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

58 files changed, 1231 insertions, 1230 deletions

diff --git a/src/gallium/drivers/i915/i915_context.h b/src/gallium/drivers/i915/i915_context.h
index da769e7b29..499a727314 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 72bd263550..e5bf4a20bd 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 7ba222c78b..3ce52cdcdb 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 b3a802b0e2..00fd0c1efe 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 66f3aad8b2..cef83ffea8 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 7226d9228b..e0f3cd2a9f 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 e38fdf1869..cc79bfc771 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 c82d00f4a4..f30c7f1813 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 f258c38cd7..936ccc444a 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 b85492114a..b9d0f003d7 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 b666ffc804..08c8f93794 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 8df3015d4b..5cc7a12c03 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 0f83eea685..dd9a8e8856 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 6759e39e01..9bb3043320 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 fe60eb0437..fbe14924cb 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 24291da91e..671e74465c 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 2533275dc1..5a3cf37d6d 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 ac3d62dc96..74b7b4ec5e 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 6a6d9f9222..b23f929b16 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 3c2f771b51..f7d10a591f 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 a7927d88df..f4a7a2bc23 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 697b1b92ff..5ef74a832d 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 e42e872de7..66c26360cb 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 85d7e1f197..62e97bcea4 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 192074e747..e2e69420ea 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 c540594b94..044437e75f 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 3f9d869d7a..e091cae602 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 1f69daec81..61b54af4dd 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 183aa17f9b..27b2e30993 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 86b98a4ba5..e0a5532327 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 0d1518a05b..c2825d5f26 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 addb28bded..51fc590e5d 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 e37d309270..70de152713 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 3c2625269b..9e71e61c30 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 770a92be74..c8420bcdd5 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 d397a8eb2b..5880eecd5f 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 <MostAwesomeDude@gmail.com>
+ * Copyright 2010 Marek Olšák <maraeo@gmail.com>
  *
  * 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 502fbfa5a2..484bde6a6b 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 <MostAwesomeDude@gmail.com>
+ * Copyright 2010 Marek Olšák <maraeo@gmail.com>
  *
  * 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 1f6f99d3e5..8c9f604622 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 9c8e907fdf..e9e40747ef 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 2f3a56e1fb..af7827820c 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 c0144f64b4..04124afd4d 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 46a5fb6188..138b62784e 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 0000000000..ec89681a3c
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_transfer.c
@@ -0,0 +1,270 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ * Copyright 2010 Marek Olšák <maraeo@gmail.com>
+ *
+ * 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 0000000000..60d1d3dc85
--- /dev/null
+++ b/src/gallium/drivers/r300/r300_transfer.h
@@ -0,0 +1,31 @@
+/*
+ * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
+ * Copyright 2010 Marek Olšák <maraeo@gmail.com>
+ *
+ * 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 60a04bbfed..379939ac75 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 18cfeee3cd..f6f0b86b68 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 40fb8a95ca..ddf2b79003 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 98c08eaffa..6749243ab4 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 9eb3ff6875..44680ea70c 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 ecd6b39863..ef7ccf4189 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 9641652da8..11d184effb 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 e8d17ce7da..1c8636d1d5 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 3b30954ac8..aedfdf1b46 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 e9792e063e..5b581debfc 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 24c1f78ca5..ca6602b436 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 b4e3af0eaf..d4bb176f9a 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 2b4915003e..c16989fa52 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 388d83eb5c..a15addeb8c 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)