break;
+
+   case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+      {
+         LLVMValueRef min, max;
+         /* min = 1.0 / (2 * length) */
+         min = lp_build_rcp(coord_bld, lp_build_mul(coord_bld, two, length_f));
+         min = lp_build_negate(coord_bld, min);
+         /* max = length - min */
+         max = lp_build_sub(coord_bld, length_f, min);
+
+         coord = lp_build_abs(coord_bld, coord);
+         coord = lp_build_mul(coord_bld, coord, length_f);
+         coord = lp_build_clamp(coord_bld, coord, min, max);
+         icoord = lp_build_ifloor(coord_bld, coord);
+      }
+      break;
+
+   default:
+      assert(0);
+      icoord = NULL;
+   }
+
+   return icoord;
+}
+
+
+/**
+ * Sample 2D texture with nearest filtering.
+ */
 static void
 lp_build_sample_2d_nearest_soa(struct lp_build_sample_context *bld,
                                LLVMValueRef s,
@@ -193,26 +742,28 @@ lp_build_sample_2d_nearest_soa(struct lp_build_sample_context *bld,
                                LLVMValueRef width,
                                LLVMValueRef height,
                                LLVMValueRef stride,
-                               LLVMValueRef data_ptr,
+                               LLVMValueRef data_array,
                                LLVMValueRef *texel)
 {
-   LLVMValueRef x;
-   LLVMValueRef y;
+   LLVMValueRef x, y;
 
-   x = lp_build_ifloor(&bld->coord_bld, s);
-   y = lp_build_ifloor(&bld->coord_bld, t);
-   lp_build_name(x, "tex.x.floor");
-   lp_build_name(y, "tex.y.floor");
+   x = lp_build_sample_wrap_nearest(bld, s, width,
+                                    bld->static_state->pot_width,
+                                    bld->static_state->wrap_s);
+   y = lp_build_sample_wrap_nearest(bld, t, height,
+                                    bld->static_state->pot_height,
+                                    bld->static_state->wrap_t);
 
-   x = lp_build_sample_wrap(bld, x, width,  bld->static_state->pot_width,  bld->static_state->wrap_s);
-   y = lp_build_sample_wrap(bld, y, height, bld->static_state->pot_height, bld->static_state->wrap_t);
    lp_build_name(x, "tex.x.wrapped");
    lp_build_name(y, "tex.y.wrapped");
 
-   lp_build_sample_texel_soa(bld, x, y, stride, data_ptr, texel);
+   lp_build_sample_texel_soa(bld, width, height, x, y, stride, data_array, texel);
 }
 
 
+/**
+ * Sample 2D texture with bilinear filtering.
+ */
 static void
 lp_build_sample_2d_linear_soa(struct lp_build_sample_context *bld,
                               LLVMValueRef s,
@@ -220,12 +771,9 @@ lp_build_sample_2d_linear_soa(struct lp_build_sample_context *bld,
                               LLVMValueRef width,
                               LLVMValueRef height,
                               LLVMValueRef stride,
-                              LLVMValueRef data_ptr,
+                              LLVMValueRef data_array,
                               LLVMValueRef *texel)
 {
-   LLVMValueRef half;
-   LLVMValueRef s_ipart;
-   LLVMValueRef t_ipart;
    LLVMValueRef s_fpart;
    LLVMValueRef t_fpart;
    LLVMValueRef x0, x1;
@@ -233,32 +781,15 @@ lp_build_sample_2d_linear_soa(struct lp_build_sample_context *bld,
    LLVMValueRef neighbors[2][2][4];
    unsigned chan;
 
-   half = lp_build_const_scalar(bld->coord_type, 0.5);
-   s = lp_build_sub(&bld->coord_bld, s, half);
-   t = lp_build_sub(&bld->coord_bld, t, half);
-
-   s_ipart = lp_build_floor(&bld->coord_bld, s);
-   t_ipart = lp_build_floor(&bld->coord_bld, t);
-
-   s_fpart = lp_build_sub(&bld->coord_bld, s, s_ipart);
-   t_fpart = lp_build_sub(&bld->coord_bld, t, t_ipart);
-
-   x0 = lp_build_itrunc(&bld->coord_bld, s_ipart);
-   y0 = lp_build_itrunc(&bld->coord_bld, t_ipart);
-
-   x0 = lp_build_sample_wrap(bld, x0, width,  bld->static_state->pot_width,  bld->static_state->wrap_s);
-   y0 = lp_build_sample_wrap(bld, y0, height, bld->static_state->pot_height, bld->static_state->wrap_t);
-
-   x1 = lp_build_add(&bld->int_coord_bld, x0, bld->int_coord_bld.one);
-   y1 = lp_build_add(&bld->int_coord_bld, y0, bld->int_coord_bld.one);
-
-   x1 = lp_build_sample_wrap(bld, x1, width,  bld->static_state->pot_width,  bld->static_state->wrap_s);
-   y1 = lp_build_sample_wrap(bld, y1, height, bld->static_state->pot_height, bld->static_state->wrap_t);
+   lp_build_sample_wrap_linear(bld, s, width, bld->static_state->pot_width,
+                               bld->static_state->wrap_s, &x0, &x1, &s_fpart);
+   lp_build_sample_wrap_linear(bld, t, height, bld->static_state->pot_height,
+                               bld->static_state->wrap_t, &y0, &y1, &t_fpart);
 
-   lp_build_sample_texel_soa(bld, x0, y0, stride, data_ptr, neighbors[0][0]);
-   lp_build_sample_texel_soa(bld, x1, y0, stride, data_ptr, neighbors[0][1]);
-   lp_build_sample_texel_soa(bld, x0, y1, stride, data_ptr, neighbors[1][0]);
-   lp_build_sample_texel_soa(bld, x1, y1, stride, data_ptr, neighbors[1][1]);
+   lp_build_sample_texel_soa(bld, width, height, x0, y0, stride, data_array, neighbors[0][0]);
+   lp_build_sample_texel_soa(bld, width, height, x1, y0, stride, data_array, neighbors[0][1]);
+   lp_build_sample_texel_soa(bld, width, height, x0, y1, stride, data_array, neighbors[1][0]);
+   lp_build_sample_texel_soa(bld, width, height, x1, y1, stride, data_array, neighbors[1][1]);
 
    /* TODO: Don't interpolate missing channels */
    for(chan = 0; chan < 4; ++chan) {
@@ -309,7 +840,7 @@ lp_build_sample_2d_linear_aos(struct lp_build_sample_context *bld,
                               LLVMValueRef width,
                               LLVMValueRef height,
                               LLVMValueRef stride,
-                              LLVMValueRef data_ptr,
+                              LLVMValueRef data_array,
                               LLVMValueRef *texel)
 {
    LLVMBuilderRef builder = bld->builder;
@@ -334,20 +865,33 @@ lp_build_sample_2d_linear_aos(struct lp_build_sample_context *bld,
    h16_vec_type = lp_build_vec_type(h16.type);
    u8n_vec_type = lp_build_vec_type(u8n.type);
 
+   if (bld->static_state->normalized_coords) {
+      LLVMTypeRef coord_vec_type = lp_build_vec_type(bld->coord_type);
+      LLVMValueRef fp_width = LLVMBuildSIToFP(bld->builder, width, coord_vec_type, "");
+      LLVMValueRef fp_height = LLVMBuildSIToFP(bld->builder, height, coord_vec_type, "");
+      s = lp_build_mul(&bld->coord_bld, s, fp_width);
+      t = lp_build_mul(&bld->coord_bld, t, fp_height);
+   }
+
+   /* scale coords by 256 (8 fractional bits) */
    s = lp_build_mul_imm(&bld->coord_bld, s, 256);
    t = lp_build_mul_imm(&bld->coord_bld, t, 256);
 
+   /* convert float to int */
    s = LLVMBuildFPToSI(builder, s, i32_vec_type, "");
    t = LLVMBuildFPToSI(builder, t, i32_vec_type, "");
 
+   /* subtract 0.5 (add -128) */
    i32_c128 = lp_build_int_const_scalar(i32.type, -128);
    s = LLVMBuildAdd(builder, s, i32_c128, "");
    t = LLVMBuildAdd(builder, t, i32_c128, "");
 
+   /* compute floor (shift right 8) */
    i32_c8 = lp_build_int_const_scalar(i32.type, 8);
    s_ipart = LLVMBuildAShr(builder, s, i32_c8, "");
    t_ipart = LLVMBuildAShr(builder, t, i32_c8, "");
 
+   /* compute fractional part (AND with 0xff) */
    i32_c255 = lp_build_int_const_scalar(i32.type, 255);
    s_fpart = LLVMBuildAnd(builder, s, i32_c255, "");
    t_fpart = LLVMBuildAnd(builder, t, i32_c255, "");
@@ -355,14 +899,18 @@ lp_build_sample_2d_linear_aos(struct lp_build_sample_context *bld,
    x0 = s_ipart;
    y0 = t_ipart;
 
-   x0 = lp_build_sample_wrap(bld, x0, width,  bld->static_state->pot_width,  bld->static_state->wrap_s);
-   y0 = lp_build_sample_wrap(bld, y0, height, bld->static_state->pot_height, bld->static_state->wrap_t);
-
    x1 = lp_build_add(&bld->int_coord_bld, x0, bld->int_coord_bld.one);
    y1 = lp_build_add(&bld->int_coord_bld, y0, bld->int_coord_bld.one);
 
-   x1 = lp_build_sample_wrap(bld, x1, width,  bld->static_state->pot_width,  bld->static_state->wrap_s);
-   y1 = lp_build_sample_wrap(bld, y1, height, bld->static_state->pot_height, bld->static_state->wrap_t);
+   x0 = lp_build_sample_wrap_int(bld, x0, width,  bld->static_state->pot_width,
+                                 bld->static_state->wrap_s);
+   y0 = lp_build_sample_wrap_int(bld, y0, height, bld->static_state->pot_height,
+                                 bld->static_state->wrap_t);
+
+   x1 = lp_build_sample_wrap_int(bld, x1, width,  bld->static_state->pot_width,
+                                 bld->static_state->wrap_s);
+   y1 = lp_build_sample_wrap_int(bld, y1, height, bld->static_state->pot_height,
+                                 bld->static_state->wrap_t);
 
    /*
     * Transform 4 x i32 in
@@ -432,10 +980,10 @@ lp_build_sample_2d_linear_aos(struct lp_build_sample_context *bld,
     * The higher 8 bits of the resulting elements will be zero.
     */
 
-   neighbors[0][0] = lp_build_sample_packed(bld, x0, y0, stride, data_ptr);
-   neighbors[0][1] = lp_build_sample_packed(bld, x1, y0, stride, data_ptr);
-   neighbors[1][0] = lp_build_sample_packed(bld, x0, y1, stride, data_ptr);
-   neighbors[1][1] = lp_build_sample_packed(bld, x1, y1, stride, data_ptr);
+   neighbors[0][0] = lp_build_sample_packed(bld, x0, y0, stride, data_array);
+   neighbors[0][1] = lp_build_sample_packed(bld, x1, y0, stride, data_array);
+   neighbors[1][0] = lp_build_sample_packed(bld, x0, y1, stride, data_array);
+   neighbors[1][1] = lp_build_sample_packed(bld, x1, y1, stride, data_array);
 
    neighbors[0][0] = LLVMBuildBitCast(builder, neighbors[0][0], u8n_vec_type, "");
    neighbors[0][1] = LLVMBuildBitCast(builder, neighbors[0][1], u8n_vec_type, "");
@@ -518,6 +1066,195 @@ lp_build_sample_compare(struct lp_build_sample_context *bld,
 }
 
 
+static int
+texture_dims(enum pipe_texture_target tex)
+{
+   switch (tex) {
+   case PIPE_TEXTURE_1D:
+      return 1;
+   case PIPE_TEXTURE_2D:
+   case PIPE_TEXTURE_CUBE:
+      return 2;
+   case PIPE_TEXTURE_3D:
+      return 3;
+   default:
+      assert(0 && "bad texture target in texture_dims()");
+      return 2;
+   }
+}
+
+
+/**
+ * Generate code to compute texture level of detail (lambda).
+ * \param s  vector of texcoord s values
+ * \param t  vector of texcoord t values
+ * \param r  vector of texcoord r values
+ * \param width  scalar int texture width
+ * \param height  scalar int texture height
+ * \param depth  scalar int texture depth
+ */
+static LLVMValueRef
+lp_build_lod_selector(struct lp_build_sample_context *bld,
+                      LLVMValueRef s,
+                      LLVMValueRef t,
+                      LLVMValueRef r,
+                      LLVMValueRef width,
+                      LLVMValueRef height,
+                      LLVMValueRef depth)
+
+{
+   const int dims = texture_dims(bld->static_state->target);
+   struct lp_build_context *coord_bld = &bld->coord_bld;
+
+   LLVMValueRef lod_bias = lp_build_const_scalar(bld->coord_bld.type,
+                                                 bld->static_state->lod_bias);
+   LLVMValueRef min_lod = lp_build_const_scalar(bld->coord_bld.type,
+                                                bld->static_state->min_lod);
+   LLVMValueRef max_lod = lp_build_const_scalar(bld->coord_bld.type,
+                                                bld->static_state->max_lod);
+
+   LLVMValueRef index0 = LLVMConstInt(LLVMInt32Type(), 0, 0);
+   LLVMValueRef index1 = LLVMConstInt(LLVMInt32Type(), 1, 0);
+   LLVMValueRef index2 = LLVMConstInt(LLVMInt32Type(), 2, 0);
+
+   LLVMValueRef s0, s1, s2;
+   LLVMValueRef t0, t1, t2;
+   LLVMValueRef r0, r1, r2;
+   LLVMValueRef dsdx, dsdy, dtdx, dtdy, drdx, drdy;
+   LLVMValueRef rho, lod;
+
+   /*
+    * dsdx = abs(s[1] - s[0]);
+    * dsdy = abs(s[2] - s[0]);
+    * dtdx = abs(t[1] - t[0]);
+    * dtdy = abs(t[2] - t[0]);
+    * drdx = abs(r[1] - r[0]);
+    * drdy = abs(r[2] - r[0]);
+    * XXX we're assuming a four-element quad in 2x2 layout here.
+    */
+   s0 = LLVMBuildExtractElement(bld->builder, s, index0, "s0");
+   s1 = LLVMBuildExtractElement(bld->builder, s, index1, "s1");
+   s2 = LLVMBuildExtractElement(bld->builder, s, index2, "s2");
+   dsdx = lp_build_abs(coord_bld, lp_build_sub(coord_bld, s1, s0));
+   dsdy = lp_build_abs(coord_bld, lp_build_sub(coord_bld, s2, s0));
+   if (dims > 1) {
+      t0 = LLVMBuildExtractElement(bld->builder, t, index0, "t0");
+      t1 = LLVMBuildExtractElement(bld->builder, t, index1, "t1");
+      t2 = LLVMBuildExtractElement(bld->builder, t, index2, "t2");
+      dtdx = lp_build_abs(coord_bld, lp_build_sub(coord_bld, t1, t0));
+      dtdy = lp_build_abs(coord_bld, lp_build_sub(coord_bld, t2, t0));
+      if (dims > 2) {
+         r0 = LLVMBuildExtractElement(bld->builder, r, index0, "r0");
+         r1 = LLVMBuildExtractElement(bld->builder, r, index1, "r1");
+         r2 = LLVMBuildExtractElement(bld->builder, r, index2, "r2");
+         drdx = lp_build_abs(coord_bld, lp_build_sub(coord_bld, r1, r0));
+         drdy = lp_build_abs(coord_bld, lp_build_sub(coord_bld, r2, r0));
+      }
+   }
+
+   /* Compute rho = max of all partial derivatives scaled by texture size.
+    * XXX this can be vectorized somewhat
+    */
+   rho = lp_build_mul(coord_bld,
+                       lp_build_max(coord_bld, dsdx, dsdy),
+                       lp_build_int_to_float(coord_bld, width));
+   if (dims > 1) {
+      LLVMValueRef max;
+      max = lp_build_mul(coord_bld,
+                         lp_build_max(coord_bld, dtdx, dtdy),
+                         lp_build_int_to_float(coord_bld, height));
+      rho = lp_build_max(coord_bld, rho, max);
+      if (dims > 2) {
+         max = lp_build_mul(coord_bld,
+                            lp_build_max(coord_bld, drdx, drdy),
+                            lp_build_int_to_float(coord_bld, depth));
+         rho = lp_build_max(coord_bld, rho, max);
+      }
+   }
+
+   /* compute lod = log2(rho) */
+   lod = lp_build_log2(coord_bld, rho);
+
+   /* add lod bias */
+   lod = lp_build_add(coord_bld, lod, lod_bias);
+
+   /* clamp lod */
+   lod = lp_build_clamp(coord_bld, lod, min_lod, max_lod);
+
+   return lod;
+}
+
+
+/**
+ * For PIPE_TEX_MIPFILTER_NEAREST, convert float LOD to integer
+ * mipmap level index.
+ * \param lod  scalar float texture level of detail
+ * \param level_out  returns integer 
+ */
+static void
+lp_build_nearest_mip_level(struct lp_build_sample_context *bld,
+                           unsigned unit,
+                           LLVMValueRef lod,
+                           LLVMValueRef *level_out)
+{
+   struct lp_build_context *coord_bld = &bld->coord_bld;
+   struct lp_build_context *int_coord_bld = &bld->int_coord_bld;
+   LLVMValueRef last_level, level;
+
+   last_level = bld->dynamic_state->last_level(bld->dynamic_state,
+                                               bld->builder, unit);
+
+   /* convert float lod to integer */
+   level = lp_build_iround(coord_bld, lod);
+
+   /* clamp level to legal range of levels */
+   *level_out = lp_build_clamp(int_coord_bld, level,
+                               int_coord_bld->zero,
+                               last_level);
+}
+
+
+/**
+ * For PIPE_TEX_MIPFILTER_LINEAR, convert float LOD to integer to
+ * two (adjacent) mipmap level indexes.  Later, we'll sample from those
+ * two mipmap levels and interpolate between them.
+ */
+static void
+lp_build_linear_mip_levels(struct lp_build_sample_context *bld,
+                           unsigned unit,
+                           LLVMValueRef lod,
+                           LLVMValueRef *level0_out,
+                           LLVMValueRef *level1_out,
+                           LLVMValueRef *weight_out)
+{
+   struct lp_build_context *coord_bld = &bld->coord_bld;
+   struct lp_build_context *int_coord_bld = &bld->int_coord_bld;
+   LLVMValueRef last_level, level;
+
+   last_level = bld->dynamic_state->last_level(bld->dynamic_state,
+                                               bld->builder, unit);
+
+   /* convert float lod to integer */
+   level = lp_build_ifloor(coord_bld, lod);
+
+   /* compute level 0 and clamp to legal range of levels */
+   *level0_out = lp_build_clamp(int_coord_bld, level,
+                                int_coord_bld->zero,
+                                last_level);
+   /* compute level 1 and clamp to legal range of levels */
+   *level1_out = lp_build_add(int_coord_bld, *level0_out, int_coord_bld->one);
+   *level1_out = lp_build_min(int_coord_bld, *level1_out, int_coord_bld->zero);
+
+   *weight_out = lp_build_fract(coord_bld, lod);
+}
+
+
+
+/**
+ * Build texture sampling code.
+ * 'texel' will return a vector of four LLVMValueRefs corresponding to
+ * R, G, B, A.
+ */
 void
 lp_build_sample_soa(LLVMBuilderRef builder,
                     const struct lp_sampler_static_state *static_state,
@@ -533,10 +1270,14 @@ lp_build_sample_soa(LLVMBuilderRef builder,
    LLVMValueRef width;
    LLVMValueRef height;
    LLVMValueRef stride;
-   LLVMValueRef data_ptr;
+   LLVMValueRef data_array;
    LLVMValueRef s;
    LLVMValueRef t;
-   LLVMValueRef p;
+   LLVMValueRef r;
+
+   (void) lp_build_lod_selector;   /* temporary to silence warning */
+   (void) lp_build_nearest_mip_level;
+   (void) lp_build_linear_mip_levels;
 
    /* Setup our build context */
    memset(&bld, 0, sizeof bld);
@@ -545,9 +1286,11 @@ lp_build_sample_soa(LLVMBuilderRef builder,
    bld.dynamic_state = dynamic_state;
    bld.format_desc = util_format_description(static_state->format);
    bld.coord_type = type;
+   bld.uint_coord_type = lp_uint_type(type);
    bld.int_coord_type = lp_int_type(type);
    bld.texel_type = type;
    lp_build_context_init(&bld.coord_bld, builder, bld.coord_type);
+   lp_build_context_init(&bld.uint_coord_bld, builder, bld.uint_coord_type);
    lp_build_context_init(&bld.int_coord_bld, builder, bld.int_coord_type);
    lp_build_context_init(&bld.texel_bld, builder, bld.texel_type);
 
@@ -555,36 +1298,34 @@ lp_build_sample_soa(LLVMBuilderRef builder,
    width = dynamic_state->width(dynamic_state, builder, unit);
    height = dynamic_state->height(dynamic_state, builder, unit);
    stride = dynamic_state->stride(dynamic_state, builder, unit);
-   data_ptr = dynamic_state->data_ptr(dynamic_state, builder, unit);
+   data_array = dynamic_state->data_ptr(dynamic_state, builder, unit);
+   /* Note that data_array is an array[level] of pointers to texture images */
 
    s = coords[0];
    t = coords[1];
-   p = coords[2];
+   r = coords[2];
 
-   width = lp_build_broadcast_scalar(&bld.int_coord_bld, width);
-   height = lp_build_broadcast_scalar(&bld.int_coord_bld, height);
-   stride = lp_build_broadcast_scalar(&bld.int_coord_bld, stride);
+   width = lp_build_broadcast_scalar(&bld.uint_coord_bld, width);
+   height = lp_build_broadcast_scalar(&bld.uint_coord_bld, height);
+   stride = lp_build_broadcast_scalar(&bld.uint_coord_bld, stride);
 
    if(static_state->target == PIPE_TEXTURE_1D)
       t = bld.coord_bld.zero;
 
-   if(static_state->normalized_coords) {
-      LLVMTypeRef coord_vec_type = lp_build_vec_type(bld.coord_type);
-      LLVMValueRef fp_width = LLVMBuildSIToFP(builder, width, coord_vec_type, "");
-      LLVMValueRef fp_height = LLVMBuildSIToFP(builder, height, coord_vec_type, "");
-      s = lp_build_mul(&bld.coord_bld, s, fp_width);
-      t = lp_build_mul(&bld.coord_bld, t, fp_height);
-   }
-
    switch (static_state->min_img_filter) {
    case PIPE_TEX_FILTER_NEAREST:
-      lp_build_sample_2d_nearest_soa(&bld, s, t, width, height, stride, data_ptr, texel);
+      lp_build_sample_2d_nearest_soa(&bld, s, t, width, height,
+                                     stride, data_array, texel);
       break;
    case PIPE_TEX_FILTER_LINEAR:
-      if(lp_format_is_rgba8(bld.format_desc))
-         lp_build_sample_2d_linear_aos(&bld, s, t, width, height, stride, data_ptr, texel);
+      if(lp_format_is_rgba8(bld.format_desc) &&
+         is_simple_wrap_mode(static_state->wrap_s) &&
+         is_simple_wrap_mode(static_state->wrap_t))
+         lp_build_sample_2d_linear_aos(&bld, s, t, width, height,
+                                       stride, data_array, texel);
       else
-         lp_build_sample_2d_linear_soa(&bld, s, t, width, height, stride, data_ptr, texel);
+         lp_build_sample_2d_linear_soa(&bld, s, t, width, height,
+                                       stride, data_array, texel);
       break;
    default:
       assert(0);
@@ -593,5 +1334,5 @@ lp_build_sample_soa(LLVMBuilderRef builder,
    /* FIXME: respect static_state->min_mip_filter */;
    /* FIXME: respect static_state->mag_img_filter */;
 
-   lp_build_sample_compare(&bld, p, texel);
+   lp_build_sample_compare(&bld, r, texel);
 }
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c b/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
index ae866243a92..fbb664d43a0 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
@@ -152,8 +152,7 @@ static void lp_exec_mask_init(struct lp_exec_mask *mask, struct lp_build_context
 static void lp_exec_mask_update(struct lp_exec_mask *mask)
 {
    mask->exec_mask = mask->cond_mask;
-   if (mask->cond_stack_size > 0)
-      mask->has_mask = TRUE;
+   mask->has_mask = (mask->cond_stack_size > 0);
 }
 
 static void lp_exec_mask_cond_push(struct lp_exec_mask *mask,
@@ -384,6 +383,11 @@ emit_store(
       assert(0);
       break;
 
+   case TGSI_FILE_PREDICATE:
+      /* FIXME */
+      assert(0);
+      break;
+
    default:
       assert( 0 );
    }
@@ -581,6 +585,17 @@ emit_instruction(
    if (indirect_temp_reference(inst))
       return FALSE;
 
+   /*
+    * Stores and write masks are handled in a general fashion after the long
+    * instruction opcode switch statement.
+    *
+    * Although not stricitly necessary, we avoid generating instructions for
+    * channels which won't be stored, in cases where's that easy. For some
+    * complex instructions, like texture sampling, it is more convenient to
+    * assume a full writemask and then let LLVM optimization passes eliminate
+    * redundant code.
+    */
+
    assert(info->num_dst <= 1);
    if(info->num_dst) {
       FOR_EACH_DST0_ENABLED_CHANNEL( inst, chan_index ) {
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_type.c b/src/gallium/auxiliary/gallivm/lp_bld_type.c
index 8270cd057f6..c327ba045a6 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_type.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_type.c
@@ -178,6 +178,25 @@ lp_build_int32_vec4_type(void)
 }
 
 
+/**
+ * Create unsigned integer type variation of given type.
+ */
+struct lp_type
+lp_uint_type(struct lp_type type)
+{
+   struct lp_type res_type;
+
+   memset(&res_type, 0, sizeof res_type);
+   res_type.width = type.width;
+   res_type.length = type.length;
+
+   return res_type;
+}
+
+
+/**
+ * Create signed integer type variation of given type.
+ */
 struct lp_type
 lp_int_type(struct lp_type type)
 {
@@ -186,6 +205,7 @@ lp_int_type(struct lp_type type)
    memset(&res_type, 0, sizeof res_type);
    res_type.width = type.width;
    res_type.length = type.length;
+   res_type.sign = 1;
 
    return res_type;
 }
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_type.h b/src/gallium/auxiliary/gallivm/lp_bld_type.h
index 62ee05be4df..16946cc28a2 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_type.h
+++ b/src/gallium/auxiliary/gallivm/lp_bld_type.h
@@ -257,6 +257,10 @@ lp_build_int32_vec4_type(void);
 
 
 struct lp_type
+lp_uint_type(struct lp_type type);
+
+
+struct lp_type
 lp_int_type(struct lp_type type);
 
 
diff --git a/src/gallium/auxiliary/util/u_format.h b/src/gallium/auxiliary/util/u_format.h
index e8fa0022b5b..b2aa5bfb188 100644
--- a/src/gallium/auxiliary/util/u_format.h
+++ b/src/gallium/auxiliary/util/u_format.h
@@ -120,9 +120,15 @@ struct util_format_channel_description
 struct util_format_description
 {
    enum pipe_format format;
+
    const char *name;
 
    /**
+    * Short name, striped of the prefix, lower case.
+    */
+   const char *short_name;
+
+   /**
     * Pixel block dimensions.
     */
    struct util_format_block block;
@@ -140,6 +146,15 @@ struct util_format_description
    unsigned is_array:1;
 
    /**
+    * Whether the pixel format can be described as a bitfield structure.
+    *
+    * In particular:
+    * - pixel depth must be 8, 16, or 32 bits;
+    * - all channels must be unsigned, signed, or void
+    */
+   unsigned is_bitmask:1;
+
+   /**
     * Whether channels have mixed types (ignoring UTIL_FORMAT_TYPE_VOID).
     */
    unsigned is_mixed:1;
diff --git a/src/gallium/auxiliary/util/u_format_pack.py b/src/gallium/auxiliary/util/u_format_pack.py
index 3f33f7cc021..b49039db39b 100644
--- a/src/gallium/auxiliary/util/u_format_pack.py
+++ b/src/gallium/auxiliary/util/u_format_pack.py
@@ -252,9 +252,6 @@ def conversion_expr(src_channel, dst_channel, dst_native_type, value, clamp=True
     if src_channel.type == FLOAT and dst_channel.type == FLOAT:
         return '(%s)%s' % (dst_native_type, value)
     
-    if not src_channel.norm and not dst_channel.norm:
-        return '(%s)%s' % (dst_native_type, value)
-
     if clamp:
         value = clamp_expr(src_channel, dst_channel, dst_native_type, value)
 
@@ -280,15 +277,15 @@ def conversion_expr(src_channel, dst_channel, dst_native_type, value, clamp=True
             value = '(%s * %s)' % (value, scale)
         return '(%s)%s' % (dst_native_type, value)
 
-    if not src_channel.norm and not dst_channel.norm:
-        # neither is normalized -- just cast
-        return '(%s)%s' % (dst_native_type, value)
-
     if src_channel.type in (SIGNED, UNSIGNED) and dst_channel.type in (SIGNED, UNSIGNED):
+        if not src_channel.norm and not dst_channel.norm:
+            # neither is normalized -- just cast
+            return '(%s)%s' % (dst_native_type, value)
+
         src_one = get_one(src_channel)
         dst_one = get_one(dst_channel)
 
-        if src_one > dst_one and src_channel.norm:
+        if src_one > dst_one and src_channel.norm and dst_channel.norm:
             # We can just bitshift
             src_shift = get_one_shift(src_channel)
             dst_shift = get_one_shift(dst_channel)
@@ -296,7 +293,7 @@ def conversion_expr(src_channel, dst_channel, dst_native_type, value, clamp=True
         else:
             # We need to rescale using an intermediate type big enough to hold the multiplication of both
             tmp_native_type = intermediate_native_type(src_channel.size + dst_channel.size, src_channel.sign and dst_channel.sign)
-            value = '(%s)%s' % (tmp_native_type, value)
+            value = '((%s)%s)' % (tmp_native_type, value)
             value = '(%s * 0x%x / 0x%x)' % (value, dst_one, src_one)
         value = '(%s)%s' % (dst_native_type, value)
         return value
@@ -307,6 +304,8 @@ def conversion_expr(src_channel, dst_channel, dst_native_type, value, clamp=True
 def generate_format_unpack(format, dst_channel, dst_native_type, dst_suffix):
     '''Generate the function to unpack pixels from a particular format'''
 
+    assert format.layout == PLAIN
+
     name = format.short_name()
 
     src_native_type = native_type(format)
@@ -314,32 +313,99 @@ def generate_format_unpack(format, dst_channel, dst_native_type, dst_suffix):
     print 'static INLINE void'
     print 'util_format_%s_unpack_%s(%s *dst, const void *src)' % (name, dst_suffix, dst_native_type)
     print '{'
-    print '   union util_format_%s pixel;' % format.short_name()
-    print '   memcpy(&pixel, src, sizeof pixel);'
-    bswap_format(format)
+    
+    if format.is_bitmask():
+        depth = format.block_size()
+        print '   uint%u_t value = *(uint%u_t *)src;' % (depth, depth) 
 
-    assert format.layout == PLAIN
+        # Declare the intermediate variables
+        for i in range(format.nr_channels()):
+            src_channel = format.channels[i]
+            if src_channel.type == UNSIGNED:
+                print '   uint%u_t %s;' % (depth, src_channel.name)
+            elif src_channel.type == SIGNED:
+                print '   int%u_t %s;' % (depth, src_channel.name)
 
-    for i in range(4):
-        swizzle = format.swizzles[i]
-        if swizzle < 4:
-            src_channel = format.channels[swizzle]
-            value = 'pixel.chan.%s' % src_channel.name 
-            value = conversion_expr(src_channel, dst_channel, dst_native_type, value)
-        elif swizzle == SWIZZLE_0:
-            value = '0'
-        elif swizzle == SWIZZLE_1:
-            value = get_one(dst_channel)
-        elif swizzle == SWIZZLE_NONE:
-            value = '0'
-        else:
-            assert False
-        if format.colorspace == ZS:
-            if i == 3:
+        print '#ifdef PIPE_ARCH_BIG_ENDIAN'
+        print '   value = util_bswap%u(value);' % depth
+        print '#endif'
+
+        # Compute the intermediate unshifted values 
+        shift = 0
+        for i in range(format.nr_channels()):
+            src_channel = format.channels[i]
+            value = 'value'
+            if src_channel.type == UNSIGNED:
+                if shift:
+                    value = '%s >> %u' % (value, shift)
+                if shift + src_channel.size < depth:
+                    value = '(%s) & 0x%x' % (value, (1 << src_channel.size) - 1)
+            elif src_channel.type == SIGNED:
+                if shift + src_channel.size < depth:
+                    # Align the sign bit
+                    lshift = depth - (shift + src_channel.size)
+                    value = '%s << %u' % (value, lshift)
+                # Cast to signed
+                value = '(int%u_t)(%s) ' % (depth, value)
+                if src_channel.size < depth:
+                    # Align the LSB bit
+                    rshift = depth - src_channel.size
+                    value = '(%s) >> %u' % (value, rshift)
+            else:
+                value = None
+                
+            if value is not None:
+                print '   %s = %s;' % (src_channel.name, value)
+                
+            shift += src_channel.size
+
+        # Convert, swizzle, and store final values
+        for i in range(4):
+            swizzle = format.swizzles[i]
+            if swizzle < 4:
+                src_channel = format.channels[swizzle]
+                value = src_channel.name 
+                value = conversion_expr(src_channel, dst_channel, dst_native_type, value)
+            elif swizzle == SWIZZLE_0:
+                value = '0'
+            elif swizzle == SWIZZLE_1:
                 value = get_one(dst_channel)
-            elif i >= 1:
-                value = 'dst[0]'
-        print '   dst[%u] = %s; /* %s */' % (i, value, 'rgba'[i])
+            elif swizzle == SWIZZLE_NONE:
+                value = '0'
+            else:
+                assert False
+            if format.colorspace == ZS:
+                if i == 3:
+                    value = get_one(dst_channel)
+                elif i >= 1:
+                    value = 'dst[0]'
+            print '   dst[%u] = %s; /* %s */' % (i, value, 'rgba'[i])
+        
+    else:
+        print '   union util_format_%s pixel;' % format.short_name()
+        print '   memcpy(&pixel, src, sizeof pixel);'
+        bswap_format(format)
+    
+        for i in range(4):
+            swizzle = format.swizzles[i]
+            if swizzle < 4:
+                src_channel = format.channels[swizzle]
+                value = 'pixel.chan.%s' % src_channel.name 
+                value = conversion_expr(src_channel, dst_channel, dst_native_type, value)
+            elif swizzle == SWIZZLE_0:
+                value = '0'
+            elif swizzle == SWIZZLE_1:
+                value = get_one(dst_channel)
+            elif swizzle == SWIZZLE_NONE:
+                value = '0'
+            else:
+                assert False
+            if format.colorspace == ZS:
+                if i == 3:
+                    value = get_one(dst_channel)
+                elif i >= 1:
+                    value = 'dst[0]'
+            print '   dst[%u] = %s; /* %s */' % (i, value, 'rgba'[i])
 
     print '}'
     print
diff --git a/src/gallium/auxiliary/util/u_format_parse.py b/src/gallium/auxiliary/util/u_format_parse.py
index 250926418ec..f74dc5e88a4 100755
--- a/src/gallium/auxiliary/util/u_format_parse.py
+++ b/src/gallium/auxiliary/util/u_format_parse.py
@@ -78,7 +78,7 @@ class Channel:
         if self.type == UNSIGNED:
             return (1 << self.size) - 1
         if self.type == SIGNED:
-            return self.size - 1
+            return (1 << (self.size - 1)) - 1
         assert False
     
     def min(self):
@@ -166,17 +166,11 @@ class Format:
         return True
 
     def is_bitmask(self):
-        if self.block_size() > 32:
-            return False
-        if not self.is_pot():
+        if self.block_size() not in (8, 16, 32):
             return False
         for channel in self.channels:
-            if not is_pot(channel.size):
-                return True
             if channel.type not in (VOID, UNSIGNED, SIGNED):
                 return False
-            if channel.size >= 32:
-                return False
         return True
 
     def inv_swizzles(self):
diff --git a/src/gallium/auxiliary/util/u_format_table.py b/src/gallium/auxiliary/util/u_format_table.py
index 4e29d15f3bb..fb68852a530 100755
--- a/src/gallium/auxiliary/util/u_format_table.py
+++ b/src/gallium/auxiliary/util/u_format_table.py
@@ -90,11 +90,13 @@ def write_format_table(formats):
     print 'util_format_none_description = {'
     print "   PIPE_FORMAT_NONE,"
     print "   \"PIPE_FORMAT_NONE\","
+    print "   \"none\","
     print "   {0, 0, 0},"
     print "   0,"
     print "   0,"
     print "   0,"
     print "   0,"
+    print "   0,"
     print "   {{0, 0, 0}, {0, 0, 0}, {0, 0, 0}, {0, 0, 0}},"
     print "   {0, 0, 0, 0},"
     print "   0"
@@ -105,10 +107,12 @@ def write_format_table(formats):
         print 'util_format_%s_description = {' % (format.short_name(),)
         print "   %s," % (format.name,)
         print "   \"%s\"," % (format.name,)
+        print "   \"%s\"," % (format.short_name(),)
         print "   {%u, %u, %u},\t/* block */" % (format.block_width, format.block_height, format.block_size())
         print "   %s," % (layout_map(format.layout),)
         print "   %u,\t/* nr_channels */" % (format.nr_channels(),)
         print "   %s,\t/* is_array */" % (bool_map(format.is_array()),)
+        print "   %s,\t/* is_bitmask */" % (bool_map(format.is_bitmask()),)
         print "   %s,\t/* is_mixed */" % (bool_map(format.is_mixed()),)
         print "   {"
         for i in range(4):
diff --git a/src/gallium/auxiliary/util/u_format_tests.c b/src/gallium/auxiliary/util/u_format_tests.c
new file mode 100644
index 00000000000..182a4740448
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_format_tests.c
@@ -0,0 +1,544 @@
+/**************************************************************************
+ *
+ * Copyright 2009-2010 VMware, 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 VMWARE 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 "u_memory.h"
+#include "u_format_tests.h"
+
+
+/*
+ * Helper macros to create the packed bytes for longer words.
+ */
+
+#define PACKED_1x8(x)          {x, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
+#define PACKED_2x8(x, y)       {x, y, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
+#define PACKED_3x8(x, y, z)    {x, y, z, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
+#define PACKED_4x8(x, y, z, w) {x, y, z, w, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}
+
+#define PACKED_1x16(x)          {(x) & 0xff, (x) >> 8,          0,        0,          0,        0,          0,        0, 0, 0, 0, 0, 0, 0, 0, 0}
+#define PACKED_2x16(x, y)       {(x) & 0xff, (x) >> 8, (y) & 0xff, (y) >> 8,          0,        0,          0,        0, 0, 0, 0, 0, 0, 0, 0, 0}
+#define PACKED_3x16(x, y, z)    {(x) & 0xff, (x) >> 8, (y) & 0xff, (y) >> 8, (z) & 0xff, (z) >> 8,          0,        0, 0, 0, 0, 0, 0, 0, 0, 0}
+#define PACKED_4x16(x, y, z, w) {(x) & 0xff, (x) >> 8, (y) & 0xff, (y) >> 8, (z) & 0xff, (z) >> 8, (w) & 0xff, (w) >> 8, 0, 0, 0, 0, 0, 0, 0, 0}
+
+#define PACKED_1x32(x)          {(x) & 0xff, ((x) >> 8) & 0xff, ((x) >> 16) & 0xff, (x) >> 24,          0,                 0,                  0,         0,          0,                 0,                  0,         0,          0,                 0,                  0,         0}
+#define PACKED_2x32(x, y)       {(x) & 0xff, ((x) >> 8) & 0xff, ((x) >> 16) & 0xff, (x) >> 24, (y) & 0xff, ((y) >> 8) & 0xff, ((y) >> 16) & 0xff, (y) >> 24,          0,                 0,                  0,         0,          0,                 0,                  0,         0}
+#define PACKED_3x32(x, y, z)    {(x) & 0xff, ((x) >> 8) & 0xff, ((x) >> 16) & 0xff, (x) >> 24, (y) & 0xff, ((y) >> 8) & 0xff, ((y) >> 16) & 0xff, (y) >> 24, (z) & 0xff, ((z) >> 8) & 0xff, ((z) >> 16) & 0xff, (z) >> 24,          0,                 0,                  0,         0}
+#define PACKED_4x32(x, y, z, w) {(x) & 0xff, ((x) >> 8) & 0xff, ((x) >> 16) & 0xff, (x) >> 24, (y) & 0xff, ((y) >> 8) & 0xff, ((y) >> 16) & 0xff, (y) >> 24, (z) & 0xff, ((z) >> 8) & 0xff, ((z) >> 16) & 0xff, (z) >> 24, (w) & 0xff, ((w) >> 8) & 0xff, ((w) >> 16) & 0xff, (w) >> 24}
+
+
+/**
+ * Test cases.
+ *
+ * These were manually entered. We could generate these
+ *
+ * To keep this to a we cover only the corner cases, which should produce
+ * good enough coverage since that pixel format transformations are afine for
+ * non SRGB formats.
+ */
+const struct util_format_test_case
+util_format_test_cases[] =
+{
+
+   /*
+    * 32-bit rendertarget formats
+    */
+
+   {PIPE_FORMAT_B8G8R8A8_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x00000000), {0.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_B8G8R8A8_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x000000ff), {0.0, 0.0, 1.0, 0.0}},
+   {PIPE_FORMAT_B8G8R8A8_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x0000ff00), {0.0, 1.0, 0.0, 0.0}},
+   {PIPE_FORMAT_B8G8R8A8_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x00ff0000), {1.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_B8G8R8A8_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0xff000000), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_B8G8R8A8_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0xffffffff), {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_B8G8R8X8_UNORM, PACKED_1x32(0x00ffffff), PACKED_1x32(0x00000000), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_B8G8R8X8_UNORM, PACKED_1x32(0x00ffffff), PACKED_1x32(0x000000ff), {0.0, 0.0, 1.0, 1.0}},
+   {PIPE_FORMAT_B8G8R8X8_UNORM, PACKED_1x32(0x00ffffff), PACKED_1x32(0x0000ff00), {0.0, 1.0, 0.0, 1.0}},
+   {PIPE_FORMAT_B8G8R8X8_UNORM, PACKED_1x32(0x00ffffff), PACKED_1x32(0x00ff0000), {1.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_B8G8R8X8_UNORM, PACKED_1x32(0x00ffffff), PACKED_1x32(0xff000000), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_B8G8R8X8_UNORM, PACKED_1x32(0x00ffffff), PACKED_1x32(0xffffffff), {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_A8R8G8B8_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x00000000), {0.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_A8R8G8B8_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x000000ff), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_A8R8G8B8_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x0000ff00), {1.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_A8R8G8B8_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x00ff0000), {0.0, 1.0, 0.0, 0.0}},
+   {PIPE_FORMAT_A8R8G8B8_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0xff000000), {0.0, 0.0, 1.0, 0.0}},
+   {PIPE_FORMAT_A8R8G8B8_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0xffffffff), {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_X8R8G8B8_UNORM, PACKED_1x32(0xffffff00), PACKED_1x32(0x00000000), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_X8R8G8B8_UNORM, PACKED_1x32(0xffffff00), PACKED_1x32(0x000000ff), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_X8R8G8B8_UNORM, PACKED_1x32(0xffffff00), PACKED_1x32(0x0000ff00), {1.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_X8R8G8B8_UNORM, PACKED_1x32(0xffffff00), PACKED_1x32(0x00ff0000), {0.0, 1.0, 0.0, 1.0}},
+   {PIPE_FORMAT_X8R8G8B8_UNORM, PACKED_1x32(0xffffff00), PACKED_1x32(0xff000000), {0.0, 0.0, 1.0, 1.0}},
+   {PIPE_FORMAT_X8R8G8B8_UNORM, PACKED_1x32(0xffffff00), PACKED_1x32(0xffffffff), {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_A8B8G8R8_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x00000000), {0.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_A8B8G8R8_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x000000ff), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_A8B8G8R8_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x0000ff00), {0.0, 0.0, 1.0, 0.0}},
+   {PIPE_FORMAT_A8B8G8R8_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x00ff0000), {0.0, 1.0, 0.0, 0.0}},
+   {PIPE_FORMAT_A8B8G8R8_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0xff000000), {1.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_A8B8G8R8_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0xffffffff), {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_X8B8G8R8_UNORM, PACKED_1x32(0xffffff00), PACKED_1x32(0x00000000), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_X8B8G8R8_UNORM, PACKED_1x32(0xffffff00), PACKED_1x32(0x000000ff), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_X8B8G8R8_UNORM, PACKED_1x32(0xffffff00), PACKED_1x32(0x0000ff00), {0.0, 0.0, 1.0, 1.0}},
+   {PIPE_FORMAT_X8B8G8R8_UNORM, PACKED_1x32(0xffffff00), PACKED_1x32(0x00ff0000), {0.0, 1.0, 0.0, 1.0}},
+   {PIPE_FORMAT_X8B8G8R8_UNORM, PACKED_1x32(0xffffff00), PACKED_1x32(0xff000000), {1.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_X8B8G8R8_UNORM, PACKED_1x32(0xffffff00), PACKED_1x32(0xffffffff), {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_R10G10B10A2_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x00000000), {0.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_R10G10B10A2_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x000003ff), {1.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_R10G10B10A2_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x000ffc00), {0.0, 1.0, 0.0, 0.0}},
+   {PIPE_FORMAT_R10G10B10A2_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x3ff00000), {0.0, 0.0, 1.0, 0.0}},
+   {PIPE_FORMAT_R10G10B10A2_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0xc0000000), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R10G10B10A2_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0xffffffff), {1.0, 1.0, 1.0, 1.0}},
+
+   /*
+    * 16-bit rendertarget formats
+    */
+
+   {PIPE_FORMAT_B5G5R5A1_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0x0000), {0.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_B5G5R5A1_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0x001f), {0.0, 0.0, 1.0, 0.0}},
+   {PIPE_FORMAT_B5G5R5A1_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0x03e0), {0.0, 1.0, 0.0, 0.0}},
+   {PIPE_FORMAT_B5G5R5A1_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0x7c00), {1.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_B5G5R5A1_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0x8000), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_B5G5R5A1_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0xffff), {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_B4G4R4A4_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0x0000), {0.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_B4G4R4A4_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0x000f), {0.0, 0.0, 1.0, 0.0}},
+   {PIPE_FORMAT_B4G4R4A4_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0x00f0), {0.0, 1.0, 0.0, 0.0}},
+   {PIPE_FORMAT_B4G4R4A4_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0x0f00), {1.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_B4G4R4A4_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0xf000), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_B4G4R4A4_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0xffff), {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_B5G6R5_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0x0000), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_B5G6R5_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0x001f), {0.0, 0.0, 1.0, 1.0}},
+   {PIPE_FORMAT_B5G6R5_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0x07e0), {0.0, 1.0, 0.0, 1.0}},
+   {PIPE_FORMAT_B5G6R5_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0xf800), {1.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_B5G6R5_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0xffff), {1.0, 1.0, 1.0, 1.0}},
+
+   /*
+    * Luminance/intensity/alpha formats
+    */
+
+   {PIPE_FORMAT_L8_UNORM, PACKED_1x8(0xff), PACKED_1x8(0x00), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_L8_UNORM, PACKED_1x8(0xff), PACKED_1x8(0xff), {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_A8_UNORM, PACKED_1x8(0xff), PACKED_1x8(0x00), {0.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_A8_UNORM, PACKED_1x8(0xff), PACKED_1x8(0xff), {0.0, 0.0, 0.0, 1.0}},
+
+   {PIPE_FORMAT_I8_UNORM, PACKED_1x8(0xff), PACKED_1x8(0x00), {0.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_I8_UNORM, PACKED_1x8(0xff), PACKED_1x8(0xff), {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_L8A8_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0x0000), {0.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_L8A8_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0x00ff), {1.0, 1.0, 1.0, 0.0}},
+   {PIPE_FORMAT_L8A8_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0xff00), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_L8A8_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0xffff), {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_L16_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0x0000), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_L16_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0xffff), {1.0, 1.0, 1.0, 1.0}},
+
+   /*
+    * TODO: SRGB formats
+    */
+
+   /*
+    * Mixed-signed formats
+    */
+
+   {PIPE_FORMAT_R8SG8SB8UX8U_NORM, PACKED_4x8(0xff, 0xff, 0xff, 0x00), PACKED_4x8(0x00, 0x00, 0x00, 0x00), { 0.0,  0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R8SG8SB8UX8U_NORM, PACKED_4x8(0xff, 0xff, 0xff, 0x00), PACKED_4x8(0x7f, 0x00, 0x00, 0x00), { 1.0,  0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R8SG8SB8UX8U_NORM, PACKED_4x8(0xff, 0xff, 0xff, 0x00), PACKED_4x8(0x81, 0x00, 0x00, 0x00), {-1.0,  0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R8SG8SB8UX8U_NORM, PACKED_4x8(0xff, 0xff, 0xff, 0x00), PACKED_4x8(0x00, 0x7f, 0x00, 0x00), { 0.0,  1.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R8SG8SB8UX8U_NORM, PACKED_4x8(0xff, 0xff, 0xff, 0x00), PACKED_4x8(0x00, 0x81, 0x00, 0x00), { 0.0, -1.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R8SG8SB8UX8U_NORM, PACKED_4x8(0xff, 0xff, 0xff, 0x00), PACKED_4x8(0x00, 0x00, 0xff, 0x00), { 0.0,  0.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_R5SG5SB6U_NORM, PACKED_1x16(0xffff), PACKED_1x16(0x0000), { 0.0,  0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R5SG5SB6U_NORM, PACKED_1x16(0xffff), PACKED_1x16(0x000f), { 1.0,  0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R5SG5SB6U_NORM, PACKED_1x16(0xffff), PACKED_1x16(0x0011), {-1.0,  0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R5SG5SB6U_NORM, PACKED_1x16(0xffff), PACKED_1x16(0x01e0), { 0.0,  1.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R5SG5SB6U_NORM, PACKED_1x16(0xffff), PACKED_1x16(0x0220), { 0.0, -1.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R5SG5SB6U_NORM, PACKED_1x16(0xffff), PACKED_1x16(0xfc00), { 0.0,  0.0, 1.0, 1.0}},
+
+   /*
+    * TODO: Depth-stencil formats
+    */
+
+   /*
+    * TODO: YUV formats
+    */
+
+   /*
+    * TODO: Compressed formats
+    */
+
+   /*
+    * Standard 8-bit integer formats
+    */
+
+   {PIPE_FORMAT_R8_UNORM, PACKED_1x8(0xff), PACKED_1x8(0x00), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R8_UNORM, PACKED_1x8(0xff), PACKED_1x8(0xff), {1.0, 0.0, 0.0, 1.0}},
+
+   {PIPE_FORMAT_R8G8_UNORM, PACKED_2x8(0xff, 0xff), PACKED_2x8(0x00, 0x00), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R8G8_UNORM, PACKED_2x8(0xff, 0xff), PACKED_2x8(0xff, 0x00), {1.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R8G8_UNORM, PACKED_2x8(0xff, 0xff), PACKED_2x8(0x00, 0xff), {0.0, 1.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R8G8_UNORM, PACKED_2x8(0xff, 0xff), PACKED_2x8(0xff, 0xff), {1.0, 1.0, 0.0, 1.0}},
+
+   {PIPE_FORMAT_R8G8B8_UNORM, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x00, 0x00, 0x00), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R8G8B8_UNORM, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0xff, 0x00, 0x00), {1.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R8G8B8_UNORM, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x00, 0xff, 0x00), {0.0, 1.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R8G8B8_UNORM, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x00, 0x00, 0xff), {0.0, 0.0, 1.0, 1.0}},
+   {PIPE_FORMAT_R8G8B8_UNORM, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0xff, 0xff, 0xff), {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_R8G8B8A8_UNORM, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x00, 0x00, 0x00), {0.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_R8G8B8A8_UNORM, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0xff, 0x00, 0x00, 0x00), {1.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_R8G8B8A8_UNORM, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0xff, 0x00, 0x00), {0.0, 1.0, 0.0, 0.0}},
+   {PIPE_FORMAT_R8G8B8A8_UNORM, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x00, 0xff, 0x00), {0.0, 0.0, 1.0, 0.0}},
+   {PIPE_FORMAT_R8G8B8A8_UNORM, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x00, 0x00, 0xff), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R8G8B8A8_UNORM, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0xff, 0xff, 0xff, 0xff), {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_R8_USCALED, PACKED_1x8(0xff), PACKED_1x8(0x00), {  0.0,   0.0,   0.0, 1.0}},
+   {PIPE_FORMAT_R8_USCALED, PACKED_1x8(0xff), PACKED_1x8(0xff), {255.0,   0.0,   0.0, 1.0}},
+
+   {PIPE_FORMAT_R8G8_USCALED, PACKED_2x8(0xff, 0xff), PACKED_2x8(0x00, 0x00), {  0.0,   0.0,   0.0, 1.0}},
+   {PIPE_FORMAT_R8G8_USCALED, PACKED_2x8(0xff, 0xff), PACKED_2x8(0xff, 0x00), {255.0,   0.0,   0.0, 1.0}},
+   {PIPE_FORMAT_R8G8_USCALED, PACKED_2x8(0xff, 0xff), PACKED_2x8(0x00, 0xff), {  0.0, 255.0,   0.0, 1.0}},
+   {PIPE_FORMAT_R8G8_USCALED, PACKED_2x8(0xff, 0xff), PACKED_2x8(0xff, 0xff), {255.0, 255.0,   0.0, 1.0}},
+
+   {PIPE_FORMAT_R8G8B8_USCALED, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x00, 0x00, 0x00), {  0.0,   0.0,   0.0, 1.0}},
+   {PIPE_FORMAT_R8G8B8_USCALED, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0xff, 0x00, 0x00), {255.0,   0.0,   0.0, 1.0}},
+   {PIPE_FORMAT_R8G8B8_USCALED, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x00, 0xff, 0x00), {  0.0, 255.0,   0.0, 1.0}},
+   {PIPE_FORMAT_R8G8B8_USCALED, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x00, 0x00, 0xff), {  0.0,   0.0, 255.0, 1.0}},
+   {PIPE_FORMAT_R8G8B8_USCALED, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0xff, 0xff, 0xff), {255.0, 255.0, 255.0, 1.0}},
+
+   {PIPE_FORMAT_R8G8B8A8_USCALED, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x00, 0x00, 0x00), {  0.0,   0.0,   0.0,   0.0}},
+   {PIPE_FORMAT_R8G8B8A8_USCALED, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0xff, 0x00, 0x00, 0x00), {255.0,   0.0,   0.0,   0.0}},
+   {PIPE_FORMAT_R8G8B8A8_USCALED, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0xff, 0x00, 0x00), {  0.0, 255.0,   0.0,   0.0}},
+   {PIPE_FORMAT_R8G8B8A8_USCALED, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x00, 0xff, 0x00), {  0.0,   0.0, 255.0,   0.0}},
+   {PIPE_FORMAT_R8G8B8A8_USCALED, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x00, 0x00, 0xff), {  0.0,   0.0,   0.0, 255.0}},
+   {PIPE_FORMAT_R8G8B8A8_USCALED, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0xff, 0xff, 0xff, 0xff), {255.0, 255.0, 255.0, 255.0}},
+
+   {PIPE_FORMAT_R8_SNORM, PACKED_1x8(0xff), PACKED_1x8(0x00), { 0.0,  0.0,  0.0,  1.0}},
+   {PIPE_FORMAT_R8_SNORM, PACKED_1x8(0xff), PACKED_1x8(0x7f), { 1.0,  0.0,  0.0,  1.0}},
+   {PIPE_FORMAT_R8_SNORM, PACKED_1x8(0xff), PACKED_1x8(0x81), {-1.0,  0.0,  0.0,  1.0}},
+
+   {PIPE_FORMAT_R8G8_SNORM, PACKED_2x8(0xff, 0xff), PACKED_2x8(0x00, 0x00), { 0.0,  0.0,  0.0,  1.0}},
+   {PIPE_FORMAT_R8G8_SNORM, PACKED_2x8(0xff, 0xff), PACKED_2x8(0x7f, 0x00), { 1.0,  0.0,  0.0,  1.0}},
+   {PIPE_FORMAT_R8G8_SNORM, PACKED_2x8(0xff, 0xff), PACKED_2x8(0x81, 0x00), {-1.0,  0.0,  0.0,  1.0}},
+   {PIPE_FORMAT_R8G8_SNORM, PACKED_2x8(0xff, 0xff), PACKED_2x8(0x00, 0x7f), { 0.0,  1.0,  0.0,  1.0}},
+   {PIPE_FORMAT_R8G8_SNORM, PACKED_2x8(0xff, 0xff), PACKED_2x8(0x00, 0x81), { 0.0, -1.0,  0.0,  1.0}},
+
+   {PIPE_FORMAT_R8G8B8_SNORM, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x00, 0x00, 0x00), { 0.0,  0.0,  0.0,  1.0}},
+   {PIPE_FORMAT_R8G8B8_SNORM, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x7f, 0x00, 0x00), { 1.0,  0.0,  0.0,  1.0}},
+   {PIPE_FORMAT_R8G8B8_SNORM, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x81, 0x00, 0x00), {-1.0,  0.0,  0.0,  1.0}},
+   {PIPE_FORMAT_R8G8B8_SNORM, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x00, 0x7f, 0x00), { 0.0,  1.0,  0.0,  1.0}},
+   {PIPE_FORMAT_R8G8B8_SNORM, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x00, 0x81, 0x00), { 0.0, -1.0,  0.0,  1.0}},
+   {PIPE_FORMAT_R8G8B8_SNORM, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x00, 0x00, 0x7f), { 0.0,  0.0,  1.0,  1.0}},
+   {PIPE_FORMAT_R8G8B8_SNORM, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x00, 0x00, 0x81), { 0.0,  0.0, -1.0,  1.0}},
+
+   {PIPE_FORMAT_R8G8B8A8_SNORM, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x00, 0x00, 0x00), { 0.0,  0.0,  0.0,  0.0}},
+   {PIPE_FORMAT_R8G8B8A8_SNORM, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x7f, 0x00, 0x00, 0x00), { 1.0,  0.0,  0.0,  0.0}},
+   {PIPE_FORMAT_R8G8B8A8_SNORM, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x81, 0x00, 0x00, 0x00), {-1.0,  0.0,  0.0,  0.0}},
+   {PIPE_FORMAT_R8G8B8A8_SNORM, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x7f, 0x00, 0x00), { 0.0,  1.0,  0.0,  0.0}},
+   {PIPE_FORMAT_R8G8B8A8_SNORM, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x81, 0x00, 0x00), { 0.0, -1.0,  0.0,  0.0}},
+   {PIPE_FORMAT_R8G8B8A8_SNORM, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x00, 0x7f, 0x00), { 0.0,  0.0,  1.0,  0.0}},
+   {PIPE_FORMAT_R8G8B8A8_SNORM, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x00, 0x81, 0x00), { 0.0,  0.0, -1.0,  0.0}},
+   {PIPE_FORMAT_R8G8B8A8_SNORM, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x00, 0x00, 0x7f), { 0.0,  0.0,  0.0,  1.0}},
+   {PIPE_FORMAT_R8G8B8A8_SNORM, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x00, 0x00, 0x81), { 0.0,  0.0,  0.0, -1.0}},
+
+   {PIPE_FORMAT_R8_SSCALED, PACKED_1x8(0xff), PACKED_1x8(0x00), {   0.0,    0.0,    0.0, 1.0}},
+   {PIPE_FORMAT_R8_SSCALED, PACKED_1x8(0xff), PACKED_1x8(0x7f), { 127.0,    0.0,    0.0, 1.0}},
+   {PIPE_FORMAT_R8_SSCALED, PACKED_1x8(0xff), PACKED_1x8(0x80), {-128.0,    0.0,    0.0, 1.0}},
+
+   {PIPE_FORMAT_R8G8_SSCALED, PACKED_2x8(0xff, 0xff), PACKED_2x8(0x00, 0x00), {   0.0,    0.0,    0.0, 1.0}},
+   {PIPE_FORMAT_R8G8_SSCALED, PACKED_2x8(0xff, 0xff), PACKED_2x8(0x7f, 0x00), { 127.0,    0.0,    0.0, 1.0}},
+   {PIPE_FORMAT_R8G8_SSCALED, PACKED_2x8(0xff, 0xff), PACKED_2x8(0x80, 0x00), {-128.0,    0.0,    0.0, 1.0}},
+   {PIPE_FORMAT_R8G8_SSCALED, PACKED_2x8(0xff, 0xff), PACKED_2x8(0x00, 0x7f), {   0.0,  127.0,    0.0, 1.0}},
+   {PIPE_FORMAT_R8G8_SSCALED, PACKED_2x8(0xff, 0xff), PACKED_2x8(0x00, 0x80), {   0.0, -128.0,    0.0, 1.0}},
+
+   {PIPE_FORMAT_R8G8B8_SSCALED, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x00, 0x00, 0x00), {   0.0,    0.0,    0.0, 1.0}},
+   {PIPE_FORMAT_R8G8B8_SSCALED, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x7f, 0x00, 0x00), { 127.0,    0.0,    0.0, 1.0}},
+   {PIPE_FORMAT_R8G8B8_SSCALED, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x80, 0x00, 0x00), {-128.0,    0.0,    0.0, 1.0}},
+   {PIPE_FORMAT_R8G8B8_SSCALED, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x00, 0x7f, 0x00), {   0.0,  127.0,    0.0, 1.0}},
+   {PIPE_FORMAT_R8G8B8_SSCALED, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x00, 0x80, 0x00), {   0.0, -128.0,    0.0, 1.0}},
+   {PIPE_FORMAT_R8G8B8_SSCALED, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x00, 0x00, 0x7f), {   0.0,    0.0,  127.0, 1.0}},
+   {PIPE_FORMAT_R8G8B8_SSCALED, PACKED_3x8(0xff, 0xff, 0xff), PACKED_3x8(0x00, 0x00, 0x80), {   0.0,    0.0, -128.0, 1.0}},
+
+   {PIPE_FORMAT_R8G8B8A8_SSCALED, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x00, 0x00, 0x00), {   0.0,    0.0,    0.0,    0.0}},
+   {PIPE_FORMAT_R8G8B8A8_SSCALED, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x7f, 0x00, 0x00, 0x00), { 127.0,    0.0,    0.0,    0.0}},
+   {PIPE_FORMAT_R8G8B8A8_SSCALED, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x80, 0x00, 0x00, 0x00), {-128.0,    0.0,    0.0,    0.0}},
+   {PIPE_FORMAT_R8G8B8A8_SSCALED, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x7f, 0x00, 0x00), {   0.0,  127.0,    0.0,    0.0}},
+   {PIPE_FORMAT_R8G8B8A8_SSCALED, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x80, 0x00, 0x00), {   0.0, -128.0,    0.0,    0.0}},
+   {PIPE_FORMAT_R8G8B8A8_SSCALED, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x00, 0x7f, 0x00), {   0.0,    0.0,  127.0,    0.0}},
+   {PIPE_FORMAT_R8G8B8A8_SSCALED, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x00, 0x80, 0x00), {   0.0,    0.0, -128.0,    0.0}},
+   {PIPE_FORMAT_R8G8B8A8_SSCALED, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x00, 0x00, 0x7f), {   0.0,    0.0,    0.0,  127.0}},
+   {PIPE_FORMAT_R8G8B8A8_SSCALED, PACKED_4x8(0xff, 0xff, 0xff, 0xff), PACKED_4x8(0x00, 0x00, 0x00, 0x80), {   0.0,    0.0,    0.0, -128.0}},
+
+   /*
+    * Standard 16-bit integer formats
+    */
+
+   {PIPE_FORMAT_R16_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0x0000), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R16_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0xffff), {1.0, 0.0, 0.0, 1.0}},
+
+   {PIPE_FORMAT_R16G16_UNORM, PACKED_2x16(0xffff, 0xffff), PACKED_2x16(0x0000, 0x0000), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R16G16_UNORM, PACKED_2x16(0xffff, 0xffff), PACKED_2x16(0xffff, 0x0000), {1.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R16G16_UNORM, PACKED_2x16(0xffff, 0xffff), PACKED_2x16(0x0000, 0xffff), {0.0, 1.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R16G16_UNORM, PACKED_2x16(0xffff, 0xffff), PACKED_2x16(0xffff, 0xffff), {1.0, 1.0, 0.0, 1.0}},
+
+   {PIPE_FORMAT_R16G16B16_UNORM, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x0000, 0x0000, 0x0000), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R16G16B16_UNORM, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0xffff, 0x0000, 0x0000), {1.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R16G16B16_UNORM, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x0000, 0xffff, 0x0000), {0.0, 1.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R16G16B16_UNORM, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x0000, 0x0000, 0xffff), {0.0, 0.0, 1.0, 1.0}},
+   {PIPE_FORMAT_R16G16B16_UNORM, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0xffff, 0xffff, 0xffff), {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_R16G16B16A16_UNORM, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x0000, 0x0000, 0x0000), {0.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_R16G16B16A16_UNORM, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0xffff, 0x0000, 0x0000, 0x0000), {1.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_R16G16B16A16_UNORM, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0xffff, 0x0000, 0x0000), {0.0, 1.0, 0.0, 0.0}},
+   {PIPE_FORMAT_R16G16B16A16_UNORM, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x0000, 0xffff, 0x0000), {0.0, 0.0, 1.0, 0.0}},
+   {PIPE_FORMAT_R16G16B16A16_UNORM, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x0000, 0x0000, 0xffff), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R16G16B16A16_UNORM, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_R16_USCALED, PACKED_1x16(0xffff), PACKED_1x16(0x0000), {    0.0,     0.0,     0.0,   1.0}},
+   {PIPE_FORMAT_R16_USCALED, PACKED_1x16(0xffff), PACKED_1x16(0xffff), {65535.0,     0.0,     0.0,   1.0}},
+
+   {PIPE_FORMAT_R16G16_USCALED, PACKED_2x16(0xffff, 0xffff), PACKED_2x16(0x0000, 0x0000), {    0.0,     0.0,     0.0,   1.0}},
+   {PIPE_FORMAT_R16G16_USCALED, PACKED_2x16(0xffff, 0xffff), PACKED_2x16(0xffff, 0x0000), {65535.0,     0.0,     0.0,   1.0}},
+   {PIPE_FORMAT_R16G16_USCALED, PACKED_2x16(0xffff, 0xffff), PACKED_2x16(0x0000, 0xffff), {    0.0, 65535.0,     0.0,   1.0}},
+   {PIPE_FORMAT_R16G16_USCALED, PACKED_2x16(0xffff, 0xffff), PACKED_2x16(0xffff, 0xffff), {65535.0, 65535.0,     0.0,   1.0}},
+
+   {PIPE_FORMAT_R16G16B16_USCALED, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x0000, 0x0000, 0x0000), {    0.0,     0.0,     0.0,   1.0}},
+   {PIPE_FORMAT_R16G16B16_USCALED, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0xffff, 0x0000, 0x0000), {65535.0,     0.0,     0.0,   1.0}},
+   {PIPE_FORMAT_R16G16B16_USCALED, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x0000, 0xffff, 0x0000), {    0.0, 65535.0,     0.0,   1.0}},
+   {PIPE_FORMAT_R16G16B16_USCALED, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x0000, 0x0000, 0xffff), {    0.0,     0.0, 65535.0,   1.0}},
+   {PIPE_FORMAT_R16G16B16_USCALED, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0xffff, 0xffff, 0xffff), {65535.0, 65535.0, 65535.0,   1.0}},
+
+   {PIPE_FORMAT_R16G16B16A16_USCALED, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x0000, 0x0000, 0x0000), {    0.0,     0.0,     0.0,     0.0}},
+   {PIPE_FORMAT_R16G16B16A16_USCALED, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0xffff, 0x0000, 0x0000, 0x0000), {65535.0,     0.0,     0.0,     0.0}},
+   {PIPE_FORMAT_R16G16B16A16_USCALED, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0xffff, 0x0000, 0x0000), {    0.0, 65535.0,     0.0,     0.0}},
+   {PIPE_FORMAT_R16G16B16A16_USCALED, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x0000, 0xffff, 0x0000), {    0.0,     0.0, 65535.0,     0.0}},
+   {PIPE_FORMAT_R16G16B16A16_USCALED, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x0000, 0x0000, 0xffff), {    0.0,     0.0,     0.0, 65535.0}},
+   {PIPE_FORMAT_R16G16B16A16_USCALED, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), {65535.0, 65535.0, 65535.0, 65535.0}},
+
+   {PIPE_FORMAT_R16_SNORM, PACKED_1x16(0xffff), PACKED_1x16(0x0000), {   0.0,    0.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R16_SNORM, PACKED_1x16(0xffff), PACKED_1x16(0x7fff), {   1.0,    0.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R16_SNORM, PACKED_1x16(0xffff), PACKED_1x16(0x8001), {  -1.0,    0.0,    0.0,    1.0}},
+
+   {PIPE_FORMAT_R16G16_SNORM, PACKED_2x16(0xffff, 0xffff), PACKED_2x16(0x0000, 0x0000), {   0.0,    0.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R16G16_SNORM, PACKED_2x16(0xffff, 0xffff), PACKED_2x16(0x7fff, 0x0000), {   1.0,    0.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R16G16_SNORM, PACKED_2x16(0xffff, 0xffff), PACKED_2x16(0x8001, 0x0000), {  -1.0,    0.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R16G16_SNORM, PACKED_2x16(0xffff, 0xffff), PACKED_2x16(0x0000, 0x7fff), {   0.0,    1.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R16G16_SNORM, PACKED_2x16(0xffff, 0xffff), PACKED_2x16(0x0000, 0x8001), {   0.0,   -1.0,    0.0,    1.0}},
+
+   {PIPE_FORMAT_R16G16B16_SNORM, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x0000, 0x0000, 0x0000), {   0.0,    0.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R16G16B16_SNORM, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x7fff, 0x0000, 0x0000), {   1.0,    0.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R16G16B16_SNORM, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x8001, 0x0000, 0x0000), {  -1.0,    0.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R16G16B16_SNORM, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x0000, 0x7fff, 0x0000), {   0.0,    1.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R16G16B16_SNORM, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x0000, 0x8001, 0x0000), {   0.0,   -1.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R16G16B16_SNORM, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x0000, 0x0000, 0x7fff), {   0.0,    0.0,    1.0,    1.0}},
+   {PIPE_FORMAT_R16G16B16_SNORM, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x0000, 0x0000, 0x8001), {   0.0,    0.0,   -1.0,    1.0}},
+
+   {PIPE_FORMAT_R16G16B16A16_SNORM, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x0000, 0x0000, 0x0000), {   0.0,    0.0,    0.0,    0.0}},
+   {PIPE_FORMAT_R16G16B16A16_SNORM, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x7fff, 0x0000, 0x0000, 0x0000), {   1.0,    0.0,    0.0,    0.0}},
+   {PIPE_FORMAT_R16G16B16A16_SNORM, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x8001, 0x0000, 0x0000, 0x0000), {  -1.0,    0.0,    0.0,    0.0}},
+   {PIPE_FORMAT_R16G16B16A16_SNORM, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x7fff, 0x0000, 0x0000), {   0.0,    1.0,    0.0,    0.0}},
+   {PIPE_FORMAT_R16G16B16A16_SNORM, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x8001, 0x0000, 0x0000), {   0.0,   -1.0,    0.0,    0.0}},
+   {PIPE_FORMAT_R16G16B16A16_SNORM, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x0000, 0x7fff, 0x0000), {   0.0,    0.0,    1.0,    0.0}},
+   {PIPE_FORMAT_R16G16B16A16_SNORM, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x0000, 0x8001, 0x0000), {   0.0,    0.0,   -1.0,    0.0}},
+   {PIPE_FORMAT_R16G16B16A16_SNORM, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x0000, 0x0000, 0x7fff), {   0.0,    0.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R16G16B16A16_SNORM, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x0000, 0x0000, 0x8001), {   0.0,    0.0,    0.0,   -1.0}},
+
+   {PIPE_FORMAT_R16_SSCALED, PACKED_1x16(0xffff), PACKED_1x16(0x0000), {     0.0,      0.0,      0.0,   1.0}},
+   {PIPE_FORMAT_R16_SSCALED, PACKED_1x16(0xffff), PACKED_1x16(0x7fff), { 32767.0,      0.0,      0.0,   1.0}},
+   {PIPE_FORMAT_R16_SSCALED, PACKED_1x16(0xffff), PACKED_1x16(0x8000), {-32768.0,      0.0,      0.0,   1.0}},
+
+   {PIPE_FORMAT_R16G16_SSCALED, PACKED_2x16(0xffff, 0xffff), PACKED_2x16(0x0000, 0x0000), {     0.0,      0.0,      0.0,   1.0}},
+   {PIPE_FORMAT_R16G16_SSCALED, PACKED_2x16(0xffff, 0xffff), PACKED_2x16(0x7fff, 0x0000), { 32767.0,      0.0,      0.0,   1.0}},
+   {PIPE_FORMAT_R16G16_SSCALED, PACKED_2x16(0xffff, 0xffff), PACKED_2x16(0x8000, 0x0000), {-32768.0,      0.0,      0.0,   1.0}},
+   {PIPE_FORMAT_R16G16_SSCALED, PACKED_2x16(0xffff, 0xffff), PACKED_2x16(0x0000, 0x7fff), {     0.0,  32767.0,      0.0,   1.0}},
+   {PIPE_FORMAT_R16G16_SSCALED, PACKED_2x16(0xffff, 0xffff), PACKED_2x16(0x0000, 0x8000), {     0.0, -32768.0,      0.0,   1.0}},
+
+   {PIPE_FORMAT_R16G16B16_SSCALED, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x0000, 0x0000, 0x0000), {     0.0,      0.0,      0.0,   1.0}},
+   {PIPE_FORMAT_R16G16B16_SSCALED, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x7fff, 0x0000, 0x0000), { 32767.0,      0.0,      0.0,   1.0}},
+   {PIPE_FORMAT_R16G16B16_SSCALED, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x8000, 0x0000, 0x0000), {-32768.0,      0.0,      0.0,   1.0}},
+   {PIPE_FORMAT_R16G16B16_SSCALED, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x0000, 0x7fff, 0x0000), {     0.0,  32767.0,      0.0,   1.0}},
+   {PIPE_FORMAT_R16G16B16_SSCALED, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x0000, 0x8000, 0x0000), {     0.0, -32768.0,      0.0,   1.0}},
+   {PIPE_FORMAT_R16G16B16_SSCALED, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x0000, 0x0000, 0x7fff), {     0.0,      0.0,  32767.0,   1.0}},
+   {PIPE_FORMAT_R16G16B16_SSCALED, PACKED_3x16(0xffff, 0xffff, 0xffff), PACKED_3x16(0x0000, 0x0000, 0x8000), {     0.0,      0.0, -32768.0,   1.0}},
+
+   {PIPE_FORMAT_R16G16B16A16_SSCALED, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x0000, 0x0000, 0x0000), {     0.0,      0.0,      0.0,      0.0}},
+   {PIPE_FORMAT_R16G16B16A16_SSCALED, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x7fff, 0x0000, 0x0000, 0x0000), { 32767.0,      0.0,      0.0,      0.0}},
+   {PIPE_FORMAT_R16G16B16A16_SSCALED, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x8000, 0x0000, 0x0000, 0x0000), {-32768.0,      0.0,      0.0,      0.0}},
+   {PIPE_FORMAT_R16G16B16A16_SSCALED, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x7fff, 0x0000, 0x0000), {     0.0,  32767.0,      0.0,      0.0}},
+   {PIPE_FORMAT_R16G16B16A16_SSCALED, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x8000, 0x0000, 0x0000), {     0.0, -32768.0,      0.0,      0.0}},
+   {PIPE_FORMAT_R16G16B16A16_SSCALED, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x0000, 0x7fff, 0x0000), {     0.0,      0.0,  32767.0,      0.0}},
+   {PIPE_FORMAT_R16G16B16A16_SSCALED, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x0000, 0x8000, 0x0000), {     0.0,      0.0, -32768.0,      0.0}},
+   {PIPE_FORMAT_R16G16B16A16_SSCALED, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x0000, 0x0000, 0x7fff), {     0.0,      0.0,      0.0,  32767.0}},
+   {PIPE_FORMAT_R16G16B16A16_SSCALED, PACKED_4x16(0xffff, 0xffff, 0xffff, 0xffff), PACKED_4x16(0x0000, 0x0000, 0x0000, 0x8000), {     0.0,      0.0,      0.0, -32768.0}},
+
+   /*
+    * Standard 32-bit integer formats
+    *
+    * NOTE: We can't accurately represent integers larger than +/-0x1000000
+    * with single precision floats, so that's as far as we test.
+    */
+
+   {PIPE_FORMAT_R32_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x00000000), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R32_UNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0xffffffff), {1.0, 0.0, 0.0, 1.0}},
+
+   {PIPE_FORMAT_R32G32_UNORM, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x00000000, 0x00000000), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R32G32_UNORM, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0xffffffff, 0x00000000), {1.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R32G32_UNORM, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x00000000, 0xffffffff), {0.0, 1.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R32G32_UNORM, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0xffffffff, 0xffffffff), {1.0, 1.0, 0.0, 1.0}},
+
+   {PIPE_FORMAT_R32G32B32_UNORM, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0x00000000, 0x00000000), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R32G32B32_UNORM, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0xffffffff, 0x00000000, 0x00000000), {1.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R32G32B32_UNORM, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0xffffffff, 0x00000000), {0.0, 1.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R32G32B32_UNORM, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0x00000000, 0xffffffff), {0.0, 0.0, 1.0, 1.0}},
+   {PIPE_FORMAT_R32G32B32_UNORM, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_R32G32B32A32_UNORM, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0x00000000, 0x00000000), {0.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_R32G32B32A32_UNORM, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0xffffffff, 0x00000000, 0x00000000, 0x00000000), {1.0, 0.0, 0.0, 0.0}},
+   {PIPE_FORMAT_R32G32B32A32_UNORM, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0xffffffff, 0x00000000, 0x00000000), {0.0, 1.0, 0.0, 0.0}},
+   {PIPE_FORMAT_R32G32B32A32_UNORM, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0xffffffff, 0x00000000), {0.0, 0.0, 1.0, 0.0}},
+   {PIPE_FORMAT_R32G32B32A32_UNORM, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0x00000000, 0xffffffff), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R32G32B32A32_UNORM, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), {1.0, 1.0, 1.0, 1.0}},
+
+   {PIPE_FORMAT_R32_USCALED, PACKED_1x32(0xffffffff), PACKED_1x32(0x00000000), {       0.0,        0.0,        0.0,   1.0}},
+   {PIPE_FORMAT_R32_USCALED, PACKED_1x32(0xffffffff), PACKED_1x32(0x01000000), {16777216.0,        0.0,        0.0,   1.0}},
+
+   {PIPE_FORMAT_R32G32_USCALED, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x00000000, 0x00000000), {       0.0,        0.0,        0.0,   1.0}},
+   {PIPE_FORMAT_R32G32_USCALED, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x01000000, 0x00000000), {16777216.0,        0.0,        0.0,   1.0}},
+   {PIPE_FORMAT_R32G32_USCALED, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x00000000, 0x01000000), {       0.0, 16777216.0,        0.0,   1.0}},
+   {PIPE_FORMAT_R32G32_USCALED, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x01000000, 0x01000000), {16777216.0, 16777216.0,        0.0,   1.0}},
+
+   {PIPE_FORMAT_R32G32B32_USCALED, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0x00000000, 0x00000000), {       0.0,        0.0,        0.0,   1.0}},
+   {PIPE_FORMAT_R32G32B32_USCALED, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x01000000, 0x00000000, 0x00000000), {16777216.0,        0.0,        0.0,   1.0}},
+   {PIPE_FORMAT_R32G32B32_USCALED, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0x01000000, 0x00000000), {       0.0, 16777216.0,        0.0,   1.0}},
+   {PIPE_FORMAT_R32G32B32_USCALED, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0x00000000, 0x01000000), {       0.0,        0.0, 16777216.0,   1.0}},
+   {PIPE_FORMAT_R32G32B32_USCALED, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x01000000, 0x01000000, 0x01000000), {16777216.0, 16777216.0, 16777216.0,   1.0}},
+
+   {PIPE_FORMAT_R32G32B32A32_USCALED, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0x00000000, 0x00000000), {       0.0,        0.0,        0.0,        0.0}},
+   {PIPE_FORMAT_R32G32B32A32_USCALED, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x01000000, 0x00000000, 0x00000000, 0x00000000), {16777216.0,        0.0,        0.0,        0.0}},
+   {PIPE_FORMAT_R32G32B32A32_USCALED, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x01000000, 0x00000000, 0x00000000), {       0.0, 16777216.0,        0.0,        0.0}},
+   {PIPE_FORMAT_R32G32B32A32_USCALED, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0x01000000, 0x00000000), {       0.0,        0.0, 16777216.0,        0.0}},
+   {PIPE_FORMAT_R32G32B32A32_USCALED, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0x00000000, 0x01000000), {       0.0,        0.0,        0.0, 16777216.0}},
+   {PIPE_FORMAT_R32G32B32A32_USCALED, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x01000000, 0x01000000, 0x01000000, 0x01000000), {16777216.0, 16777216.0, 16777216.0, 16777216.0}},
+
+   {PIPE_FORMAT_R32_SNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x00000000), {   0.0,    0.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R32_SNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x7fffffff), {   1.0,    0.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R32_SNORM, PACKED_1x32(0xffffffff), PACKED_1x32(0x80000001), {  -1.0,    0.0,    0.0,    1.0}},
+
+   {PIPE_FORMAT_R32G32_SNORM, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x00000000, 0x00000000), {   0.0,    0.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R32G32_SNORM, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x7fffffff, 0x00000000), {   1.0,    0.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R32G32_SNORM, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x80000001, 0x00000000), {  -1.0,    0.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R32G32_SNORM, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x00000000, 0x7fffffff), {   0.0,    1.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R32G32_SNORM, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x00000000, 0x80000001), {   0.0,   -1.0,    0.0,    1.0}},
+
+   {PIPE_FORMAT_R32G32B32_SNORM, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0x00000000, 0x00000000), {   0.0,    0.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R32G32B32_SNORM, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x7fffffff, 0x00000000, 0x00000000), {   1.0,    0.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R32G32B32_SNORM, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x80000001, 0x00000000, 0x00000000), {  -1.0,    0.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R32G32B32_SNORM, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0x7fffffff, 0x00000000), {   0.0,    1.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R32G32B32_SNORM, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0x80000001, 0x00000000), {   0.0,   -1.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R32G32B32_SNORM, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0x00000000, 0x7fffffff), {   0.0,    0.0,    1.0,    1.0}},
+   {PIPE_FORMAT_R32G32B32_SNORM, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0x00000000, 0x80000001), {   0.0,    0.0,   -1.0,    1.0}},
+
+   {PIPE_FORMAT_R32G32B32A32_SNORM, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0x00000000, 0x00000000), {   0.0,    0.0,    0.0,    0.0}},
+   {PIPE_FORMAT_R32G32B32A32_SNORM, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x7fffffff, 0x00000000, 0x00000000, 0x00000000), {   1.0,    0.0,    0.0,    0.0}},
+   {PIPE_FORMAT_R32G32B32A32_SNORM, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x80000001, 0x00000000, 0x00000000, 0x00000000), {  -1.0,    0.0,    0.0,    0.0}},
+   {PIPE_FORMAT_R32G32B32A32_SNORM, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x7fffffff, 0x00000000, 0x00000000), {   0.0,    1.0,    0.0,    0.0}},
+   {PIPE_FORMAT_R32G32B32A32_SNORM, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x80000001, 0x00000000, 0x00000000), {   0.0,   -1.0,    0.0,    0.0}},
+   {PIPE_FORMAT_R32G32B32A32_SNORM, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0x7fffffff, 0x00000000), {   0.0,    0.0,    1.0,    0.0}},
+   {PIPE_FORMAT_R32G32B32A32_SNORM, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0x80000001, 0x00000000), {   0.0,    0.0,   -1.0,    0.0}},
+   {PIPE_FORMAT_R32G32B32A32_SNORM, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0x00000000, 0x7fffffff), {   0.0,    0.0,    0.0,    1.0}},
+   {PIPE_FORMAT_R32G32B32A32_SNORM, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0x00000000, 0x80000001), {   0.0,    0.0,    0.0,   -1.0}},
+
+   {PIPE_FORMAT_R32_SSCALED, PACKED_1x32(0xffffffff), PACKED_1x32(0x00000000), {        0.0,         0.0,         0.0,   1.0}},
+   {PIPE_FORMAT_R32_SSCALED, PACKED_1x32(0xffffffff), PACKED_1x32(0x01000000), { 16777216.0,         0.0,         0.0,   1.0}},
+   {PIPE_FORMAT_R32_SSCALED, PACKED_1x32(0xffffffff), PACKED_1x32(0xff000000), {-16777216.0,         0.0,         0.0,   1.0}},
+
+   {PIPE_FORMAT_R32G32_SSCALED, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x00000000, 0x00000000), {        0.0,         0.0,         0.0,   1.0}},
+   {PIPE_FORMAT_R32G32_SSCALED, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x01000000, 0x00000000), { 16777216.0,         0.0,         0.0,   1.0}},
+   {PIPE_FORMAT_R32G32_SSCALED, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0xff000000, 0x00000000), {-16777216.0,         0.0,         0.0,   1.0}},
+   {PIPE_FORMAT_R32G32_SSCALED, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x00000000, 0x01000000), {        0.0,  16777216.0,         0.0,   1.0}},
+   {PIPE_FORMAT_R32G32_SSCALED, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x00000000, 0xff000000), {        0.0, -16777216.0,         0.0,   1.0}},
+
+   {PIPE_FORMAT_R32G32B32_SSCALED, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0x00000000, 0x00000000), {        0.0,         0.0,         0.0,   1.0}},
+   {PIPE_FORMAT_R32G32B32_SSCALED, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x01000000, 0x00000000, 0x00000000), { 16777216.0,         0.0,         0.0,   1.0}},
+   {PIPE_FORMAT_R32G32B32_SSCALED, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0xff000000, 0x00000000, 0x00000000), {-16777216.0,         0.0,         0.0,   1.0}},
+   {PIPE_FORMAT_R32G32B32_SSCALED, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0x01000000, 0x00000000), {        0.0,  16777216.0,         0.0,   1.0}},
+   {PIPE_FORMAT_R32G32B32_SSCALED, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0xff000000, 0x00000000), {        0.0, -16777216.0,         0.0,   1.0}},
+   {PIPE_FORMAT_R32G32B32_SSCALED, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0x00000000, 0x01000000), {        0.0,         0.0,  16777216.0,   1.0}},
+   {PIPE_FORMAT_R32G32B32_SSCALED, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0x00000000, 0xff000000), {        0.0,         0.0, -16777216.0,   1.0}},
+
+   {PIPE_FORMAT_R32G32B32A32_SSCALED, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0x00000000, 0x00000000), {        0.0,         0.0,         0.0,         0.0}},
+   {PIPE_FORMAT_R32G32B32A32_SSCALED, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x01000000, 0x00000000, 0x00000000, 0x00000000), { 16777216.0,         0.0,         0.0,         0.0}},
+   {PIPE_FORMAT_R32G32B32A32_SSCALED, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0xff000000, 0x00000000, 0x00000000, 0x00000000), {-16777216.0,         0.0,         0.0,         0.0}},
+   {PIPE_FORMAT_R32G32B32A32_SSCALED, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x01000000, 0x00000000, 0x00000000), {        0.0,  16777216.0,         0.0,         0.0}},
+   {PIPE_FORMAT_R32G32B32A32_SSCALED, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0xff000000, 0x00000000, 0x00000000), {        0.0, -16777216.0,         0.0,         0.0}},
+   {PIPE_FORMAT_R32G32B32A32_SSCALED, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0x01000000, 0x00000000), {        0.0,         0.0,  16777216.0,         0.0}},
+   {PIPE_FORMAT_R32G32B32A32_SSCALED, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0xff000000, 0x00000000), {        0.0,         0.0, -16777216.0,         0.0}},
+   {PIPE_FORMAT_R32G32B32A32_SSCALED, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0x00000000, 0x01000000), {        0.0,         0.0,         0.0,  16777216.0}},
+   {PIPE_FORMAT_R32G32B32A32_SSCALED, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0x00000000, 0xff000000), {        0.0,         0.0,         0.0, -16777216.0}},
+
+   /*
+    * Standard 32-bit float formats
+    */
+
+   {PIPE_FORMAT_R32_FLOAT, PACKED_1x32(0xffffffff), PACKED_1x32(0x00000000), {  0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R32_FLOAT, PACKED_1x32(0xffffffff), PACKED_1x32(0x3f800000), {  1.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R32_FLOAT, PACKED_1x32(0xffffffff), PACKED_1x32(0xbf800000), { -1.0, 0.0, 0.0, 1.0}},
+
+   {PIPE_FORMAT_R32G32_FLOAT, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x00000000, 0x00000000), { 0.0,  0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R32G32_FLOAT, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x3f800000, 0x00000000), { 1.0,  0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R32G32_FLOAT, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0xbf800000, 0x00000000), {-1.0,  0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R32G32_FLOAT, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x00000000, 0x3f800000), { 0.0,  1.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R32G32_FLOAT, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x00000000, 0xbf800000), { 0.0, -1.0, 0.0, 1.0}},
+   {PIPE_FORMAT_R32G32_FLOAT, PACKED_2x32(0xffffffff, 0xffffffff), PACKED_2x32(0x3f800000, 0x3f800000), { 1.0,  1.0, 0.0, 1.0}},
+
+   {PIPE_FORMAT_R32G32B32_FLOAT, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0x00000000, 0x00000000), { 0.0,  0.0,  0.0, 1.0}},
+   {PIPE_FORMAT_R32G32B32_FLOAT, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x3f800000, 0x00000000, 0x00000000), { 1.0,  0.0,  0.0, 1.0}},
+   {PIPE_FORMAT_R32G32B32_FLOAT, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0xbf800000, 0x00000000, 0x00000000), {-1.0,  0.0,  0.0, 1.0}},
+   {PIPE_FORMAT_R32G32B32_FLOAT, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0x3f800000, 0x00000000), { 0.0,  1.0,  0.0, 1.0}},
+   {PIPE_FORMAT_R32G32B32_FLOAT, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0xbf800000, 0x00000000), { 0.0, -1.0,  0.0, 1.0}},
+   {PIPE_FORMAT_R32G32B32_FLOAT, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0x00000000, 0x3f800000), { 0.0,  0.0,  1.0, 1.0}},
+   {PIPE_FORMAT_R32G32B32_FLOAT, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x00000000, 0x00000000, 0xbf800000), { 0.0,  0.0, -1.0, 1.0}},
+   {PIPE_FORMAT_R32G32B32_FLOAT, PACKED_3x32(0xffffffff, 0xffffffff, 0xffffffff), PACKED_3x32(0x3f800000, 0x3f800000, 0x3f800000), { 1.0,  1.0,  1.0, 1.0}},
+
+   {PIPE_FORMAT_R32G32B32A32_FLOAT, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0x00000000, 0x00000000), { 0.0,  0.0,  0.0,  0.0}},
+   {PIPE_FORMAT_R32G32B32A32_FLOAT, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x3f800000, 0x00000000, 0x00000000, 0x00000000), { 1.0,  0.0,  0.0,  0.0}},
+   {PIPE_FORMAT_R32G32B32A32_FLOAT, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0xbf800000, 0x00000000, 0x00000000, 0x00000000), {-1.0,  0.0,  0.0,  0.0}},
+   {PIPE_FORMAT_R32G32B32A32_FLOAT, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x3f800000, 0x00000000, 0x00000000), { 0.0,  1.0,  0.0,  0.0}},
+   {PIPE_FORMAT_R32G32B32A32_FLOAT, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0xbf800000, 0x00000000, 0x00000000), { 0.0, -1.0,  0.0,  0.0}},
+   {PIPE_FORMAT_R32G32B32A32_FLOAT, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0x3f800000, 0x00000000), { 0.0,  0.0,  1.0,  0.0}},
+   {PIPE_FORMAT_R32G32B32A32_FLOAT, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0xbf800000, 0x00000000), { 0.0,  0.0, -1.0,  0.0}},
+   {PIPE_FORMAT_R32G32B32A32_FLOAT, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0x00000000, 0x3f800000), { 0.0,  0.0,  0.0,  1.0}},
+   {PIPE_FORMAT_R32G32B32A32_FLOAT, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x00000000, 0x00000000, 0x00000000, 0xbf800000), { 0.0,  0.0,  0.0, -1.0}},
+   {PIPE_FORMAT_R32G32B32A32_FLOAT, PACKED_4x32(0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff), PACKED_4x32(0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000), { 1.0,  1.0,  1.0,  1.0}},
+};
+
+
+const unsigned util_format_nr_test_cases = Elements(util_format_test_cases);
diff --git a/src/gallium/auxiliary/util/u_format_tests.h b/src/gallium/auxiliary/util/u_format_tests.h
new file mode 100644
index 00000000000..2d4d9d5fa9d
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_format_tests.h
@@ -0,0 +1,69 @@
+/**************************************************************************
+ *
+ * Copyright 2010 VMware, 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 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
+ * THE COPYRIGHT HOLDERS, AUTHORS 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.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ **************************************************************************/
+
+
+#ifndef U_FORMAT_TESTS_H_
+#define U_FORMAT_TESTS_H_
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+
+
+#define UTIL_FORMAT_MAX_PACKED_BYTES 16
+
+
+/**
+ * A (packed, unpacked) color pair.
+ */
+struct util_format_test_case
+{
+   enum pipe_format format;
+
+   /**
+    * Mask of the bits that actually meaningful data. Used to mask out the
+    * "X" channels.
+    */
+   uint8_t mask[UTIL_FORMAT_MAX_PACKED_BYTES];
+
+   uint8_t packed[UTIL_FORMAT_MAX_PACKED_BYTES];
+
+   /**
+    * RGBA.
+    */
+   double unpacked[4];
+};
+
+
+extern const struct util_format_test_case
+util_format_test_cases[];
+
+
+extern const unsigned util_format_nr_test_cases;
+
+
+#endif /* U_FORMAT_TESTS_H_ */
diff --git a/src/gallium/auxiliary/util/u_simple_screen.c b/src/gallium/auxiliary/util/u_simple_screen.c
index 53f3c16dbcc..9203cb6580c 100644
--- a/src/gallium/auxiliary/util/u_simple_screen.c
+++ b/src/gallium/auxiliary/util/u_simple_screen.c
@@ -59,22 +59,7 @@ pass_user_buffer_create(struct pipe_screen *screen,
    return buffer;
 }
 
-static struct pipe_buffer *
-pass_surface_buffer_create(struct pipe_screen *screen,
-                           unsigned width, unsigned height,
-                           enum pipe_format format,
-                           unsigned usage,
-                           unsigned tex_usage,
-                           unsigned *stride)
-{
-   struct pipe_buffer *buffer =
-      screen->winsys->surface_buffer_create(screen->winsys, width, height,
-                                            format, usage, tex_usage, stride);
 
-   buffer->screen = screen;
-
-   return buffer;
-}
 
 static void *
 pass_buffer_map(struct pipe_screen *screen,
@@ -135,7 +120,6 @@ u_simple_screen_init(struct pipe_screen *screen)
 {
    screen->buffer_create = pass_buffer_create;
    screen->user_buffer_create = pass_user_buffer_create;
-   screen->surface_buffer_create = pass_surface_buffer_create;
 
    screen->buffer_map = pass_buffer_map;
    screen->buffer_unmap = pass_buffer_unmap;
diff --git a/src/gallium/docs/source/cso/sampler.rst b/src/gallium/docs/source/cso/sampler.rst
index 77979fc44d1..9bbb784de8e 100644
--- a/src/gallium/docs/source/cso/sampler.rst
+++ b/src/gallium/docs/source/cso/sampler.rst
@@ -13,38 +13,97 @@ Members
 -------
 
 wrap_s
-    How to wrap the S coordinate. One of PIPE_TEX_WRAP.
+    How to wrap the S coordinate. One of PIPE_TEX_WRAP_*.
 wrap_t
-    How to wrap the T coordinate. One of PIPE_TEX_WRAP.
+    How to wrap the T coordinate. One of PIPE_TEX_WRAP_*.
 wrap_r
-    How to wrap the R coordinate. One of PIPE_TEX_WRAP.
+    How to wrap the R coordinate. One of PIPE_TEX_WRAP_*.
+
+The wrap modes are:
+
+* ``PIPE_TEX_WRAP_REPEAT``: Standard coord repeat/wrap-around mode.
+* ``PIPE_TEX_WRAP_CLAMP_TO_EDGE``: Clamp coord to edge of texture, the border
+  color is never sampled.
+* ``PIPE_TEX_WRAP_CLAMP_TO_BORDER``: Clamp coord to border of texture, the
+  border color is sampled when coords go outside the range [0,1].
+* ``PIPE_TEX_WRAP_CLAMP``: The coord is clamped to the range [0,1] before
+  scaling to the texture size.  This corresponds to the legacy OpenGL GL_CLAMP
+  texture wrap mode.  Historically, this mode hasn't acted consistantly across
+  all graphics hardware.  It sometimes acts like CLAMP_TO_EDGE or
+  CLAMP_TO_BORDER.  The behaviour may also vary depending on linear vs.
+  nearest sampling mode.
+* ``PIPE_TEX_WRAP_MIRROR_REPEAT``: If the integer part of the coordinate
+  is odd, the coord becomes (1 - coord).  Then, normal texture REPEAT is
+  applied to the coord.
+* ``PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE``: First, the absolute value of the
+  coordinate is computed.  Then, regular CLAMP_TO_EDGE is applied to the coord.
+* ``PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER``: First, the absolute value of the
+  coordinate is computed.  Then, regular CLAMP_TO_BORDER is applied to the
+  coord.
+* ``PIPE_TEX_WRAP_MIRROR_CLAMP``: First, the absolute value of the coord is
+  computed.  Then, regular CLAMP is applied to the coord.
+
+
 min_img_filter
-    The filter to use when minifying texels. One of PIPE_TEX_FILTER.
+    The image filter to use when minifying texels. One of PIPE_TEX_FILTER_*.
+mag_img_filter
+    The image filter to use when magnifying texels. One of PIPE_TEX_FILTER_*.
+
+The texture image filter modes are:
+
+* ``PIPE_TEX_FILTER_NEAREST``: One texel is fetched from the texture image
+  at the texture coordinate.
+* ``PIPE_TEX_FILTER_LINEAR``: Two, four or eight texels (depending on the
+  texture dimensions; 1D/2D/3D) are fetched from the texture image and
+  linearly weighted and blended together.
+
 min_mip_filter
     The filter to use when minifying mipmapped textures. One of
-    PIPE_TEX_FILTER.
-mag_img_filter
-    The filter to use when magnifying texels. One of PIPE_TEX_FILTER.
+    PIPE_TEX_MIPFILTER_*.
+
+The texture mip filter modes are:
+
+* ``PIPE_TEX_MIPFILTER_NEAREST``: A single mipmap level/image is selected
+  according to the texture LOD (lambda) value.
+* ``PIPE_TEX_MIPFILTER_LINEAR``: The two mipmap levels/images above/below
+  the texture LOD value are sampled from.  The results of sampling from
+  those two images are blended together with linear interpolation.
+* ``PIPE_TEX_MIPFILTER_NONE``: Mipmap filtering is disabled.  All texels
+  are taken from the level 0 image.
+
+
 compare_mode
-    If set to PIPE_TEX_COMPARE_R_TO_TEXTURE, texture output is computed
-    according to compare_func, using r coord and the texture value as operands.
+    If set to PIPE_TEX_COMPARE_R_TO_TEXTURE, the result of texture sampling
+    is not a color but a true/false value which is the result of comparing the
+    sampled texture value (typically a Z value from a depth texture) to the
+    texture coordinate's R component.
     If set to PIPE_TEX_COMPARE_NONE, no comparison calculation is performed.
 compare_func
-    How the comparison is computed. One of PIPE_FUNC.
+    The inequality operator used when compare_mode=1.  One of PIPE_FUNC_x.
 normalized_coords
-    Whether the texture coordinates are normalized. If normalized, they will
-    always be in [0, 1]. If not, they will be in the range of each dimension
-    of the loaded texture.
+    If set, the incoming texture coordinates (nominally in the range [0,1])
+    will be scaled by the texture width, height, depth to compute texel
+    addresses.  Otherwise, the texture coords are used as-is (they are not
+    scaled by the texture dimensions).
+    When normalized_coords=0, only a subset of the texture wrap modes are
+    allowed: PIPE_TEX_WRAP_CLAMP, PIPE_TEX_WRAP_CLAMP_TO_EDGE and
+    PIPE_TEX_WRAP_CLAMP_TO_BORDER.
 lod_bias
-    The bias to apply to the level of detail.
+    Bias factor which is added to the computed level of detail.
+    The normal level of detail is computed from the partial derivatives of
+    the texture coordinates and/or the fragment shader TEX/TXB/TXL
+    instruction.
 min_lod
-    Minimum level of detail, used to clamp LoD after bias.
+    Minimum level of detail, used to clamp LOD after bias.  The LOD values
+    correspond to mipmap levels where LOD=0 is the level 0 mipmap image.
 max_lod
-    Maximum level of detail, used to clamp LoD after bias.
+    Maximum level of detail, used to clamp LOD after bias.
 border_color
-    RGBA color used for out-of-bounds coordinates.
+    RGBA color used for texel coordinates that are outside the [0,width-1],
+    [0, height-1] or [0, depth-1] ranges.
 max_anisotropy
-    Maximum filtering to apply anisotropically to textures. Setting this to
-    0 disables anisotropic filtering. Any other setting enables anisotropic
-    filtering, however it's not unexpected some drivers only will change their
-    filtering with a setting of 2 and higher.
+    Maximum anistropy ratio to use when sampling from textures.  For example,
+    if max_anistropy=4, a region of up to 1 by 4 texels will be sampled.
+    Set to zero to disable anisotropic filtering.  Any other setting enables
+    anisotropic filtering, however it's not unexpected some drivers only will
+    change their filtering with a setting of 2 and higher.
diff --git a/src/gallium/drivers/i915/i915_context.h b/src/gallium/drivers/i915/i915_context.h
index da769e7b290..499a727314d 100644
--- a/src/gallium/drivers/i915/i915_context.h
+++ b/src/gallium/drivers/i915/i915_context.h
@@ -187,6 +187,9 @@ struct i915_sampler_state {
    unsigned maxlod;
 };
 
+#define I915_MAX_TEXTURE_2D_LEVELS 11  /* max 1024x1024 */
+#define I915_MAX_TEXTURE_3D_LEVELS  8  /* max 128x128x128 */
+
 struct i915_texture {
    struct pipe_texture base;
 
@@ -199,7 +202,7 @@ struct i915_texture {
    unsigned sw_tiled; /**< tiled with software flags */
    unsigned hw_tiled; /**< tiled with hardware fences */
 
-   unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS];
+   unsigned nr_images[I915_MAX_TEXTURE_2D_LEVELS];
 
    /* Explicitly store the offset of each image for each cube face or
     * depth value.  Pretty much have to accept that hardware formats
@@ -207,7 +210,7 @@ struct i915_texture {
     * compute the offsets of depth/cube images within a mipmap level,
     * so have to store them as a lookup table:
     */
-   unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS];   /**< array [depth] of offsets */
+   unsigned *image_offset[I915_MAX_TEXTURE_2D_LEVELS];   /**< array [depth] of offsets */
 
    /* The data is held here:
     */
diff --git a/src/gallium/drivers/i915/i915_screen.c b/src/gallium/drivers/i915/i915_screen.c
index 72bd2635506..e5bf4a20bd0 100644
--- a/src/gallium/drivers/i915/i915_screen.c
+++ b/src/gallium/drivers/i915/i915_screen.c
@@ -116,11 +116,11 @@ i915_get_param(struct pipe_screen *screen, int param)
    case PIPE_CAP_TEXTURE_SHADOW_MAP:
       return 1;
    case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
-      return 11; /* max 1024x1024 */
+      return I915_MAX_TEXTURE_2D_LEVELS;
    case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
-      return 8;  /* max 128x128x128 */
+      return I915_MAX_TEXTURE_3D_LEVELS;
    case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
-      return 11; /* max 1024x1024 */
+      return I915_MAX_TEXTURE_2D_LEVELS;
    case PIPE_CAP_TGSI_FS_COORD_ORIGIN_UPPER_LEFT:
    case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_HALF_INTEGER:
       return 1;
diff --git a/src/gallium/drivers/i915/i915_texture.c b/src/gallium/drivers/i915/i915_texture.c
index 7ba222c78b7..3ce52cdcdbd 100644
--- a/src/gallium/drivers/i915/i915_texture.c
+++ b/src/gallium/drivers/i915/i915_texture.c
@@ -96,7 +96,7 @@ i915_miptree_set_level_info(struct i915_texture *tex,
                              unsigned nr_images,
                              unsigned w, unsigned h, unsigned d)
 {
-   assert(level < PIPE_MAX_TEXTURE_LEVELS);
+   assert(level < Elements(tex->nr_images));
 
    tex->nr_images[level] = nr_images;
 
@@ -219,12 +219,12 @@ i915_miptree_layout_2d(struct i915_texture *tex)
    unsigned nblocksy = util_format_get_nblocksy(pt->format, pt->width0);
 
    /* used for scanouts that need special layouts */
-   if (pt->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY)
+   if (pt->tex_usage & PIPE_TEXTURE_USAGE_SCANOUT)
       if (i915_scanout_layout(tex))
          return;
 
-   /* for shared buffers we use something very like scanout */
-   if (pt->tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET)
+   /* shared buffers needs to be compatible with X servers */
+   if (pt->tex_usage & PIPE_TEXTURE_USAGE_SHARED)
       if (i915_display_target_layout(tex))
          return;
 
@@ -369,12 +369,12 @@ i945_miptree_layout_2d(struct i915_texture *tex)
    unsigned nblocksy = util_format_get_nblocksy(pt->format, pt->height0);
 
    /* used for scanouts that need special layouts */
-   if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_PRIMARY)
+   if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_SCANOUT)
       if (i915_scanout_layout(tex))
          return;
 
-   /* for shared buffers we use some very like scanout */
-   if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET)
+   /* shared buffers needs to be compatible with X servers */
+   if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_SHARED)
       if (i915_display_target_layout(tex))
          return;
 
@@ -642,7 +642,7 @@ i915_texture_create(struct pipe_screen *screen,
 
 
    /* for scanouts and cursors, cursors arn't scanouts */
-   if (templat->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY && templat->width0 != 64)
+   if (templat->tex_usage & PIPE_TEXTURE_USAGE_SCANOUT && templat->width0 != 64)
       buf_usage = INTEL_NEW_SCANOUT;
    else
       buf_usage = INTEL_NEW_TEXTURE;
@@ -673,19 +673,24 @@ fail:
 }
 
 static struct pipe_texture *
-i915_texture_blanket(struct pipe_screen * screen,
-                     const struct pipe_texture *base,
-                     const unsigned *stride,
-                     struct pipe_buffer *buffer)
+i915_texture_from_handle(struct pipe_screen * screen,
+                         const struct pipe_texture *templat,
+                         struct winsys_handle *whandle)
 {
-#if 0
+   struct i915_screen *is = i915_screen(screen);
    struct i915_texture *tex;
+   struct intel_winsys *iws = is->iws;
+   struct intel_buffer *buffer;
+   unsigned stride;
+
    assert(screen);
 
+   buffer = iws->buffer_from_handle(iws, whandle, &stride);
+
    /* Only supports one type */
-   if (base->target != PIPE_TEXTURE_2D ||
-       base->last_level != 0 ||
-       base->depth0 != 1) {
+   if (templat->target != PIPE_TEXTURE_2D ||
+       templat->last_level != 0 ||
+       templat->depth0 != 1) {
       return NULL;
    }
 
@@ -693,23 +698,33 @@ i915_texture_blanket(struct pipe_screen * screen,
    if (!tex)
       return NULL;
 
-   tex->base = *base;
+   tex->base = *templat;
    pipe_reference_init(&tex->base.reference, 1);
    tex->base.screen = screen;
 
-   tex->stride = stride[0];
+   tex->stride = stride;
 
-   i915_miptree_set_level_info(tex, 0, 1, base->width0, base->height0, 1);
+   i915_miptree_set_level_info(tex, 0, 1, templat->width0, templat->height0, 1);
    i915_miptree_set_image_offset(tex, 0, 0, 0, 0);
 
-   pipe_buffer_reference(&tex->buffer, buffer);
+   tex->buffer = buffer;
 
    return &tex->base;
-#else
-   return NULL;
-#endif
 }
 
+static boolean
+i915_texture_get_handle(struct pipe_screen * screen,
+                        struct pipe_texture *texture,
+                        struct winsys_handle *whandle)
+{
+   struct i915_screen *is = i915_screen(screen);
+   struct i915_texture *tex = (struct i915_texture *)texture;
+   struct intel_winsys *iws = is->iws;
+
+   return iws->buffer_get_handle(iws, tex->buffer, whandle, tex->stride);
+}
+
+
 static void
 i915_texture_destroy(struct pipe_texture *pt)
 {
@@ -723,7 +738,7 @@ i915_texture_destroy(struct pipe_texture *pt)
 
    iws->buffer_destroy(iws, tex->buffer);
 
-   for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++)
+   for (i = 0; i < Elements(tex->image_offset); i++)
       if (tex->image_offset[i])
          FREE(tex->image_offset[i]);
 
@@ -869,7 +884,8 @@ void
 i915_init_screen_texture_functions(struct i915_screen *is)
 {
    is->base.texture_create = i915_texture_create;
-   is->base.texture_blanket = i915_texture_blanket;
+   is->base.texture_from_handle = i915_texture_from_handle;
+   is->base.texture_get_handle = i915_texture_get_handle;
    is->base.texture_destroy = i915_texture_destroy;
    is->base.get_tex_surface = i915_get_tex_surface;
    is->base.tex_surface_destroy = i915_tex_surface_destroy;
@@ -878,53 +894,3 @@ i915_init_screen_texture_functions(struct i915_screen *is)
    is->base.transfer_unmap = i915_transfer_unmap;
    is->base.tex_transfer_destroy = i915_tex_transfer_destroy;
 }
-
-struct pipe_texture *
-i915_texture_blanket_intel(struct pipe_screen *screen,
-                           struct pipe_texture *base,
-                           unsigned stride,
-                           struct intel_buffer *buffer)
-{
-   struct i915_texture *tex;
-   assert(screen);
-
-   /* Only supports one type */
-   if (base->target != PIPE_TEXTURE_2D ||
-       base->last_level != 0 ||
-       base->depth0 != 1) {
-      return NULL;
-   }
-
-   tex = CALLOC_STRUCT(i915_texture);
-   if (!tex)
-      return NULL;
-
-   tex->base = *base;
-   pipe_reference_init(&tex->base.reference, 1);
-   tex->base.screen = screen;
-
-   tex->stride = stride;
-
-   i915_miptree_set_level_info(tex, 0, 1, base->width0, base->height0, 1);
-   i915_miptree_set_image_offset(tex, 0, 0, 0, 0);
-
-   tex->buffer = buffer;
-
-   return &tex->base;
-}
-
-boolean
-i915_get_texture_buffer_intel(struct pipe_texture *texture,
-                              struct intel_buffer **buffer,
-                              unsigned *stride)
-{
-   struct i915_texture *tex = (struct i915_texture *)texture;
-
-   if (!texture)
-      return FALSE;
-
-   *stride = tex->stride;
-   *buffer = tex->buffer;
-
-   return TRUE;
-}
diff --git a/src/gallium/drivers/i915/intel_winsys.h b/src/gallium/drivers/i915/intel_winsys.h
index b3a802b0e29..00fd0c1efea 100644
--- a/src/gallium/drivers/i915/intel_winsys.h
+++ b/src/gallium/drivers/i915/intel_winsys.h
@@ -33,6 +33,7 @@ struct intel_buffer;
 struct intel_batchbuffer;
 struct pipe_texture;
 struct pipe_fence_handle;
+struct winsys_handle;
 
 enum intel_buffer_usage
 {
@@ -129,6 +130,25 @@ struct intel_winsys {
                                          enum intel_buffer_type type);
 
    /**
+    * Creates a buffer from a handle.
+    * Used to implement pipe_screen::texture_from_handle.
+    * Also provides the stride information needed for the
+    * texture via the stride argument.
+    */
+   struct intel_buffer *(*buffer_from_handle)(struct intel_winsys *iws,
+                                              struct winsys_handle *whandle,
+                                              unsigned *stride);
+
+   /**
+    * Used to implement pipe_screen::texture_get_handle.
+    * The winsys might need the stride information.
+    */
+   boolean (*buffer_get_handle)(struct intel_winsys *iws,
+                                struct intel_buffer *buffer,
+                                struct winsys_handle *whandle,
+                                unsigned stride);
+
+   /**
     * Fence a buffer with a fence reg.
     * Not to be confused with pipe_fence_handle.
     */
@@ -204,23 +224,4 @@ struct intel_winsys {
 struct pipe_screen *i915_create_screen(struct intel_winsys *iws, unsigned pci_id);
 
 
-/**
- * Get the intel_winsys buffer backing the texture.
- *
- * TODO UGLY
- */
-boolean i915_get_texture_buffer_intel(struct pipe_texture *texture,
-                                      struct intel_buffer **buffer,
-                                      unsigned *stride);
-
-/**
- * Wrap a intel_winsys buffer with a texture blanket.
- *
- * TODO UGLY
- */
-struct pipe_texture * i915_texture_blanket_intel(struct pipe_screen *screen,
-                                                 struct pipe_texture *tmplt,
-                                                 unsigned pitch,
-                                                 struct intel_buffer *buffer);
-
 #endif
diff --git a/src/gallium/drivers/i965/brw_screen.c b/src/gallium/drivers/i965/brw_screen.c
index 66f3aad8b21..cef83ffea80 100644
--- a/src/gallium/drivers/i965/brw_screen.c
+++ b/src/gallium/drivers/i965/brw_screen.c
@@ -174,11 +174,11 @@ brw_get_param(struct pipe_screen *screen, int param)
    case PIPE_CAP_TEXTURE_SHADOW_MAP:
       return 1;
    case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
-      return 11; /* max 1024x1024 */
+      return BRW_MAX_TEXTURE_2D_LEVELS;
    case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
-      return 8;  /* max 128x128x128 */
+      return BRW_MAX_TEXTURE_3D_LEVELS;
    case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
-      return 11; /* max 1024x1024 */
+      return BRW_MAX_TEXTURE_2D_LEVELS;
    case PIPE_CAP_TGSI_FS_COORD_ORIGIN_UPPER_LEFT:
    case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_HALF_INTEGER:
       return 1;
diff --git a/src/gallium/drivers/i965/brw_screen.h b/src/gallium/drivers/i965/brw_screen.h
index 7226d9228b7..e0f3cd2a9fe 100644
--- a/src/gallium/drivers/i965/brw_screen.h
+++ b/src/gallium/drivers/i965/brw_screen.h
@@ -100,6 +100,9 @@ struct brw_surface
 };
 
 
+#define BRW_MAX_TEXTURE_2D_LEVELS 11  /* max 1024x1024 */
+#define BRW_MAX_TEXTURE_3D_LEVELS  8  /* max 128x128x128 */
+
 
 struct brw_texture
 {
@@ -107,9 +110,9 @@ struct brw_texture
    struct brw_winsys_buffer *bo;
    struct brw_surface_state ss;
 
-   unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS];
-   unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS];
-   unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS];
+   unsigned *image_offset[BRW_MAX_TEXTURE_2D_LEVELS];
+   unsigned nr_images[BRW_MAX_TEXTURE_2D_LEVELS];
+   unsigned level_offset[BRW_MAX_TEXTURE_2D_LEVELS];
 
    boolean compressed;
    unsigned brw_target;
diff --git a/src/gallium/drivers/i965/brw_screen_texture.c b/src/gallium/drivers/i965/brw_screen_texture.c
index e38fdf1869b..cc79bfc7715 100644
--- a/src/gallium/drivers/i965/brw_screen_texture.c
+++ b/src/gallium/drivers/i965/brw_screen_texture.c
@@ -231,8 +231,8 @@ static struct pipe_texture *brw_texture_create( struct pipe_screen *screen,
       goto fail;
 
    
-   if (templ->tex_usage & (PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
-                           PIPE_TEXTURE_USAGE_PRIMARY)) {
+   if (templ->tex_usage & (PIPE_TEXTURE_USAGE_SCANOUT |
+                           PIPE_TEXTURE_USAGE_SHARED)) {
       buffer_type = BRW_BUFFER_TYPE_SCANOUT;
    }
    else {
@@ -303,14 +303,121 @@ fail:
    return NULL;
 }
 
-static struct pipe_texture *brw_texture_blanket(struct pipe_screen *screen,
-						const struct pipe_texture *templ,
-						const unsigned *stride,
-						struct pipe_buffer *buffer)
+static struct pipe_texture * 
+brw_texture_from_handle(struct pipe_screen *screen,
+                        const struct pipe_texture *templ,
+                        struct winsys_handle *whandle)
 {
+   struct brw_screen *bscreen = brw_screen(screen);
+   struct brw_texture *tex;
+   struct brw_winsys_buffer *buffer;
+   unsigned tiling;
+   unsigned pitch;
+
+   if (templ->target != PIPE_TEXTURE_2D ||
+       templ->last_level != 0 ||
+       templ->depth0 != 1)
+      return NULL;
+
+   if (util_format_is_compressed(templ->format))
+      return NULL;
+
+   tex = CALLOC_STRUCT(brw_texture);
+   if (!tex)
+      return NULL;
+
+   if (bscreen->sws->bo_from_handle(bscreen->sws, whandle, &pitch, &tiling, &buffer) != PIPE_OK)
+      goto fail;
+
+   memcpy(&tex->base, templ, sizeof *templ);
+   pipe_reference_init(&tex->base.reference, 1);
+   tex->base.screen = screen;
+
+   /* XXX: cpp vs. blocksize
+    */
+   tex->cpp = util_format_get_blocksize(tex->base.format);
+   tex->tiling = tiling;
+
+   make_empty_list(&tex->views[0]);
+   make_empty_list(&tex->views[1]);
+
+   if (!brw_texture_layout(bscreen, tex))
+      goto fail;
+
+   /* XXX Maybe some more checks? */
+   if ((pitch / tex->cpp) < tex->pitch)
+      goto fail;
+
+   tex->pitch = pitch / tex->cpp;
+
+   tex->bo = buffer;
+
+   /* fix this warning */
+#if 0
+   if (tex->size > buffer->size)
+      goto fail;
+#endif
+
+   tex->ss.ss0.mipmap_layout_mode = BRW_SURFACE_MIPMAPLAYOUT_BELOW;
+   tex->ss.ss0.surface_type = translate_tex_target(tex->base.target);
+   tex->ss.ss0.surface_format = translate_tex_format(tex->base.format);
+   assert(tex->ss.ss0.surface_format != BRW_SURFACEFORMAT_INVALID);
+
+   /* This is ok for all textures with channel width 8bit or less:
+    */
+/*    tex->ss.ss0.data_return_format = BRW_SURFACERETURNFORMAT_S1; */
+
+
+   /* XXX: what happens when tex->bo->offset changes???
+    */
+   tex->ss.ss1.base_addr = 0; /* reloc */
+   tex->ss.ss2.mip_count = tex->base.last_level;
+   tex->ss.ss2.width = tex->base.width0 - 1;
+   tex->ss.ss2.height = tex->base.height0 - 1;
+
+   switch (tex->tiling) {
+   case BRW_TILING_NONE:
+      tex->ss.ss3.tiled_surface = 0;
+      tex->ss.ss3.tile_walk = 0;
+      break;
+   case BRW_TILING_X:
+      tex->ss.ss3.tiled_surface = 1;
+      tex->ss.ss3.tile_walk = BRW_TILEWALK_XMAJOR;
+      break;
+   case BRW_TILING_Y:
+      tex->ss.ss3.tiled_surface = 1;
+      tex->ss.ss3.tile_walk = BRW_TILEWALK_YMAJOR;
+      break;
+   }
+
+   tex->ss.ss3.pitch = (tex->pitch * tex->cpp) - 1;
+   tex->ss.ss3.depth = tex->base.depth0 - 1;
+
+   tex->ss.ss4.min_lod = 0;
+
+   return &tex->base;
+
+fail:
+   FREE(tex);
    return NULL;
 }
 
+static boolean
+brw_texture_get_handle(struct pipe_screen *screen,
+                       struct pipe_texture *texture,
+                       struct winsys_handle *whandle)
+{
+   struct brw_screen *bscreen = brw_screen(screen);
+   struct brw_texture *tex = brw_texture(texture);
+   unsigned stride;
+
+   stride = tex->pitch * tex->cpp;
+
+   return bscreen->sws->bo_get_handle(tex->bo, whandle, stride);
+}
+
+
+
 static void brw_texture_destroy(struct pipe_texture *pt)
 {
    struct brw_texture *tex = brw_texture(pt);
@@ -451,127 +558,13 @@ brw_tex_transfer_destroy(struct pipe_transfer *trans)
 }
 
 
-/*
- * Functions exported to the winsys
- */
-
-boolean brw_texture_get_winsys_buffer(struct pipe_texture *texture,
-                                      struct brw_winsys_buffer **buffer,
-                                      unsigned *stride)
-{
-   struct brw_texture *tex = brw_texture(texture);
-
-   *buffer = tex->bo;
-   if (stride)
-      *stride = tex->pitch * tex->cpp;
-
-   return TRUE;
-}
-
-struct pipe_texture * 
-brw_texture_blanket_winsys_buffer(struct pipe_screen *screen,
-                                  const struct pipe_texture *templ,
-                                  unsigned pitch,
-				  unsigned tiling,
-                                  struct brw_winsys_buffer *buffer)
-{
-   struct brw_screen *bscreen = brw_screen(screen);
-   struct brw_texture *tex;
-   GLuint format;
-
-   if (templ->target != PIPE_TEXTURE_2D ||
-       templ->last_level != 0 ||
-       templ->depth0 != 1)
-      return NULL;
-
-   if (util_format_is_compressed(templ->format))
-      return NULL;
-
-   tex = CALLOC_STRUCT(brw_texture);
-   if (!tex)
-      return NULL;
-
-   memcpy(&tex->base, templ, sizeof *templ);
-   pipe_reference_init(&tex->base.reference, 1);
-   tex->base.screen = screen;
-
-   /* XXX: cpp vs. blocksize
-    */
-   tex->cpp = util_format_get_blocksize(tex->base.format);
-   tex->tiling = tiling;
-
-   make_empty_list(&tex->views[0]);
-   make_empty_list(&tex->views[1]);
-
-   if (!brw_texture_layout(bscreen, tex))
-      goto fail;
-
-   /* XXX Maybe some more checks? */
-   if ((pitch / tex->cpp) < tex->pitch)
-      goto fail;
-
-   tex->pitch = pitch / tex->cpp;
-
-   tex->bo = buffer;
-
-   /* fix this warning */
-#if 0
-   if (tex->size > buffer->size)
-      goto fail;
-#endif
-
-   tex->ss.ss0.mipmap_layout_mode = BRW_SURFACE_MIPMAPLAYOUT_BELOW;
-   tex->ss.ss0.surface_type = translate_tex_target(tex->base.target);
-
-   format = translate_tex_format(tex->base.format);
-   assert(format != BRW_SURFACEFORMAT_INVALID);
-   tex->ss.ss0.surface_format = format;
-
-   /* This is ok for all textures with channel width 8bit or less:
-    */
-/*    tex->ss.ss0.data_return_format = BRW_SURFACERETURNFORMAT_S1; */
-
-
-   /* XXX: what happens when tex->bo->offset changes???
-    */
-   tex->ss.ss1.base_addr = 0; /* reloc */
-   tex->ss.ss2.mip_count = tex->base.last_level;
-   tex->ss.ss2.width = tex->base.width0 - 1;
-   tex->ss.ss2.height = tex->base.height0 - 1;
-
-   switch (tex->tiling) {
-   case BRW_TILING_NONE:
-      tex->ss.ss3.tiled_surface = 0;
-      tex->ss.ss3.tile_walk = 0;
-      break;
-   case BRW_TILING_X:
-      tex->ss.ss3.tiled_surface = 1;
-      tex->ss.ss3.tile_walk = BRW_TILEWALK_XMAJOR;
-      break;
-   case BRW_TILING_Y:
-      tex->ss.ss3.tiled_surface = 1;
-      tex->ss.ss3.tile_walk = BRW_TILEWALK_YMAJOR;
-      break;
-   }
-
-   tex->ss.ss3.pitch = (tex->pitch * tex->cpp) - 1;
-   tex->ss.ss3.depth = tex->base.depth0 - 1;
-
-   tex->ss.ss4.min_lod = 0;
-
-   return &tex->base;
-
-fail:
-   FREE(tex);
-   return NULL;
-}
-
 void brw_screen_tex_init( struct brw_screen *brw_screen )
 {
    brw_screen->base.is_format_supported = brw_is_format_supported;
    brw_screen->base.texture_create = brw_texture_create;
+   brw_screen->base.texture_from_handle = brw_texture_from_handle;
+   brw_screen->base.texture_get_handle = brw_texture_get_handle;
    brw_screen->base.texture_destroy = brw_texture_destroy;
-   brw_screen->base.texture_blanket = brw_texture_blanket;
    brw_screen->base.get_tex_transfer = brw_get_tex_transfer;
    brw_screen->base.transfer_map = brw_transfer_map;
    brw_screen->base.transfer_unmap = brw_transfer_unmap;
diff --git a/src/gallium/drivers/i965/brw_winsys.h b/src/gallium/drivers/i965/brw_winsys.h
index c82d00f4a47..f30c7f18132 100644
--- a/src/gallium/drivers/i965/brw_winsys.h
+++ b/src/gallium/drivers/i965/brw_winsys.h
@@ -162,6 +162,16 @@ struct brw_winsys_screen {
                                unsigned alignment,
                                struct brw_winsys_buffer **bo_out);
 
+   enum pipe_error (*bo_from_handle)(struct brw_winsys_screen *sws,
+                                     struct winsys_handle *whandle,
+                                     unsigned *stride,
+                                     unsigned *tiling,
+                                     struct brw_winsys_buffer **bo_out);
+
+   enum pipe_error (*bo_get_handle)(struct brw_winsys_buffer *buffer,
+                                    struct winsys_handle *whandle,
+                                    unsigned stride);
+
    /* Destroy a buffer when our refcount goes to zero:
     */
    void (*bo_destroy)(struct brw_winsys_buffer *buffer);
@@ -257,28 +267,6 @@ bo_reference(struct brw_winsys_buffer **ptr, struct brw_winsys_buffer *buf)
 struct pipe_screen *brw_create_screen(struct brw_winsys_screen *iws, unsigned pci_id);
 
 
-/**
- * Get the brw_winsys buffer backing the texture.
- *
- * TODO UGLY
- */
-struct pipe_texture;
-boolean brw_texture_get_winsys_buffer(struct pipe_texture *texture,
-                                      struct brw_winsys_buffer **buffer,
-                                      unsigned *stride);
-
-/**
- * Wrap a brw_winsys buffer with a texture blanket.
- *
- * TODO UGLY
- */
-struct pipe_texture * 
-brw_texture_blanket_winsys_buffer(struct pipe_screen *screen,
-                                  const struct pipe_texture *template,
-                                  unsigned pitch,
-				  unsigned tiling,
-                                  struct brw_winsys_buffer *buffer);
-
 
 /*************************************************************************
  * Cooperative dumping between winsys and driver.  TODO: make this
diff --git a/src/gallium/drivers/identity/id_drm.c b/src/gallium/drivers/identity/id_drm.c
index f258c38cd71..936ccc444a8 100644
--- a/src/gallium/drivers/identity/id_drm.c
+++ b/src/gallium/drivers/identity/id_drm.c
@@ -63,62 +63,6 @@ identity_drm_create_screen(struct drm_api *_api, int fd,
    return identity_screen_create(screen);
 }
 
-
-static struct pipe_texture *
-identity_drm_texture_from_shared_handle(struct drm_api *_api,
-                                        struct pipe_screen *_screen,
-                                        struct pipe_texture *templ,
-                                        const char *name,
-                                        unsigned stride,
-                                        unsigned handle)
-{
-   struct identity_screen *id_screen = identity_screen(_screen);
-   struct identity_drm_api *id_api = identity_drm_api(_api);
-   struct pipe_screen *screen = id_screen->screen;
-   struct drm_api *api = id_api->api;
-   struct pipe_texture *result;
-
-   result = api->texture_from_shared_handle(api, screen, templ, name, stride, handle);
-
-   result = identity_texture_create(identity_screen(_screen), result);
-
-   return result;
-}
-
-static boolean
-identity_drm_shared_handle_from_texture(struct drm_api *_api,
-                                        struct pipe_screen *_screen,
-                                        struct pipe_texture *_texture,
-                                        unsigned *stride,
-                                        unsigned *handle)
-{
-   struct identity_screen *id_screen = identity_screen(_screen);
-   struct identity_texture *id_texture = identity_texture(_texture);
-   struct identity_drm_api *id_api = identity_drm_api(_api);
-   struct pipe_screen *screen = id_screen->screen;
-   struct pipe_texture *texture = id_texture->texture;
-   struct drm_api *api = id_api->api;
-
-   return api->shared_handle_from_texture(api, screen, texture, stride, handle);
-}
-
-static boolean
-identity_drm_local_handle_from_texture(struct drm_api *_api,
-                                       struct pipe_screen *_screen,
-                                       struct pipe_texture *_texture,
-                                       unsigned *stride,
-                                       unsigned *handle)
-{
-   struct identity_screen *id_screen = identity_screen(_screen);
-   struct identity_texture *id_texture = identity_texture(_texture);
-   struct identity_drm_api *id_api = identity_drm_api(_api);
-   struct pipe_screen *screen = id_screen->screen;
-   struct pipe_texture *texture = id_texture->texture;
-   struct drm_api *api = id_api->api;
-
-   return api->local_handle_from_texture(api, screen, texture, stride, handle);
-}
-
 static void
 identity_drm_destroy(struct drm_api *_api)
 {
@@ -145,9 +89,6 @@ identity_drm_create(struct drm_api *api)
    id_api->base.name = api->name;
    id_api->base.driver_name = api->driver_name;
    id_api->base.create_screen = identity_drm_create_screen;
-   id_api->base.texture_from_shared_handle = identity_drm_texture_from_shared_handle;
-   id_api->base.shared_handle_from_texture = identity_drm_shared_handle_from_texture;
-   id_api->base.local_handle_from_texture = identity_drm_local_handle_from_texture;
    id_api->base.destroy = identity_drm_destroy;
    id_api->api = api;
 
diff --git a/src/gallium/drivers/identity/id_screen.c b/src/gallium/drivers/identity/id_screen.c
index b85492114a3..b9d0f003d74 100644
--- a/src/gallium/drivers/identity/id_screen.c
+++ b/src/gallium/drivers/identity/id_screen.c
@@ -135,27 +135,40 @@ identity_screen_texture_create(struct pipe_screen *_screen,
 }
 
 static struct pipe_texture *
-identity_screen_texture_blanket(struct pipe_screen *_screen,
-                                const struct pipe_texture *templat,
-                                const unsigned *stride,
-                                struct pipe_buffer *_buffer)
+identity_screen_texture_from_handle(struct pipe_screen *_screen,
+                                    const struct pipe_texture *templ,
+                                    struct winsys_handle *handle)
 {
    struct identity_screen *id_screen = identity_screen(_screen);
-   struct identity_buffer *id_buffer = identity_buffer(_buffer);
    struct pipe_screen *screen = id_screen->screen;
-   struct pipe_buffer *buffer = id_buffer->buffer;
    struct pipe_texture *result;
 
-   result = screen->texture_blanket(screen,
-                                    templat,
-                                    stride,
-                                    buffer);
+   /* TODO trace call */
 
-   if (result)
-      return identity_texture_create(id_screen, result);
-   return NULL;
+   result = screen->texture_from_handle(screen, templ, handle);
+
+   result = identity_texture_create(identity_screen(_screen), result);
+
+   return result;
+}
+
+static boolean
+identity_screen_texture_get_handle(struct pipe_screen *_screen,
+                                   struct pipe_texture *_texture,
+                                   struct winsys_handle *handle)
+{
+   struct identity_screen *id_screen = identity_screen(_screen);
+   struct identity_texture *id_texture = identity_texture(_texture);
+   struct pipe_screen *screen = id_screen->screen;
+   struct pipe_texture *texture = id_texture->texture;
+
+   /* TODO trace call */
+
+   return screen->texture_get_handle(screen, texture, handle);
 }
 
+
+
 static void
 identity_screen_texture_destroy(struct pipe_texture *_texture)
 {
@@ -298,31 +311,6 @@ identity_screen_user_buffer_create(struct pipe_screen *_screen,
    return NULL;
 }
 
-static struct pipe_buffer *
-identity_screen_surface_buffer_create(struct pipe_screen *_screen,
-                                      unsigned width,
-                                      unsigned height,
-                                      enum pipe_format format,
-                                      unsigned usage,
-                                      unsigned tex_usage,
-                                      unsigned *stride)
-{
-   struct identity_screen *id_screen = identity_screen(_screen);
-   struct pipe_screen *screen = id_screen->screen;
-   struct pipe_buffer *result;
-
-   result = screen->surface_buffer_create(screen,
-                                          width,
-                                          height,
-                                          format,
-                                          usage,
-                                          tex_usage,
-                                          stride);
-
-   if (result)
-      return identity_buffer_create(id_screen, result);
-   return NULL;
-}
 
 static void *
 identity_screen_buffer_map(struct pipe_screen *_screen,
@@ -495,7 +483,8 @@ identity_screen_create(struct pipe_screen *screen)
    id_screen->base.is_format_supported = identity_screen_is_format_supported;
    id_screen->base.context_create = identity_screen_context_create;
    id_screen->base.texture_create = identity_screen_texture_create;
-   id_screen->base.texture_blanket = identity_screen_texture_blanket;
+   id_screen->base.texture_from_handle = identity_screen_texture_from_handle;
+   id_screen->base.texture_get_handle = identity_screen_texture_get_handle;
    id_screen->base.texture_destroy = identity_screen_texture_destroy;
    id_screen->base.get_tex_surface = identity_screen_get_tex_surface;
    id_screen->base.tex_surface_destroy = identity_screen_tex_surface_destroy;
@@ -505,7 +494,6 @@ identity_screen_create(struct pipe_screen *screen)
    id_screen->base.transfer_unmap = identity_screen_transfer_unmap;
    id_screen->base.buffer_create = identity_screen_buffer_create;
    id_screen->base.user_buffer_create = identity_screen_user_buffer_create;
-   id_screen->base.surface_buffer_create = identity_screen_surface_buffer_create;
    if (screen->buffer_map)
       id_screen->base.buffer_map = identity_screen_buffer_map;
    if (screen->buffer_map_range)
diff --git a/src/gallium/drivers/llvmpipe/lp_jit.c b/src/gallium/drivers/llvmpipe/lp_jit.c
index b666ffc8043..08c8f93794d 100644
--- a/src/gallium/drivers/llvmpipe/lp_jit.c
+++ b/src/gallium/drivers/llvmpipe/lp_jit.c
@@ -51,12 +51,16 @@ lp_jit_init_globals(struct llvmpipe_screen *screen)
 
    /* struct lp_jit_texture */
    {
-      LLVMTypeRef elem_types[4];
+      LLVMTypeRef elem_types[6];
 
       elem_types[LP_JIT_TEXTURE_WIDTH]  = LLVMInt32Type();
       elem_types[LP_JIT_TEXTURE_HEIGHT] = LLVMInt32Type();
+      elem_types[LP_JIT_TEXTURE_DEPTH] = LLVMInt32Type();
+      elem_types[LP_JIT_TEXTURE_LAST_LEVEL] = LLVMInt32Type();
       elem_types[LP_JIT_TEXTURE_STRIDE] = LLVMInt32Type();
-      elem_types[LP_JIT_TEXTURE_DATA]   = LLVMPointerType(LLVMInt8Type(), 0);
+      elem_types[LP_JIT_TEXTURE_DATA] =
+         LLVMArrayType(LLVMPointerType(LLVMInt8Type(), 0),
+                       LP_MAX_TEXTURE_2D_LEVELS);
 
       texture_type = LLVMStructType(elem_types, Elements(elem_types), 0);
 
@@ -66,6 +70,12 @@ lp_jit_init_globals(struct llvmpipe_screen *screen)
       LP_CHECK_MEMBER_OFFSET(struct lp_jit_texture, height,
                              screen->target, texture_type,
                              LP_JIT_TEXTURE_HEIGHT);
+      LP_CHECK_MEMBER_OFFSET(struct lp_jit_texture, depth,
+                             screen->target, texture_type,
+                             LP_JIT_TEXTURE_DEPTH);
+      LP_CHECK_MEMBER_OFFSET(struct lp_jit_texture, last_level,
+                             screen->target, texture_type,
+                             LP_JIT_TEXTURE_LAST_LEVEL);
       LP_CHECK_MEMBER_OFFSET(struct lp_jit_texture, stride,
                              screen->target, texture_type,
                              LP_JIT_TEXTURE_STRIDE);
diff --git a/src/gallium/drivers/llvmpipe/lp_jit.h b/src/gallium/drivers/llvmpipe/lp_jit.h
index 8df3015d4b4..5cc7a12c03b 100644
--- a/src/gallium/drivers/llvmpipe/lp_jit.h
+++ b/src/gallium/drivers/llvmpipe/lp_jit.h
@@ -39,6 +39,7 @@
 #include "gallivm/lp_bld_struct.h"
 
 #include "pipe/p_state.h"
+#include "lp_texture.h"
 
 
 struct llvmpipe_screen;
@@ -48,14 +49,18 @@ struct lp_jit_texture
 {
    uint32_t width;
    uint32_t height;
+   uint32_t depth;
+   uint32_t last_level;
    uint32_t stride;
-   const void *data;
+   const void *data[LP_MAX_TEXTURE_2D_LEVELS];
 };
 
 
 enum {
    LP_JIT_TEXTURE_WIDTH = 0,
    LP_JIT_TEXTURE_HEIGHT,
+   LP_JIT_TEXTURE_DEPTH,
+   LP_JIT_TEXTURE_LAST_LEVEL,
    LP_JIT_TEXTURE_STRIDE,
    LP_JIT_TEXTURE_DATA
 };
diff --git a/src/gallium/drivers/llvmpipe/lp_rast.c b/src/gallium/drivers/llvmpipe/lp_rast.c
index 0f83eea6853..dd9a8e8856f 100644
--- a/src/gallium/drivers/llvmpipe/lp_rast.c
+++ b/src/gallium/drivers/llvmpipe/lp_rast.c
@@ -846,6 +846,12 @@ void lp_rast_destroy( struct lp_rasterizer *rast )
       pipe_semaphore_signal(&rast->tasks[i].work_ready);
    }
 
+   /* Wait for threads to terminate before cleaning up per-thread data */
+   for (i = 0; i < rast->num_threads; i++) {
+      pipe_thread_wait(rast->threads[i]);
+   }
+
+   /* Clean up per-thread data */
    for (i = 0; i < rast->num_threads; i++) {
       pipe_semaphore_destroy(&rast->tasks[i].work_ready);
       pipe_semaphore_destroy(&rast->tasks[i].work_done);
diff --git a/src/gallium/drivers/llvmpipe/lp_screen.c b/src/gallium/drivers/llvmpipe/lp_screen.c
index 6759e39e015..9bb3043320c 100644
--- a/src/gallium/drivers/llvmpipe/lp_screen.c
+++ b/src/gallium/drivers/llvmpipe/lp_screen.c
@@ -108,11 +108,11 @@ llvmpipe_get_param(struct pipe_screen *screen, int param)
    case PIPE_CAP_TEXTURE_SHADOW_MAP:
       return 1;
    case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
-      return 13; /* max 4Kx4K */
+      return LP_MAX_TEXTURE_2D_LEVELS;
    case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
-      return 9;  /* max 256x256x256 */
+      return LP_MAX_TEXTURE_3D_LEVELS;
    case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
-      return 13; /* max 4Kx4K */
+      return LP_MAX_TEXTURE_2D_LEVELS;
    case PIPE_CAP_TGSI_CONT_SUPPORTED:
       return 1;
    case PIPE_CAP_BLEND_EQUATION_SEPARATE:
@@ -239,18 +239,6 @@ llvmpipe_is_format_supported( struct pipe_screen *_screen,
 }
 
 
-static struct pipe_buffer *
-llvmpipe_surface_buffer_create(struct pipe_screen *screen,
-                               unsigned width, unsigned height,
-                               enum pipe_format format,
-                               unsigned tex_usage,
-                               unsigned usage,
-                               unsigned *stride)
-{
-   /* This function should never be used */
-   assert(0);
-   return NULL;
-}
 
 
 static void
@@ -310,7 +298,6 @@ llvmpipe_create_screen(struct sw_winsys *winsys)
    screen->base.get_paramf = llvmpipe_get_paramf;
    screen->base.is_format_supported = llvmpipe_is_format_supported;
 
-   screen->base.surface_buffer_create = llvmpipe_surface_buffer_create;
    screen->base.context_create = llvmpipe_create_context;
    screen->base.flush_frontbuffer = llvmpipe_flush_frontbuffer;
 
diff --git a/src/gallium/drivers/llvmpipe/lp_setup.c b/src/gallium/drivers/llvmpipe/lp_setup.c
index fe60eb0437e..fbe14924cb1 100644
--- a/src/gallium/drivers/llvmpipe/lp_setup.c
+++ b/src/gallium/drivers/llvmpipe/lp_setup.c
@@ -469,20 +469,28 @@ lp_setup_set_sampler_textures( struct setup_context *setup,
          jit_tex = &setup->fs.current.jit_context.textures[i];
          jit_tex->width = tex->width0;
          jit_tex->height = tex->height0;
+         jit_tex->depth = tex->depth0;
+         jit_tex->last_level = tex->last_level;
          jit_tex->stride = lp_tex->stride[0];
-         if(!lp_tex->dt) {
-            jit_tex->data = lp_tex->data;
+         if (!lp_tex->dt) {
+            /* regular texture - setup array of mipmap level pointers */
+            int j;
+            for (j = 0; j < LP_MAX_TEXTURE_2D_LEVELS; j++) {
+               jit_tex->data[j] =
+                  (ubyte *) lp_tex->data + lp_tex->level_offset[j];
+            }
          }
          else {
+            /* display target texture/surface */
             /*
              * XXX: Where should this be unmapped?
              */
 
             struct llvmpipe_screen *screen = llvmpipe_screen(tex->screen);
             struct sw_winsys *winsys = screen->winsys;
-            jit_tex->data = winsys->displaytarget_map(winsys, lp_tex->dt,
+            jit_tex->data[0] = winsys->displaytarget_map(winsys, lp_tex->dt,
                                                       PIPE_BUFFER_USAGE_CPU_READ);
-            assert(jit_tex->data);
+            assert(jit_tex->data[0]);
          }
 
          /* the scene references this texture */
diff --git a/src/gallium/drivers/llvmpipe/lp_setup_vbuf.c b/src/gallium/drivers/llvmpipe/lp_setup_vbuf.c
index 24291da91e4..671e74465c0 100644
--- a/src/gallium/drivers/llvmpipe/lp_setup_vbuf.c
+++ b/src/gallium/drivers/llvmpipe/lp_setup_vbuf.c
@@ -231,57 +231,29 @@ lp_setup_draw(struct vbuf_render *vbr, const ushort *indices, uint nr)
       break;
 
    case PIPE_PRIM_QUADS:
-      if (setup->flatshade_first) {
-         for (i = 3; i < nr; i += 4) {
-            setup->triangle( setup,
-                             get_vert(vertex_buffer, indices[i-2], stride),
-                             get_vert(vertex_buffer, indices[i-1], stride),
-                             get_vert(vertex_buffer, indices[i-3], stride) );
-            setup->triangle( setup,
-                             get_vert(vertex_buffer, indices[i-1], stride),
-                             get_vert(vertex_buffer, indices[i-0], stride),
-                             get_vert(vertex_buffer, indices[i-3], stride) );
-         }
-      }
-      else {
-         for (i = 3; i < nr; i += 4) {
-            setup->triangle( setup,
-                             get_vert(vertex_buffer, indices[i-3], stride),
-                             get_vert(vertex_buffer, indices[i-2], stride),
-                             get_vert(vertex_buffer, indices[i-0], stride) );
+      for (i = 3; i < nr; i += 4) {
+         setup->triangle( setup,
+                          get_vert(vertex_buffer, indices[i-3], stride),
+                          get_vert(vertex_buffer, indices[i-2], stride),
+                          get_vert(vertex_buffer, indices[i-0], stride) );
 
-            setup->triangle( setup,
-                             get_vert(vertex_buffer, indices[i-2], stride),
-                             get_vert(vertex_buffer, indices[i-1], stride),
-                             get_vert(vertex_buffer, indices[i-0], stride) );
-         }
+         setup->triangle( setup,
+                          get_vert(vertex_buffer, indices[i-2], stride),
+                          get_vert(vertex_buffer, indices[i-1], stride),
+                          get_vert(vertex_buffer, indices[i-0], stride) );
       }
       break;
 
    case PIPE_PRIM_QUAD_STRIP:
-      if (setup->flatshade_first) {
-         for (i = 3; i < nr; i += 2) {
-            setup->triangle( setup,
-                             get_vert(vertex_buffer, indices[i-0], stride),
-                             get_vert(vertex_buffer, indices[i-1], stride),
-                             get_vert(vertex_buffer, indices[i-3], stride));
-            setup->triangle( setup,
-                             get_vert(vertex_buffer, indices[i-2], stride),
-                             get_vert(vertex_buffer, indices[i-0], stride),
-                             get_vert(vertex_buffer, indices[i-3], stride) );
-         }
-      }
-      else {
-         for (i = 3; i < nr; i += 2) {
-            setup->triangle( setup,
-                             get_vert(vertex_buffer, indices[i-3], stride),
-                             get_vert(vertex_buffer, indices[i-2], stride),
-                             get_vert(vertex_buffer, indices[i-0], stride) );
-            setup->triangle( setup,
-                             get_vert(vertex_buffer, indices[i-1], stride),
-                             get_vert(vertex_buffer, indices[i-3], stride),
-                             get_vert(vertex_buffer, indices[i-0], stride) );
-         }
+      for (i = 3; i < nr; i += 2) {
+         setup->triangle( setup,
+                          get_vert(vertex_buffer, indices[i-3], stride),
+                          get_vert(vertex_buffer, indices[i-2], stride),
+                          get_vert(vertex_buffer, indices[i-0], stride) );
+         setup->triangle( setup,
+                          get_vert(vertex_buffer, indices[i-1], stride),
+                          get_vert(vertex_buffer, indices[i-3], stride),
+                          get_vert(vertex_buffer, indices[i-0], stride) );
       }
       break;
 
@@ -415,57 +387,28 @@ lp_setup_draw_arrays(struct vbuf_render *vbr, uint start, uint nr)
       break;
 
    case PIPE_PRIM_QUADS:
-      if (setup->flatshade_first) {
-         for (i = 3; i < nr; i += 4) {
-            setup->triangle( setup,
-                             get_vert(vertex_buffer, i-2, stride),
-                             get_vert(vertex_buffer, i-1, stride),
-                             get_vert(vertex_buffer, i-3, stride) );
-            setup->triangle( setup,
-                             get_vert(vertex_buffer, i-1, stride),
-                             get_vert(vertex_buffer, i-0, stride),
-                             get_vert(vertex_buffer, i-3, stride) );
-         }
-      }
-      else {
-         for (i = 3; i < nr; i += 4) {
-            setup->triangle( setup,
-                             get_vert(vertex_buffer, i-3, stride),
-                             get_vert(vertex_buffer, i-2, stride),
-                             get_vert(vertex_buffer, i-0, stride) );
-            setup->triangle( setup,
-                             get_vert(vertex_buffer, i-2, stride),
-                             get_vert(vertex_buffer, i-1, stride),
-                             get_vert(vertex_buffer, i-0, stride) );
-         }
+      for (i = 3; i < nr; i += 4) {
+         setup->triangle( setup,
+                          get_vert(vertex_buffer, i-3, stride),
+                          get_vert(vertex_buffer, i-2, stride),
+                          get_vert(vertex_buffer, i-0, stride) );
+         setup->triangle( setup,
+                          get_vert(vertex_buffer, i-2, stride),
+                          get_vert(vertex_buffer, i-1, stride),
+                          get_vert(vertex_buffer, i-0, stride) );
       }
       break;
 
    case PIPE_PRIM_QUAD_STRIP:
-      if (setup->flatshade_first) {
-         for (i = 3; i < nr; i += 2) {
-            setup->triangle( setup,
-                             get_vert(vertex_buffer, i-0, stride),
-                             get_vert(vertex_buffer, i-1, stride),
-                             get_vert(vertex_buffer, i-3, stride) );
-            setup->triangle( setup,
-
-                             get_vert(vertex_buffer, i-2, stride),
-                             get_vert(vertex_buffer, i-0, stride),
-                             get_vert(vertex_buffer, i-3, stride) );
-         }
-      }
-      else {
-         for (i = 3; i < nr; i += 2) {
-            setup->triangle( setup,
-                             get_vert(vertex_buffer, i-3, stride),
-                             get_vert(vertex_buffer, i-2, stride),
-                             get_vert(vertex_buffer, i-0, stride) );
-            setup->triangle( setup,
-                             get_vert(vertex_buffer, i-1, stride),
-                             get_vert(vertex_buffer, i-3, stride),
-                             get_vert(vertex_buffer, i-0, stride) );
-         }
+      for (i = 3; i < nr; i += 2) {
+         setup->triangle( setup,
+                          get_vert(vertex_buffer, i-3, stride),
+                          get_vert(vertex_buffer, i-2, stride),
+                          get_vert(vertex_buffer, i-0, stride) );
+         setup->triangle( setup,
+                          get_vert(vertex_buffer, i-1, stride),
+                          get_vert(vertex_buffer, i-3, stride),
+                          get_vert(vertex_buffer, i-0, stride) );
       }
       break;
 
diff --git a/src/gallium/drivers/llvmpipe/lp_tex_sample_llvm.c b/src/gallium/drivers/llvmpipe/lp_tex_sample_llvm.c
index 2533275dc18..5a3cf37d6d8 100644
--- a/src/gallium/drivers/llvmpipe/lp_tex_sample_llvm.c
+++ b/src/gallium/drivers/llvmpipe/lp_tex_sample_llvm.c
@@ -51,10 +51,11 @@
 
 
 /**
- * This provides the bridge between the sampler state store in lp_jit_context
- * and lp_jit_texture and the sampler code generator. It provides the
- * texture layout information required by the texture sampler code generator
- * in terms of the state stored in lp_jit_context and lp_jit_texture in runtime.
+ * This provides the bridge between the sampler state store in
+ * lp_jit_context and lp_jit_texture and the sampler code
+ * generator. It provides the texture layout information required by
+ * the texture sampler code generator in terms of the state stored in
+ * lp_jit_context and lp_jit_texture in runtime.
  */
 struct llvmpipe_sampler_dynamic_state
 {
@@ -79,6 +80,9 @@ struct lp_llvm_sampler_soa
 
 /**
  * Fetch the specified member of the lp_jit_texture structure.
+ * \param emit_load  if TRUE, emit the LLVM load instruction to actually
+ *                   fetch the field's value.  Otherwise, just emit the
+ *                   GEP code to address the field.
  *
  * @sa http://llvm.org/docs/GetElementPtr.html
  */
@@ -87,9 +91,11 @@ lp_llvm_texture_member(struct lp_sampler_dynamic_state *base,
                        LLVMBuilderRef builder,
                        unsigned unit,
                        unsigned member_index,
-                       const char *member_name)
+                       const char *member_name,
+                       boolean emit_load)
 {
-   struct llvmpipe_sampler_dynamic_state *state = (struct llvmpipe_sampler_dynamic_state *)base;
+   struct llvmpipe_sampler_dynamic_state *state =
+      (struct llvmpipe_sampler_dynamic_state *)base;
    LLVMValueRef indices[4];
    LLVMValueRef ptr;
    LLVMValueRef res;
@@ -107,7 +113,10 @@ lp_llvm_texture_member(struct lp_sampler_dynamic_state *base,
 
    ptr = LLVMBuildGEP(builder, state->context_ptr, indices, Elements(indices), "");
 
-   res = LLVMBuildLoad(builder, ptr, "");
+   if (emit_load)
+      res = LLVMBuildLoad(builder, ptr, "");
+   else
+      res = ptr;
 
    lp_build_name(res, "context.texture%u.%s", unit, member_name);
 
@@ -116,26 +125,30 @@ lp_llvm_texture_member(struct lp_sampler_dynamic_state *base,
 
 
 /**
- * Helper macro to instantiate the functions that generate the code to fetch
- * the members of lp_jit_texture to fulfill the sampler code generator requests.
+ * Helper macro to instantiate the functions that generate the code to
+ * fetch the members of lp_jit_texture to fulfill the sampler code
+ * generator requests.
  *
- * This complexity is the price we have to pay to keep the texture sampler code
- * generator a reusable module without dependencies to llvmpipe internals.
+ * This complexity is the price we have to pay to keep the texture
+ * sampler code generator a reusable module without dependencies to
+ * llvmpipe internals.
  */
-#define LP_LLVM_TEXTURE_MEMBER(_name, _index) \
+#define LP_LLVM_TEXTURE_MEMBER(_name, _index, _emit_load)  \
    static LLVMValueRef \
    lp_llvm_texture_##_name( struct lp_sampler_dynamic_state *base, \
                             LLVMBuilderRef builder, \
                             unsigned unit) \
    { \
-      return lp_llvm_texture_member(base, builder, unit, _index, #_name ); \
+      return lp_llvm_texture_member(base, builder, unit, _index, #_name, _emit_load ); \
    }
 
 
-LP_LLVM_TEXTURE_MEMBER(width,    LP_JIT_TEXTURE_WIDTH)
-LP_LLVM_TEXTURE_MEMBER(height,   LP_JIT_TEXTURE_HEIGHT)
-LP_LLVM_TEXTURE_MEMBER(stride,   LP_JIT_TEXTURE_STRIDE)
-LP_LLVM_TEXTURE_MEMBER(data_ptr, LP_JIT_TEXTURE_DATA)
+LP_LLVM_TEXTURE_MEMBER(width,      LP_JIT_TEXTURE_WIDTH, TRUE)
+LP_LLVM_TEXTURE_MEMBER(height,     LP_JIT_TEXTURE_HEIGHT, TRUE)
+LP_LLVM_TEXTURE_MEMBER(depth,      LP_JIT_TEXTURE_DEPTH, TRUE)
+LP_LLVM_TEXTURE_MEMBER(last_level, LP_JIT_TEXTURE_LAST_LEVEL, TRUE)
+LP_LLVM_TEXTURE_MEMBER(stride,     LP_JIT_TEXTURE_STRIDE, TRUE)
+LP_LLVM_TEXTURE_MEMBER(data_ptr,   LP_JIT_TEXTURE_DATA, FALSE)
 
 
 static void
@@ -145,6 +158,10 @@ lp_llvm_sampler_soa_destroy(struct lp_build_sampler_soa *sampler)
 }
 
 
+/**
+ * Fetch filtered values from texture.
+ * The 'texel' parameter returns four vectors corresponding to R, G, B, A.
+ */
 static void
 lp_llvm_sampler_soa_emit_fetch_texel(struct lp_build_sampler_soa *base,
                                      LLVMBuilderRef builder,
@@ -185,6 +202,8 @@ lp_llvm_sampler_soa_create(const struct lp_sampler_static_state *static_state,
    sampler->base.emit_fetch_texel = lp_llvm_sampler_soa_emit_fetch_texel;
    sampler->dynamic_state.base.width = lp_llvm_texture_width;
    sampler->dynamic_state.base.height = lp_llvm_texture_height;
+   sampler->dynamic_state.base.depth = lp_llvm_texture_depth;
+   sampler->dynamic_state.base.last_level = lp_llvm_texture_last_level;
    sampler->dynamic_state.base.stride = lp_llvm_texture_stride;
    sampler->dynamic_state.base.data_ptr = lp_llvm_texture_data_ptr;
    sampler->dynamic_state.static_state = static_state;
diff --git a/src/gallium/drivers/llvmpipe/lp_texture.c b/src/gallium/drivers/llvmpipe/lp_texture.c
index ac3d62dc967..74b7b4ec5e1 100644
--- a/src/gallium/drivers/llvmpipe/lp_texture.c
+++ b/src/gallium/drivers/llvmpipe/lp_texture.c
@@ -125,7 +125,8 @@ llvmpipe_texture_create(struct pipe_screen *_screen,
    lpt->base.screen = &screen->base;
 
    if (lpt->base.tex_usage & (PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
-                              PIPE_TEXTURE_USAGE_PRIMARY)) {
+                              PIPE_TEXTURE_USAGE_SCANOUT |
+                              PIPE_TEXTURE_USAGE_SHARED)) {
       if (!llvmpipe_displaytarget_layout(screen, lpt))
          goto fail;
    }
@@ -142,43 +143,6 @@ llvmpipe_texture_create(struct pipe_screen *_screen,
 }
 
 
-static struct pipe_texture *
-llvmpipe_texture_blanket(struct pipe_screen * screen,
-                         const struct pipe_texture *base,
-                         const unsigned *stride,
-                         struct pipe_buffer *buffer)
-{
-   /* FIXME */
-#if 0
-   struct llvmpipe_texture *lpt;
-   assert(screen);
-
-   /* Only supports one type */
-   if (base->target != PIPE_TEXTURE_2D ||
-       base->last_level != 0 ||
-       base->depth0 != 1) {
-      return NULL;
-   }
-
-   lpt = CALLOC_STRUCT(llvmpipe_texture);
-   if (!lpt)
-      return NULL;
-
-   lpt->base = *base;
-   pipe_reference_init(&lpt->base.reference, 1);
-   lpt->base.screen = screen;
-   lpt->stride[0] = stride[0];
-
-   pipe_buffer_reference(&lpt->buffer, buffer);
-
-   return &lpt->base;
-#else
-   debug_printf("llvmpipe_texture_blanket() not implemented!");
-   return NULL;
-#endif
-}
-
-
 static void
 llvmpipe_texture_destroy(struct pipe_texture *pt)
 {
@@ -408,7 +372,6 @@ void
 llvmpipe_init_screen_texture_funcs(struct pipe_screen *screen)
 {
    screen->texture_create = llvmpipe_texture_create;
-   screen->texture_blanket = llvmpipe_texture_blanket;
    screen->texture_destroy = llvmpipe_texture_destroy;
 
    screen->get_tex_surface = llvmpipe_get_tex_surface;
diff --git a/src/gallium/drivers/llvmpipe/lp_texture.h b/src/gallium/drivers/llvmpipe/lp_texture.h
index 6a6d9f9222f..b23f929b167 100644
--- a/src/gallium/drivers/llvmpipe/lp_texture.h
+++ b/src/gallium/drivers/llvmpipe/lp_texture.h
@@ -32,6 +32,10 @@
 #include "pipe/p_state.h"
 
 
+#define LP_MAX_TEXTURE_2D_LEVELS 13  /* 4K x 4K for now */
+#define LP_MAX_TEXTURE_3D_LEVELS 10  /* 512 x 512 x 512 for now */
+
+
 struct pipe_context;
 struct pipe_screen;
 struct llvmpipe_context;
@@ -43,8 +47,8 @@ struct llvmpipe_texture
 {
    struct pipe_texture base;
 
-   unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS];
-   unsigned stride[PIPE_MAX_TEXTURE_LEVELS];
+   unsigned long level_offset[LP_MAX_TEXTURE_2D_LEVELS];
+   unsigned stride[LP_MAX_TEXTURE_2D_LEVELS];
 
    /**
     * Display target, for textures with the PIPE_TEXTURE_USAGE_DISPLAY_TARGET
diff --git a/src/gallium/drivers/nouveau/nouveau_screen.c b/src/gallium/drivers/nouveau/nouveau_screen.c
index 3c2f771b51e..f7d10a591f6 100644
--- a/src/gallium/drivers/nouveau/nouveau_screen.c
+++ b/src/gallium/drivers/nouveau/nouveau_screen.c
@@ -4,6 +4,7 @@
 
 #include "util/u_memory.h"
 #include "util/u_inlines.h"
+#include "util/u_format.h"
 
 #include <stdio.h>
 #include <errno.h>
@@ -12,6 +13,9 @@
 #include "nouveau_winsys.h"
 #include "nouveau_screen.h"
 
+/* XXX this should go away */
+#include "state_tracker/drm_api.h"
+
 static const char *
 nouveau_screen_get_name(struct pipe_screen *pscreen)
 {
@@ -231,6 +235,72 @@ nouveau_screen_fence_finish(struct pipe_screen *screen,
 	return 0;
 }
 
+
+/*
+ * Both texture_{from|get}_handle use drm api defines directly which they
+ * shouldn't do. The problem is that from|get are pipe functions and as
+ * such they should be defined in the pipe level. If nouveau had a propper
+ * winsys interface we would have added from|get to that interface using
+ * the winsys_handle struct as done with other drivers. However this code
+ * calls directly into the libdrm_nouveau.so functions (nouveau_bo_*). So
+ * we need to translate the handle into something they understand.
+ */
+static struct pipe_texture *
+nouveau_screen_texture_from_handle(struct pipe_screen *pscreen,
+				   const struct pipe_texture *templ,
+				   struct winsys_handle *whandle)
+{
+	struct nouveau_device *dev = nouveau_screen(pscreen)->device;
+	struct pipe_texture *pt;
+	struct pipe_buffer *pb;
+	int ret;
+
+	pb = CALLOC(1, sizeof(struct pipe_buffer) + sizeof(struct nouveau_bo*));
+	if (!pb)
+		return NULL;
+
+	ret = nouveau_bo_handle_ref(dev, whandle->handle, (struct nouveau_bo**)(pb+1));
+	if (ret) {
+		debug_printf("%s: ref name 0x%08x failed with %d\n",
+			     __func__, whandle->handle, ret);
+		FREE(pb);
+		return NULL;
+	}
+
+	pipe_reference_init(&pb->reference, 1);
+	pb->screen = pscreen;
+	pb->alignment = 0;
+	pb->usage = PIPE_BUFFER_USAGE_GPU_READ_WRITE |
+		    PIPE_BUFFER_USAGE_CPU_READ_WRITE;
+	pb->size = nouveau_bo(pb)->size;
+	pt = nouveau_screen(pscreen)->texture_blanket(pscreen, templ,
+                                                      &whandle->stride, pb);
+	pipe_buffer_reference(&pb, NULL);
+	return pt;
+}
+
+static boolean
+nouveau_screen_texture_get_handle(struct pipe_screen *pscreen,
+				  struct pipe_texture *pt,
+				  struct winsys_handle *whandle)
+{
+	struct nouveau_miptree *mt = nouveau_miptree(pt);
+
+	if (!mt || !mt->bo)
+		return false;
+
+	whandle->stride = util_format_get_stride(mt->base.format, mt->base.width0);
+
+	if (whandle->type == DRM_API_HANDLE_TYPE_SHARED) { 
+		return nouveau_bo_handle_get(mt->bo, &whandle->handle) == 0;
+	} else if (whandle->type == DRM_API_HANDLE_TYPE_KMS) {
+		whandle->handle = mt->bo->handle;
+		return TRUE;
+	} else {
+		return FALSE;
+	}
+}
+
 int
 nouveau_screen_init(struct nouveau_screen *screen, struct nouveau_device *dev)
 {
@@ -258,6 +328,9 @@ nouveau_screen_init(struct nouveau_screen *screen, struct nouveau_device *dev)
 	pscreen->fence_signalled = nouveau_screen_fence_signalled;
 	pscreen->fence_finish = nouveau_screen_fence_finish;
 
+	pscreen->texture_from_handle = nouveau_screen_texture_from_handle;
+	pscreen->texture_get_handle = nouveau_screen_texture_get_handle;
+
 	return 0;
 }
 
diff --git a/src/gallium/drivers/nouveau/nouveau_screen.h b/src/gallium/drivers/nouveau/nouveau_screen.h
index a7927d88dfc..f4a7a2bc234 100644
--- a/src/gallium/drivers/nouveau/nouveau_screen.h
+++ b/src/gallium/drivers/nouveau/nouveau_screen.h
@@ -6,6 +6,18 @@ struct nouveau_screen {
 	struct nouveau_device *device;
 	struct nouveau_channel *channel;
 
+        /**
+         * Create a new texture object, using the given template info, but on top of
+         * existing memory.
+         * 
+         * It is assumed that the buffer data is layed out according to the expected
+         * by the hardware. NULL will be returned if any inconsistency is found.  
+         */
+        struct pipe_texture * (*texture_blanket)(struct pipe_screen *,
+                                                 const struct pipe_texture *templat,
+                                                 const unsigned *stride,
+                                                 struct pipe_buffer *buffer);
+
 	int (*pre_pipebuffer_map_callback) (struct pipe_screen *pscreen,
 		struct pipe_buffer *pb, unsigned usage);
 };
diff --git a/src/gallium/drivers/nv30/nv30_miptree.c b/src/gallium/drivers/nv30/nv30_miptree.c
index 697b1b92ff6..5ef74a832dc 100644
--- a/src/gallium/drivers/nv30/nv30_miptree.c
+++ b/src/gallium/drivers/nv30/nv30_miptree.c
@@ -18,7 +18,7 @@ nv30_miptree_layout(struct nv30_miptree *nv30mt)
 		                           PIPE_TEXTURE_USAGE_DEPTH_STENCIL |
 		                           PIPE_TEXTURE_USAGE_RENDER_TARGET |
 		                           PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
-		                           PIPE_TEXTURE_USAGE_PRIMARY);
+		                           PIPE_TEXTURE_USAGE_SCANOUT);
 
 	if (pt->target == PIPE_TEXTURE_CUBE) {
 		nr_faces = 6;
@@ -78,7 +78,7 @@ nv30_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt)
 	    pt->height0 & (pt->height0 - 1))
 		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
 	else
-	if (pt->tex_usage & (PIPE_TEXTURE_USAGE_PRIMARY |
+	if (pt->tex_usage & (PIPE_TEXTURE_USAGE_SCANOUT |
 	                     PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
 	                     PIPE_TEXTURE_USAGE_DEPTH_STENCIL))
 		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
@@ -232,8 +232,9 @@ void
 nv30_screen_init_miptree_functions(struct pipe_screen *pscreen)
 {
 	pscreen->texture_create = nv30_miptree_create;
-	pscreen->texture_blanket = nv30_miptree_blanket;
 	pscreen->texture_destroy = nv30_miptree_destroy;
 	pscreen->get_tex_surface = nv30_miptree_surface_new;
 	pscreen->tex_surface_destroy = nv30_miptree_surface_del;
+
+	nouveau_screen(pscreen)->texture_blanket = nv50_miptree_blanket;
 }
diff --git a/src/gallium/drivers/nv30/nv30_state.h b/src/gallium/drivers/nv30/nv30_state.h
index e42e872de75..66c26360cb2 100644
--- a/src/gallium/drivers/nv30/nv30_state.h
+++ b/src/gallium/drivers/nv30/nv30_state.h
@@ -70,6 +70,8 @@ struct nv30_fragment_program {
 	struct nouveau_stateobj *so;
 };
 
+#define NV30_MAX_TEXTURE_LEVELS  16
+
 struct nv30_miptree {
 	struct pipe_texture base;
 	struct nouveau_bo *bo;
@@ -80,7 +82,7 @@ struct nv30_miptree {
 	struct {
 		uint pitch;
 		uint *image_offset;
-	} level[PIPE_MAX_TEXTURE_LEVELS];
+	} level[NV30_MAX_TEXTURE_LEVELS];
 };
 
 #endif
diff --git a/src/gallium/drivers/nv40/nv40_miptree.c b/src/gallium/drivers/nv40/nv40_miptree.c
index 85d7e1f1972..62e97bcea48 100644
--- a/src/gallium/drivers/nv40/nv40_miptree.c
+++ b/src/gallium/drivers/nv40/nv40_miptree.c
@@ -20,7 +20,7 @@ nv40_miptree_layout(struct nv40_miptree *mt)
 		                           PIPE_TEXTURE_USAGE_DEPTH_STENCIL |
 		                           PIPE_TEXTURE_USAGE_RENDER_TARGET |
 		                           PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
-		                           PIPE_TEXTURE_USAGE_PRIMARY);
+		                           PIPE_TEXTURE_USAGE_SCANOUT);
 
 	if (pt->target == PIPE_TEXTURE_CUBE) {
 		nr_faces = 6;
@@ -80,7 +80,7 @@ nv40_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *pt)
 	    pt->height0 & (pt->height0 - 1))
 		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
 	else
-	if (pt->tex_usage & (PIPE_TEXTURE_USAGE_PRIMARY |
+	if (pt->tex_usage & (PIPE_TEXTURE_USAGE_SCANOUT |
 	                     PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
 	                     PIPE_TEXTURE_USAGE_DEPTH_STENCIL))
 		mt->base.tex_usage |= NOUVEAU_TEXTURE_USAGE_LINEAR;
@@ -227,9 +227,10 @@ void
 nv40_screen_init_miptree_functions(struct pipe_screen *pscreen)
 {
 	pscreen->texture_create = nv40_miptree_create;
-	pscreen->texture_blanket = nv40_miptree_blanket;
 	pscreen->texture_destroy = nv40_miptree_destroy;
 	pscreen->get_tex_surface = nv40_miptree_surface_new;
 	pscreen->tex_surface_destroy = nv40_miptree_surface_del;
+
+	nouveau_screen(pscreen)->texture_blanket = nv40_miptree_blanket;
 }
 
diff --git a/src/gallium/drivers/nv40/nv40_state.h b/src/gallium/drivers/nv40/nv40_state.h
index 192074e7471..e2e69420eae 100644
--- a/src/gallium/drivers/nv40/nv40_state.h
+++ b/src/gallium/drivers/nv40/nv40_state.h
@@ -73,6 +73,8 @@ struct nv40_fragment_program {
 	struct nouveau_stateobj *so;
 };
 
+#define NV40_MAX_TEXTURE_LEVELS  16
+
 struct nv40_miptree {
 	struct pipe_texture base;
 	struct nouveau_bo *bo;
@@ -83,7 +85,7 @@ struct nv40_miptree {
 	struct {
 		uint pitch;
 		uint *image_offset;
-	} level[PIPE_MAX_TEXTURE_LEVELS];
+	} level[NV40_MAX_TEXTURE_LEVELS];
 };
 
 #endif
diff --git a/src/gallium/drivers/nv50/nv50_context.h b/src/gallium/drivers/nv50/nv50_context.h
index c540594b949..044437e75fa 100644
--- a/src/gallium/drivers/nv50/nv50_context.h
+++ b/src/gallium/drivers/nv50/nv50_context.h
@@ -90,10 +90,12 @@ struct nv50_miptree_level {
 	unsigned tile_mode;
 };
 
+#define NV50_MAX_TEXTURE_LEVELS 16
+
 struct nv50_miptree {
 	struct nouveau_miptree base;
 
-	struct nv50_miptree_level level[PIPE_MAX_TEXTURE_LEVELS];
+	struct nv50_miptree_level level[NV50_MAX_TEXTURE_LEVELS];
 	int image_nr;
 	int total_size;
 };
diff --git a/src/gallium/drivers/nv50/nv50_miptree.c b/src/gallium/drivers/nv50/nv50_miptree.c
index 3f9d869d7a4..e091cae6024 100644
--- a/src/gallium/drivers/nv50/nv50_miptree.c
+++ b/src/gallium/drivers/nv50/nv50_miptree.c
@@ -104,7 +104,7 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp)
 		tile_flags = 0x7400;
 		break;
 	default:
-		if ((pt->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) &&
+		if ((pt->tex_usage & PIPE_TEXTURE_USAGE_SCANOUT) &&
 		    util_format_get_blocksizebits(pt->format) == 32)
 			tile_flags = 0x7a00;
 		else
@@ -255,9 +255,10 @@ void
 nv50_screen_init_miptree_functions(struct pipe_screen *pscreen)
 {
 	pscreen->texture_create = nv50_miptree_create;
-	pscreen->texture_blanket = nv50_miptree_blanket;
 	pscreen->texture_destroy = nv50_miptree_destroy;
 	pscreen->get_tex_surface = nv50_miptree_surface_new;
 	pscreen->tex_surface_destroy = nv50_miptree_surface_del;
+
+	nouveau_screen(pscreen)->texture_blanket = nv50_miptree_blanket;
 }
 
diff --git a/src/gallium/drivers/r300/Makefile b/src/gallium/drivers/r300/Makefile
index 1f69daec819..61b54af4ddf 100644
--- a/src/gallium/drivers/r300/Makefile
+++ b/src/gallium/drivers/r300/Makefile
@@ -19,7 +19,8 @@ C_SOURCES = \
 	r300_state_invariant.c \
 	r300_vs.c \
 	r300_texture.c \
-	r300_tgsi_to_rc.c
+	r300_tgsi_to_rc.c \
+	r300_transfer.c
 
 LIBRARY_INCLUDES = \
 	-I$(TOP)/src/mesa/drivers/dri/r300/compiler \
diff --git a/src/gallium/drivers/r300/SConscript b/src/gallium/drivers/r300/SConscript
index 183aa17f9b3..27b2e309932 100644
--- a/src/gallium/drivers/r300/SConscript
+++ b/src/gallium/drivers/r300/SConscript
@@ -30,6 +30,7 @@ r300 = env.ConvenienceLibrary(
         'r300_vs.c',
         'r300_texture.c',
         'r300_tgsi_to_rc.c',
+        'r300_transfer.c',
     ] + r300compiler) + r300compiler
 
 Export('r300')
diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
index 86b98a4ba52..e0a55323273 100644
--- a/src/gallium/drivers/r300/r300_context.c
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -72,11 +72,8 @@ r300_is_texture_referenced(struct pipe_context *pipe,
                            struct pipe_texture *texture,
                            unsigned face, unsigned level)
 {
-    struct pipe_buffer* buf = 0;
-
-    r300_get_texture_buffer(pipe->screen, texture, &buf, NULL);
-
-    return pipe->is_buffer_referenced(pipe, buf);
+    return pipe->is_buffer_referenced(pipe,
+                                      ((struct r300_texture *)texture)->buffer);
 }
 
 static unsigned int
@@ -86,7 +83,14 @@ r300_is_buffer_referenced(struct pipe_context *pipe,
     /* This only checks to see whether actual hardware buffers are
      * referenced. Since we use managed BOs and transfers, it's actually not
      * possible for pipe_buffers to ever reference the actual hardware, so
-     * buffers are never referenced. */
+     * buffers are never referenced. 
+     */
+
+    /* XXX: that doesn't make sense given that
+     * r300_is_texture_referenced is implemented on top of this
+     * function and hardware can certainly refer to textures
+     * directly...
+     */
     return 0;
 }
 
@@ -159,6 +163,8 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
     if (!r300)
         return NULL;
 
+    r300screen->ctx = (struct pipe_context*)r300;
+
     r300->winsys = radeon_winsys;
 
     r300->context.winsys = (struct pipe_winsys*)radeon_winsys;
@@ -212,7 +218,6 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen,
     r300->invariant_state.dirty = TRUE;
 
     r300->winsys->set_flush_cb(r300->winsys, r300_flush_cb, r300);
-    r300->dirty_state = R300_NEW_KITCHEN_SINK;
     r300->dirty_hw++;
 
     r300->blitter = util_blitter_create(&r300->context);
diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index 0d1518a05bc..c2825d5f267 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -124,13 +124,15 @@ struct r300_texture_format_state {
     uint32_t format2; /* R300_TX_FORMAT2: 0x4500 */
 };
 
+#define R300_MAX_TEXTURE_LEVELS 13
+
 struct r300_texture_fb_state {
     /* Colorbuffer. */
-    uint32_t colorpitch[PIPE_MAX_TEXTURE_LEVELS]; /* R300_RB3D_COLORPITCH[0-3]*/
+    uint32_t colorpitch[R300_MAX_TEXTURE_LEVELS]; /* R300_RB3D_COLORPITCH[0-3]*/
     uint32_t us_out_fmt; /* R300_US_OUT_FMT[0-3] */
 
     /* Zbuffer. */
-    uint32_t depthpitch[PIPE_MAX_TEXTURE_LEVELS]; /* R300_RB3D_DEPTHPITCH */
+    uint32_t depthpitch[R300_MAX_TEXTURE_LEVELS]; /* R300_RB3D_DEPTHPITCH */
     uint32_t zb_format; /* R300_ZB_FORMAT */
 };
 
@@ -236,16 +238,16 @@ struct r300_texture {
     struct pipe_texture tex;
 
     /* Offsets into the buffer. */
-    unsigned offset[PIPE_MAX_TEXTURE_LEVELS];
+    unsigned offset[R300_MAX_TEXTURE_LEVELS];
 
     /* A pitch for each mip-level */
-    unsigned pitch[PIPE_MAX_TEXTURE_LEVELS];
+    unsigned pitch[R300_MAX_TEXTURE_LEVELS];
 
     /* Size of one zslice or face based on the texture target */
-    unsigned layer_size[PIPE_MAX_TEXTURE_LEVELS];
+    unsigned layer_size[R300_MAX_TEXTURE_LEVELS];
 
     /* Whether the mipmap level is macrotiled. */
-    enum r300_buffer_tiling mip_macrotile[PIPE_MAX_TEXTURE_LEVELS];
+    enum r300_buffer_tiling mip_macrotile[R300_MAX_TEXTURE_LEVELS];
 
     /**
      * If non-zero, override the natural texture layout with
@@ -368,6 +370,8 @@ struct r300_context {
     boolean polygon_offset_enabled;
     /* Z buffer bit depth. */
     uint32_t zbuffer_bpp;
+    /* Whether scissor is enabled. */
+    boolean scissor_enabled;
 };
 
 /* Convenience cast wrapper. */
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index addb28bded3..51fc590e5d9 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -675,7 +675,7 @@ void r300_emit_scissor_state(struct r300_context* r300,
     maxx = fb->width;
     maxy = fb->height;
 
-    if (((struct r300_rs_state*)r300->rs_state.state)->rs.scissor) {
+    if (r300->scissor_enabled) {
         minx = MAX2(minx, scissor->minx);
         miny = MAX2(miny, scissor->miny);
         maxx = MIN2(maxx, scissor->maxx);
@@ -794,6 +794,30 @@ void r300_emit_aos(struct r300_context* r300, unsigned offset)
     END_CS;
 }
 
+void r300_emit_vertex_buffer(struct r300_context* r300)
+{
+    CS_LOCALS(r300);
+
+    DBG(r300, DBG_DRAW, "r300: Preparing vertex buffer %p for render, "
+            "vertex size %d\n", r300->vbo,
+            r300->vertex_info.size);
+    /* Set the pointer to our vertex buffer. The emitted values are this:
+     * PACKET3 [3D_LOAD_VBPNTR]
+     * COUNT   [1]
+     * FORMAT  [size | stride << 8]
+     * OFFSET  [offset into BO]
+     * VBPNTR  [relocated BO]
+     */
+    BEGIN_CS(7);
+    OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, 3);
+    OUT_CS(1);
+    OUT_CS(r300->vertex_info.size |
+            (r300->vertex_info.size << 8));
+    OUT_CS(r300->vbo_offset);
+    OUT_CS_RELOC(r300->vbo, 0, RADEON_GEM_DOMAIN_GTT, 0, 0);
+    END_CS;
+}
+
 void r300_emit_vertex_stream_state(struct r300_context* r300,
                                    unsigned size, void* state)
 {
@@ -868,7 +892,7 @@ void r300_emit_vs_state(struct r300_context* r300, unsigned size, void* state)
     CS_LOCALS(r300);
 
     if (!r300screen->caps->has_tcl) {
-        debug_printf("r300: Implementation error: emit_vertex_shader called,"
+        debug_printf("r300: Implementation error: emit_vs_state called,"
                 " but has_tcl is FALSE!\n");
         return;
     }
@@ -907,7 +931,7 @@ void r300_emit_vs_constant_buffer(struct r300_context* r300,
     CS_LOCALS(r300);
 
     if (!r300screen->caps->has_tcl) {
-        debug_printf("r300: Implementation error: emit_vertex_shader called,"
+        debug_printf("r300: Implementation error: emit_vs_constant_buffer called,"
         " but has_tcl is FALSE!\n");
         return;
     }
@@ -1135,8 +1159,10 @@ void r300_emit_dirty_state(struct r300_context* r300)
     assert(r300->dirty_state == 0);
     */
 
-    /* Finally, emit the VBO. */
-    /* r300_emit_vertex_buffer(r300); */
+    /* Emit the VBO for SWTCL. */
+    if (!r300screen->caps->has_tcl) {
+        r300_emit_vertex_buffer(r300);
+    }
 
     r300->dirty_hw++;
 }
diff --git a/src/gallium/drivers/r300/r300_flush.c b/src/gallium/drivers/r300/r300_flush.c
index e37d3092703..70de152713d 100644
--- a/src/gallium/drivers/r300/r300_flush.c
+++ b/src/gallium/drivers/r300/r300_flush.c
@@ -61,6 +61,12 @@ static void r300_flush(struct pipe_context* pipe,
                 atom->dirty = TRUE;
             }
         }
+
+        /* Unmark HWTCL state for SWTCL. */
+        if (!r300_screen(pipe->screen)->caps->has_tcl) {
+            r300->vs_state.dirty = FALSE;
+            r300->dirty_state &= ~R300_NEW_VERTEX_SHADER_CONSTANTS;
+        }
     }
 
     /* reset flushed query */
diff --git a/src/gallium/drivers/r300/r300_fs.c b/src/gallium/drivers/r300/r300_fs.c
index 3c2625269b8..9e71e61c303 100644
--- a/src/gallium/drivers/r300/r300_fs.c
+++ b/src/gallium/drivers/r300/r300_fs.c
@@ -207,6 +207,7 @@ static void r300_translate_fragment_shader(
         DBG(r300, DBG_FP, "r300: Error compiling fragment program: %s\n",
             compiler.Base.ErrorMsg);
         assert(0);
+        abort();
     }
 
     /* And, finally... */
diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c
index 770a92be74f..c8420bcdd5b 100644
--- a/src/gallium/drivers/r300/r300_render.c
+++ b/src/gallium/drivers/r300/r300_render.c
@@ -186,7 +186,7 @@ static void r300_emit_draw_arrays_immediate(struct r300_context *r300,
     dwords = 10 + count * vertex_size;
 
     r300_reserve_cs_space(r300, r300_get_num_dirty_dwords(r300) + dwords);
-    r300_emit_buffer_validate(r300, FALSE, 0);
+    r300_emit_buffer_validate(r300, FALSE, NULL);
     r300_emit_dirty_state(r300);
 
     BEGIN_CS(dwords);
@@ -273,9 +273,14 @@ static void r300_emit_draw_elements(struct r300_context *r300,
     CS_LOCALS(r300);
 
     assert((start * indexSize)  % 4 == 0);
+    assert(count < (1 << 24));
+
+    DBG(r300, DBG_DRAW, "r300: Indexbuf of %u indices, min %u max %u\n",
+        count, minIndex, maxIndex);
+
+    maxIndex = MIN2(maxIndex, r300->vertex_buffer_max_index);
 
     if (alt_num_verts) {
-        assert(count < (1 << 24));
         BEGIN_CS(16);
         OUT_CS_REG(R500_VAP_ALT_NUM_VERTICES, count);
     } else {
@@ -445,7 +450,7 @@ void r300_draw_arrays(struct pipe_context* pipe, unsigned mode,
         /* Make sure there are at least 128 spare dwords in the command buffer.
          * (most of it being consumed by emit_aos) */
         r300_reserve_cs_space(r300, r300_get_num_dirty_dwords(r300) + 128);
-        r300_emit_buffer_validate(r300, TRUE, 0);
+        r300_emit_buffer_validate(r300, TRUE, NULL);
         r300_emit_dirty_state(r300);
 
         if (alt_num_verts || count <= 65535) {
@@ -463,7 +468,7 @@ void r300_draw_arrays(struct pipe_context* pipe, unsigned mode,
                 /* Again, we emit both AOS and draw_arrays so there should be
                  * at least 128 spare dwords. */
                 if (count && r300_reserve_cs_space(r300, 128)) {
-                    r300_emit_buffer_validate(r300, TRUE, 0);
+                    r300_emit_buffer_validate(r300, TRUE, NULL);
                     r300_emit_dirty_state(r300);
                 }
             } while (count);
@@ -683,6 +688,7 @@ static void r300_render_draw_arrays(struct vbuf_render* render,
     CS_LOCALS(r300);
 
     r300_reserve_cs_space(r300, r300_get_num_dirty_dwords(r300) + 2);
+    r300_emit_buffer_validate(r300, FALSE, NULL);
     r300_emit_dirty_state(r300);
 
     DBG(r300, DBG_DRAW, "r300: Doing vbuf render, count %d\n", count);
@@ -706,6 +712,7 @@ static void r300_render_draw(struct vbuf_render* render,
     CS_LOCALS(r300);
 
     r300_reserve_cs_space(r300, r300_get_num_dirty_dwords(r300) + dwords);
+    r300_emit_buffer_validate(r300, FALSE, NULL);
     r300_emit_dirty_state(r300);
 
     BEGIN_CS(dwords);
diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c
index d397a8eb2b4..5880eecd5fe 100644
--- a/src/gallium/drivers/r300/r300_screen.c
+++ b/src/gallium/drivers/r300/r300_screen.c
@@ -1,5 +1,6 @@
 /*
  * Copyright 2008 Corbin Simpson <[email protected]>
+ * Copyright 2010 Marek Olšák <[email protected]>
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -20,14 +21,13 @@
  * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
  * USE OR OTHER DEALINGS IN THE SOFTWARE. */
 
-#include "util/u_inlines.h"
 #include "util/u_format.h"
 #include "util/u_memory.h"
 #include "util/u_simple_screen.h"
 
 #include "r300_context.h"
-#include "r300_screen.h"
 #include "r300_texture.h"
+#include "r300_transfer.h"
 
 #include "radeon_winsys.h"
 #include "r300_winsys.h"
@@ -231,14 +231,16 @@ static boolean r300_is_format_supported(struct pipe_screen* screen,
     /* Check colorbuffer format support. */
     if ((usage & (PIPE_TEXTURE_USAGE_RENDER_TARGET |
                   PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
-                  PIPE_TEXTURE_USAGE_PRIMARY)) &&
+                  PIPE_TEXTURE_USAGE_SCANOUT |
+                  PIPE_TEXTURE_USAGE_SHARED)) &&
         /* 2101010 cannot be rendered to on non-r5xx. */
         (is_r500 || !is_color2101010) &&
         r300_is_colorbuffer_format_supported(format)) {
         retval |= usage &
             (PIPE_TEXTURE_USAGE_RENDER_TARGET |
              PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
-             PIPE_TEXTURE_USAGE_PRIMARY);
+             PIPE_TEXTURE_USAGE_SCANOUT |
+             PIPE_TEXTURE_USAGE_SHARED);
     }
 
     /* Check depth-stencil format support. */
@@ -250,70 +252,6 @@ static boolean r300_is_format_supported(struct pipe_screen* screen,
     return retval == usage;
 }
 
-static struct pipe_transfer*
-r300_get_tex_transfer(struct pipe_screen *screen,
-                      struct pipe_texture *texture,
-                      unsigned face, unsigned level, unsigned zslice,
-                      enum pipe_transfer_usage usage, unsigned x, unsigned y,
-                      unsigned w, unsigned h)
-{
-    struct r300_texture *tex = (struct r300_texture *)texture;
-    struct r300_transfer *trans;
-    struct r300_screen *rscreen = r300_screen(screen);
-    unsigned offset;
-
-    offset = r300_texture_get_offset(tex, level, zslice, face);  /* in bytes */
-
-    trans = CALLOC_STRUCT(r300_transfer);
-    if (trans) {
-        pipe_texture_reference(&trans->transfer.texture, texture);
-        trans->transfer.x = x;
-        trans->transfer.y = y;
-        trans->transfer.width = w;
-        trans->transfer.height = h;
-        trans->transfer.stride = r300_texture_get_stride(rscreen, tex, level);
-        trans->transfer.usage = usage;
-        trans->transfer.zslice = zslice;
-        trans->transfer.face = face;
-
-        trans->offset = offset;
-    }
-    return &trans->transfer;
-}
-
-static void
-r300_tex_transfer_destroy(struct pipe_transfer *trans)
-{
-   pipe_texture_reference(&trans->texture, NULL);
-   FREE(trans);
-}
-
-static void* r300_transfer_map(struct pipe_screen* screen,
-                              struct pipe_transfer* transfer)
-{
-    struct r300_texture* tex = (struct r300_texture*)transfer->texture;
-    char* map;
-    enum pipe_format format = tex->tex.format;
-
-    map = pipe_buffer_map(screen, tex->buffer,
-                          pipe_transfer_buffer_flags(transfer));
-
-    if (!map) {
-        return NULL;
-    }
-
-    return map + r300_transfer(transfer)->offset +
-        transfer->y / util_format_get_blockheight(format) * transfer->stride +
-        transfer->x / util_format_get_blockwidth(format) * util_format_get_blocksize(format);
-}
-
-static void r300_transfer_unmap(struct pipe_screen* screen,
-                                struct pipe_transfer* transfer)
-{
-    struct r300_texture* tex = (struct r300_texture*)transfer->texture;
-    pipe_buffer_unmap(screen, tex->buffer);
-}
-
 static void r300_destroy_screen(struct pipe_screen* pscreen)
 {
     struct r300_screen* r300screen = r300_screen(pscreen);
@@ -350,13 +288,11 @@ struct pipe_screen* r300_create_screen(struct radeon_winsys* radeon_winsys)
     r300screen->screen.get_paramf = r300_get_paramf;
     r300screen->screen.is_format_supported = r300_is_format_supported;
     r300screen->screen.context_create = r300_create_context;
-    r300screen->screen.get_tex_transfer = r300_get_tex_transfer;
-    r300screen->screen.tex_transfer_destroy = r300_tex_transfer_destroy;
-    r300screen->screen.transfer_map = r300_transfer_map;
-    r300screen->screen.transfer_unmap = r300_transfer_unmap;
 
     r300_init_screen_texture_functions(&r300screen->screen);
+    r300_init_screen_transfer_functions(&r300screen->screen);
     u_simple_screen_init(&r300screen->screen);
 
     return &r300screen->screen;
 }
+
diff --git a/src/gallium/drivers/r300/r300_screen.h b/src/gallium/drivers/r300/r300_screen.h
index 502fbfa5a24..484bde6a6be 100644
--- a/src/gallium/drivers/r300/r300_screen.h
+++ b/src/gallium/drivers/r300/r300_screen.h
@@ -1,5 +1,6 @@
 /*
  * Copyright 2008 Corbin Simpson <[email protected]>
+ * Copyright 2010 Marek Olšák <[email protected]>
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -27,6 +28,8 @@
 
 #include "r300_chipset.h"
 
+#define R300_TEXTURE_USAGE_TRANSFER PIPE_TEXTURE_USAGE_CUSTOM
+
 struct radeon_winsys;
 
 struct r300_screen {
@@ -35,6 +38,10 @@ struct r300_screen {
 
     struct radeon_winsys* radeon_winsys;
 
+    /* XXX This hack will be removed once texture transfers become part of
+     * pipe_context. */
+    struct pipe_context* ctx;
+
     /* Chipset capabilities */
     struct r300_capabilities* caps;
 
@@ -42,25 +49,14 @@ struct r300_screen {
     unsigned debug;
 };
 
-struct r300_transfer {
-    /* Parent class */
-    struct pipe_transfer transfer;
-
-    /* Offset from start of buffer. */
-    unsigned offset;
-};
 
 /* Convenience cast wrapper. */
 static INLINE struct r300_screen* r300_screen(struct pipe_screen* screen) {
     return (struct r300_screen*)screen;
 }
 
-/* Convenience cast wrapper. */
-static INLINE struct r300_transfer*
-r300_transfer(struct pipe_transfer* transfer)
-{
-    return (struct r300_transfer*)transfer;
-}
+/* Creates a new r300 screen. */
+struct pipe_screen* r300_create_screen(struct radeon_winsys* radeon_winsys);
 
 /* Debug functionality. */
 
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index 1f6f99d3e52..8c9f6046228 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -571,6 +571,7 @@ static void
 {
     struct r300_context* r300 = r300_context(pipe);
     struct r300_screen* r300screen = r300_screen(pipe->screen);
+    struct pipe_framebuffer_state *old_state = r300->fb_state.state;
     unsigned max_width, max_height;
     uint32_t zbuffer_bpp = 0;
 
@@ -595,23 +596,30 @@ static void
         return;
     }
 
-
     if (r300->draw) {
         draw_flush(r300->draw);
     }
 
-    memcpy(r300->fb_state.state, state, sizeof(struct pipe_framebuffer_state));
+    r300->fb_state.dirty = TRUE;
 
-    r300->fb_state.size = (10 * state->nr_cbufs) + (2 * (4 - state->nr_cbufs)) +
-                          (state->zsbuf ? 10 : 0) + 8;
+    /* If nr_cbufs is changed from zero to non-zero or vice versa... */
+    if (!!old_state->nr_cbufs != !!state->nr_cbufs) {
+        r300->blend_state.dirty = TRUE;
+    }
+    /* If zsbuf is set from NULL to non-NULL or vice versa.. */
+    if (!!old_state->zsbuf != !!state->zsbuf) {
+        r300->dsa_state.dirty = TRUE;
+    }
+    if (!r300->scissor_enabled) {
+        r300->scissor_state.dirty = TRUE;
+    }
 
     r300_fb_update_tiling_flags(r300, r300->fb_state.state, state);
 
-    /* XXX wait what */
-    r300->blend_state.dirty = TRUE;
-    r300->dsa_state.dirty = TRUE;
-    r300->fb_state.dirty = TRUE;
-    r300->scissor_state.dirty = TRUE;
+    memcpy(r300->fb_state.state, state, sizeof(struct pipe_framebuffer_state));
+
+    r300->fb_state.size = (10 * state->nr_cbufs) + (2 * (4 - state->nr_cbufs)) +
+                          (state->zsbuf ? 10 : 0) + 8;
 
     /* Polygon offset depends on the zbuffer bit depth. */
     if (state->zsbuf && r300->polygon_offset_enabled) {
@@ -716,8 +724,7 @@ static void* r300_create_rs_state(struct pipe_context* pipe,
     rs->vap_control_status = R300_VC_32BIT_SWAP;
 #endif
 
-    /* If bypassing TCL, or if no TCL engine is present, turn off the HW TCL.
-     * Else, enable HW TCL and force Draw's TCL off. */
+    /* If no TCL engine is present, turn off the HW TCL. */
     if (!r300screen->caps->has_tcl) {
         rs->vap_control_status |= R300_VAP_TCL_BYPASS;
     }
@@ -807,6 +814,7 @@ static void r300_bind_rs_state(struct pipe_context* pipe, void* state)
 {
     struct r300_context* r300 = r300_context(pipe);
     struct r300_rs_state* rs = (struct r300_rs_state*)state;
+    boolean scissor_was_enabled = r300->scissor_enabled;
 
     if (r300->draw) {
         draw_flush(r300->draw);
@@ -815,20 +823,17 @@ static void r300_bind_rs_state(struct pipe_context* pipe, void* state)
 
     if (rs) {
         r300->polygon_offset_enabled = rs->rs.offset_cw || rs->rs.offset_ccw;
+        r300->scissor_enabled = rs->rs.scissor;
     } else {
         r300->polygon_offset_enabled = FALSE;
+        r300->scissor_enabled = FALSE;
     }
 
     UPDATE_STATE(state, r300->rs_state);
     r300->rs_state.size = 17 + (r300->polygon_offset_enabled ? 5 : 0);
 
-    /* XXX Why is this still needed, dammit!? */
-    r300->scissor_state.dirty = TRUE;
-    r300->viewport_state.dirty = TRUE;
-
-    /* XXX Clean these up when we move to atom emits */
-    if (r300->fs && r300->fs->inputs.wpos != ATTR_UNUSED) {
-        r300->dirty_state |= R300_NEW_FRAGMENT_SHADER_CONSTANTS;
+    if (scissor_was_enabled != r300->scissor_enabled) {
+        r300->scissor_state.dirty = TRUE;
     }
 }
 
@@ -973,7 +978,9 @@ static void r300_set_scissor_state(struct pipe_context* pipe,
     memcpy(r300->scissor_state.state, state,
         sizeof(struct pipe_scissor_state));
 
-    r300->scissor_state.dirty = TRUE;
+    if (r300->scissor_enabled) {
+        r300->scissor_state.dirty = TRUE;
+    }
 }
 
 static void r300_set_viewport_state(struct pipe_context* pipe,
@@ -1088,68 +1095,71 @@ static void* r300_create_vs_state(struct pipe_context* pipe,
 {
     struct r300_context* r300 = r300_context(pipe);
 
-    if (r300_screen(pipe->screen)->caps->has_tcl) {
-        struct r300_vertex_shader* vs = CALLOC_STRUCT(r300_vertex_shader);
-        /* Copy state directly into shader. */
-        vs->state = *shader;
-        vs->state.tokens = tgsi_dup_tokens(shader->tokens);
-
-        tgsi_scan_shader(shader->tokens, &vs->info);
+    struct r300_vertex_shader* vs = CALLOC_STRUCT(r300_vertex_shader);
+    r300_vertex_shader_common_init(vs, shader);
 
-        return (void*)vs;
+    if (r300_screen(pipe->screen)->caps->has_tcl) {
+        r300_translate_vertex_shader(r300, vs);
     } else {
-        return draw_create_vertex_shader(r300->draw, shader);
+        vs->draw_vs = draw_create_vertex_shader(r300->draw, shader);
     }
+
+    return vs;
 }
 
 static void r300_bind_vs_state(struct pipe_context* pipe, void* shader)
 {
     struct r300_context* r300 = r300_context(pipe);
+    struct r300_vertex_shader* vs = (struct r300_vertex_shader*)shader;
 
-    if (r300_screen(pipe->screen)->caps->has_tcl) {
-        struct r300_vertex_shader* vs = (struct r300_vertex_shader*)shader;
+    if (vs == NULL) {
+        r300->vs_state.state = NULL;
+        return;
+    }
+    if (vs == r300->vs_state.state) {
+        return;
+    }
+    r300->vs_state.state = vs;
 
-        if (vs == NULL) {
-            r300->vs_state.state = NULL;
-            return;
-        } else if (!vs->translated) {
-            r300_translate_vertex_shader(r300, vs);
-        }
+    // VS output mapping for HWTCL or stream mapping for SWTCL to the RS block
+    if (r300->fs) {
+        r300_vertex_shader_setup_wpos(r300);
+    }
+    memcpy(r300->vap_output_state.state, &vs->vap_out,
+           sizeof(struct r300_vap_output_state));
+    r300->vap_output_state.dirty = TRUE;
+
+    /* The majority of the RS block bits is dependent on the vertex shader. */
+    r300->rs_block_state.dirty = TRUE; /* Will be updated before the emission. */
 
-        UPDATE_STATE(shader, r300->vs_state);
+    if (r300_screen(pipe->screen)->caps->has_tcl) {
+        r300->vs_state.dirty = TRUE;
         r300->vs_state.size = vs->code.length + 9;
 
-        r300->rs_block_state.dirty = TRUE; /* Will be updated before the emission. */
-        r300->vap_output_state.dirty = TRUE;
-        r300->vertex_stream_state.dirty = TRUE; /* XXX needed for TCL bypass */
         r300->pvs_flush.dirty = TRUE;
 
-        if (r300->fs) {
-            r300_vertex_shader_setup_wpos(r300);
-        }
-
         r300->dirty_state |= R300_NEW_VERTEX_SHADER_CONSTANTS;
     } else {
         draw_flush(r300->draw);
         draw_bind_vertex_shader(r300->draw,
-                (struct draw_vertex_shader*)shader);
+                (struct draw_vertex_shader*)vs->draw_vs);
     }
 }
 
 static void r300_delete_vs_state(struct pipe_context* pipe, void* shader)
 {
     struct r300_context* r300 = r300_context(pipe);
+    struct r300_vertex_shader* vs = (struct r300_vertex_shader*)shader;
 
     if (r300_screen(pipe->screen)->caps->has_tcl) {
-        struct r300_vertex_shader* vs = (struct r300_vertex_shader*)shader;
-
         rc_constants_destroy(&vs->code.constants);
-        FREE((void*)vs->state.tokens);
-        FREE(shader);
     } else {
         draw_delete_vertex_shader(r300->draw,
-                (struct draw_vertex_shader*)shader);
+                (struct draw_vertex_shader*)vs->draw_vs);
     }
+
+    FREE((void*)vs->state.tokens);
+    FREE(shader);
 }
 
 static void r300_set_constant_buffer(struct pipe_context *pipe,
@@ -1201,8 +1211,10 @@ static void r300_set_constant_buffer(struct pipe_context *pipe,
     pipe_buffer_unmap(pipe->screen, buf);
 
     if (shader == PIPE_SHADER_VERTEX) {
-        r300->dirty_state |= R300_NEW_VERTEX_SHADER_CONSTANTS;
-        r300->pvs_flush.dirty = TRUE;
+        if (r300screen->caps->has_tcl) {
+            r300->dirty_state |= R300_NEW_VERTEX_SHADER_CONSTANTS;
+            r300->pvs_flush.dirty = TRUE;
+        }
     }
     else if (shader == PIPE_SHADER_FRAGMENT)
         r300->dirty_state |= R300_NEW_FRAGMENT_SHADER_CONSTANTS;
diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c
index 9c8e907fdf7..e9e40747ef1 100644
--- a/src/gallium/drivers/r300/r300_state_derived.c
+++ b/src/gallium/drivers/r300/r300_state_derived.c
@@ -433,18 +433,13 @@ static void r300_update_derived_shader_state(struct r300_context* r300)
 {
     struct r300_vertex_shader* vs = r300->vs_state.state;
     struct r300_screen* r300screen = r300_screen(r300->context.screen);
-    struct r300_vap_output_state *vap_out =
-        (struct r300_vap_output_state*)r300->vap_output_state.state;
-
-    /* XXX Mmm, delicious hax */
-    memset(&r300->vertex_info, 0, sizeof(struct vertex_info));
-    memcpy(vap_out, vs->hwfmt, sizeof(uint)*4);
 
     r300_update_rs_block(r300, &vs->outputs, &r300->fs->inputs);
 
     if (r300screen->caps->has_tcl) {
         r300_vertex_psc(r300);
     } else {
+        memset(&r300->vertex_info, 0, sizeof(struct vertex_info));
         r300_draw_emit_all_attribs(r300);
         draw_compute_vertex_size(&r300->vertex_info);
         r300_swtcl_vertex_psc(r300);
@@ -582,8 +577,7 @@ static void r300_merge_textures_and_samplers(struct r300_context* r300)
 void r300_update_derived_state(struct r300_context* r300)
 {
     if (r300->rs_block_state.dirty ||
-        r300->vertex_stream_state.dirty || /* XXX put updating this state out of this file */
-        r300->rs_state.dirty) {  /* XXX and remove this one (tcl_bypass dependency) */
+        r300->vertex_stream_state.dirty) { /* XXX put updating PSC out of this file */
         r300_update_derived_shader_state(r300);
     }
 
diff --git a/src/gallium/drivers/r300/r300_state_inlines.h b/src/gallium/drivers/r300/r300_state_inlines.h
index 2f3a56e1fbc..af7827820cc 100644
--- a/src/gallium/drivers/r300/r300_state_inlines.h
+++ b/src/gallium/drivers/r300/r300_state_inlines.h
@@ -453,7 +453,6 @@ r300_translate_vertex_data_type(enum pipe_format format) {
 static INLINE uint16_t
 r300_translate_vertex_data_swizzle(enum pipe_format format) {
     const struct util_format_description *desc = util_format_description(format);
-    unsigned swizzle[4], i;
 
     assert(format);
 
@@ -463,25 +462,10 @@ r300_translate_vertex_data_swizzle(enum pipe_format format) {
         return 0;
     }
 
-    /* Swizzles for 8bits formats are in the reversed order, not sure why. */
-    if (desc->channel[0].size == 8) {
-        for (i = 0; i < 4; i++) {
-            if (desc->swizzle[i] <= 3) {
-                swizzle[i] = 3 - desc->swizzle[i];
-            } else {
-                swizzle[i] = desc->swizzle[i];
-            }
-        }
-    } else {
-        for (i = 0; i < 4; i++) {
-            swizzle[i] = desc->swizzle[i];
-        }
-    }
-
-    return ((swizzle[0] << R300_SWIZZLE_SELECT_X_SHIFT) |
-            (swizzle[1] << R300_SWIZZLE_SELECT_Y_SHIFT) |
-            (swizzle[2] << R300_SWIZZLE_SELECT_Z_SHIFT) |
-            (swizzle[3] << R300_SWIZZLE_SELECT_W_SHIFT) |
+    return ((desc->swizzle[0] << R300_SWIZZLE_SELECT_X_SHIFT) |
+            (desc->swizzle[1] << R300_SWIZZLE_SELECT_Y_SHIFT) |
+            (desc->swizzle[2] << R300_SWIZZLE_SELECT_Z_SHIFT) |
+            (desc->swizzle[3] << R300_SWIZZLE_SELECT_W_SHIFT) |
             (0xf << R300_WRITE_ENA_SHIFT));
 }
 
diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c
index c0144f64b4a..04124afd4dc 100644
--- a/src/gallium/drivers/r300/r300_texture.c
+++ b/src/gallium/drivers/r300/r300_texture.c
@@ -617,18 +617,23 @@ static unsigned r300_texture_get_tile_size(struct r300_texture* tex,
 /* Return true if macrotiling should be enabled on the miplevel. */
 static boolean r300_texture_macro_switch(struct r300_texture *tex,
                                          unsigned level,
-                                         boolean rv350_mode)
+                                         boolean rv350_mode,
+                                         int dim)
 {
-    unsigned tile_width, width;
+    unsigned tile, texdim;
 
-    tile_width = r300_texture_get_tile_size(tex, TILE_WIDTH, TRUE);
-    width = u_minify(tex->tex.width0, level);
+    tile = r300_texture_get_tile_size(tex, dim, TRUE);
+    if (dim == TILE_WIDTH) {
+        texdim = u_minify(tex->tex.width0, level);
+    } else {
+        texdim = u_minify(tex->tex.height0, level);
+    }
 
     /* See TX_FILTER1_n.MACRO_SWITCH. */
     if (rv350_mode) {
-        return width >= tile_width;
+        return texdim >= tile;
     } else {
-        return width > tile_width;
+        return texdim > tile;
     }
 }
 
@@ -692,9 +697,10 @@ static void r300_setup_miptree(struct r300_screen* screen,
 
     for (i = 0; i <= base->last_level; i++) {
         /* Let's see if this miplevel can be macrotiled. */
-        tex->mip_macrotile[i] = (tex->macrotile == R300_BUFFER_TILED &&
-                                 r300_texture_macro_switch(tex, i, rv350_mode)) ?
-                                 R300_BUFFER_TILED : R300_BUFFER_LINEAR;
+        tex->mip_macrotile[i] =
+            (tex->macrotile == R300_BUFFER_TILED &&
+             r300_texture_macro_switch(tex, i, rv350_mode, TILE_WIDTH)) ?
+             R300_BUFFER_TILED : R300_BUFFER_LINEAR;
 
         stride = r300_texture_get_stride(screen, tex, i);
         nblocksy = r300_texture_get_nblocksy(tex, i);
@@ -724,10 +730,46 @@ static void r300_setup_flags(struct r300_texture* tex)
                    !util_is_power_of_two(tex->tex.height0);
 }
 
+static void r300_setup_tiling(struct pipe_screen *screen,
+                              struct r300_texture *tex)
+{
+    enum pipe_format format = tex->tex.format;
+    boolean rv350_mode = r300_screen(screen)->caps->family >= CHIP_FAMILY_RV350;
+
+    if (util_format_is_compressed(format)) {
+        return;
+    }
+
+    if (tex->tex.width0 == 1 ||
+        tex->tex.height0 == 1) {
+        return;
+    }
+
+    /* Set microtiling. */
+    switch (util_format_get_blocksize(format)) {
+        case 1:
+        case 4:
+            tex->microtile = R300_BUFFER_TILED;
+            break;
+
+        /* XXX Square-tiling doesn't work with kernel older than 2.6.34,
+         * XXX need to check the DRM version */
+        /*case 2:
+        case 8:
+            tex->microtile = R300_BUFFER_SQUARETILED;
+            break;*/
+    }
+
+    /* Set macrotiling. */
+    if (r300_texture_macro_switch(tex, 0, rv350_mode, TILE_WIDTH) &&
+        r300_texture_macro_switch(tex, 0, rv350_mode, TILE_HEIGHT)) {
+        tex->macrotile = R300_BUFFER_TILED;
+    }
+}
+
 /* Create a new texture. */
-static struct pipe_texture*
-    r300_texture_create(struct pipe_screen* screen,
-                        const struct pipe_texture* template)
+static struct pipe_texture* r300_texture_create(struct pipe_screen* screen,
+                                         const struct pipe_texture* template)
 {
     struct r300_texture* tex = CALLOC_STRUCT(r300_texture);
     struct r300_screen* rscreen = r300_screen(screen);
@@ -742,6 +784,9 @@ static struct pipe_texture*
     tex->tex.screen = screen;
 
     r300_setup_flags(tex);
+    if (!(template->tex_usage & R300_TEXTURE_USAGE_TRANSFER)) {
+        r300_setup_tiling(screen, tex);
+    }
     r300_setup_miptree(rscreen, tex);
     r300_setup_texture_state(rscreen, tex);
 
@@ -806,14 +851,17 @@ static void r300_tex_surface_destroy(struct pipe_surface* s)
     FREE(s);
 }
 
+
 static struct pipe_texture*
-    r300_texture_blanket(struct pipe_screen* screen,
-                         const struct pipe_texture* base,
-                         const unsigned* stride,
-                         struct pipe_buffer* buffer)
+    r300_texture_from_handle(struct pipe_screen* screen,
+                             const struct pipe_texture* base,
+                             struct winsys_handle *whandle)
 {
-    struct r300_texture* tex;
+    struct radeon_winsys* winsys = (struct radeon_winsys*)screen->winsys;
     struct r300_screen* rscreen = r300_screen(screen);
+    struct pipe_buffer *buffer;
+    struct r300_texture* tex;
+    unsigned stride;
 
     /* Support only 2D textures without mipmaps */
     if (base->target != PIPE_TEXTURE_2D ||
@@ -822,6 +870,11 @@ static struct pipe_texture*
         return NULL;
     }
 
+    buffer = winsys->buffer_from_handle(winsys, screen, whandle, &stride);
+    if (!buffer) {
+        return NULL;
+    }
+
     tex = CALLOC_STRUCT(r300_texture);
     if (!tex) {
         return NULL;
@@ -831,17 +884,38 @@ static struct pipe_texture*
     pipe_reference_init(&tex->tex.reference, 1);
     tex->tex.screen = screen;
 
-    tex->stride_override = *stride;
-    tex->pitch[0] = *stride / util_format_get_blocksize(base->format);
+    tex->stride_override = stride;
+    tex->pitch[0] = stride / util_format_get_blocksize(base->format);
 
     r300_setup_flags(tex);
     r300_setup_texture_state(rscreen, tex);
 
-    pipe_buffer_reference(&tex->buffer, buffer);
+    /* one ref already taken */
+    tex->buffer = buffer;
 
     return (struct pipe_texture*)tex;
 }
 
+static boolean
+    r300_texture_get_handle(struct pipe_screen* screen,
+                            struct pipe_texture *texture,
+                            struct winsys_handle *whandle)
+{
+    struct radeon_winsys* winsys = (struct radeon_winsys*)screen->winsys;
+    struct r300_texture* tex = (struct r300_texture*)texture;
+    unsigned stride;
+
+    if (!tex) {
+        return FALSE;
+    }
+
+    stride = r300_texture_get_stride(r300_screen(screen), tex, 0);
+
+    winsys->buffer_get_handle(winsys, tex->buffer, stride, whandle);
+
+    return TRUE;
+}
+
 static struct pipe_video_surface *
 r300_video_surface_create(struct pipe_screen *screen,
                           enum pipe_video_chroma_format chroma_format,
@@ -893,30 +967,13 @@ static void r300_video_surface_destroy(struct pipe_video_surface *vsfc)
 void r300_init_screen_texture_functions(struct pipe_screen* screen)
 {
     screen->texture_create = r300_texture_create;
+    screen->texture_from_handle = r300_texture_from_handle;
+    screen->texture_get_handle = r300_texture_get_handle;
     screen->texture_destroy = r300_texture_destroy;
     screen->get_tex_surface = r300_get_tex_surface;
     screen->tex_surface_destroy = r300_tex_surface_destroy;
-    screen->texture_blanket = r300_texture_blanket;
 
     screen->video_surface_create = r300_video_surface_create;
     screen->video_surface_destroy= r300_video_surface_destroy;
 }
 
-boolean r300_get_texture_buffer(struct pipe_screen* screen,
-                                struct pipe_texture* texture,
-                                struct pipe_buffer** buffer,
-                                unsigned* stride)
-{
-    struct r300_texture* tex = (struct r300_texture*)texture;
-    if (!tex) {
-        return FALSE;
-    }
-
-    pipe_buffer_reference(buffer, tex->buffer);
-
-    if (stride) {
-        *stride = r300_texture_get_stride(r300_screen(screen), tex, 0);
-    }
-
-    return TRUE;
-}
diff --git a/src/gallium/drivers/r300/r300_texture.h b/src/gallium/drivers/r300/r300_texture.h
index 46a5fb6188b..138b62784e6 100644
--- a/src/gallium/drivers/r300/r300_texture.h
+++ b/src/gallium/drivers/r300/r300_texture.h
@@ -60,13 +60,11 @@ r300_video_surface(struct pipe_video_surface *pvs)
     return (struct r300_video_surface *)pvs;
 }
 
-#ifndef R300_WINSYS_H
-
+/* Used internally for texture_is_referenced()
+ */
 boolean r300_get_texture_buffer(struct pipe_screen* screen,
                                 struct pipe_texture* texture,
                                 struct pipe_buffer** buffer,
                                 unsigned* stride);
 
-#endif /* R300_WINSYS_H */
-
 #endif /* R300_TEXTURE_H */
diff --git a/src/gallium/drivers/r300/r300_transfer.c b/src/gallium/drivers/r300/r300_transfer.c
new file mode 100644
index 00000000000..ec89681a3c0
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_transfer.c
@@ -0,0 +1,270 @@
+/*
+ * Copyright 2008 Corbin Simpson <[email protected]>
+ * Copyright 2010 Marek Olšák <[email protected]>
+ *
+ * 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
+ * on 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
+ * THE AUTHOR(S) AND/OR THEIR 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 "r300_context.h"
+#include "r300_transfer.h"
+#include "r300_texture.h"
+#include "r300_screen.h"
+
+#include "util/u_memory.h"
+#include "util/u_format.h"
+
+struct r300_transfer {
+    /* Parent class */
+    struct pipe_transfer transfer;
+
+    /* Pipe context. */
+    struct pipe_context *ctx;
+
+    /* Parameters of get_tex_transfer. */
+    unsigned x, y, level, zslice, face;
+
+    /* Offset from start of buffer. */
+    unsigned offset;
+
+    /* Detiled texture. */
+    struct r300_texture *detiled_texture;
+
+    /* Transfer and format flags. */
+    unsigned buffer_usage, render_target_usage;
+};
+
+/* Convenience cast wrapper. */
+static INLINE struct r300_transfer*
+r300_transfer(struct pipe_transfer* transfer)
+{
+    return (struct r300_transfer*)transfer;
+}
+
+/* Copy from a tiled texture to a detiled one. */
+static void r300_copy_from_tiled_texture(struct pipe_context *ctx,
+                                         struct r300_transfer *r300transfer)
+{
+    struct pipe_screen *screen = ctx->screen;
+    struct pipe_transfer *transfer = (struct pipe_transfer*)r300transfer;
+    struct pipe_texture *tex = transfer->texture;
+    struct pipe_surface *src, *dst;
+
+    src = screen->get_tex_surface(screen, tex, r300transfer->face,
+                                  r300transfer->level, r300transfer->zslice,
+                                  PIPE_BUFFER_USAGE_GPU_READ |
+                                  PIPE_BUFFER_USAGE_PIXEL);
+
+    dst = screen->get_tex_surface(screen, &r300transfer->detiled_texture->tex,
+                                  0, 0, 0,
+                                  PIPE_BUFFER_USAGE_GPU_WRITE |
+                                  PIPE_BUFFER_USAGE_PIXEL |
+                                  r300transfer->buffer_usage);
+
+    ctx->surface_copy(ctx, dst, 0, 0, src, r300transfer->x, r300transfer->y,
+                      transfer->width, transfer->height);
+
+    pipe_surface_reference(&src, NULL);
+    pipe_surface_reference(&dst, NULL);
+}
+
+/* Copy a detiled texture to a tiled one. */
+static void r300_copy_into_tiled_texture(struct pipe_context *ctx,
+                                         struct r300_transfer *r300transfer)
+{
+    struct pipe_screen *screen = ctx->screen;
+    struct pipe_transfer *transfer = (struct pipe_transfer*)r300transfer;
+    struct pipe_texture *tex = transfer->texture;
+    struct pipe_surface *src, *dst;
+
+    src = screen->get_tex_surface(screen, &r300transfer->detiled_texture->tex,
+                                  0, 0, 0,
+                                  PIPE_BUFFER_USAGE_GPU_READ |
+                                  PIPE_BUFFER_USAGE_PIXEL);
+
+    dst = screen->get_tex_surface(screen, tex, r300transfer->face,
+                                  r300transfer->level, r300transfer->zslice,
+                                  PIPE_BUFFER_USAGE_GPU_WRITE |
+                                  PIPE_BUFFER_USAGE_PIXEL);
+
+    /* XXX this flush prevents the following DRM error from occuring:
+     * [drm:radeon_cs_ioctl] *ERROR* Failed to parse relocation !
+     * Reproducible with perf/copytex. */
+    ctx->flush(ctx, 0, NULL);
+
+    ctx->surface_copy(ctx, dst, r300transfer->x, r300transfer->y, src, 0, 0,
+                      transfer->width, transfer->height);
+
+    /* XXX this flush fixes a few piglit tests (e.g. glean/pixelFormats). */
+    ctx->flush(ctx, 0, NULL);
+
+    pipe_surface_reference(&src, NULL);
+    pipe_surface_reference(&dst, NULL);
+}
+
+static struct pipe_transfer*
+r300_get_tex_transfer(struct pipe_screen *screen,
+                      struct pipe_texture *texture,
+                      unsigned face, unsigned level, unsigned zslice,
+                      enum pipe_transfer_usage usage, unsigned x, unsigned y,
+                      unsigned w, unsigned h)
+{
+    struct r300_texture *tex = (struct r300_texture *)texture;
+    struct r300_transfer *trans;
+    struct r300_screen *r300screen = r300_screen(screen);
+    struct pipe_texture template;
+
+    trans = CALLOC_STRUCT(r300_transfer);
+    if (trans) {
+        /* Initialize the transfer object. */
+        pipe_texture_reference(&trans->transfer.texture, texture);
+        trans->transfer.usage = usage;
+        trans->transfer.width = w;
+        trans->transfer.height = h;
+        trans->ctx = r300screen->ctx;
+        trans->x = x;
+        trans->y = y;
+        trans->level = level;
+        trans->zslice = zslice;
+        trans->face = face;
+
+        /* If the texture is tiled, we must create a temporary detiled texture
+         * for this transfer. */
+        if (tex->microtile || tex->macrotile) {
+            trans->buffer_usage = pipe_transfer_buffer_flags(&trans->transfer);
+            trans->render_target_usage =
+                util_format_is_depth_or_stencil(texture->format) ?
+                PIPE_TEXTURE_USAGE_DEPTH_STENCIL :
+                PIPE_TEXTURE_USAGE_RENDER_TARGET;
+
+            template.target = PIPE_TEXTURE_2D;
+            template.format = texture->format;
+            template.width0 = w;
+            template.height0 = h;
+            template.depth0 = 0;
+            template.last_level = 0;
+            template.nr_samples = 0;
+            template.tex_usage = PIPE_TEXTURE_USAGE_DYNAMIC |
+                                 R300_TEXTURE_USAGE_TRANSFER;
+
+            /* For texture reading, the temporary (detiled) texture is used as
+             * a render target when blitting from a tiled texture. */
+            if (usage & PIPE_TRANSFER_READ) {
+                template.tex_usage |= trans->render_target_usage;
+            }
+            /* For texture writing, the temporary texture is used as a sampler
+             * when blitting into a tiled texture. */
+            if (usage & PIPE_TRANSFER_WRITE) {
+                template.tex_usage |= PIPE_TEXTURE_USAGE_SAMPLER;
+            }
+
+            /* Create the temporary texture. */
+            trans->detiled_texture =
+                (struct r300_texture*)screen->texture_create(screen, &template);
+            assert(!trans->detiled_texture->microtile &&
+                   !trans->detiled_texture->macrotile);
+
+            /* Set the stride.
+             * Parameters x, y, level, zslice, and face remain zero. */
+            trans->transfer.stride =
+                r300_texture_get_stride(r300screen, trans->detiled_texture, 0);
+
+            if (usage & PIPE_TRANSFER_READ) {
+                /* We cannot map a tiled texture directly because the data is
+                 * in a different order, therefore we do detiling using a blit. */
+                r300_copy_from_tiled_texture(r300screen->ctx, trans);
+            }
+        } else {
+            trans->transfer.x = x;
+            trans->transfer.y = y;
+            trans->transfer.stride =
+                r300_texture_get_stride(r300screen, tex, level);
+            trans->transfer.level = level;
+            trans->transfer.zslice = zslice;
+            trans->transfer.face = face;
+            trans->offset = r300_texture_get_offset(tex, level, zslice, face);
+        }
+    }
+    return &trans->transfer;
+}
+
+static void r300_tex_transfer_destroy(struct pipe_transfer *trans)
+{
+    struct r300_transfer *r300transfer = r300_transfer(trans);
+
+    if (r300transfer->detiled_texture) {
+        if (trans->usage & PIPE_TRANSFER_WRITE) {
+            r300_copy_into_tiled_texture(r300transfer->ctx, r300transfer);
+        }
+
+        pipe_texture_reference(
+            (struct pipe_texture**)&r300transfer->detiled_texture, NULL);
+    }
+    pipe_texture_reference(&trans->texture, NULL);
+    FREE(trans);
+}
+
+static void* r300_transfer_map(struct pipe_screen *screen,
+                               struct pipe_transfer *transfer)
+{
+    struct r300_transfer *r300transfer = r300_transfer(transfer);
+    struct r300_texture *tex = (struct r300_texture*)transfer->texture;
+    char *map;
+    enum pipe_format format = tex->tex.format;
+
+    if (r300transfer->detiled_texture) {
+        /* The detiled texture is of the same size as the region being mapped
+         * (no offset needed). */
+        return pipe_buffer_map(screen,
+                               r300transfer->detiled_texture->buffer,
+                               pipe_transfer_buffer_flags(transfer));
+    } else {
+        /* Tiling is disabled. */
+        map = pipe_buffer_map(screen, tex->buffer,
+                              pipe_transfer_buffer_flags(transfer));
+
+        if (!map) {
+            return NULL;
+        }
+
+        return map + r300_transfer(transfer)->offset +
+            transfer->y / util_format_get_blockheight(format) * transfer->stride +
+            transfer->x / util_format_get_blockwidth(format) * util_format_get_blocksize(format);
+    }
+}
+
+static void r300_transfer_unmap(struct pipe_screen *screen,
+                                struct pipe_transfer *transfer)
+{
+    struct r300_transfer *r300transfer = r300_transfer(transfer);
+    struct r300_texture *tex = (struct r300_texture*)transfer->texture;
+
+    if (r300transfer->detiled_texture) {
+        pipe_buffer_unmap(screen, r300transfer->detiled_texture->buffer);
+    } else {
+        pipe_buffer_unmap(screen, tex->buffer);
+    }
+}
+
+void r300_init_screen_transfer_functions(struct pipe_screen *screen)
+{
+    screen->get_tex_transfer = r300_get_tex_transfer;
+    screen->tex_transfer_destroy = r300_tex_transfer_destroy;
+    screen->transfer_map = r300_transfer_map;
+    screen->transfer_unmap = r300_transfer_unmap;
+}
diff --git a/src/gallium/drivers/r300/r300_transfer.h b/src/gallium/drivers/r300/r300_transfer.h
new file mode 100644
index 00000000000..60d1d3dc85c
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_transfer.h
@@ -0,0 +1,31 @@
+/*
+ * Copyright 2008 Corbin Simpson <[email protected]>
+ * Copyright 2010 Marek Olšák <[email protected]>
+ *
+ * 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
+ * on 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
+ * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE. */
+
+#ifndef R300_TRANSFER
+#define R300_TRANSFER
+
+#include "pipe/p_screen.h"
+
+void r300_init_screen_transfer_functions(struct pipe_screen *screen);
+
+#endif
diff --git a/src/gallium/drivers/r300/r300_vs.c b/src/gallium/drivers/r300/r300_vs.c
index 60a04bbfeda..379939ac750 100644
--- a/src/gallium/drivers/r300/r300_vs.c
+++ b/src/gallium/drivers/r300/r300_vs.c
@@ -89,95 +89,41 @@ static void r300_shader_read_vs_outputs(
                 assert(0);
         }
     }
+
+    /* WPOS is a straight copy of POSITION and it's always emitted. */
+    vs_outputs->wpos = i;
 }
 
-static void r300_shader_vap_output_fmt(struct r300_vertex_shader* vs)
+/* This function sets up:
+ * - VAP mapping, which maps VS registers to output semantics and
+ *   at the same time it indicates which attributes are enabled and should
+ *   be rasterized.
+ * - Stream mapping to VS outputs if TCL is not present. */
+static void r300_init_vs_output_mapping(struct r300_vertex_shader* vs)
 {
     struct r300_shader_semantics* vs_outputs = &vs->outputs;
-    uint32_t* hwfmt = vs->hwfmt;
-    int i, gen_count;
+    struct r300_vap_output_state *vap_out = &vs->vap_out;
+    int *stream_loc = vs->stream_loc_notcl;
+    int i, gen_count, tabi = 0;
     boolean any_bcolor_used = vs_outputs->bcolor[0] != ATTR_UNUSED ||
                               vs_outputs->bcolor[1] != ATTR_UNUSED;
 
-    /* Do the actual vertex_info setup.
-     *
-     * vertex_info has four uints of hardware-specific data in it.
-     * vinfo.hwfmt[0] is R300_VAP_VTX_STATE_CNTL
-     * vinfo.hwfmt[1] is R300_VAP_VSM_VTX_ASSM
-     * vinfo.hwfmt[2] is R300_VAP_OUTPUT_VTX_FMT_0
-     * vinfo.hwfmt[3] is R300_VAP_OUTPUT_VTX_FMT_1 */
-
-    hwfmt[0] = 0x5555; /* XXX this is classic Mesa bonghits */
+    vap_out->vap_vtx_state_cntl = 0x5555; /* XXX this is classic Mesa bonghits */
 
     /* Position. */
     if (vs_outputs->pos != ATTR_UNUSED) {
-        hwfmt[1] |= R300_INPUT_CNTL_POS;
-        hwfmt[2] |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT;
+        vap_out->vap_vsm_vtx_assm |= R300_INPUT_CNTL_POS;
+        vap_out->vap_out_vtx_fmt[0] |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT;
+
+        stream_loc[tabi++] = 0;
     } else {
         assert(0);
     }
 
     /* Point size. */
     if (vs_outputs->psize != ATTR_UNUSED) {
-        hwfmt[2] |= R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT;
-    }
-
-    /* Colors. */
-    for (i = 0; i < ATTR_COLOR_COUNT; i++) {
-        if (vs_outputs->color[i] != ATTR_UNUSED || any_bcolor_used ||
-            vs_outputs->color[1] != ATTR_UNUSED) {
-            hwfmt[1] |= R300_INPUT_CNTL_COLOR;
-            hwfmt[2] |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT << i;
-        }
-    }
-
-    /* Back-face colors. */
-    if (any_bcolor_used) {
-        for (i = 0; i < ATTR_COLOR_COUNT; i++) {
-            hwfmt[1] |= R300_INPUT_CNTL_COLOR;
-            hwfmt[2] |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT << (2+i);
-        }
-    }
+        vap_out->vap_out_vtx_fmt[0] |= R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT;
 
-    /* Texture coordinates. */
-    gen_count = 0;
-    for (i = 0; i < ATTR_GENERIC_COUNT; i++) {
-        if (vs_outputs->generic[i] != ATTR_UNUSED) {
-            hwfmt[1] |= (R300_INPUT_CNTL_TC0 << gen_count);
-            hwfmt[3] |= (4 << (3 * gen_count));
-            gen_count++;
-        }
-    }
-
-    /* Fog coordinates. */
-    if (vs_outputs->fog != ATTR_UNUSED) {
-        hwfmt[1] |= (R300_INPUT_CNTL_TC0 << gen_count);
-        hwfmt[3] |= (4 << (3 * gen_count));
-        gen_count++;
-    }
-
-    /* XXX magic */
-    assert(gen_count <= 8);
-
-    /* WPOS. */
-    vs->wpos_tex_output = gen_count;
-}
-
-/* Sets up stream mapping to equivalent VS outputs if TCL is bypassed
- * or isn't present. */
-static void r300_stream_locations_notcl(
-    struct r300_shader_semantics* vs_outputs,
-    int* stream_loc)
-{
-    int i, tabi = 0, gen_count;
-    boolean any_bcolor_used = vs_outputs->bcolor[0] != ATTR_UNUSED ||
-                              vs_outputs->bcolor[1] != ATTR_UNUSED;
-
-    /* Position. */
-    stream_loc[tabi++] = 0;
-
-    /* Point size. */
-    if (vs_outputs->psize != ATTR_UNUSED) {
         stream_loc[tabi++] = 1;
     }
 
@@ -185,6 +131,9 @@ static void r300_stream_locations_notcl(
     for (i = 0; i < ATTR_COLOR_COUNT; i++) {
         if (vs_outputs->color[i] != ATTR_UNUSED || any_bcolor_used ||
             vs_outputs->color[1] != ATTR_UNUSED) {
+            vap_out->vap_vsm_vtx_assm |= R300_INPUT_CNTL_COLOR;
+            vap_out->vap_out_vtx_fmt[0] |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT << i;
+
             stream_loc[tabi++] = 2 + i;
         }
     }
@@ -192,6 +141,9 @@ static void r300_stream_locations_notcl(
     /* Back-face colors. */
     if (any_bcolor_used) {
         for (i = 0; i < ATTR_COLOR_COUNT; i++) {
+            vap_out->vap_vsm_vtx_assm |= R300_INPUT_CNTL_COLOR;
+            vap_out->vap_out_vtx_fmt[0] |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT << (2+i);
+
             stream_loc[tabi++] = 4 + i;
         }
     }
@@ -200,6 +152,9 @@ static void r300_stream_locations_notcl(
     gen_count = 0;
     for (i = 0; i < ATTR_GENERIC_COUNT; i++) {
         if (vs_outputs->generic[i] != ATTR_UNUSED) {
+            vap_out->vap_vsm_vtx_assm |= (R300_INPUT_CNTL_TC0 << gen_count);
+            vap_out->vap_out_vtx_fmt[1] |= (4 << (3 * gen_count));
+
             assert(tabi < 16);
             stream_loc[tabi++] = 6 + gen_count;
             gen_count++;
@@ -208,17 +163,22 @@ static void r300_stream_locations_notcl(
 
     /* Fog coordinates. */
     if (vs_outputs->fog != ATTR_UNUSED) {
+        vap_out->vap_vsm_vtx_assm |= (R300_INPUT_CNTL_TC0 << gen_count);
+        vap_out->vap_out_vtx_fmt[1] |= (4 << (3 * gen_count));
+
         assert(tabi < 16);
         stream_loc[tabi++] = 6 + gen_count;
         gen_count++;
     }
 
+    /* XXX magic */
+    assert(gen_count <= 8);
+
     /* WPOS. */
-    if (vs_outputs->wpos != ATTR_UNUSED) {
-        assert(tabi < 16);
-        stream_loc[tabi++] = 6 + gen_count;
-        gen_count++;
-    }
+    vs->wpos_tex_output = gen_count;
+
+    assert(tabi < 16);
+    stream_loc[tabi++] = 6 + gen_count;
 
     for (; tabi < 16;) {
         stream_loc[tabi++] = -1;
@@ -294,26 +254,16 @@ static void set_vertex_inputs_outputs(struct r300_vertex_program_compiler * c)
     }
 }
 
-static void r300_insert_wpos(struct r300_vertex_program_compiler* c,
-                             struct r300_shader_semantics* outputs)
+void r300_vertex_shader_common_init(struct r300_vertex_shader *vs,
+                                    const struct pipe_shader_state *shader)
 {
-    int i, lastOutput = 0;
-
-    /* Find the max output index. */
-    lastOutput = MAX2(lastOutput, outputs->psize);
-    for (i = 0; i < ATTR_COLOR_COUNT; i++) {
-        lastOutput = MAX2(lastOutput, outputs->color[i]);
-        lastOutput = MAX2(lastOutput, outputs->bcolor[i]);
-    }
-    for (i = 0; i < ATTR_GENERIC_COUNT; i++) {
-        lastOutput = MAX2(lastOutput, outputs->generic[i]);
-    }
-    lastOutput = MAX2(lastOutput, outputs->fog);
+    /* Copy state directly into shader. */
+    vs->state = *shader;
+    vs->state.tokens = tgsi_dup_tokens(shader->tokens);
+    tgsi_scan_shader(shader->tokens, &vs->info);
 
-    /* Set WPOS after the last output. */
-    lastOutput++;
-    rc_copy_output(&c->Base, 0, lastOutput); /* out[lastOutput] = out[0]; */
-    outputs->wpos = lastOutput;
+    r300_shader_read_vs_outputs(&vs->info, &vs->outputs);
+    r300_init_vs_output_mapping(vs);
 }
 
 void r300_translate_vertex_shader(struct r300_context* r300,
@@ -322,9 +272,6 @@ void r300_translate_vertex_shader(struct r300_context* r300,
     struct r300_vertex_program_compiler compiler;
     struct tgsi_to_rc ttr;
 
-    /* Initialize. */
-    r300_shader_read_vs_outputs(&vs->info, &vs->outputs);
-
     /* Setup the compiler */
     rc_init(&compiler.Base);
 
@@ -348,10 +295,7 @@ void r300_translate_vertex_shader(struct r300_context* r300,
     compiler.SetHwInputOutput = &set_vertex_inputs_outputs;
 
     /* Insert the WPOS output. */
-    r300_insert_wpos(&compiler, &vs->outputs);
-
-    r300_shader_vap_output_fmt(vs);
-    r300_stream_locations_notcl(&vs->outputs, vs->stream_loc_notcl);
+    rc_copy_output(&compiler.Base, 0, vs->outputs.wpos);
 
     /* Invoke the compiler */
     r3xx_compile_vertex_program(&compiler);
@@ -363,30 +307,29 @@ void r300_translate_vertex_shader(struct r300_context* r300,
 
     /* And, finally... */
     rc_destroy(&compiler.Base);
-    vs->translated = TRUE;
 }
 
 boolean r300_vertex_shader_setup_wpos(struct r300_context* r300)
 {
     struct r300_vertex_shader* vs = r300->vs_state.state;
+    struct r300_vap_output_state *vap_out = &vs->vap_out;
     int tex_output = vs->wpos_tex_output;
     uint32_t tex_fmt = R300_INPUT_CNTL_TC0 << tex_output;
-    uint32_t* hwfmt = vs->hwfmt;
 
     if (r300->fs->inputs.wpos != ATTR_UNUSED) {
         /* Enable WPOS in VAP. */
-        if (!(hwfmt[1] & tex_fmt)) {
-            hwfmt[1] |= tex_fmt;
-            hwfmt[3] |= (4 << (3 * tex_output));
+        if (!(vap_out->vap_vsm_vtx_assm & tex_fmt)) {
+            vap_out->vap_vsm_vtx_assm |= tex_fmt;
+            vap_out->vap_out_vtx_fmt[1] |= (4 << (3 * tex_output));
 
             assert(tex_output < 8);
             return TRUE;
         }
     } else {
         /* Disable WPOS in VAP. */
-        if (hwfmt[1] & tex_fmt) {
-            hwfmt[1] &= ~tex_fmt;
-            hwfmt[3] &= ~(4 << (3 * tex_output));
+        if (vap_out->vap_vsm_vtx_assm & tex_fmt) {
+            vap_out->vap_vsm_vtx_assm &= ~tex_fmt;
+            vap_out->vap_out_vtx_fmt[1] &= ~(4 << (3 * tex_output));
             return TRUE;
         }
     }
diff --git a/src/gallium/drivers/r300/r300_vs.h b/src/gallium/drivers/r300/r300_vs.h
index 18cfeee3cd4..f6f0b86b683 100644
--- a/src/gallium/drivers/r300/r300_vs.h
+++ b/src/gallium/drivers/r300/r300_vs.h
@@ -28,6 +28,7 @@
 #include "tgsi/tgsi_scan.h"
 #include "radeon_code.h"
 
+#include "r300_context.h"
 #include "r300_shader_semantics.h"
 
 struct r300_context;
@@ -38,7 +39,7 @@ struct r300_vertex_shader {
 
     struct tgsi_shader_info info;
     struct r300_shader_semantics outputs;
-    uint hwfmt[4];
+    struct r300_vap_output_state vap_out;
 
     /* Stream locations for SWTCL or if TCL is bypassed. */
     int stream_loc_notcl[16];
@@ -46,13 +47,17 @@ struct r300_vertex_shader {
     /* Output stream location for WPOS. */
     int wpos_tex_output;
 
-    /* Has this shader been translated yet? */
-    boolean translated;
-
+    /* HWTCL-specific.  */
     /* Machine code (if translated) */
     struct r300_vertex_program_code code;
+
+    /* SWTCL-specific. */
+    void *draw_vs;
 };
 
+void r300_vertex_shader_common_init(struct r300_vertex_shader *vs,
+                                    const struct pipe_shader_state *shader);
+
 void r300_translate_vertex_shader(struct r300_context* r300,
                                   struct r300_vertex_shader* vs);
 
diff --git a/src/gallium/drivers/r300/r300_winsys.h b/src/gallium/drivers/r300/r300_winsys.h
index 40fb8a95ca5..ddf2b790039 100644
--- a/src/gallium/drivers/r300/r300_winsys.h
+++ b/src/gallium/drivers/r300/r300_winsys.h
@@ -40,11 +40,6 @@ struct radeon_winsys;
 struct pipe_screen* r300_create_screen(struct radeon_winsys* radeon_winsys);
 
 
-boolean r300_get_texture_buffer(struct pipe_screen* screen,
-                                struct pipe_texture* texture,
-                                struct pipe_buffer** buffer,
-                                unsigned* stride);
-
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c
index 98c08eaffaf..6749243ab41 100644
--- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c
+++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c
@@ -264,57 +264,29 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr)
       break;
 
    case PIPE_PRIM_QUADS:
-      if (softpipe->rasterizer->flatshade_first) {
-         for (i = 3; i < nr; i += 4) {
-            sp_setup_tri( setup_ctx,
-                       get_vert(vertex_buffer, indices[i-2], stride),
-                       get_vert(vertex_buffer, indices[i-1], stride),
-                       get_vert(vertex_buffer, indices[i-3], stride) );
-            sp_setup_tri( setup_ctx,
-                       get_vert(vertex_buffer, indices[i-1], stride),
-                       get_vert(vertex_buffer, indices[i-0], stride),
-                       get_vert(vertex_buffer, indices[i-3], stride) );
-         }
-      }
-      else {
-         for (i = 3; i < nr; i += 4) {
-            sp_setup_tri( setup_ctx,
+      for (i = 3; i < nr; i += 4) {
+         sp_setup_tri( setup_ctx,
                        get_vert(vertex_buffer, indices[i-3], stride),
                        get_vert(vertex_buffer, indices[i-2], stride),
                        get_vert(vertex_buffer, indices[i-0], stride) );
 
-            sp_setup_tri( setup_ctx,
+         sp_setup_tri( setup_ctx,
                        get_vert(vertex_buffer, indices[i-2], stride),
                        get_vert(vertex_buffer, indices[i-1], stride),
                        get_vert(vertex_buffer, indices[i-0], stride) );
-         }
       }
       break;
 
    case PIPE_PRIM_QUAD_STRIP:
-      if (softpipe->rasterizer->flatshade_first) {
-         for (i = 3; i < nr; i += 2) {
-            sp_setup_tri( setup_ctx,
-                       get_vert(vertex_buffer, indices[i-0], stride),
-                       get_vert(vertex_buffer, indices[i-1], stride),
-                       get_vert(vertex_buffer, indices[i-3], stride));
-            sp_setup_tri( setup_ctx,
-                       get_vert(vertex_buffer, indices[i-2], stride),
-                       get_vert(vertex_buffer, indices[i-0], stride),
-                       get_vert(vertex_buffer, indices[i-3], stride) );
-         }
-      }
-      else {
-         for (i = 3; i < nr; i += 2) {
-            sp_setup_tri( setup_ctx,
+      for (i = 3; i < nr; i += 2) {
+         sp_setup_tri( setup_ctx,
                        get_vert(vertex_buffer, indices[i-3], stride),
                        get_vert(vertex_buffer, indices[i-2], stride),
                        get_vert(vertex_buffer, indices[i-0], stride) );
-            sp_setup_tri( setup_ctx,
+         sp_setup_tri( setup_ctx,
                        get_vert(vertex_buffer, indices[i-1], stride),
                        get_vert(vertex_buffer, indices[i-3], stride),
                        get_vert(vertex_buffer, indices[i-0], stride) );
-         }
       }
       break;
 
@@ -448,56 +420,28 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr)
       break;
 
    case PIPE_PRIM_QUADS:
-      if (softpipe->rasterizer->flatshade_first) {
-         for (i = 3; i < nr; i += 4) {
-            sp_setup_tri( setup_ctx,
-                       get_vert(vertex_buffer, i-2, stride),
-                       get_vert(vertex_buffer, i-1, stride),
-                       get_vert(vertex_buffer, i-3, stride) );
-            sp_setup_tri( setup_ctx,
-                       get_vert(vertex_buffer, i-1, stride),
-                       get_vert(vertex_buffer, i-0, stride),
-                       get_vert(vertex_buffer, i-3, stride) );
-         }
-      }
-      else {
-         for (i = 3; i < nr; i += 4) {
-            sp_setup_tri( setup_ctx,
+      for (i = 3; i < nr; i += 4) {
+         sp_setup_tri( setup_ctx,
                        get_vert(vertex_buffer, i-3, stride),
                        get_vert(vertex_buffer, i-2, stride),
                        get_vert(vertex_buffer, i-0, stride) );
-            sp_setup_tri( setup_ctx,
+         sp_setup_tri( setup_ctx,
                        get_vert(vertex_buffer, i-2, stride),
                        get_vert(vertex_buffer, i-1, stride),
                        get_vert(vertex_buffer, i-0, stride) );
-         }
       }
       break;
 
    case PIPE_PRIM_QUAD_STRIP:
-      if (softpipe->rasterizer->flatshade_first) {
-         for (i = 3; i < nr; i += 2) {
-            sp_setup_tri( setup_ctx,
-                       get_vert(vertex_buffer, i-0, stride),
-                       get_vert(vertex_buffer, i-1, stride),
-                       get_vert(vertex_buffer, i-3, stride) );
-            sp_setup_tri( setup_ctx,
-                       get_vert(vertex_buffer, i-2, stride),
-                       get_vert(vertex_buffer, i-0, stride),
-                       get_vert(vertex_buffer, i-3, stride) );
-         }
-      }
-      else {
-         for (i = 3; i < nr; i += 2) {
-            sp_setup_tri( setup_ctx,
+      for (i = 3; i < nr; i += 2) {
+         sp_setup_tri( setup_ctx,
                        get_vert(vertex_buffer, i-3, stride),
                        get_vert(vertex_buffer, i-2, stride),
                        get_vert(vertex_buffer, i-0, stride) );
-            sp_setup_tri( setup_ctx,
+         sp_setup_tri( setup_ctx,
                        get_vert(vertex_buffer, i-1, stride),
                        get_vert(vertex_buffer, i-3, stride),
                        get_vert(vertex_buffer, i-0, stride) );
-         }
       }
       break;
 
diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c
index 9eb3ff68756..44680ea70c8 100644
--- a/src/gallium/drivers/softpipe/sp_screen.c
+++ b/src/gallium/drivers/softpipe/sp_screen.c
@@ -83,11 +83,11 @@ softpipe_get_param(struct pipe_screen *screen, int param)
    case PIPE_CAP_TEXTURE_SHADOW_MAP:
       return 1;
    case PIPE_CAP_MAX_TEXTURE_2D_LEVELS:
-      return 13; /* max 4Kx4K */
+      return SP_MAX_TEXTURE_2D_LEVELS;
    case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
-      return 9;  /* max 256x256x256 */
+      return SP_MAX_TEXTURE_3D_LEVELS;
    case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
-      return 13; /* max 4Kx4K */
+      return SP_MAX_TEXTURE_2D_LEVELS;
    case PIPE_CAP_TGSI_CONT_SUPPORTED:
       return 1;
    case PIPE_CAP_BLEND_EQUATION_SEPARATE:
diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c
index ecd6b39863c..ef7ccf41898 100644
--- a/src/gallium/drivers/softpipe/sp_tex_sample.c
+++ b/src/gallium/drivers/softpipe/sp_tex_sample.c
@@ -55,7 +55,7 @@
 static INLINE float
 frac(float f)
 {
-   return f - util_ifloor(f);
+   return f - floorf(f);
 }
 
 
@@ -201,11 +201,9 @@ wrap_nearest_mirror_repeat(const float s[4], unsigned size, int icoord[4])
    const float max = 1.0F - min;
    for (ch = 0; ch < 4; ch++) {
       const int flr = util_ifloor(s[ch]);
-      float u;
+      float u = frac(s[ch]);
       if (flr & 1)
-         u = 1.0F - (s[ch] - (float) flr);
-      else
-         u = s[ch] - (float) flr;
+         u = 1.0F - u;
       if (u < min)
          icoord[ch] = 0;
       else if (u > max)
@@ -358,11 +356,9 @@ wrap_linear_mirror_repeat(const float s[4], unsigned size,
    uint ch;
    for (ch = 0; ch < 4; ch++) {
       const int flr = util_ifloor(s[ch]);
-      float u;
+      float u = frac(s[ch]);
       if (flr & 1)
-         u = 1.0F - (s[ch] - (float) flr);
-      else
-         u = s[ch] - (float) flr;
+         u = 1.0F - u;
       u = u * size - 0.5F;
       icoord0[ch] = util_ifloor(u);
       icoord1[ch] = icoord0[ch] + 1;
@@ -441,8 +437,7 @@ wrap_linear_mirror_clamp_to_border(const float s[4], unsigned size,
 
 
 /**
- * For RECT textures / unnormalized texcoords
- * Only a subset of wrap modes supported.
+ * PIPE_TEX_WRAP_CLAMP for nearest sampling, unnormalized coords.
  */
 static void
 wrap_nearest_unorm_clamp(const float s[4], unsigned size, int icoord[4])
@@ -456,7 +451,7 @@ wrap_nearest_unorm_clamp(const float s[4], unsigned size, int icoord[4])
 
 
 /**
- * Handles clamp_to_edge and clamp_to_border:
+ * PIPE_TEX_WRAP_CLAMP_TO_BORDER for nearest sampling, unnormalized coords.
  */
 static void
 wrap_nearest_unorm_clamp_to_border(const float s[4], unsigned size,
@@ -464,14 +459,27 @@ wrap_nearest_unorm_clamp_to_border(const float s[4], unsigned size,
 {
    uint ch;
    for (ch = 0; ch < 4; ch++) {
+      icoord[ch]= util_ifloor( CLAMP(s[ch], -0.5F, (float) size + 0.5F) );
+   }
+}
+
+
+/**
+ * PIPE_TEX_WRAP_CLAMP_TO_EDGE for nearest sampling, unnormalized coords.
+ */
+static void
+wrap_nearest_unorm_clamp_to_edge(const float s[4], unsigned size,
+                                 int icoord[4])
+{
+   uint ch;
+   for (ch = 0; ch < 4; ch++) {
       icoord[ch]= util_ifloor( CLAMP(s[ch], 0.5F, (float) size - 0.5F) );
    }
 }
 
 
 /**
- * For RECT textures / unnormalized texcoords.
- * Only a subset of wrap modes supported.
+ * PIPE_TEX_WRAP_CLAMP for linear sampling, unnormalized coords.
  */
 static void
 wrap_linear_unorm_clamp(const float s[4], unsigned size,
@@ -488,13 +496,36 @@ wrap_linear_unorm_clamp(const float s[4], unsigned size,
 }
 
 
+/**
+ * PIPE_TEX_WRAP_CLAMP_TO_BORDER for linear sampling, unnormalized coords.
+ */
 static void
 wrap_linear_unorm_clamp_to_border(const float s[4], unsigned size,
                                   int icoord0[4], int icoord1[4], float w[4])
 {
    uint ch;
    for (ch = 0; ch < 4; ch++) {
-      float u = CLAMP(s[ch], 0.5F, (float) size - 0.5F);
+      float u = CLAMP(s[ch], -0.5F, (float) size + 0.5F);
+      u -= 0.5F;
+      icoord0[ch] = util_ifloor(u);
+      icoord1[ch] = icoord0[ch] + 1;
+      if (icoord1[ch] > (int) size - 1)
+         icoord1[ch] = size - 1;
+      w[ch] = frac(u);
+   }
+}
+
+
+/**
+ * PIPE_TEX_WRAP_CLAMP_TO_EDGE for linear sampling, unnormalized coords.
+ */
+static void
+wrap_linear_unorm_clamp_to_edge(const float s[4], unsigned size,
+                                int icoord0[4], int icoord1[4], float w[4])
+{
+   uint ch;
+   for (ch = 0; ch < 4; ch++) {
+      float u = CLAMP(s[ch], +0.5F, (float) size - 0.5F);
       u -= 0.5F;
       icoord0[ch] = util_ifloor(u);
       icoord1[ch] = icoord0[ch] + 1;
@@ -1711,6 +1742,7 @@ get_nearest_unorm_wrap(unsigned mode)
    case PIPE_TEX_WRAP_CLAMP:
       return wrap_nearest_unorm_clamp;
    case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+      return wrap_nearest_unorm_clamp_to_edge;
    case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
       return wrap_nearest_unorm_clamp_to_border;
    default:
@@ -1754,6 +1786,7 @@ get_linear_unorm_wrap(unsigned mode)
    case PIPE_TEX_WRAP_CLAMP:
       return wrap_linear_unorm_clamp;
    case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+      return wrap_linear_unorm_clamp_to_edge;
    case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
       return wrap_linear_unorm_clamp_to_border;
    default:
diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c
index 9641652da85..11d184effb9 100644
--- a/src/gallium/drivers/softpipe/sp_texture.c
+++ b/src/gallium/drivers/softpipe/sp_texture.c
@@ -36,6 +36,7 @@
 #include "util/u_format.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
+#include "util/u_simple_screen.h"
 
 #include "sp_context.h"
 #include "sp_texture.h"
@@ -121,7 +122,8 @@ softpipe_texture_create(struct pipe_screen *screen,
                util_is_power_of_two(template->depth0));
 
    if (spt->base.tex_usage & (PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
-                              PIPE_TEXTURE_USAGE_PRIMARY)) {
+                              PIPE_TEXTURE_USAGE_SCANOUT |
+                              PIPE_TEXTURE_USAGE_SHARED)) {
       if (!softpipe_displaytarget_layout(screen, spt))
          goto fail;
    }
@@ -138,20 +140,6 @@ softpipe_texture_create(struct pipe_screen *screen,
 }
 
 
-/**
- * Create a new pipe_texture which wraps an existing buffer.
- */
-static struct pipe_texture *
-softpipe_texture_blanket(struct pipe_screen * screen,
-                         const struct pipe_texture *base,
-                         const unsigned *stride,
-                         struct pipe_buffer *buffer)
-{
-   /* Texture blanket is going away.
-    */
-   debug_printf("softpipe_texture_blanket() not implemented!");
-   return NULL;
-}
 
 
 static void
@@ -463,7 +451,6 @@ void
 softpipe_init_screen_texture_funcs(struct pipe_screen *screen)
 {
    screen->texture_create = softpipe_texture_create;
-   screen->texture_blanket = softpipe_texture_blanket;
    screen->texture_destroy = softpipe_texture_destroy;
 
    screen->get_tex_surface = softpipe_get_tex_surface;
diff --git a/src/gallium/drivers/softpipe/sp_texture.h b/src/gallium/drivers/softpipe/sp_texture.h
index e8d17ce7daf..1c8636d1d56 100644
--- a/src/gallium/drivers/softpipe/sp_texture.h
+++ b/src/gallium/drivers/softpipe/sp_texture.h
@@ -33,6 +33,10 @@
 #include "pipe/p_video_state.h"
 
 
+#define SP_MAX_TEXTURE_2D_LEVELS 13  /* 4K x 4K */
+#define SP_MAX_TEXTURE_3D_LEVELS 9   /* 512 x 512 x 512 */
+
+
 struct pipe_context;
 struct pipe_screen;
 struct softpipe_context;
@@ -42,8 +46,8 @@ struct softpipe_texture
 {
    struct pipe_texture base;
 
-   unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS];
-   unsigned stride[PIPE_MAX_TEXTURE_LEVELS];
+   unsigned long level_offset[SP_MAX_TEXTURE_2D_LEVELS];
+   unsigned stride[SP_MAX_TEXTURE_2D_LEVELS];
 
    /**
     * Display target, for textures with the PIPE_TEXTURE_USAGE_DISPLAY_TARGET
diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c
index 3b30954ac8d..aedfdf1b469 100644
--- a/src/gallium/drivers/softpipe/sp_tile_cache.c
+++ b/src/gallium/drivers/softpipe/sp_tile_cache.c
@@ -299,13 +299,14 @@ sp_tile_cache_flush_clear(struct softpipe_tile_cache *tc)
                               x, y, TILE_SIZE, TILE_SIZE,
                               tc->tile.data.color32, 0/*STRIDE*/);
 
-            /* do this? */
-            clear_clear_flag(tc->clear_flags, addr);
-
             numCleared++;
          }
       }
    }
+
+   /* reset all clear flags to zero */
+   memset(tc->clear_flags, 0, sizeof(tc->clear_flags));
+
 #if 0
    debug_printf("num cleared: %u\n", numCleared);
 #endif
diff --git a/src/gallium/drivers/svga/svga_screen_texture.c b/src/gallium/drivers/svga/svga_screen_texture.c
index e9792e063e0..5b581debfc7 100644
--- a/src/gallium/drivers/svga/svga_screen_texture.c
+++ b/src/gallium/drivers/svga/svga_screen_texture.c
@@ -315,7 +315,11 @@ svga_texture_create(struct pipe_screen *screen,
       tex->key.cachable = 0;
    }
 
-   if(templat->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) {
+   if(templat->tex_usage & PIPE_TEXTURE_USAGE_SHARED) {
+      tex->key.cachable = 0;
+   }
+
+   if(templat->tex_usage & PIPE_TEXTURE_USAGE_SCANOUT) {
       tex->key.flags |= SVGA3D_SURFACE_HINT_SCANOUT;
       tex->key.cachable = 0;
    }
@@ -355,80 +359,18 @@ error1:
 }
 
 
-static struct pipe_texture *
-svga_texture_blanket(struct pipe_screen * screen,
-                     const struct pipe_texture *base,
-                     const unsigned *stride,
-                     struct pipe_buffer *buffer)
-{
-   struct svga_texture *tex;
-   struct svga_buffer *sbuf = svga_buffer(buffer);
-   struct svga_winsys_screen *sws = svga_winsys_screen(screen);
-   assert(screen);
-
-   /* Only supports one type */
-   if (base->target != PIPE_TEXTURE_2D ||
-       base->last_level != 0 ||
-       base->depth0 != 1) {
-      return NULL;
-   }
-
-   /**
-    * We currently can't do texture blanket on
-    * SVGA3D_BUFFER. Need to blit to a temporary surface?
-    */
-
-   assert(sbuf->handle);
-   if (!sbuf->handle)
-      return NULL;
-
-   if (svga_translate_format(base->format) != sbuf->key.format) {
-      unsigned f1 = svga_translate_format(base->format);
-      unsigned f2 = sbuf->key.format;
-
-      /* It's okay for XRGB and ARGB or depth with/out stencil to get mixed up */
-      if ( !( (f1 == SVGA3D_X8R8G8B8 && f2 == SVGA3D_A8R8G8B8) ||
-              (f1 == SVGA3D_A8R8G8B8 && f2 == SVGA3D_X8R8G8B8) ||
-              (f1 == SVGA3D_Z_D24X8 && f2 == SVGA3D_Z_D24S8) ) ) {
-         debug_printf("%s wrong format %u != %u\n", __FUNCTION__, f1, f2);
-         return NULL;
-      }
-   }
-
-   tex = CALLOC_STRUCT(svga_texture);
-   if (!tex)
-      return NULL;
-
-   tex->base = *base;
-   
-
-   if (sbuf->key.format == 1)
-      tex->base.format = PIPE_FORMAT_B8G8R8X8_UNORM;
-   else if (sbuf->key.format == 2)
-      tex->base.format = PIPE_FORMAT_B8G8R8A8_UNORM;
-
-   pipe_reference_init(&tex->base.reference, 1);
-   tex->base.screen = screen;
 
-   SVGA_DBG(DEBUG_DMA, "blanket sid %p\n", sbuf->handle);
 
-   /* We don't own this storage, so don't try to cache it.
-    */
-   assert(sbuf->key.cachable == 0);
-   tex->key.cachable = 0;
-   sws->surface_reference(sws, &tex->handle, sbuf->handle);
 
-   return &tex->base;
-}
-
-
-struct pipe_texture *
-svga_screen_texture_wrap_surface(struct pipe_screen *screen,
-				 struct pipe_texture *base,
-				 enum SVGA3dSurfaceFormat format,
-				 struct svga_winsys_surface *srf)
+static struct pipe_texture *
+svga_screen_texture_from_handle(struct pipe_screen *screen,
+                                const struct pipe_texture *base,
+                                struct winsys_handle *whandle)
 {
+   struct svga_winsys_screen *sws = svga_winsys_screen(screen);
+   struct svga_winsys_surface *srf;
    struct svga_texture *tex;
+   enum SVGA3dSurfaceFormat format = 0;
    assert(screen);
 
    /* Only supports one type */
@@ -438,6 +380,8 @@ svga_screen_texture_wrap_surface(struct pipe_screen *screen,
       return NULL;
    }
 
+   srf = sws->surface_from_handle(sws, whandle, &format);
+
    if (!srf)
       return NULL;
 
@@ -478,6 +422,22 @@ svga_screen_texture_wrap_surface(struct pipe_screen *screen,
 }
 
 
+static boolean 
+svga_screen_texture_get_handle(struct pipe_screen *screen,
+                               struct pipe_texture *texture,
+                               struct winsys_handle *whandle)
+{
+   struct svga_winsys_screen *sws = svga_winsys_screen(texture->screen);
+   unsigned stride;
+
+   assert(svga_texture(texture)->key.cachable == 0);
+   svga_texture(texture)->key.cachable = 0;
+   stride = util_format_get_nblocksx(texture->format, texture->width0) *
+            util_format_get_blocksize(texture->format);
+   return sws->surface_get_handle(sws, svga_texture(texture)->handle, stride, whandle);
+}
+
+
 static void
 svga_texture_destroy(struct pipe_texture *pt)
 {
@@ -955,10 +915,11 @@ void
 svga_screen_init_texture_functions(struct pipe_screen *screen)
 {
    screen->texture_create = svga_texture_create;
+   screen->texture_from_handle = svga_screen_texture_from_handle;
+   screen->texture_get_handle = svga_screen_texture_get_handle;
    screen->texture_destroy = svga_texture_destroy;
    screen->get_tex_surface = svga_get_tex_surface;
    screen->tex_surface_destroy = svga_tex_surface_destroy;
-   screen->texture_blanket = svga_texture_blanket;
    screen->get_tex_transfer = svga_get_tex_transfer;
    screen->transfer_map = svga_transfer_map;
    screen->transfer_unmap = svga_transfer_unmap;
@@ -1120,33 +1081,3 @@ svga_destroy_sampler_view_priv(struct svga_sampler_view *v)
    pipe_texture_reference(&v->texture, NULL);
    FREE(v);
 }
-
-boolean
-svga_screen_buffer_from_texture(struct pipe_texture *texture,
-				struct pipe_buffer **buffer,
-				unsigned *stride)
-{
-   struct svga_texture *stex = svga_texture(texture);
-
-   *buffer = svga_screen_buffer_wrap_surface
-      (texture->screen,
-       svga_translate_format(texture->format),
-       stex->handle);
-
-   *stride = util_format_get_stride(texture->format, texture->width0);
-
-   return *buffer != NULL;
-}
-
-
-struct svga_winsys_surface *
-svga_screen_texture_get_winsys_surface(struct pipe_texture *texture)
-{
-   struct svga_winsys_screen *sws = svga_winsys_screen(texture->screen);
-   struct svga_winsys_surface *vsurf = NULL;
-
-   assert(svga_texture(texture)->key.cachable == 0);
-   svga_texture(texture)->key.cachable = 0;
-   sws->surface_reference(sws, &vsurf, svga_texture(texture)->handle);
-   return vsurf;
-}
diff --git a/src/gallium/drivers/svga/svga_screen_texture.h b/src/gallium/drivers/svga/svga_screen_texture.h
index 24c1f78ca55..ca6602b4369 100644
--- a/src/gallium/drivers/svga/svga_screen_texture.h
+++ b/src/gallium/drivers/svga/svga_screen_texture.h
@@ -78,7 +78,7 @@ struct svga_texture
 {
    struct pipe_texture base;
 
-   boolean defined[6][PIPE_MAX_TEXTURE_LEVELS];
+   boolean defined[6][SVGA_MAX_TEXTURE_LEVELS];
    
    struct svga_sampler_view *cached_view;
 
diff --git a/src/gallium/drivers/svga/svga_winsys.h b/src/gallium/drivers/svga/svga_winsys.h
index b4e3af0eafc..d4bb176f9a8 100644
--- a/src/gallium/drivers/svga/svga_winsys.h
+++ b/src/gallium/drivers/svga/svga_winsys.h
@@ -51,6 +51,7 @@ struct pipe_context;
 struct pipe_fence_handle;
 struct pipe_texture;
 struct svga_region;
+struct winsys_handle;
 
 
 #define SVGA_BUFFER_USAGE_PINNED  (PIPE_BUFFER_USAGE_CUSTOM << 0)
@@ -187,6 +188,25 @@ struct svga_winsys_screen
                      uint32 numMipLevels);
 
    /**
+    * Creates a surface from a winsys handle.
+    * Used to implement pipe_screen::texture_from_handle.
+    */
+   struct svga_winsys_surface *
+   (*surface_from_handle)(struct svga_winsys_screen *sws,
+                          struct winsys_handle *whandle,
+                          SVGA3dSurfaceFormat *format);
+
+   /**
+    * Get a winsys_handle from a surface.
+    * Used to implement pipe_screen::texture_get_handle.
+    */
+   boolean
+   (*surface_get_handle)(struct svga_winsys_screen *sws,
+                         struct svga_winsys_surface *surface,
+                         unsigned stride,
+                         struct winsys_handle *whandle);
+
+   /**
     * Whether this surface is sitting in a validate list
     */
    boolean
@@ -284,19 +304,6 @@ svga_screen_buffer_wrap_surface(struct pipe_screen *screen,
 				struct svga_winsys_surface *srf);
 
 struct svga_winsys_surface *
-svga_screen_texture_get_winsys_surface(struct pipe_texture *texture);
-struct svga_winsys_surface *
 svga_screen_buffer_get_winsys_surface(struct pipe_buffer *buffer);
 
-boolean
-svga_screen_buffer_from_texture(struct pipe_texture *texture,
-				struct pipe_buffer **buffer,
-				unsigned *stride);
-
-struct pipe_texture *
-svga_screen_texture_wrap_surface(struct pipe_screen *screen,
-				 struct pipe_texture *base,
-				 enum SVGA3dSurfaceFormat format,
-				 struct svga_winsys_surface *srf);
-
 #endif /* SVGA_WINSYS_H_ */
diff --git a/src/gallium/drivers/trace/tr_drm.c b/src/gallium/drivers/trace/tr_drm.c
index 2b4915003e2..c16989fa529 100644
--- a/src/gallium/drivers/trace/tr_drm.c
+++ b/src/gallium/drivers/trace/tr_drm.c
@@ -62,69 +62,8 @@ trace_drm_create_screen(struct drm_api *_api, int fd,
 
    screen = api->create_screen(api, fd, arg);
 
-   return trace_screen_create(screen);
-}
-
-
-static struct pipe_texture *
-trace_drm_texture_from_shared_handle(struct drm_api *_api,
-                                     struct pipe_screen *_screen,
-                                     struct pipe_texture *templ,
-                                     const char *name,
-                                     unsigned stride,
-                                     unsigned handle)
-{
-   struct trace_screen *tr_screen = trace_screen(_screen);
-   struct trace_drm_api *tr_api = trace_drm_api(_api);
-   struct pipe_screen *screen = tr_screen->screen;
-   struct drm_api *api = tr_api->api;
-   struct pipe_texture *result;
-
-   /* TODO trace call */
-
-   result = api->texture_from_shared_handle(api, screen, templ, name, stride, handle);
-
-   result = trace_texture_create(trace_screen(_screen), result);
-
-   return result;
-}
-
-static boolean
-trace_drm_shared_handle_from_texture(struct drm_api *_api,
-                                     struct pipe_screen *_screen,
-                                     struct pipe_texture *_texture,
-                                     unsigned *stride,
-                                     unsigned *handle)
-{
-   struct trace_screen *tr_screen = trace_screen(_screen);
-   struct trace_texture *tr_texture = trace_texture(_texture);
-   struct trace_drm_api *tr_api = trace_drm_api(_api);
-   struct pipe_screen *screen = tr_screen->screen;
-   struct pipe_texture *texture = tr_texture->texture;
-   struct drm_api *api = tr_api->api;
-
-   /* TODO trace call */
-
-   return api->shared_handle_from_texture(api, screen, texture, stride, handle);
-}
 
-static boolean
-trace_drm_local_handle_from_texture(struct drm_api *_api,
-                                    struct pipe_screen *_screen,
-                                    struct pipe_texture *_texture,
-                                    unsigned *stride,
-                                    unsigned *handle)
-{
-   struct trace_screen *tr_screen = trace_screen(_screen);
-   struct trace_texture *tr_texture = trace_texture(_texture);
-   struct trace_drm_api *tr_api = trace_drm_api(_api);
-   struct pipe_screen *screen = tr_screen->screen;
-   struct pipe_texture *texture = tr_texture->texture;
-   struct drm_api *api = tr_api->api;
-
-   /* TODO trace call */
-
-   return api->local_handle_from_texture(api, screen, texture, stride, handle);
+   return trace_screen_create(screen);
 }
 
 static void
@@ -158,9 +97,6 @@ trace_drm_create(struct drm_api *api)
    tr_api->base.name = api->name;
    tr_api->base.driver_name = api->driver_name;
    tr_api->base.create_screen = trace_drm_create_screen;
-   tr_api->base.texture_from_shared_handle = trace_drm_texture_from_shared_handle;
-   tr_api->base.shared_handle_from_texture = trace_drm_shared_handle_from_texture;
-   tr_api->base.local_handle_from_texture = trace_drm_local_handle_from_texture;
    tr_api->base.destroy = trace_drm_destroy;
    tr_api->api = api;
 
diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c
index 388d83eb5c2..a15addeb8cf 100644
--- a/src/gallium/drivers/trace/tr_screen.c
+++ b/src/gallium/drivers/trace/tr_screen.c
@@ -236,39 +236,41 @@ trace_screen_texture_create(struct pipe_screen *_screen,
    return result;
 }
 
-
 static struct pipe_texture *
-trace_screen_texture_blanket(struct pipe_screen *_screen,
-                             const struct pipe_texture *templat,
-                             const unsigned *ppitch,
-                             struct pipe_buffer *_buffer)
+trace_screen_texture_from_handle(struct pipe_screen *_screen,
+                                 const struct pipe_texture *templ,
+                                 struct winsys_handle *handle)
 {
-   struct trace_screen *tr_scr = trace_screen(_screen);
-   struct trace_buffer *tr_buf = trace_buffer(_buffer);
-   struct pipe_screen *screen = tr_scr->screen;
-   struct pipe_buffer *buffer = tr_buf->buffer;
-   unsigned pitch = *ppitch;
+   struct trace_screen *tr_screen = trace_screen(_screen);
+   struct pipe_screen *screen = tr_screen->screen;
    struct pipe_texture *result;
 
-   trace_dump_call_begin("pipe_screen", "texture_blanket");
+   /* TODO trace call */
 
-   trace_dump_arg(ptr, screen);
-   trace_dump_arg(template, templat);
-   trace_dump_arg(uint, pitch);
-   trace_dump_arg(ptr, buffer);
+   result = screen->texture_from_handle(screen, templ, handle);
 
-   result = screen->texture_blanket(screen, templat, ppitch, buffer);
+   result = trace_texture_create(trace_screen(_screen), result);
 
-   trace_dump_ret(ptr, result);
+   return result;
+}
 
-   trace_dump_call_end();
+static boolean
+trace_screen_texture_get_handle(struct pipe_screen *_screen,
+                                struct pipe_texture *_texture,
+                                struct winsys_handle *handle)
+{
+   struct trace_screen *tr_screen = trace_screen(_screen);
+   struct trace_texture *tr_texture = trace_texture(_texture);
+   struct pipe_screen *screen = tr_screen->screen;
+   struct pipe_texture *texture = tr_texture->texture;
 
-   result = trace_texture_create(tr_scr, result);
+   /* TODO trace call */
 
-   return result;
+   return screen->texture_get_handle(screen, texture, handle);
 }
 
 
+
 static void
 trace_screen_texture_destroy(struct pipe_texture *_texture)
 {
@@ -484,45 +486,7 @@ trace_screen_transfer_unmap(struct pipe_screen *_screen,
  */
 
 
-static struct pipe_buffer *
-trace_screen_surface_buffer_create(struct pipe_screen *_screen,
-                                   unsigned width, unsigned height,
-                                   enum pipe_format format,
-                                   unsigned usage,
-                                   unsigned tex_usage,
-                                   unsigned *pstride)
-{
-   struct trace_screen *tr_scr = trace_screen(_screen);
-   struct pipe_screen *screen = tr_scr->screen;
-   unsigned stride;
-   struct pipe_buffer *result;
-
-   trace_dump_call_begin("pipe_screen", "surface_buffer_create");
-
-   trace_dump_arg(ptr, screen);
-   trace_dump_arg(uint, width);
-   trace_dump_arg(uint, height);
-   trace_dump_arg(format, format);
-   trace_dump_arg(uint, usage);
-   trace_dump_arg(uint, tex_usage);
 
-   result = screen->surface_buffer_create(screen,
-                                          width, height,
-                                          format,
-                                          usage,
-                                          tex_usage,
-                                          pstride);
-
-   stride = *pstride;
-
-   trace_dump_arg(uint, stride);
-
-   trace_dump_ret(ptr, result);
-
-   trace_dump_call_end();
-
-   return trace_buffer_create(tr_scr, result);
-}
 
 
 static struct pipe_buffer *
@@ -931,7 +895,8 @@ trace_screen_create(struct pipe_screen *screen)
    assert(screen->context_create);
    tr_scr->base.context_create = trace_screen_context_create;
    tr_scr->base.texture_create = trace_screen_texture_create;
-   tr_scr->base.texture_blanket = trace_screen_texture_blanket;
+   tr_scr->base.texture_from_handle = trace_screen_texture_from_handle;
+   tr_scr->base.texture_get_handle = trace_screen_texture_get_handle;
    tr_scr->base.texture_destroy = trace_screen_texture_destroy;
    tr_scr->base.get_tex_surface = trace_screen_get_tex_surface;
    tr_scr->base.tex_surface_destroy = trace_screen_tex_surface_destroy;
@@ -941,7 +906,6 @@ trace_screen_create(struct pipe_screen *screen)
    tr_scr->base.transfer_unmap = trace_screen_transfer_unmap;
    tr_scr->base.buffer_create = trace_screen_buffer_create;
    tr_scr->base.user_buffer_create = trace_screen_user_buffer_create;
-   tr_scr->base.surface_buffer_create = trace_screen_surface_buffer_create;
    if (screen->buffer_map)
       tr_scr->base.buffer_map = trace_screen_buffer_map;
    if (screen->buffer_map_range)
diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h
index 5cebd43ace2..5c97dc87e82 100644
--- a/src/gallium/include/pipe/p_defines.h
+++ b/src/gallium/include/pipe/p_defines.h
@@ -176,11 +176,12 @@ enum pipe_texture_target {
 #define PIPE_TEX_COMPARE_R_TO_TEXTURE  1
 
 #define PIPE_TEXTURE_USAGE_RENDER_TARGET   0x1
-#define PIPE_TEXTURE_USAGE_DISPLAY_TARGET  0x2 /* ie a backbuffer */
-#define PIPE_TEXTURE_USAGE_PRIMARY         0x4 /* ie a frontbuffer */
+#define PIPE_TEXTURE_USAGE_DISPLAY_TARGET  0x2  /* windows presentable buffer, ie a backbuffer */
+#define PIPE_TEXTURE_USAGE_SCANOUT         0x4  /* ie a frontbuffer */
 #define PIPE_TEXTURE_USAGE_DEPTH_STENCIL   0x8
 #define PIPE_TEXTURE_USAGE_SAMPLER         0x10
 #define PIPE_TEXTURE_USAGE_DYNAMIC         0x20
+#define PIPE_TEXTURE_USAGE_SHARED          0x40
 /** Pipe driver custom usage flags should be greater or equal to this value */
 #define PIPE_TEXTURE_USAGE_CUSTOM          (1 << 16)
 
diff --git a/src/gallium/include/pipe/p_screen.h b/src/gallium/include/pipe/p_screen.h
index e4a92228093..690455f7222 100644
--- a/src/gallium/include/pipe/p_screen.h
+++ b/src/gallium/include/pipe/p_screen.h
@@ -50,6 +50,8 @@ extern "C" {
 
 
 /** Opaque type */
+struct winsys_handle;
+/** Opaque type */
 struct pipe_fence_handle;
 struct pipe_winsys;
 struct pipe_buffer;
@@ -108,16 +110,23 @@ struct pipe_screen {
                                            const struct pipe_texture *templat);
 
    /**
-    * Create a new texture object, using the given template info, but on top of
-    * existing memory.
-    * 
-    * It is assumed that the buffer data is layed out according to the expected
-    * by the hardware. NULL will be returned if any inconsistency is found.  
+    * Create a texture from a winsys_handle. The handle is often created in
+    * another process by first creating a pipe texture and then calling
+    * texture_get_handle.
+    */
+   struct pipe_texture * (*texture_from_handle)(struct pipe_screen *,
+                                                const struct pipe_texture *templat,
+                                                struct winsys_handle *handle);
+
+   /**
+    * Get a winsys_handle from a texture. Some platforms/winsys requires
+    * that the texture is created with a special usage flag like
+    * DISPLAYTARGET or PRIMARY.
     */
-   struct pipe_texture * (*texture_blanket)(struct pipe_screen *,
-                                            const struct pipe_texture *templat,
-                                            const unsigned *stride,
-                                            struct pipe_buffer *buffer);
+   boolean (*texture_get_handle)(struct pipe_screen *,
+                                 struct pipe_texture *tex,
+                                 struct winsys_handle *handle);
+
 
    void (*texture_destroy)(struct pipe_texture *pt);
 
@@ -187,23 +196,6 @@ struct pipe_screen {
                                              void *ptr,
                                              unsigned bytes);
 
-   /**
-    * Allocate storage for a display target surface.
-    *
-    * Often surfaces which are meant to be blitted to the front screen (i.e.,
-    * display targets) must be allocated with special characteristics, memory
-    * pools, or obtained directly from the windowing system.
-    *
-    * This callback is invoked by the pipe_screenwhen creating a texture marked
-    * with the PIPE_TEXTURE_USAGE_DISPLAY_TARGET flag  to get the underlying
-    * buffer storage.
-    */
-   struct pipe_buffer *(*surface_buffer_create)(struct pipe_screen *screen,
-						unsigned width, unsigned height,
-						enum pipe_format format,
-						unsigned usage,
-						unsigned tex_usage,
-						unsigned *stride);
 
 
    /**
diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h
index 02558520bfe..72635d1031b 100644
--- a/src/gallium/include/pipe/p_state.h
+++ b/src/gallium/include/pipe/p_state.h
@@ -249,7 +249,7 @@ struct pipe_framebuffer_state
 {
    unsigned width, height;
 
-   /** multiple colorbuffers for multiple render targets */
+   /** multiple color buffers for multiple render targets */
    unsigned nr_cbufs;
    struct pipe_surface *cbufs[PIPE_MAX_COLOR_BUFS];
 
@@ -285,12 +285,12 @@ struct pipe_sampler_state
 struct pipe_surface
 {
    struct pipe_reference reference;
-   enum pipe_format format;      /**< PIPE_FORMAT_x */
+   enum pipe_format format;
    unsigned width;               /**< logical width in pixels */
    unsigned height;              /**< logical height in pixels */
    unsigned layout;              /**< PIPE_SURFACE_LAYOUT_x */
    unsigned offset;              /**< offset from start of buffer, in bytes */
-   unsigned usage;               /**< PIPE_BUFFER_USAGE_*  */
+   unsigned usage;               /**< bitmask of PIPE_BUFFER_USAGE_x */
 
    unsigned zslice;
    struct pipe_texture *texture; /**< texture into which this is a view  */
@@ -336,7 +336,7 @@ struct pipe_texture
 
    unsigned nr_samples:8;    /**< for multisampled surfaces, nr of samples */
 
-   unsigned tex_usage;       /* PIPE_TEXTURE_USAGE_* */
+   unsigned tex_usage;       /**< bitmask of PIPE_TEXTURE_USAGE_* */
 
    struct pipe_screen *screen; /**< screen that this texture belongs to */
 };
@@ -375,7 +375,7 @@ struct pipe_vertex_element
    unsigned vertex_buffer_index:8;
    unsigned nr_components:8;
  
-   enum pipe_format src_format; 	   /**< PIPE_FORMAT_* */
+   enum pipe_format src_format;
 };
 
 
diff --git a/src/gallium/include/state_tracker/drm_api.h b/src/gallium/include/state_tracker/drm_api.h
index e9fa9b4d2a3..fe7ef253ef0 100644
--- a/src/gallium/include/state_tracker/drm_api.h
+++ b/src/gallium/include/state_tracker/drm_api.h
@@ -17,6 +17,32 @@ enum drm_create_screen_mode {
 	DRM_CREATE_MAX
 };
 
+#define DRM_API_HANDLE_TYPE_SHARED 0
+#define DRM_API_HANDLE_TYPE_KMS    1
+
+/**
+ * For use with pipe_screen::{texture_from_handle|texture_get_handle}.
+ */
+struct winsys_handle
+{
+	/**
+	 * Unused for texture_from_handle, always
+	 * DRM_API_HANDLE_TYPE_SHARED.  Input to texture_get_handle,
+	 * use TEXTURE_USAGE to select handle for kms or ipc.
+	 */
+	unsigned type;
+	/**
+	 * Input to texture_from_handle.
+	 * Output for texture_get_handle.
+	 */
+	unsigned handle;
+	/**
+	 * Input to texture_from_handle.
+	 * Output for texture_get_handle.
+	 */
+	unsigned stride;
+};
+
 /**
  * Modes other than DRM_CREATE_NORMAL derive from this struct.
  */
@@ -28,6 +54,8 @@ struct drm_create_screen_arg {
 
 struct drm_api
 {
+	void (*destroy)(struct drm_api *api);
+
         const char *name;
 
 	/**
@@ -36,37 +64,10 @@ struct drm_api
 	const char *driver_name;
 
 	/**
-	 * Special buffer functions
+	 * Create a pipe srcreen.
 	 */
-	/*@{*/
 	struct pipe_screen*  (*create_screen)(struct drm_api *api, int drm_fd,
 	                                      struct drm_create_screen_arg *arg);
-	/*@}*/
-
-	/**
-	 * Special buffer functions
-	 */
-	/*@{*/
-	struct pipe_texture*
-	    (*texture_from_shared_handle)(struct drm_api *api,
-	                                  struct pipe_screen *screen,
-	                                  struct pipe_texture *templ,
-	                                  const char *name,
-	                                  unsigned stride,
-	                                  unsigned handle);
-	boolean (*shared_handle_from_texture)(struct drm_api *api,
-	                                      struct pipe_screen *screen,
-	                                      struct pipe_texture *texture,
-	                                      unsigned *stride,
-	                                      unsigned *handle);
-	boolean (*local_handle_from_texture)(struct drm_api *api,
-	                                     struct pipe_screen *screen,
-	                                     struct pipe_texture *texture,
-	                                     unsigned *stride,
-	                                     unsigned *handle);
-	/*@}*/
-
-	void (*destroy)(struct drm_api *api);
 };
 
 extern struct drm_api * drm_api_create(void);
diff --git a/src/gallium/state_trackers/dri/dri_drawable.c b/src/gallium/state_trackers/dri/dri_drawable.c
index 173f4041c8c..458473853cf 100644
--- a/src/gallium/state_trackers/dri/dri_drawable.c
+++ b/src/gallium/state_trackers/dri/dri_drawable.c
@@ -58,6 +58,7 @@ dri_surface_from_handle(struct drm_api *api,
    struct pipe_surface *surface = NULL;
    struct pipe_texture *texture = NULL;
    struct pipe_texture templat;
+   struct winsys_handle whandle;
 
    memset(&templat, 0, sizeof(templat));
    templat.tex_usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET;
@@ -68,8 +69,11 @@ dri_surface_from_handle(struct drm_api *api,
    templat.width0 = width;
    templat.height0 = height;
 
-   texture = api->texture_from_shared_handle(api, screen, &templat,
-                                             "dri2 buffer", pitch, handle);
+   memset(&whandle, 0, sizeof(whandle));
+   whandle.handle = handle;
+   whandle.stride = pitch;
+
+   texture = screen->texture_from_handle(screen, &templat, &whandle);
 
    if (!texture) {
       debug_printf("%s: Failed to blanket the buffer with a texture\n", __func__);
diff --git a/src/gallium/state_trackers/egl/common/egl_g3d.c b/src/gallium/state_trackers/egl/common/egl_g3d.c
index 2f7a8a8f7eb..f54c51a6681 100644
--- a/src/gallium/state_trackers/egl/common/egl_g3d.c
+++ b/src/gallium/state_trackers/egl/common/egl_g3d.c
@@ -36,6 +36,7 @@
 
 #include "native.h"
 #include "egl_g3d.h"
+#include "egl_g3d_image.h"
 #include "egl_st.h"
 
 /**
@@ -517,6 +518,24 @@ egl_g3d_update_buffer(struct pipe_screen *screen, void *context_private)
    egl_g3d_validate_context(gctx->base.Resource.Display, &gctx->base);
 }
 
+static void
+egl_g3d_invalid_surface(struct native_display *ndpy,
+                        struct native_surface *nsurf,
+                        unsigned int seq_num)
+{
+   /* XXX not thread safe? */
+   struct egl_g3d_surface *gsurf = egl_g3d_surface(nsurf->user_data);
+   struct egl_g3d_context *gctx = egl_g3d_context(gsurf->base.CurrentContext);
+
+   /* set force_validate to skip an unnecessary check */
+   if (gctx)
+      gctx->force_validate = TRUE;
+}
+
+static struct native_event_handler egl_g3d_native_event_handler = {
+   .invalid_surface = egl_g3d_invalid_surface
+};
+
 static EGLBoolean
 egl_g3d_terminate(_EGLDriver *drv, _EGLDisplay *dpy)
 {
@@ -561,12 +580,14 @@ egl_g3d_initialize(_EGLDriver *drv, _EGLDisplay *dpy,
    }
    dpy->DriverData = gdpy;
 
-   gdpy->native = native_create_display(dpy->NativeDisplay);
+   gdpy->native = native_create_display(dpy->NativeDisplay,
+         &egl_g3d_native_event_handler);
    if (!gdpy->native) {
       _eglError(EGL_NOT_INITIALIZED, "eglInitialize(no usable display)");
       goto fail;
    }
 
+   gdpy->native->user_data = (void *) dpy;
    gdpy->native->screen->update_buffer = egl_g3d_update_buffer;
 
    egl_g3d_init_st(&gdrv->base);
@@ -580,6 +601,10 @@ egl_g3d_initialize(_EGLDriver *drv, _EGLDisplay *dpy,
    }
 #endif
 
+   dpy->Extensions.KHR_image_base = EGL_TRUE;
+   if (gdpy->native->get_param(gdpy->native, NATIVE_PARAM_USE_NATIVE_BUFFER))
+      dpy->Extensions.KHR_image_pixmap = EGL_TRUE;
+
    if (egl_g3d_add_configs(drv, dpy, 1) == 1) {
       _eglError(EGL_NOT_INITIALIZED, "eglInitialize(unable to add configs)");
       goto fail;
@@ -761,6 +786,7 @@ egl_g3d_create_surface(_EGLDriver *drv, _EGLDisplay *dpy, _EGLConfig *conf,
       return NULL;
    }
 
+   nsurf->user_data = &gsurf->base;
    gsurf->native = nsurf;
 
    gsurf->render_att = (gsurf->base.RenderBuffer == EGL_SINGLE_BUFFER) ?
@@ -922,8 +948,8 @@ egl_g3d_swap_buffers(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surf)
 /**
  * Find a config that supports the pixmap.
  */
-static _EGLConfig *
-find_pixmap_config(_EGLDisplay *dpy, EGLNativePixmapType pix)
+_EGLConfig *
+egl_g3d_find_pixmap_config(_EGLDisplay *dpy, EGLNativePixmapType pix)
 {
    struct egl_g3d_display *gdpy = egl_g3d_display(dpy);
    struct egl_g3d_config *gconf;
@@ -975,7 +1001,7 @@ egl_g3d_copy_buffers(_EGLDriver *drv, _EGLDisplay *dpy, _EGLSurface *surf,
    if (!gsurf->render_surface)
       return EGL_TRUE;
 
-   gconf = egl_g3d_config(find_pixmap_config(dpy, target));
+   gconf = egl_g3d_config(egl_g3d_find_pixmap_config(dpy, target));
    if (!gconf)
       return _eglError(EGL_BAD_NATIVE_PIXMAP, "eglCopyBuffers");
 
@@ -1286,6 +1312,9 @@ _eglMain(const char *args)
    gdrv->base.API.BindTexImage = egl_g3d_bind_tex_image;
    gdrv->base.API.ReleaseTexImage = egl_g3d_release_tex_image;
 
+   gdrv->base.API.CreateImageKHR = egl_g3d_create_image;
+   gdrv->base.API.DestroyImageKHR = egl_g3d_destroy_image;
+
 #ifdef EGL_MESA_screen_surface
    gdrv->base.API.CreateScreenSurfaceMESA = egl_g3d_create_screen_surface;
    gdrv->base.API.ShowScreenSurfaceMESA = egl_g3d_show_screen_surface;
diff --git a/src/gallium/state_trackers/egl/common/egl_g3d.h b/src/gallium/state_trackers/egl/common/egl_g3d.h
index 5d2d9c481ab..e3e55e46d3b 100644
--- a/src/gallium/state_trackers/egl/common/egl_g3d.h
+++ b/src/gallium/state_trackers/egl/common/egl_g3d.h
@@ -34,6 +34,7 @@
 #include "eglcontext.h"
 #include "eglsurface.h"
 #include "eglconfig.h"
+#include "eglimage.h"
 #include "eglscreen.h"
 #include "eglmode.h"
 
@@ -81,6 +82,14 @@ struct egl_g3d_config {
    const struct native_config *native;
 };
 
+struct egl_g3d_image {
+   _EGLImage base;
+   struct pipe_texture *texture;
+   unsigned face;
+   unsigned level;
+   unsigned zslice;
+};
+
 struct egl_g3d_screen {
    _EGLScreen base;
    const struct native_connector *native;
@@ -90,5 +99,10 @@ struct egl_g3d_screen {
 /* standard typecasts */
 _EGL_DRIVER_STANDARD_TYPECASTS(egl_g3d)
 _EGL_DRIVER_TYPECAST(egl_g3d_screen, _EGLScreen, obj)
+_EGL_DRIVER_TYPECAST(egl_g3d_image, _EGLImage, obj)
+
+
+_EGLConfig *
+egl_g3d_find_pixmap_config(_EGLDisplay *dpy, EGLNativePixmapType pix);
 
 #endif /* _EGL_G3D_H_ */
diff --git a/src/gallium/state_trackers/egl/common/egl_g3d_image.c b/src/gallium/state_trackers/egl/common/egl_g3d_image.c
new file mode 100644
index 00000000000..d701f9c9a88
--- /dev/null
+++ b/src/gallium/state_trackers/egl/common/egl_g3d_image.c
@@ -0,0 +1,136 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.8
+ *
+ * Copyright (C) 2010 LunarG Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Chia-I Wu <[email protected]>
+ */
+
+#include <assert.h>
+#include "pipe/p_screen.h"
+#include "util/u_memory.h"
+#include "util/u_rect.h"
+#include "util/u_inlines.h"
+#include "eglcurrent.h"
+#include "egllog.h"
+
+#include "native.h"
+#include "egl_g3d.h"
+#include "egl_g3d_image.h"
+
+/**
+ * Reference and return the front left buffer of the native pixmap.
+ */
+static struct pipe_texture *
+egl_g3d_reference_native_pixmap(_EGLDisplay *dpy, EGLNativePixmapType pix)
+{
+   struct egl_g3d_display *gdpy = egl_g3d_display(dpy);
+   struct egl_g3d_config *gconf;
+   struct native_surface *nsurf;
+   struct pipe_texture *textures[NUM_NATIVE_ATTACHMENTS];
+   enum native_attachment natt;
+
+   gconf = egl_g3d_config(egl_g3d_find_pixmap_config(dpy, pix));
+   if (!gconf)
+      return NULL;
+
+   nsurf = gdpy->native->create_pixmap_surface(gdpy->native,
+         pix, gconf->native);
+   if (!nsurf)
+      return NULL;
+
+   natt = NATIVE_ATTACHMENT_FRONT_LEFT;
+   if (!nsurf->validate(nsurf, 1 << natt, NULL, textures, NULL, NULL))
+      textures[natt] = NULL;
+
+   nsurf->destroy(nsurf);
+
+   return textures[natt];
+}
+
+_EGLImage *
+egl_g3d_create_image(_EGLDriver *drv, _EGLDisplay *dpy, _EGLContext *ctx,
+                     EGLenum target, EGLClientBuffer buffer,
+                     const EGLint *attribs)
+{
+   struct pipe_texture *ptex;
+   struct egl_g3d_image *gimg;
+   unsigned face = 0, level = 0, zslice = 0;
+
+   gimg = CALLOC_STRUCT(egl_g3d_image);
+   if (!gimg) {
+      _eglError(EGL_BAD_ALLOC, "eglCreatePbufferSurface");
+      return NULL;
+   }
+
+   if (!_eglInitImage(&gimg->base, dpy, attribs)) {
+      free(gimg);
+      return NULL;
+   }
+
+   switch (target) {
+   case EGL_NATIVE_PIXMAP_KHR:
+      ptex = egl_g3d_reference_native_pixmap(dpy,
+            (EGLNativePixmapType) buffer);
+      break;
+   default:
+      ptex = NULL;
+      break;
+   }
+
+   if (!ptex) {
+      free(gimg);
+      return NULL;
+   }
+
+   if (level > ptex->last_level) {
+      _eglError(EGL_BAD_MATCH, "eglCreateEGLImageKHR");
+      pipe_texture_reference(&gimg->texture, NULL);
+      free(gimg);
+      return NULL;
+   }
+   if (zslice > ptex->depth0) {
+      _eglError(EGL_BAD_PARAMETER, "eglCreateEGLImageKHR");
+      pipe_texture_reference(&gimg->texture, NULL);
+      free(gimg);
+      return NULL;
+   }
+
+   /* transfer the ownership to the image */
+   gimg->texture = ptex;
+   gimg->face = face;
+   gimg->level = level;
+   gimg->zslice = zslice;
+
+   return &gimg->base;
+}
+
+EGLBoolean
+egl_g3d_destroy_image(_EGLDriver *drv, _EGLDisplay *dpy, _EGLImage *img)
+{
+   struct egl_g3d_image *gimg = egl_g3d_image(img);
+
+   pipe_texture_reference(&gimg->texture, NULL);
+   free(gimg);
+
+   return EGL_TRUE;
+}
diff --git a/src/gallium/state_trackers/egl/common/egl_g3d_image.h b/src/gallium/state_trackers/egl/common/egl_g3d_image.h
new file mode 100644
index 00000000000..c199c466456
--- /dev/null
+++ b/src/gallium/state_trackers/egl/common/egl_g3d_image.h
@@ -0,0 +1,41 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.8
+ *
+ * Copyright (C) 2010 LunarG Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Chia-I Wu <[email protected]>
+ */
+
+#ifndef _EGL_G3D_IMAGE_H_
+#define _EGL_G3D_IMAGE_H_
+
+#include "egl_g3d.h"
+
+_EGLImage *
+egl_g3d_create_image(_EGLDriver *drv, _EGLDisplay *dpy, _EGLContext *ctx,
+                     EGLenum target, EGLClientBuffer buffer,
+                     const EGLint *attribs);
+
+EGLBoolean
+egl_g3d_destroy_image(_EGLDriver *drv, _EGLDisplay *dpy, _EGLImage *image);
+
+#endif /* _EGL_G3D_IMAGE_H_ */
diff --git a/src/gallium/state_trackers/egl/common/native.h b/src/gallium/state_trackers/egl/common/native.h
index 4c9217e1a5b..93c81b26e10 100644
--- a/src/gallium/state_trackers/egl/common/native.h
+++ b/src/gallium/state_trackers/egl/common/native.h
@@ -34,6 +34,8 @@
 #include "pipe/p_context.h"
 #include "pipe/p_state.h"
 
+#include "native_modeset.h"
+
 /**
  * Only color buffers are listed.  The others are allocated privately through,
  * for example, st_renderbuffer_alloc_storage().
@@ -47,6 +49,14 @@ enum native_attachment {
    NUM_NATIVE_ATTACHMENTS
 };
 
+enum native_param_type {
+   /*
+    * Return TRUE if window/pixmap surfaces use the buffers of the native
+    * types.
+    */
+   NATIVE_PARAM_USE_NATIVE_BUFFER
+};
+
 /**
  * Enumerations for probe results.
  */
@@ -69,6 +79,11 @@ struct native_probe {
 };
 
 struct native_surface {
+   /**
+    * Available for caller's use.
+    */
+   void *user_data;
+
    void (*destroy)(struct native_surface *nsurf);
 
    /**
@@ -117,18 +132,6 @@ struct native_config {
    boolean scanout_bit;
 };
 
-struct native_connector {
-   int dummy;
-};
-
-struct native_mode {
-   const char *desc;
-   int width, height;
-   int refresh_rate;
-};
-
-struct native_display_modeset;
-
 /**
  * A pipe winsys abstracts the OS.  A pipe screen abstracts the graphcis
  * hardware.  A native display consists of a pipe winsys, a pipe screen, and
@@ -140,9 +143,22 @@ struct native_display {
     */
    struct pipe_screen *screen;
 
+   /**
+    * Available for caller's use.
+    */
+   void *user_data;
+
    void (*destroy)(struct native_display *ndpy);
 
    /**
+    * Query the parameters of the native display.
+    *
+    * The return value is defined by the parameter.
+    */
+   int (*get_param)(struct native_display *ndpy,
+                    enum native_param_type param);
+
+   /**
     * Get the supported configs.  The configs are owned by the display, but
     * the returned array should be free()ed.
     *
@@ -193,46 +209,17 @@ struct native_display {
 };
 
 /**
- * Mode setting interface of the native display.  It exposes the mode setting
- * capabilities of the underlying graphics hardware.
+ * The handler for events that a native display may generate.  The events are
+ * generated asynchronously and the handler may be called by any thread at any
+ * time.
  */
-struct native_display_modeset {
-   /**
-    * Get the available physical connectors and the number of CRTCs.
-    */
-   const struct native_connector **(*get_connectors)(struct native_display *ndpy,
-                                                     int *num_connectors,
-                                                     int *num_crtcs);
-
+struct native_event_handler {
    /**
-    * Get the current supported modes of a connector.  The returned modes may
-    * change every time this function is called and those from previous calls
-    * might become invalid.
-    */
-   const struct native_mode **(*get_modes)(struct native_display *ndpy,
-                                           const struct native_connector *nconn,
-                                           int *num_modes);
-
-   /**
-    * Create a scan-out surface.  Required unless no config has
-    * GLX_SCREEN_BIT_MESA set.
-    */
-   struct native_surface *(*create_scanout_surface)(struct native_display *ndpy,
-                                                    const struct native_config *nconf,
-                                                    uint width, uint height);
-
-   /**
-    * Program the CRTC to output the surface to the given connectors with the
-    * given mode.  When surface is not given, the CRTC is disabled.
-    *
-    * This interface does not export a way to query capabilities of the CRTCs.
-    * The native display usually needs to dynamically map the index to a CRTC
-    * that supports the given connectors.
+    * This function is called when a surface needs to be validated.
     */
-   boolean (*program)(struct native_display *ndpy, int crtc_idx,
-                      struct native_surface *nsurf, uint x, uint y,
-                      const struct native_connector **nconns, int num_nconns,
-                      const struct native_mode *nmode);
+   void (*invalid_surface)(struct native_display *ndpy,
+                           struct native_surface *nsurf,
+                           unsigned int seq_num);
 };
 
 /**
@@ -264,6 +251,7 @@ const char *
 native_get_name(void);
 
 struct native_display *
-native_create_display(EGLNativeDisplayType dpy);
+native_create_display(EGLNativeDisplayType dpy,
+                      struct native_event_handler *handler);
 
 #endif /* _NATIVE_H_ */
diff --git a/src/gallium/state_trackers/egl/common/native_modeset.h b/src/gallium/state_trackers/egl/common/native_modeset.h
new file mode 100644
index 00000000000..71dc3ec860e
--- /dev/null
+++ b/src/gallium/state_trackers/egl/common/native_modeset.h
@@ -0,0 +1,87 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.8
+ *
+ * Copyright (C) 2009-2010 Chia-I Wu <[email protected]>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _NATIVE_MODESET_H_
+#define _NATIVE_MODESET_H_
+
+#include "pipe/p_compiler.h"
+
+struct native_display;
+struct native_surface;
+struct native_config;
+
+struct native_connector {
+   int dummy;
+};
+
+struct native_mode {
+   const char *desc;
+   int width, height;
+   int refresh_rate;
+};
+
+/**
+ * Mode setting interface of the native display.  It exposes the mode setting
+ * capabilities of the underlying graphics hardware.
+ */
+struct native_display_modeset {
+   /**
+    * Get the available physical connectors and the number of CRTCs.
+    */
+   const struct native_connector **(*get_connectors)(struct native_display *ndpy,
+                                                     int *num_connectors,
+                                                     int *num_crtcs);
+
+   /**
+    * Get the current supported modes of a connector.  The returned modes may
+    * change every time this function is called and those from previous calls
+    * might become invalid.
+    */
+   const struct native_mode **(*get_modes)(struct native_display *ndpy,
+                                           const struct native_connector *nconn,
+                                           int *num_modes);
+
+   /**
+    * Create a scan-out surface.  Required unless no config has
+    * GLX_SCREEN_BIT_MESA set.
+    */
+   struct native_surface *(*create_scanout_surface)(struct native_display *ndpy,
+                                                    const struct native_config *nconf,
+                                                    uint width, uint height);
+
+   /**
+    * Program the CRTC to output the surface to the given connectors with the
+    * given mode.  When surface is not given, the CRTC is disabled.
+    *
+    * This interface does not export a way to query capabilities of the CRTCs.
+    * The native display usually needs to dynamically map the index to a CRTC
+    * that supports the given connectors.
+    */
+   boolean (*program)(struct native_display *ndpy, int crtc_idx,
+                      struct native_surface *nsurf, uint x, uint y,
+                      const struct native_connector **nconns, int num_nconns,
+                      const struct native_mode *nmode);
+};
+
+#endif /* _NATIVE_MODESET_H_ */
diff --git a/src/gallium/state_trackers/egl/kms/native_kms.c b/src/gallium/state_trackers/egl/kms/native_kms.c
index aedf3d430ca..73222408565 100644
--- a/src/gallium/state_trackers/egl/kms/native_kms.c
+++ b/src/gallium/state_trackers/egl/kms/native_kms.c
@@ -55,7 +55,7 @@ kms_surface_validate(struct native_surface *nsurf, uint attachment_mask,
       templ.format = ksurf->color_format;
       templ.tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET;
       if (ksurf->type == KMS_SURFACE_TYPE_SCANOUT)
-         templ.tex_usage |= PIPE_TEXTURE_USAGE_PRIMARY;
+         templ.tex_usage |= PIPE_TEXTURE_USAGE_SCANOUT;
    }
 
    /* create textures */
@@ -100,7 +100,7 @@ kms_surface_init_framebuffers(struct native_surface *nsurf, boolean need_back)
    for (i = 0; i < num_framebuffers; i++) {
       struct kms_framebuffer *fb;
       enum native_attachment natt;
-      unsigned int handle, stride;
+      struct winsys_handle whandle;
       uint block_bits;
 
       if (i == 0) {
@@ -128,13 +128,17 @@ kms_surface_init_framebuffers(struct native_surface *nsurf, boolean need_back)
       /* TODO detect the real value */
       fb->is_passive = TRUE;
 
-      if (!kdpy->api->local_handle_from_texture(kdpy->api,
-               kdpy->base.screen, fb->texture, &stride, &handle))
+      memset(&whandle, 0, sizeof(whandle));
+      whandle.type = DRM_API_HANDLE_TYPE_KMS;
+
+      if (!kdpy->base.screen->texture_get_handle(kdpy->base.screen,
+               fb->texture, &whandle))
          return FALSE;
 
       block_bits = util_format_get_blocksizebits(ksurf->color_format);
       err = drmModeAddFB(kdpy->fd, ksurf->width, ksurf->height,
-            block_bits, block_bits, stride, handle, &fb->buffer_id);
+            block_bits, block_bits, whandle.stride, whandle.handle,
+            &fb->buffer_id);
       if (err) {
          fb->buffer_id = 0;
          return FALSE;
@@ -201,6 +205,8 @@ kms_surface_swap_buffers(struct native_surface *nsurf)
 
    /* the front/back textures are swapped */
    ksurf->sequence_number++;
+   kdpy->event_handler->invalid_surface(&kdpy->base,
+         &ksurf->base, ksurf->sequence_number);
 
    return TRUE;
 }
@@ -663,6 +669,21 @@ kms_display_get_configs(struct native_display *ndpy, int *num_configs)
    return configs;
 }
 
+static int
+kms_display_get_param(struct native_display *ndpy,
+                      enum native_param_type param)
+{
+   int val;
+
+   switch (param) {
+   default:
+      val = 0;
+      break;
+   }
+
+   return val;
+}
+
 static void
 kms_display_destroy(struct native_display *ndpy)
 {
@@ -762,7 +783,9 @@ static struct native_display_modeset kms_display_modeset = {
 };
 
 static struct native_display *
-kms_create_display(EGLNativeDisplayType dpy, struct drm_api *api)
+kms_create_display(EGLNativeDisplayType dpy,
+                   struct native_event_handler *event_handler,
+                   struct drm_api *api)
 {
    struct kms_display *kdpy;
 
@@ -770,6 +793,8 @@ kms_create_display(EGLNativeDisplayType dpy, struct drm_api *api)
    if (!kdpy)
       return NULL;
 
+   kdpy->event_handler = event_handler;
+
    kdpy->api = api;
    if (!kdpy->api) {
       _eglLog(_EGL_WARNING, "failed to create DRM API");
@@ -805,6 +830,7 @@ kms_create_display(EGLNativeDisplayType dpy, struct drm_api *api)
    }
 
    kdpy->base.destroy = kms_display_destroy;
+   kdpy->base.get_param = kms_display_get_param;
    kdpy->base.get_configs = kms_display_get_configs;
    kdpy->base.create_pbuffer_surface = kms_display_create_pbuffer_surface;
 
@@ -845,7 +871,8 @@ native_get_name(void)
 }
 
 struct native_display *
-native_create_display(EGLNativeDisplayType dpy)
+native_create_display(EGLNativeDisplayType dpy,
+                      struct native_event_handler *event_handler)
 {
    struct native_display *ndpy = NULL;
 
@@ -853,7 +880,7 @@ native_create_display(EGLNativeDisplayType dpy)
       drm_api = drm_api_create();
 
    if (drm_api)
-      ndpy = kms_create_display(dpy, drm_api);
+      ndpy = kms_create_display(dpy, event_handler, drm_api);
 
    return ndpy;
 }
diff --git a/src/gallium/state_trackers/egl/kms/native_kms.h b/src/gallium/state_trackers/egl/kms/native_kms.h
index 095186e3cf3..f9cbcb158b5 100644
--- a/src/gallium/state_trackers/egl/kms/native_kms.h
+++ b/src/gallium/state_trackers/egl/kms/native_kms.h
@@ -53,6 +53,8 @@ struct kms_crtc {
 struct kms_display {
    struct native_display base;
 
+   struct native_event_handler *event_handler;
+
    int fd;
    struct drm_api *api;
    drmModeResPtr resources;
diff --git a/src/gallium/state_trackers/egl/x11/native_dri2.c b/src/gallium/state_trackers/egl/x11/native_dri2.c
index 74d3d104b94..98399792311 100644
--- a/src/gallium/state_trackers/egl/x11/native_dri2.c
+++ b/src/gallium/state_trackers/egl/x11/native_dri2.c
@@ -48,6 +48,8 @@ struct dri2_display {
    Display *dpy;
    boolean own_dpy;
 
+   struct native_event_handler *event_handler;
+
    struct drm_api *api;
    struct x11_screen *xscr;
    int xscr_number;
@@ -112,6 +114,7 @@ dri2_surface_process_drawable_buffers(struct native_surface *nsurf,
    struct dri2_surface *dri2surf = dri2_surface(nsurf);
    struct dri2_display *dri2dpy = dri2surf->dri2dpy;
    struct pipe_texture templ;
+   struct winsys_handle whandle;
    uint valid_mask;
    int i;
 
@@ -169,9 +172,11 @@ dri2_surface_process_drawable_buffers(struct native_surface *nsurf,
          continue;
       }
 
-      dri2surf->textures[natt] =
-         dri2dpy->api->texture_from_shared_handle(dri2dpy->api,
-               dri2dpy->base.screen, &templ, desc, xbuf->pitch, xbuf->name);
+      memset(&whandle, 0, sizeof(whandle));
+      whandle.stride = xbuf->pitch;
+      whandle.handle = xbuf->name;
+      dri2surf->textures[natt] = dri2dpy->base.screen->texture_from_handle(
+         dri2dpy->base.screen, &templ, &whandle);
       if (dri2surf->textures[natt])
          valid_mask |= 1 << natt;
    }
@@ -324,8 +329,11 @@ dri2_surface_flush_frontbuffer(struct native_surface *nsurf)
             DRI2BufferFakeFrontLeft, DRI2BufferFrontLeft);
 
    /* force buffers to be updated in next validation call */
-   if (!dri2_surface_receive_events(&dri2surf->base))
+   if (!dri2_surface_receive_events(&dri2surf->base)) {
       dri2surf->server_stamp++;
+      dri2dpy->event_handler->invalid_surface(&dri2dpy->base,
+            &dri2surf->base, dri2surf->server_stamp);
+   }
 
    return TRUE;
 }
@@ -353,8 +361,11 @@ dri2_surface_swap_buffers(struct native_surface *nsurf)
             DRI2BufferFrontLeft, DRI2BufferFakeFrontLeft);
 
    /* force buffers to be updated in next validation call */
-   if (!dri2_surface_receive_events(&dri2surf->base))
+   if (!dri2_surface_receive_events(&dri2surf->base)) {
       dri2surf->server_stamp++;
+      dri2dpy->event_handler->invalid_surface(&dri2dpy->base,
+            &dri2surf->base, dri2surf->server_stamp);
+   }
 
    return TRUE;
 }
@@ -699,6 +710,25 @@ dri2_display_is_pixmap_supported(struct native_display *ndpy,
    return (depth == nconf_depth || (depth == 24 && depth + 8 == nconf_depth));
 }
 
+static int
+dri2_display_get_param(struct native_display *ndpy,
+                       enum native_param_type param)
+{
+   int val;
+
+   switch (param) {
+   case NATIVE_PARAM_USE_NATIVE_BUFFER:
+      /* DRI2GetBuffers use the native buffers */
+      val = TRUE;
+      break;
+   default:
+      val = 0;
+      break;
+   }
+
+   return val;
+}
+
 static void
 dri2_display_destroy(struct native_display *ndpy)
 {
@@ -737,7 +767,10 @@ dri2_display_invalidate_buffers(struct x11_screen *xscr, Drawable drawable,
       return;
 
    dri2surf = dri2_surface(nsurf);
+
    dri2surf->server_stamp++;
+   dri2dpy->event_handler->invalid_surface(&dri2dpy->base,
+         &dri2surf->base, dri2surf->server_stamp);
 }
 
 /**
@@ -796,7 +829,9 @@ dri2_display_hash_table_compare(void *key1, void *key2)
 }
 
 struct native_display *
-x11_create_dri2_display(EGLNativeDisplayType dpy, struct drm_api *api)
+x11_create_dri2_display(EGLNativeDisplayType dpy,
+                        struct native_event_handler *event_handler,
+                        struct drm_api *api)
 {
    struct dri2_display *dri2dpy;
 
@@ -804,6 +839,7 @@ x11_create_dri2_display(EGLNativeDisplayType dpy, struct drm_api *api)
    if (!dri2dpy)
       return NULL;
 
+   dri2dpy->event_handler = event_handler;
    dri2dpy->api = api;
 
    dri2dpy->dpy = dpy;
@@ -836,6 +872,7 @@ x11_create_dri2_display(EGLNativeDisplayType dpy, struct drm_api *api)
    }
 
    dri2dpy->base.destroy = dri2_display_destroy;
+   dri2dpy->base.get_param = dri2_display_get_param;
    dri2dpy->base.get_configs = dri2_display_get_configs;
    dri2dpy->base.is_pixmap_supported = dri2_display_is_pixmap_supported;
    dri2dpy->base.create_window_surface = dri2_display_create_window_surface;
diff --git a/src/gallium/state_trackers/egl/x11/native_x11.c b/src/gallium/state_trackers/egl/x11/native_x11.c
index ef7490503b0..2b3edbc006a 100644
--- a/src/gallium/state_trackers/egl/x11/native_x11.c
+++ b/src/gallium/state_trackers/egl/x11/native_x11.c
@@ -126,7 +126,8 @@ native_get_name(void)
 }
 
 struct native_display *
-native_create_display(EGLNativeDisplayType dpy)
+native_create_display(EGLNativeDisplayType dpy,
+                      struct native_event_handler *event_handler)
 {
    struct native_display *ndpy = NULL;
    boolean force_sw;
@@ -136,7 +137,7 @@ native_create_display(EGLNativeDisplayType dpy)
 
    force_sw = debug_get_bool_option("EGL_SOFTWARE", FALSE);
    if (api && !force_sw) {
-      ndpy = x11_create_dri2_display(dpy, api);
+      ndpy = x11_create_dri2_display(dpy, event_handler, api);
    }
 
    if (!ndpy) {
diff --git a/src/gallium/state_trackers/egl/x11/native_x11.h b/src/gallium/state_trackers/egl/x11/native_x11.h
index 2e71de1885b..21ed6a2e063 100644
--- a/src/gallium/state_trackers/egl/x11/native_x11.h
+++ b/src/gallium/state_trackers/egl/x11/native_x11.h
@@ -32,6 +32,8 @@ struct native_display *
 x11_create_ximage_display(EGLNativeDisplayType dpy);
 
 struct native_display *
-x11_create_dri2_display(EGLNativeDisplayType dpy, struct drm_api *api);
+x11_create_dri2_display(EGLNativeDisplayType dpy,
+                        struct native_event_handler *event_handler,
+                        struct drm_api *api);
 
 #endif /* _NATIVE_X11_H_ */
diff --git a/src/gallium/state_trackers/egl/x11/native_ximage.c b/src/gallium/state_trackers/egl/x11/native_ximage.c
index 6c6ba24e7b3..0006b0caca4 100644
--- a/src/gallium/state_trackers/egl/x11/native_ximage.c
+++ b/src/gallium/state_trackers/egl/x11/native_ximage.c
@@ -196,6 +196,16 @@ ximage_surface_update_geometry(struct native_surface *nsurf)
    return updated;
 }
 
+static void
+ximage_surface_notify_invalid(struct native_surface *nsurf)
+{
+   struct ximage_surface *xsurf = ximage_surface(nsurf);
+   struct ximage_display *xdpy = xsurf->xdpy;
+
+   xdpy->event_handler->invalid_surface(&xdpy->base,
+         &xsurf->base, xsurf->server_stamp);
+}
+
 /**
  * Update the buffers of the surface.  It is a slow function due to the
  * round-trip to the server.
@@ -278,6 +288,7 @@ ximage_surface_flush_frontbuffer(struct native_surface *nsurf)
          NATIVE_ATTACHMENT_FRONT_LEFT);
    /* force buffers to be updated in next validation call */
    xsurf->server_stamp++;
+   ximage_surface_notify_invalid(&xsurf->base);
 
    return ret;
 }
@@ -294,6 +305,7 @@ ximage_surface_swap_buffers(struct native_surface *nsurf)
          NATIVE_ATTACHMENT_BACK_LEFT);
    /* force buffers to be updated in next validation call */
    xsurf->server_stamp++;
+   ximage_surface_notify_invalid(&xsurf->base);
 
    xfront = &xsurf->buffers[NATIVE_ATTACHMENT_FRONT_LEFT];
    xback = &xsurf->buffers[NATIVE_ATTACHMENT_BACK_LEFT];
@@ -579,6 +591,25 @@ ximage_display_is_pixmap_supported(struct native_display *ndpy,
    return (fmt == nconf->color_format);
 }
 
+static int
+ximage_display_get_param(struct native_display *ndpy,
+                         enum native_param_type param)
+{
+   int val;
+
+   switch (param) {
+   case NATIVE_PARAM_USE_NATIVE_BUFFER:
+      /* private buffers are allocated */
+      val = FALSE;
+      break;
+   default:
+      val = 0;
+      break;
+   }
+
+   return val;
+}
+
 static void
 ximage_display_destroy(struct native_display *ndpy)
 {
@@ -625,6 +656,7 @@ x11_create_ximage_display(EGLNativeDisplayType dpy)
    xdpy->base.screen = xdpy->driver->create_pipe_screen(xdpy->dpy);
 
    xdpy->base.destroy = ximage_display_destroy;
+   xdpy->base.get_param = ximage_display_get_param;
 
    xdpy->base.get_configs = ximage_display_get_configs;
    xdpy->base.is_pixmap_supported = ximage_display_is_pixmap_supported;
diff --git a/src/gallium/state_trackers/vega/Makefile b/src/gallium/state_trackers/vega/Makefile
index 037d8dc911a..7f04b2aa832 100644
--- a/src/gallium/state_trackers/vega/Makefile
+++ b/src/gallium/state_trackers/vega/Makefile
@@ -53,7 +53,7 @@ INCLUDE_DIRS = \
 
 
 .c.o:
-	$(CC) -c $(INCLUDE_DIRS) $(DEFINES) $(CFLAGS) $< -o $@
+	$(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@
 
 default: depend $(TOP)/$(LIB_DIR)/$(VG_LIB_NAME)
 
diff --git a/src/gallium/state_trackers/vega/paint.c b/src/gallium/state_trackers/vega/paint.c
index caf0c14b745..cdb87d3bf68 100644
--- a/src/gallium/state_trackers/vega/paint.c
+++ b/src/gallium/state_trackers/vega/paint.c
@@ -639,9 +639,6 @@ VGint paint_bind_samplers(struct vg_paint *paint, struct pipe_sampler_state **sa
    }
       break;
    default:
-      samplers[0] = &paint->pattern.sampler; /* dummy */
-      textures[0] = 0;
-      return 0;
       break;
    }
    return 0;
diff --git a/src/gallium/state_trackers/xorg/xorg_crtc.c b/src/gallium/state_trackers/xorg/xorg_crtc.c
index 000ec9048f5..44f7da0f960 100644
--- a/src/gallium/state_trackers/xorg/xorg_crtc.c
+++ b/src/gallium/state_trackers/xorg/xorg_crtc.c
@@ -197,11 +197,11 @@ crtc_load_cursor_argb_ga3d(xf86CrtcPtr crtc, CARD32 * image)
 
     if (!crtcp->cursor_tex) {
 	struct pipe_texture templat;
-	unsigned pitch;
+	struct winsys_handle whandle;
 
 	memset(&templat, 0, sizeof(templat));
 	templat.tex_usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET;
-	templat.tex_usage |= PIPE_TEXTURE_USAGE_PRIMARY;
+	templat.tex_usage |= PIPE_TEXTURE_USAGE_SCANOUT;
 	templat.target = PIPE_TEXTURE_2D;
 	templat.last_level = 0;
 	templat.depth0 = 1;
@@ -209,13 +209,14 @@ crtc_load_cursor_argb_ga3d(xf86CrtcPtr crtc, CARD32 * image)
 	templat.width0 = 64;
 	templat.height0 = 64;
 
+	memset(&whandle, 0, sizeof(whandle));
+	whandle.type = DRM_API_HANDLE_TYPE_KMS;
+
 	crtcp->cursor_tex = ms->screen->texture_create(ms->screen,
 						       &templat);
-	ms->api->local_handle_from_texture(ms->api,
-					   ms->screen,
-					   crtcp->cursor_tex,
-					   &pitch,
-					   &crtcp->cursor_handle);
+	ms->screen->texture_get_handle(ms->screen, crtcp->cursor_tex, &whandle);
+
+	crtcp->cursor_handle = whandle.handle;
     }
 
     transfer = ms->screen->get_tex_transfer(ms->screen, crtcp->cursor_tex,
diff --git a/src/gallium/state_trackers/xorg/xorg_dri2.c b/src/gallium/state_trackers/xorg/xorg_dri2.c
index 5fc85c0e98c..f23e4c6cc7c 100644
--- a/src/gallium/state_trackers/xorg/xorg_dri2.c
+++ b/src/gallium/state_trackers/xorg/xorg_dri2.c
@@ -67,7 +67,7 @@ dri2_do_create_buffer(DrawablePtr pDraw, DRI2BufferPtr buffer, unsigned int form
     struct exa_pixmap_priv *exa_priv;
     BufferPrivatePtr private = buffer->driverPrivate;
     PixmapPtr pPixmap;
-    unsigned stride, handle;
+    struct winsys_handle whandle;
 
     if (pDraw->type == DRAWABLE_PIXMAP)
 	pPixmap = (PixmapPtr) pDraw;
@@ -75,6 +75,7 @@ dri2_do_create_buffer(DrawablePtr pDraw, DRI2BufferPtr buffer, unsigned int form
 	pPixmap = (*pScreen->GetWindowPixmap)((WindowPtr) pDraw);
     exa_priv = exaGetPixmapDriverPrivate(pPixmap);
 
+
     switch (buffer->attachment) {
     default:
 	if (buffer->attachment != DRI2BufferFakeFrontLeft ||
@@ -128,7 +129,7 @@ dri2_do_create_buffer(DrawablePtr pDraw, DRI2BufferPtr buffer, unsigned int form
 	    template.depth0 = 1;
 	    template.last_level = 0;
 	    template.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL |
-		PIPE_TEXTURE_USAGE_DISPLAY_TARGET;
+		PIPE_TEXTURE_USAGE_SHARED;
 	    tex = ms->screen->texture_create(ms->screen, &template);
 	    pipe_texture_reference(&exa_priv->depth_stencil_tex, tex);
 	}
@@ -153,10 +154,13 @@ dri2_do_create_buffer(DrawablePtr pDraw, DRI2BufferPtr buffer, unsigned int form
     if (!tex)
 	FatalError("NO TEXTURE IN DRI2\n");
 
-    ms->api->shared_handle_from_texture(ms->api, ms->screen, tex, &stride, &handle);
+    memset(&whandle, 0, sizeof(whandle));
+    whandle.type = DRM_API_HANDLE_TYPE_SHARED;
+
+    ms->screen->texture_get_handle(ms->screen, tex, &whandle);
 
-    buffer->name = handle;
-    buffer->pitch = stride;
+    buffer->name = whandle.handle;
+    buffer->pitch = whandle.stride;
     buffer->cpp = 4;
     buffer->driverPrivate = private;
     buffer->flags = 0; /* not tiled */
diff --git a/src/gallium/state_trackers/xorg/xorg_driver.c b/src/gallium/state_trackers/xorg/xorg_driver.c
index 8fb6e5a96dd..004a28f00e1 100644
--- a/src/gallium/state_trackers/xorg/xorg_driver.c
+++ b/src/gallium/state_trackers/xorg/xorg_driver.c
@@ -989,8 +989,9 @@ static Bool
 drv_create_front_buffer_ga3d(ScrnInfoPtr pScrn)
 {
     modesettingPtr ms = modesettingPTR(pScrn);
-    unsigned handle, stride, fb_id;
     struct pipe_texture *tex;
+    struct winsys_handle whandle;
+    unsigned fb_id;
     int ret;
 
     ms->noEvict = TRUE;
@@ -1001,10 +1002,10 @@ drv_create_front_buffer_ga3d(ScrnInfoPtr pScrn)
     if (!tex)
 	return FALSE;
 
-    if (!ms->api->local_handle_from_texture(ms->api, ms->screen,
-					    tex,
-					    &stride,
-					    &handle))
+    memset(&whandle, 0, sizeof(whandle));
+    whandle.type = DRM_API_HANDLE_TYPE_KMS;
+
+    if (!ms->screen->texture_get_handle(ms->screen, tex, &whandle))
 	goto err_destroy;
 
     ret = drmModeAddFB(ms->fd,
@@ -1012,8 +1013,8 @@ drv_create_front_buffer_ga3d(ScrnInfoPtr pScrn)
 		       pScrn->virtualY,
 		       pScrn->depth,
 		       pScrn->bitsPerPixel,
-		       stride,
-		       handle,
+		       whandle.stride,
+		       whandle.handle,
 		       &fb_id);
     if (ret) {
 	debug_printf("%s: failed to create framebuffer (%i, %s)\n",
diff --git a/src/gallium/state_trackers/xorg/xorg_exa.c b/src/gallium/state_trackers/xorg/xorg_exa.c
index a242e02ee77..a7ffe3f4992 100644
--- a/src/gallium/state_trackers/xorg/xorg_exa.c
+++ b/src/gallium/state_trackers/xorg/xorg_exa.c
@@ -789,7 +789,7 @@ xorg_exa_set_displayed_usage(PixmapPtr pPixmap)
 	return 0;
     }
 
-    priv->flags |= PIPE_TEXTURE_USAGE_PRIMARY;
+    priv->flags |= PIPE_TEXTURE_USAGE_SCANOUT;
 
     return 0;
 }
@@ -805,7 +805,7 @@ xorg_exa_set_shared_usage(PixmapPtr pPixmap)
 	return 0;
     }
 
-    priv->flags |= PIPE_TEXTURE_USAGE_DISPLAY_TARGET;
+    priv->flags |= PIPE_TEXTURE_USAGE_SHARED;
 
     return 0;
 }
@@ -943,7 +943,7 @@ xorg_exa_set_texture(PixmapPtr pPixmap, struct  pipe_texture *tex)
 {
     struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap);
 
-    int mask = PIPE_TEXTURE_USAGE_PRIMARY | PIPE_TEXTURE_USAGE_DISPLAY_TARGET;
+    int mask = PIPE_TEXTURE_USAGE_SHARED | PIPE_TEXTURE_USAGE_SCANOUT;
 
     if (!priv)
 	return FALSE;
@@ -976,8 +976,8 @@ xorg_exa_create_root_texture(ScrnInfoPtr pScrn,
     template.depth0 = 1;
     template.last_level = 0;
     template.tex_usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET;
-    template.tex_usage |= PIPE_TEXTURE_USAGE_PRIMARY;
-    template.tex_usage |= PIPE_TEXTURE_USAGE_DISPLAY_TARGET;
+    template.tex_usage |= PIPE_TEXTURE_USAGE_SCANOUT;
+    template.tex_usage |= PIPE_TEXTURE_USAGE_SHARED;
 
     return exa->scrn->texture_create(exa->scrn, &template);
 }
diff --git a/src/gallium/state_trackers/xorg/xvmc/surface.c b/src/gallium/state_trackers/xorg/xvmc/surface.c
index 87d1dfaace2..12d94e0c5ca 100644
--- a/src/gallium/state_trackers/xorg/xvmc/surface.c
+++ b/src/gallium/state_trackers/xorg/xvmc/surface.c
@@ -106,7 +106,7 @@ CreateOrResizeBackBuffer(struct pipe_video_context *vpipe, unsigned int width, u
    template.width0 = width;
    template.height0 = height;
    template.depth0 = 1;
-   template.tex_usage = PIPE_TEXTURE_USAGE_DISPLAY_TARGET;
+   template.tex_usage = PIPE_TEXTURE_USAGE_SHARED;
 
    tex = vpipe->screen->texture_create(vpipe->screen, &template);
    if (!tex)
diff --git a/src/gallium/winsys/drm/i965/gem/i965_drm_api.c b/src/gallium/winsys/drm/i965/gem/i965_drm_api.c
index a061eef0beb..21e82303f72 100644
--- a/src/gallium/winsys/drm/i965/gem/i965_drm_api.c
+++ b/src/gallium/winsys/drm/i965/gem/i965_drm_api.c
@@ -37,129 +37,6 @@ i965_libdrm_get_device_id(unsigned int *device_id)
    fclose(file);
 }
 
-static struct i965_libdrm_buffer *
-i965_libdrm_buffer_from_handle(struct i965_libdrm_winsys *idws,
-                               const char* name, unsigned handle)
-{
-   struct i965_libdrm_buffer *buf = CALLOC_STRUCT(i965_libdrm_buffer);
-   uint32_t swizzle = 0;
-
-   if (BRW_DUMP)
-      debug_printf("%s\n", __FUNCTION__);
-
-   if (!buf)
-      return NULL;
-   pipe_reference_init(&buf->base.reference, 1);
-   buf->bo = drm_intel_bo_gem_create_from_name(idws->gem, name, handle);
-   buf->base.size = buf->bo->size;
-   buf->base.sws = &idws->base;
-   buf->flinked = TRUE;
-   buf->flink = handle;
-
-
-   if (!buf->bo)
-      goto err;
-
-   drm_intel_bo_get_tiling(buf->bo, &buf->tiling, &swizzle);
-   if (buf->tiling != 0)
-      buf->map_gtt = TRUE;
-
-   return buf;
-
-err:
-   FREE(buf);
-   return NULL;
-}
-
-
-/*
- * Exported functions
- */
-
-
-static struct pipe_texture *
-i965_libdrm_texture_from_shared_handle(struct drm_api *api,
-                                       struct pipe_screen *screen,
-                                       struct pipe_texture *template,
-                                       const char* name,
-                                       unsigned pitch,
-                                       unsigned handle)
-{
-   /* XXX: this is silly -- there should be a way to get directly from
-    * the "drm_api" struct to ourselves, without peering into
-    * unrelated code:
-    */
-   struct i965_libdrm_winsys *idws = i965_libdrm_winsys(brw_screen(screen)->sws);
-   struct i965_libdrm_buffer *buffer;
-
-   if (BRW_DUMP)
-      debug_printf("%s %s pitch %d handle 0x%x\n", __FUNCTION__,
-		   name, pitch, handle);
-
-   buffer = i965_libdrm_buffer_from_handle(idws, name, handle);
-   if (!buffer)
-      return NULL;
-
-   return brw_texture_blanket_winsys_buffer(screen, template, pitch,
-					    buffer->tiling,
-					    &buffer->base);
-}
-
-
-static boolean
-i965_libdrm_shared_handle_from_texture(struct drm_api *api,
-                                       struct pipe_screen *screen,
-                                       struct pipe_texture *texture,
-                                       unsigned *pitch,
-                                       unsigned *handle)
-{
-   struct i965_libdrm_buffer *buf = NULL;
-   struct brw_winsys_buffer *buffer = NULL;
-
-   if (BRW_DUMP)
-      debug_printf("%s\n", __FUNCTION__);
-
-   if (!brw_texture_get_winsys_buffer(texture, &buffer, pitch))
-      return FALSE;
-
-   buf = i965_libdrm_buffer(buffer);
-   if (!buf->flinked) {
-      if (drm_intel_bo_flink(buf->bo, &buf->flink))
-         return FALSE;
-      buf->flinked = TRUE;
-   }
-
-   *handle = buf->flink;
-
-   if (BRW_DUMP)
-      debug_printf("   -> pitch %d handle 0x%x\n", *pitch, *handle);
-
-   return TRUE;
-}
-
-static boolean
-i965_libdrm_local_handle_from_texture(struct drm_api *api,
-                                      struct pipe_screen *screen,
-                                      struct pipe_texture *texture,
-                                      unsigned *pitch,
-                                      unsigned *handle)
-{
-   struct brw_winsys_buffer *buffer = NULL;
-
-   if (BRW_DUMP)
-      debug_printf("%s\n", __FUNCTION__);
-
-   if (!brw_texture_get_winsys_buffer(texture, &buffer, pitch))
-      return FALSE;
-
-   *handle = i965_libdrm_buffer(buffer)->bo->handle;
-
-   if (BRW_DUMP)
-      debug_printf("   -> pitch %d handle 0x%x\n", *pitch, *handle);
-
-   return TRUE;
-}
-
 static void
 i965_libdrm_winsys_destroy(struct brw_winsys_screen *iws)
 {
@@ -225,9 +102,6 @@ struct drm_api i965_libdrm_api =
 {
    .name = "i965",
    .create_screen = i965_libdrm_create_screen,
-   .texture_from_shared_handle = i965_libdrm_texture_from_shared_handle,
-   .shared_handle_from_texture = i965_libdrm_shared_handle_from_texture,
-   .local_handle_from_texture = i965_libdrm_local_handle_from_texture,
    .destroy = destroy,
 };
 
diff --git a/src/gallium/winsys/drm/i965/gem/i965_drm_buffer.c b/src/gallium/winsys/drm/i965/gem/i965_drm_buffer.c
index 07be1df87f0..33a17496b2b 100644
--- a/src/gallium/winsys/drm/i965/gem/i965_drm_buffer.c
+++ b/src/gallium/winsys/drm/i965/gem/i965_drm_buffer.c
@@ -1,4 +1,5 @@
 
+#include "state_tracker/drm_api.h"
 #include "i965_drm_winsys.h"
 #include "util/u_memory.h"
 #include "util/u_inlines.h"
@@ -122,6 +123,78 @@ err:
    return PIPE_ERROR_OUT_OF_MEMORY;
 }
 
+static enum pipe_error
+i965_libdrm_bo_from_handle(struct brw_winsys_screen *sws,
+                           struct winsys_handle *whandle,
+                           unsigned *stride,
+                           unsigned *tile,
+                           struct brw_winsys_buffer **bo_out)
+{
+   struct i965_libdrm_winsys *idws = i965_libdrm_winsys(sws);
+   struct i965_libdrm_buffer *buf = CALLOC_STRUCT(i965_libdrm_buffer);
+   uint32_t swizzle = 0;
+
+   if (BRW_DUMP)
+      debug_printf("%s\n", __FUNCTION__);
+
+   if (!buf)
+      return PIPE_ERROR_OUT_OF_MEMORY;
+
+   pipe_reference_init(&buf->base.reference, 1);
+   buf->bo = drm_intel_bo_gem_create_from_name(idws->gem, "FROM_HANDLE", whandle->handle);
+   buf->base.size = buf->bo->size;
+   buf->base.sws = &idws->base;
+   buf->flinked = TRUE;
+   buf->flink = whandle->handle;
+
+
+   if (!buf->bo)
+      goto err;
+
+   drm_intel_bo_get_tiling(buf->bo, &buf->tiling, &swizzle);
+   if (buf->tiling != 0)
+      buf->map_gtt = TRUE;
+
+   *tile = buf->tiling;
+   *stride = whandle->stride;
+
+   *bo_out = &buf->base;
+   return PIPE_OK;
+
+err:
+   FREE(buf);
+   return PIPE_ERROR_OUT_OF_MEMORY;
+}
+
+static enum pipe_error
+i965_libdrm_bo_get_handle(struct brw_winsys_buffer *buffer,
+                          struct winsys_handle *whandle,
+                          unsigned stride)
+{
+   struct i965_libdrm_buffer *buf = i965_libdrm_buffer(buffer);
+
+   if (BRW_DUMP)
+      debug_printf("%s\n", __FUNCTION__);
+
+   if (whandle->type == DRM_API_HANDLE_TYPE_SHARED) {
+      if (!buf->flinked) {
+         if (drm_intel_bo_flink(buf->bo, &buf->flink))
+            return PIPE_ERROR_BAD_INPUT;
+         buf->flinked = TRUE;
+      }
+
+      whandle->handle = buf->flink;
+   } else if (whandle->type == DRM_API_HANDLE_TYPE_KMS) {
+      whandle->handle = buf->bo->handle;
+   } else {
+      assert(!"unknown usage");
+      return PIPE_ERROR_BAD_INPUT;
+   }
+
+   whandle->stride = stride;
+   return PIPE_OK;
+}
+
 static void 
 i965_libdrm_bo_destroy(struct brw_winsys_buffer *buffer)
 {
@@ -415,6 +488,8 @@ void
 i965_libdrm_winsys_init_buffer_functions(struct i965_libdrm_winsys *idws)
 {
    idws->base.bo_alloc             = i965_libdrm_bo_alloc;
+   idws->base.bo_from_handle       = i965_libdrm_bo_from_handle;
+   idws->base.bo_get_handle        = i965_libdrm_bo_get_handle;
    idws->base.bo_destroy           = i965_libdrm_bo_destroy;
    idws->base.bo_emit_reloc        = i965_libdrm_bo_emit_reloc;
    idws->base.bo_exec              = i965_libdrm_bo_exec;
diff --git a/src/gallium/winsys/drm/intel/gem/intel_drm_api.c b/src/gallium/winsys/drm/intel/gem/intel_drm_api.c
index 377ed255131..e3b980a832b 100644
--- a/src/gallium/winsys/drm/intel/gem/intel_drm_api.c
+++ b/src/gallium/winsys/drm/intel/gem/intel_drm_api.c
@@ -38,99 +38,6 @@ intel_drm_get_device_id(unsigned int *device_id)
    fclose(file);
 }
 
-static struct intel_buffer *
-intel_drm_buffer_from_handle(struct intel_drm_winsys *idws,
-                             const char* name, unsigned handle)
-{
-   struct intel_drm_buffer *buf = CALLOC_STRUCT(intel_drm_buffer);
-   uint32_t tile = 0, swizzle = 0;
-
-   if (!buf)
-      return NULL;
-
-   buf->magic = 0xDEAD1337;
-   buf->bo = drm_intel_bo_gem_create_from_name(idws->pools.gem, name, handle);
-   buf->flinked = TRUE;
-   buf->flink = handle;
-
-   if (!buf->bo)
-      goto err;
-
-   drm_intel_bo_get_tiling(buf->bo, &tile, &swizzle);
-   if (tile != INTEL_TILE_NONE)
-      buf->map_gtt = TRUE;
-
-   return (struct intel_buffer *)buf;
-
-err:
-   FREE(buf);
-   return NULL;
-}
-
-
-/*
- * Exported functions
- */
-
-
-static struct pipe_texture *
-intel_drm_texture_from_shared_handle(struct drm_api *api,
-                                     struct pipe_screen *screen,
-                                     struct pipe_texture *templ,
-                                     const char* name,
-                                     unsigned pitch,
-                                     unsigned handle)
-{
-   struct intel_drm_winsys *idws = intel_drm_winsys(i915_screen(screen)->iws);
-   struct intel_buffer *buffer;
-
-   buffer = intel_drm_buffer_from_handle(idws, name, handle);
-   if (!buffer)
-      return NULL;
-
-   return i915_texture_blanket_intel(screen, templ, pitch, buffer);
-}
-
-static boolean
-intel_drm_shared_handle_from_texture(struct drm_api *api,
-                                     struct pipe_screen *screen,
-                                     struct pipe_texture *texture,
-                                     unsigned *pitch,
-                                     unsigned *handle)
-{
-   struct intel_drm_buffer *buf = NULL;
-   struct intel_buffer *buffer = NULL;
-   if (!i915_get_texture_buffer_intel(texture, &buffer, pitch))
-      return FALSE;
-
-   buf = intel_drm_buffer(buffer);
-   if (!buf->flinked) {
-      if (drm_intel_bo_flink(buf->bo, &buf->flink))
-         return FALSE;
-      buf->flinked = TRUE;
-   }
-
-   *handle = buf->flink;
-
-   return TRUE;
-}
-
-static boolean
-intel_drm_local_handle_from_texture(struct drm_api *api,
-                                    struct pipe_screen *screen,
-                                    struct pipe_texture *texture,
-                                    unsigned *pitch,
-                                    unsigned *handle)
-{
-   struct intel_buffer *buffer = NULL;
-   if (!i915_get_texture_buffer_intel(texture, &buffer, pitch))
-      return FALSE;
-
-   *handle = intel_drm_buffer(buffer)->bo->handle;
-
-   return TRUE;
-}
-
 static void
 intel_drm_winsys_destroy(struct intel_winsys *iws)
 {
@@ -192,9 +99,6 @@ struct drm_api intel_drm_api =
    .name = "i915",
    .driver_name = "i915",
    .create_screen = intel_drm_create_screen,
-   .texture_from_shared_handle = intel_drm_texture_from_shared_handle,
-   .shared_handle_from_texture = intel_drm_shared_handle_from_texture,
-   .local_handle_from_texture = intel_drm_local_handle_from_texture,
    .destroy = destroy,
 };
 
diff --git a/src/gallium/winsys/drm/intel/gem/intel_drm_buffer.c b/src/gallium/winsys/drm/intel/gem/intel_drm_buffer.c
index ac4dd6e00e9..cb4f92a3b17 100644
--- a/src/gallium/winsys/drm/intel/gem/intel_drm_buffer.c
+++ b/src/gallium/winsys/drm/intel/gem/intel_drm_buffer.c
@@ -1,4 +1,5 @@
 
+#include "state_tracker/drm_api.h"
 #include "intel_drm_winsys.h"
 #include "util/u_memory.h"
 
@@ -52,6 +53,66 @@ err:
    return NULL;
 }
 
+static struct intel_buffer *
+intel_drm_buffer_from_handle(struct intel_winsys *iws,
+                             struct winsys_handle *whandle,
+                             unsigned *stride)
+{
+   struct intel_drm_winsys *idws = intel_drm_winsys(iws);
+   struct intel_drm_buffer *buf = CALLOC_STRUCT(intel_drm_buffer);
+   uint32_t tile = 0, swizzle = 0;
+
+   if (!buf)
+      return NULL;
+
+   buf->magic = 0xDEAD1337;
+   buf->bo = drm_intel_bo_gem_create_from_name(idws->pools.gem, "gallium3d_from_handle", whandle->handle);
+   buf->flinked = TRUE;
+   buf->flink = whandle->handle;
+
+   if (!buf->bo)
+      goto err;
+
+   drm_intel_bo_get_tiling(buf->bo, &tile, &swizzle);
+   if (tile != INTEL_TILE_NONE)
+      buf->map_gtt = TRUE;
+
+   *stride = whandle->stride;
+
+   return (struct intel_buffer *)buf;
+
+err:
+   FREE(buf);
+   return NULL;
+}
+
+static boolean
+intel_drm_buffer_get_handle(struct intel_winsys *iws,
+                            struct intel_buffer *buffer,
+                            struct winsys_handle *whandle,
+                            unsigned stride)
+{
+   struct intel_drm_buffer *buf = intel_drm_buffer(buffer);
+
+   if (whandle->type == DRM_API_HANDLE_TYPE_SHARED) {
+      if (!buf->flinked) {
+         if (drm_intel_bo_flink(buf->bo, &buf->flink))
+            return FALSE;
+         buf->flinked = TRUE;
+      }
+
+      whandle->handle = buf->flink;
+   } else if (whandle->type == DRM_API_HANDLE_TYPE_KMS) {
+      whandle->handle = buf->bo->handle;
+   } else {
+      assert(!"unknown usage");
+      return FALSE;
+   }
+
+   whandle->stride = stride;
+   return TRUE;
+}
+
 static int
 intel_drm_buffer_set_fence_reg(struct intel_winsys *iws,
                                struct intel_buffer *buffer,
@@ -146,6 +207,8 @@ void
 intel_drm_winsys_init_buffer_functions(struct intel_drm_winsys *idws)
 {
    idws->base.buffer_create = intel_drm_buffer_create;
+   idws->base.buffer_from_handle = intel_drm_buffer_from_handle;
+   idws->base.buffer_get_handle = intel_drm_buffer_get_handle;
    idws->base.buffer_set_fence_reg = intel_drm_buffer_set_fence_reg;
    idws->base.buffer_map = intel_drm_buffer_map;
    idws->base.buffer_unmap = intel_drm_buffer_unmap;
diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
index 0d05f316c42..21517b4bb54 100644
--- a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
+++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
@@ -21,9 +21,10 @@ dri_surface_from_handle(struct drm_api *api, struct pipe_screen *pscreen,
 	struct pipe_surface *ps = NULL;
 	struct pipe_texture *pt = NULL;
 	struct pipe_texture tmpl;
+	struct winsys_handle whandle;
 
 	memset(&tmpl, 0, sizeof(tmpl));
-	tmpl.tex_usage = PIPE_TEXTURE_USAGE_PRIMARY;
+	tmpl.tex_usage = PIPE_TEXTURE_USAGE_SCANOUT;
 	tmpl.target = PIPE_TEXTURE_2D;
 	tmpl.last_level = 0;
 	tmpl.depth0 = 1;
@@ -31,8 +32,11 @@ dri_surface_from_handle(struct drm_api *api, struct pipe_screen *pscreen,
 	tmpl.width0 = width;
 	tmpl.height0 = height;
 
-	pt = api->texture_from_shared_handle(api, pscreen, &tmpl,
-					     "front buffer", pitch, handle);
+	memset(&whandle, 0, sizeof(whandle));
+	whandle.stride = pitch;
+	whandle.handle = handle;
+
+	pt = pscreen->texture_from_handle(pscreen, &tmpl, &whandle);
 	if (!pt)
 		return NULL;
 
@@ -142,74 +146,10 @@ nouveau_drm_create_screen(struct drm_api *api, int fd,
 	return nvws->pscreen;
 }
 
-static struct pipe_texture *
-nouveau_drm_pt_from_name(struct drm_api *api, struct pipe_screen *pscreen,
-			 struct pipe_texture *templ, const char *name,
-			 unsigned stride, unsigned handle)
-{
-	struct nouveau_device *dev = nouveau_screen(pscreen)->device;
-	struct pipe_texture *pt;
-	struct pipe_buffer *pb;
-	int ret;
-
-	pb = CALLOC(1, sizeof(struct pipe_buffer) + sizeof(struct nouveau_bo*));
-	if (!pb)
-		return NULL;
-
-	ret = nouveau_bo_handle_ref(dev, handle, (struct nouveau_bo**)(pb+1));
-	if (ret) {
-		debug_printf("%s: ref name 0x%08x failed with %d\n",
-			     __func__, handle, ret);
-		FREE(pb);
-		return NULL;
-	}
-
-	pipe_reference_init(&pb->reference, 1);
-	pb->screen = pscreen;
-	pb->alignment = 0;
-	pb->usage = PIPE_BUFFER_USAGE_GPU_READ_WRITE |
-		    PIPE_BUFFER_USAGE_CPU_READ_WRITE;
-	pb->size = nouveau_bo(pb)->size;
-	pt = pscreen->texture_blanket(pscreen, templ, &stride, pb);
-	pipe_buffer_reference(&pb, NULL);
-	return pt;
-}
-
-static boolean
-nouveau_drm_name_from_pt(struct drm_api *api, struct pipe_screen *pscreen,
-			 struct pipe_texture *pt, unsigned *stride,
-			 unsigned *handle)
-{
-	struct nouveau_miptree *mt = nouveau_miptree(pt);
-
-	if (!mt || !mt->bo)
-		return false;
-
-	return nouveau_bo_handle_get(mt->bo, handle) == 0;
-}
-
-static boolean
-nouveau_drm_handle_from_pt(struct drm_api *api, struct pipe_screen *pscreen,
-			   struct pipe_texture *pt, unsigned *stride,
-			   unsigned *handle)
-{
-	struct nouveau_miptree *mt = nouveau_miptree(pt);
-
-	if (!mt || !mt->bo)
-		return false;
-
-	*handle = mt->bo->handle;
-	*stride = util_format_get_stride(mt->base.format, mt->base.width0);
-	return true;
-}
-
 struct drm_api drm_api_hooks = {
 	.name = "nouveau",
 	.driver_name = "nouveau",
 	.create_screen = nouveau_drm_create_screen,
-	.texture_from_shared_handle = nouveau_drm_pt_from_name,
-	.shared_handle_from_texture = nouveau_drm_name_from_pt,
-	.local_handle_from_texture = nouveau_drm_handle_from_pt,
 };
 
 struct drm_api *
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
index e5c69199330..daa032af6f2 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
+++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c
@@ -264,6 +264,75 @@ static int radeon_fence_finish(struct pipe_winsys *ws,
     return 0;
 }
 
+/* Create a buffer from a handle. */
+static struct pipe_buffer* radeon_buffer_from_handle(struct radeon_winsys *radeon_ws,
+                                                     struct pipe_screen *screen,
+                                                     struct winsys_handle *whandle,
+                                                     unsigned *stride)
+{
+    struct radeon_bo_manager* bom = radeon_ws->priv->bom;
+    struct radeon_pipe_buffer* radeon_buffer;
+    struct radeon_bo* bo = NULL;
+
+    bo = radeon_bo_open(bom, whandle->handle, 0, 0, 0, 0);
+    if (bo == NULL) {
+        return NULL;
+    }
+
+    radeon_buffer = CALLOC_STRUCT(radeon_pipe_buffer);
+    if (radeon_buffer == NULL) {
+        radeon_bo_unref(bo);
+        return NULL;
+    }
+
+    pipe_reference_init(&radeon_buffer->base.reference, 1);
+    radeon_buffer->base.screen = screen;
+    radeon_buffer->base.usage = PIPE_BUFFER_USAGE_PIXEL;
+    radeon_buffer->bo = bo;
+
+    *stride = whandle->stride;
+
+    return &radeon_buffer->base;
+}
+
+static boolean radeon_buffer_get_handle(struct radeon_winsys *radeon_ws,
+                                        struct pipe_buffer *buffer,
+                                        unsigned stride,
+                                        struct winsys_handle *whandle)
+{
+    int retval, fd;
+    struct drm_gem_flink flink;
+    struct radeon_pipe_buffer* radeon_buffer;
+
+    radeon_buffer = (struct radeon_pipe_buffer*)buffer;
+
+
+    if (whandle->type == DRM_API_HANDLE_TYPE_SHARED) {
+        if (!radeon_buffer->flinked) {
+            fd = radeon_ws->priv->fd;
+
+            flink.handle = radeon_buffer->bo->handle;
+
+            retval = ioctl(fd, DRM_IOCTL_GEM_FLINK, &flink);
+            if (retval) {
+                debug_printf("radeon: DRM_IOCTL_GEM_FLINK failed, error %d\n",
+                             retval);
+                return FALSE;
+            }
+
+            radeon_buffer->flink = flink.name;
+            radeon_buffer->flinked = TRUE;
+        }
+
+        whandle->handle = radeon_buffer->flink;
+    } else if (whandle->type == DRM_API_HANDLE_TYPE_KMS) {
+        whandle->handle = ((struct radeon_pipe_buffer*)buffer)->bo->handle;
+    }
+    whandle->stride = stride;
+
+    return TRUE;
+}
+
 struct radeon_winsys* radeon_pipe_winsys(int fd)
 {
     struct radeon_winsys* radeon_ws;
@@ -298,6 +367,8 @@ struct radeon_winsys* radeon_pipe_winsys(int fd)
     radeon_ws->base.get_name = radeon_get_name;
 
     radeon_ws->buffer_set_tiling = radeon_buffer_set_tiling;
+    radeon_ws->buffer_from_handle = radeon_buffer_from_handle;
+    radeon_ws->buffer_get_handle = radeon_buffer_get_handle;
 
     return radeon_ws;
 }
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_drm.c b/src/gallium/winsys/drm/radeon/core/radeon_drm.c
index e817a26da6d..97edb6a47e3 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_drm.c
+++ b/src/gallium/winsys/drm/radeon/core/radeon_drm.c
@@ -148,124 +148,6 @@ struct pipe_screen* radeon_create_screen(struct drm_api* api,
     }
 }
 
-
-boolean radeon_buffer_from_texture(struct drm_api* api,
-                                   struct pipe_screen* screen,
-                                   struct pipe_texture* texture,
-                                   struct pipe_buffer** buffer,
-                                   unsigned* stride)
-{
-    /* XXX fix this */
-    return r300_get_texture_buffer(screen, texture, buffer, stride);
-}
-
-/* Create a buffer from a handle. */
-/* XXX what's up with name? */
-struct pipe_buffer* radeon_buffer_from_handle(struct drm_api* api,
-                                              struct pipe_screen* screen,
-                                              const char* name,
-                                              unsigned handle)
-{
-    struct radeon_bo_manager* bom =
-        ((struct radeon_winsys*)screen->winsys)->priv->bom;
-    struct radeon_pipe_buffer* radeon_buffer;
-    struct radeon_bo* bo = NULL;
-
-    bo = radeon_bo_open(bom, handle, 0, 0, 0, 0);
-    if (bo == NULL) {
-        return NULL;
-    }
-
-    radeon_buffer = CALLOC_STRUCT(radeon_pipe_buffer);
-    if (radeon_buffer == NULL) {
-        radeon_bo_unref(bo);
-        return NULL;
-    }
-
-    pipe_reference_init(&radeon_buffer->base.reference, 1);
-    radeon_buffer->base.screen = screen;
-    radeon_buffer->base.usage = PIPE_BUFFER_USAGE_PIXEL;
-    radeon_buffer->bo = bo;
-    return &radeon_buffer->base;
-}
-
-static struct pipe_texture*
-radeon_texture_from_shared_handle(struct drm_api *api,
-                                  struct pipe_screen *screen,
-                                  struct pipe_texture *templ,
-                                  const char *name,
-                                  unsigned stride,
-                                  unsigned handle)
-{
-    struct pipe_buffer *buffer;
-    struct pipe_texture *blanket;
-
-    buffer = radeon_buffer_from_handle(api, screen, name, handle);
-    if (!buffer) {
-        return NULL;
-    }
-
-    blanket = screen->texture_blanket(screen, templ, &stride, buffer);
-
-    pipe_buffer_reference(&buffer, NULL);
-
-    return blanket;
-}
-
-static boolean radeon_shared_handle_from_texture(struct drm_api *api,
-                                                 struct pipe_screen *screen,
-                                                 struct pipe_texture *texture,
-                                                 unsigned *stride,
-                                                 unsigned *handle)
-{
-    int retval, fd;
-    struct drm_gem_flink flink;
-    struct radeon_pipe_buffer* radeon_buffer;
-    struct pipe_buffer *buffer = NULL;
-
-    if (!radeon_buffer_from_texture(api, screen, texture, &buffer, stride)) {
-        return FALSE;
-    }
-
-    radeon_buffer = (struct radeon_pipe_buffer*)buffer;
-    if (!radeon_buffer->flinked) {
-        fd = ((struct radeon_winsys*)screen->winsys)->priv->fd;
-
-        flink.handle = radeon_buffer->bo->handle;
-
-        retval = ioctl(fd, DRM_IOCTL_GEM_FLINK, &flink);
-        if (retval) {
-            debug_printf("radeon: DRM_IOCTL_GEM_FLINK failed, error %d\n",
-                    retval);
-            return FALSE;
-        }
-
-        radeon_buffer->flink = flink.name;
-        radeon_buffer->flinked = TRUE;
-    }
-
-    *handle = radeon_buffer->flink;
-    return TRUE;
-}
-
-static boolean radeon_local_handle_from_texture(struct drm_api *api,
-                                                struct pipe_screen *screen,
-                                                struct pipe_texture *texture,
-                                                unsigned *stride,
-                                                unsigned *handle)
-{
-    struct pipe_buffer *buffer = NULL;
-    if (!radeon_buffer_from_texture(api, screen, texture, &buffer, stride)) {
-        return FALSE;
-    }
-
-    *handle = ((struct radeon_pipe_buffer*)buffer)->bo->handle;
-
-    pipe_buffer_reference(&buffer, NULL);
-
-    return TRUE;
-}
-
 static void radeon_drm_api_destroy(struct drm_api *api)
 {
     return;
@@ -275,9 +157,6 @@ struct drm_api drm_api_hooks = {
     .name = "radeon",
     .driver_name = "radeon",
     .create_screen = radeon_create_screen,
-    .texture_from_shared_handle = radeon_texture_from_shared_handle,
-    .shared_handle_from_texture = radeon_shared_handle_from_texture,
-    .local_handle_from_texture = radeon_local_handle_from_texture,
     .destroy = radeon_drm_api_destroy,
 };
 
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_drm.h b/src/gallium/winsys/drm/radeon/core/radeon_drm.h
index f62a9b80485..78451b6f011 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_drm.h
+++ b/src/gallium/winsys/drm/radeon/core/radeon_drm.h
@@ -37,28 +37,6 @@ struct pipe_screen* radeon_create_screen(struct drm_api* api,
                                          int drmFB,
 					 struct drm_create_screen_arg *arg);
 
-
-boolean radeon_buffer_from_texture(struct drm_api* api,
-                                   struct pipe_screen* screen,
-                                   struct pipe_texture* texture,
-                                   struct pipe_buffer** buffer,
-                                   unsigned* stride);
-
-struct pipe_buffer* radeon_buffer_from_handle(struct drm_api* api,
-                                              struct pipe_screen* screen,
-                                              const char* name,
-                                              unsigned handle);
-
-boolean radeon_handle_from_buffer(struct drm_api* api,
-                                  struct pipe_screen* screen,
-                                  struct pipe_buffer* buffer,
-                                  unsigned* handle);
-
-boolean radeon_global_handle_from_buffer(struct drm_api* api,
-                                         struct pipe_screen* screen,
-                                         struct pipe_buffer* buffer,
-                                         unsigned* handle);
-
 void radeon_destroy_drm_api(struct drm_api* api);
 
 /* Guess at whether this chipset should use r300g.
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_winsys.h b/src/gallium/winsys/drm/radeon/core/radeon_winsys.h
index 4901080ca7b..37eeb459791 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_winsys.h
+++ b/src/gallium/winsys/drm/radeon/core/radeon_winsys.h
@@ -56,6 +56,18 @@ struct radeon_winsys {
     /* VRAM size. */
     uint32_t vram_size;
 
+    /* Create a buffer from a winsys handle. */
+    struct pipe_buffer *(*buffer_from_handle)(struct radeon_winsys *winsys,
+                                              struct pipe_screen *screen,
+                                              struct winsys_handle *whandle,
+                                              unsigned *stride);
+
+    /* Get the handle from a buffer. */
+    boolean (*buffer_get_handle)(struct radeon_winsys *winsys,
+                                 struct pipe_buffer *buffer,
+                                 unsigned stride,
+                                 struct winsys_handle *whandle);
+
     /* Add a pipe_buffer to the list of buffer objects to validate. */
     boolean (*add_buffer)(struct radeon_winsys* winsys,
                           struct pipe_buffer* pbuffer,
diff --git a/src/gallium/winsys/drm/vmware/core/vmw_screen_dri.c b/src/gallium/winsys/drm/vmware/core/vmw_screen_dri.c
index a7c6e88b9eb..657544dcb21 100644
--- a/src/gallium/winsys/drm/vmware/core/vmw_screen_dri.c
+++ b/src/gallium/winsys/drm/vmware/core/vmw_screen_dri.c
@@ -44,6 +44,16 @@
 
 #include <stdio.h>
 
+static struct svga_winsys_surface *
+vmw_drm_surface_from_handle(struct svga_winsys_screen *sws,
+			    struct winsys_handle *whandle,
+			    SVGA3dSurfaceFormat *format);
+static boolean
+vmw_drm_surface_get_handle(struct svga_winsys_screen *sws,
+			   struct svga_winsys_surface *surface,
+			   unsigned stride,
+			   struct winsys_handle *whandle);
+
 static struct dri1_api dri1_api_hooks;
 static struct dri1_api_version ddx_required = { 0, 1, 0 };
 static struct dri1_api_version ddx_compat = { 0, 0, 0 };
@@ -129,7 +139,12 @@ vmw_drm_create_screen(struct drm_api *drm_api,
 				     &drm_compat, "use old scanout field (not a error)"))
 	    use_old_scanout_flag = TRUE;
 	 dri1->api = &dri1_api_hooks;
+#if 0
 	 break;
+#else
+	 assert(!"No dri 1 support for now\n");
+	 return NULL;
+#endif
       default:
 	 return NULL;
       }
@@ -139,6 +154,10 @@ vmw_drm_create_screen(struct drm_api *drm_api,
    if (!vws)
       goto out_no_vws;
 
+   /* XXX do this properly */
+   vws->base.surface_from_handle = vmw_drm_surface_from_handle;
+   vws->base.surface_get_handle = vmw_drm_surface_get_handle;
+
    screen = svga_screen_create( &vws->base );
    if (!screen)
       goto out_no_screen;
@@ -200,6 +219,7 @@ vmw_dri1_present_locked(struct pipe_context *locked_pipe,
 			const struct drm_clip_rect *bbox,
 			struct pipe_fence_handle **p_fence)
 {
+#if 0
    struct svga_winsys_surface *srf =
       svga_screen_texture_get_winsys_surface(surf->texture);
    struct vmw_svga_winsys_surface *vsrf = vmw_svga_winsys_surface(srf);
@@ -246,21 +266,19 @@ vmw_dri1_present_locked(struct pipe_context *locked_pipe,
 
    *p_fence = (visible) ? vmw_pipe_fence(fence_seq) : NULL;
    vmw_svga_winsys_surface_reference(&vsrf, NULL);
+#else
+   assert(!"No dri 1 support for now\n");
+#endif
 }
 
-static struct pipe_texture *
-vmw_drm_texture_from_handle(struct drm_api *drm_api,
-			    struct pipe_screen *screen,
-			    struct pipe_texture *templat,
-			    const char *name,
-			    unsigned stride,
-			    unsigned handle)
+static struct svga_winsys_surface *
+vmw_drm_surface_from_handle(struct svga_winsys_screen *sws,
+			    struct winsys_handle *whandle,
+			    SVGA3dSurfaceFormat *format)
 {
     struct vmw_svga_winsys_surface *vsrf;
     struct svga_winsys_surface *ssrf;
-    struct vmw_winsys_screen *vws =
-	vmw_winsys_screen(svga_winsys_screen(screen));
-    struct pipe_texture *tex;
+    struct vmw_winsys_screen *vws = vmw_winsys_screen(sws);
     union drm_vmw_surface_reference_arg arg;
     struct drm_vmw_surface_arg *req = &arg.req;
     struct drm_vmw_surface_create_req *rep = &arg.rep;
@@ -273,7 +291,7 @@ vmw_drm_texture_from_handle(struct drm_api *drm_api,
      */
 
     memset(&arg, 0, sizeof(arg));
-    req->sid = handle;
+    req->sid = whandle->handle;
 
     ret = drmCommandWriteRead(vws->ioctl.drm_fd, DRM_VMW_REF_SURFACE,
 			      &arg, sizeof(arg));
@@ -281,14 +299,14 @@ vmw_drm_texture_from_handle(struct drm_api *drm_api,
     if (ret) {
 	fprintf(stderr, "Failed referencing shared surface. SID %d.\n"
 		"Error %d (%s).\n",
-		handle, ret, strerror(-ret));
+		whandle->handle, ret, strerror(-ret));
 	return NULL;
     }
 
     if (rep->mip_levels[0] != 1) {
 	fprintf(stderr, "Incorrect number of mipmap levels on shared surface."
 		" SID %d, levels %d\n",
-		handle, rep->mip_levels[0]);
+		whandle->handle, rep->mip_levels[0]);
 	goto out_mip;
     }
 
@@ -296,7 +314,7 @@ vmw_drm_texture_from_handle(struct drm_api *drm_api,
 	if (rep->mip_levels[i] != 0) {
 	    fprintf(stderr, "Incorrect number of faces levels on shared surface."
 		    " SID %d, face %d present.\n",
-		    handle, i);
+		    whandle->handle, i);
 	    goto out_mip;
 	}
     }
@@ -308,38 +326,32 @@ vmw_drm_texture_from_handle(struct drm_api *drm_api,
     pipe_reference_init(&vsrf->refcnt, 1);
     p_atomic_set(&vsrf->validated, 0);
     vsrf->screen = vws;
-    vsrf->sid = handle;
+    vsrf->sid = whandle->handle;
     ssrf = svga_winsys_surface(vsrf);
-    tex = svga_screen_texture_wrap_surface(screen, templat, rep->format, ssrf);
-    if (!tex)
-	vmw_svga_winsys_surface_reference(&vsrf, NULL);
+    *format = rep->format;
+
+    return ssrf;
 
-    return tex;
-  out_mip:
-    vmw_ioctl_surface_destroy(vws, handle);
+out_mip:
+    vmw_ioctl_surface_destroy(vws, whandle->handle);
     return NULL;
 }
 
 static boolean
-vmw_drm_handle_from_texture(struct drm_api *drm_api,
-                           struct pipe_screen *screen,
-			   struct pipe_texture *texture,
-			   unsigned *stride,
-			   unsigned *handle)
+vmw_drm_surface_get_handle(struct svga_winsys_screen *sws,
+			   struct svga_winsys_surface *surface,
+			   unsigned stride,
+			   struct winsys_handle *whandle)
 {
-    struct svga_winsys_surface *surface =
-	svga_screen_texture_get_winsys_surface(texture);
     struct vmw_svga_winsys_surface *vsrf;
 
     if (!surface)
 	return FALSE;
 
     vsrf = vmw_svga_winsys_surface(surface);
-    *handle = vsrf->sid;
-    *stride = util_format_get_nblocksx(texture->format, texture->width0) *
-       util_format_get_blocksize(texture->format);
+    whandle->handle = vsrf->sid;
+    whandle->stride = stride;
 
-    vmw_svga_winsys_surface_reference(&vsrf, NULL);
     return TRUE;
 }
 
@@ -353,9 +365,6 @@ static struct drm_api vmw_drm_api_hooks = {
    .name = "vmwgfx",
    .driver_name = "vmwgfx",
    .create_screen = vmw_drm_create_screen,
-   .texture_from_shared_handle = vmw_drm_texture_from_handle,
-   .shared_handle_from_texture = vmw_drm_handle_from_texture,
-   .local_handle_from_texture = vmw_drm_handle_from_texture,
 };
 
 struct drm_api* drm_api_create()