Merge master and fix conflicts

33 files changed, 1621 insertions, 730 deletions

diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c
index f79092291b..ca15ce1b47 100644
--- a/src/mesa/state_tracker/st_atom.c
+++ b/src/mesa/state_tracker/st_atom.c
@@ -37,11 +37,8 @@
 
        
 
-/* This is used to initialize st->atoms[].  We could use this list
- * directly except for a single atom, st_update_constants, which has a
- * .dirty value which changes according to the parameters of the
- * current fragment and vertex programs, and so cannot be a static
- * value.
+/**
+ * This is used to initialize st->atoms[].
  */
 static const struct st_tracked_state *atoms[] =
 {
@@ -67,34 +64,13 @@ static const struct st_tracked_state *atoms[] =
 
 void st_init_atoms( struct st_context *st )
 {
-   GLuint i;
-
-   st->atoms = _mesa_malloc(sizeof(atoms));
-   st->nr_atoms = sizeof(atoms)/sizeof(*atoms);
-   memcpy(st->atoms, atoms, sizeof(atoms));
-
-   /* Patch in a pointer to the dynamic state atom:
-    */
-   for (i = 0; i < st->nr_atoms; i++) {
-      if (st->atoms[i] == &st_update_vs_constants) {
-	 st->atoms[i] = &st->constants.tracked_state[PIPE_SHADER_VERTEX];
-	 st->atoms[i][0] = st_update_vs_constants;
-      }
-
-      if (st->atoms[i] == &st_update_fs_constants) {
-	 st->atoms[i] = &st->constants.tracked_state[PIPE_SHADER_FRAGMENT];
-	 st->atoms[i][0] = st_update_fs_constants;
-      }
-   }
+   /* no-op */
 }
 
 
 void st_destroy_atoms( struct st_context *st )
 {
-   if (st->atoms) {
-      _mesa_free(st->atoms);
-      st->atoms = NULL;
-   }
+   /* no-op */
 }
 
 
@@ -172,8 +148,8 @@ void st_validate_state( struct st_context *st )
       memset(&examined, 0, sizeof(examined));
       prev = *state;
 
-      for (i = 0; i < st->nr_atoms; i++) {	 
-	 const struct st_tracked_state *atom = st->atoms[i];
+      for (i = 0; i < Elements(atoms); i++) {	 
+	 const struct st_tracked_state *atom = atoms[i];
 	 struct st_state_flags generated;
 	 
 //	 _mesa_printf("atom %s %x/%x\n", atom->name, atom->dirty.mesa, atom->dirty.st);
@@ -185,7 +161,7 @@ void st_validate_state( struct st_context *st )
 	 }
 
 	 if (check_state(state, &atom->dirty)) {
-	    st->atoms[i]->update( st );
+	    atoms[i]->update( st );
 //	    _mesa_printf("after: %x\n", atom->dirty.mesa);
 	 }
 
@@ -203,11 +179,9 @@ void st_validate_state( struct st_context *st )
 
    }
    else {
-      const GLuint nr = st->nr_atoms;
-
-      for (i = 0; i < nr; i++) {	 
-	 if (check_state(state, &st->atoms[i]->dirty))
-	    st->atoms[i]->update( st );
+      for (i = 0; i < Elements(atoms); i++) {	 
+	 if (check_state(state, &atoms[i]->dirty))
+	    atoms[i]->update( st );
       }
    }
 
diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c
index fd81ac36d2..5d4d8eee02 100644
--- a/src/mesa/state_tracker/st_atom_constbuf.c
+++ b/src/mesa/state_tracker/st_atom_constbuf.c
@@ -24,11 +24,12 @@
  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  * 
  **************************************************************************/
- /*
-  * Authors:
-  *   Keith Whitwell <keith@tungstengraphics.com>
-  *   Brian Paul
-  */
+
+/*
+ * Authors:
+ *   Keith Whitwell <keith@tungstengraphics.com>
+ *   Brian Paul
+ */
 
 #include "main/imports.h"
 #include "shader/prog_parameter.h"
@@ -42,63 +43,61 @@
 #include "st_atom.h"
 #include "st_atom_constbuf.h"
 #include "st_program.h"
+#include "st_inlines.h"
 
 
 /**
  * Pass the given program parameters to the graphics pipe as a
  * constant buffer.
- * \param id  either PIPE_SHADER_VERTEX or PIPE_SHADER_FRAGMENT
+ * \param shader_type  either PIPE_SHADER_VERTEX or PIPE_SHADER_FRAGMENT
  */
 void st_upload_constants( struct st_context *st,
                           struct gl_program_parameter_list *params,
-                          unsigned id)
+                          unsigned shader_type)
 {
    struct pipe_context *pipe = st->pipe;
-   struct pipe_constant_buffer *cbuf = &st->state.constants[id];
+   struct pipe_constant_buffer *cbuf = &st->state.constants[shader_type];
 
-   assert(id == PIPE_SHADER_VERTEX || id == PIPE_SHADER_FRAGMENT);
+   assert(shader_type == PIPE_SHADER_VERTEX ||
+          shader_type == PIPE_SHADER_FRAGMENT);
 
    /* update constants */
    if (params && params->NumParameters) {
       const uint paramBytes = params->NumParameters * sizeof(GLfloat) * 4;
 
-      /* Update our own dependency flags.  This works because this
-       * function will also be called whenever the program changes.
-       */
-      st->constants.tracked_state[id].dirty.mesa =
-         (params->StateFlags | _NEW_PROGRAM);
-
       _mesa_load_state_parameters(st->ctx, params);
 
       /* We always need to get a new buffer, to keep the drivers simple and
        * avoid gratuitous rendering synchronization.
        */
       pipe_buffer_reference(&cbuf->buffer, NULL );
-      cbuf->buffer = pipe_buffer_create(pipe->screen, 16, PIPE_BUFFER_USAGE_CONSTANT,
+      cbuf->buffer = pipe_buffer_create(pipe->screen, 16,
+                                        PIPE_BUFFER_USAGE_CONSTANT,
 					paramBytes );
 
-      if (0)
-      {
-	 printf("%s(shader=%d, numParams=%d, stateFlags=0x%x)\n", 
-                __FUNCTION__, id, params->NumParameters, params->StateFlags);
+      if (0) {
+	 debug_printf("%s(shader=%d, numParams=%d, stateFlags=0x%x)\n", 
+                      __FUNCTION__, shader_type, params->NumParameters,
+                      params->StateFlags);
          _mesa_print_parameter_list(params);
       }
 
       /* load Mesa constants into the constant buffer */
       if (cbuf->buffer)
-         pipe_buffer_write(pipe->screen, cbuf->buffer, 
-                           0, paramBytes, 
-                           params->ParameterValues);
+         st_no_flush_pipe_buffer_write(st, cbuf->buffer,
+				       0, paramBytes,
+				       params->ParameterValues);
 
-      st->pipe->set_constant_buffer(st->pipe, id, 0, cbuf);
+      st->pipe->set_constant_buffer(st->pipe, shader_type, 0, cbuf);
    }
    else {
-      st->constants.tracked_state[id].dirty.mesa = 0;
-      //  st->pipe->set_constant_buffer(st->pipe, id, 0, NULL);
+      st->constants.tracked_state[shader_type].dirty.mesa = 0x0;
    }
 }
 
-/* Vertex shader:
+
+/**
+ * Vertex shader:
  */
 static void update_vs_constants(struct st_context *st )
 {
@@ -108,16 +107,20 @@ static void update_vs_constants(struct st_context *st )
    st_upload_constants( st, params, PIPE_SHADER_VERTEX );
 }
 
+
 const struct st_tracked_state st_update_vs_constants = {
    "st_update_vs_constants",				/* name */
    {							/* dirty */
-      0,  /* set dynamically above */			/* mesa */
+      (_NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS),          /* mesa */
       ST_NEW_VERTEX_PROGRAM,				/* st */
    },
    update_vs_constants					/* update */
 };
 
-/* Fragment shader:
+
+
+/**
+ * Fragment shader:
  */
 static void update_fs_constants(struct st_context *st )
 {
@@ -127,10 +130,11 @@ static void update_fs_constants(struct st_context *st )
    st_upload_constants( st, params, PIPE_SHADER_FRAGMENT );
 }
 
+
 const struct st_tracked_state st_update_fs_constants = {
    "st_update_fs_constants",				/* name */
    {							/* dirty */
-      0,  /* set dynamically above */			/* mesa */
+      (_NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS),          /* mesa */
       ST_NEW_FRAGMENT_PROGRAM,				/* st */
    },
    update_fs_constants					/* update */
diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c
index df0f0931ea..4d897b677e 100644
--- a/src/mesa/state_tracker/st_atom_framebuffer.c
+++ b/src/mesa/state_tracker/st_atom_framebuffer.c
@@ -98,8 +98,6 @@ update_framebuffer_state( struct st_context *st )
    struct st_renderbuffer *strb;
    GLuint i;
 
-   memset(framebuffer, 0, sizeof(*framebuffer));
-
    framebuffer->width = fb->Width;
    framebuffer->height = fb->Height;
 
@@ -120,12 +118,20 @@ update_framebuffer_state( struct st_context *st )
          }
 
          if (strb->surface) {
-            framebuffer->cbufs[framebuffer->nr_cbufs] = strb->surface;
+            pipe_surface_reference(&framebuffer->cbufs[framebuffer->nr_cbufs],
+                                   strb->surface);
             framebuffer->nr_cbufs++;
          }
+         strb->defined = GL_TRUE; /* we'll be drawing something */
       }
    }
+   for (i = framebuffer->nr_cbufs; i < PIPE_MAX_COLOR_BUFS; i++) {
+      pipe_surface_reference(&framebuffer->cbufs[i], NULL);
+   }
 
+   /*
+    * Depth/Stencil renderbuffer/surface.
+    */
    strb = st_renderbuffer(fb->Attachment[BUFFER_DEPTH].Renderbuffer);
    if (strb) {
       strb = st_renderbuffer(strb->Base.Wrapped);
@@ -133,15 +139,14 @@ update_framebuffer_state( struct st_context *st )
          /* rendering to a GL texture, may have to update surface */
          update_renderbuffer_surface(st, strb);
       }
-
-      framebuffer->zsbuf = strb->surface;
+      pipe_surface_reference(&framebuffer->zsbuf, strb->surface);
    }
    else {
       strb = st_renderbuffer(fb->Attachment[BUFFER_STENCIL].Renderbuffer);
       if (strb) {
          strb = st_renderbuffer(strb->Base.Wrapped);
          assert(strb->surface);
-         framebuffer->zsbuf = strb->surface;
+         pipe_surface_reference(&framebuffer->zsbuf, strb->surface);
       }
    }
 
diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c
index 8d0029dde5..eff3666ca8 100644
--- a/src/mesa/state_tracker/st_atom_pixeltransfer.c
+++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c
@@ -45,6 +45,7 @@
 #include "st_format.h"
 #include "st_program.h"
 #include "st_texture.h"
+#include "st_inlines.h"
 
 #include "pipe/p_screen.h"
 #include "pipe/p_context.h"
@@ -125,8 +126,7 @@ create_color_map_texture(GLcontext *ctx)
 
    /* create texture for color map/table */
    pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0,
-                          texSize, texSize, 1, 0,
-                          PIPE_TEXTURE_USAGE_SAMPLER);
+                          texSize, texSize, 1, PIPE_TEXTURE_USAGE_SAMPLER);
    return pt;
 }
 
@@ -148,8 +148,9 @@ load_color_map_texture(GLcontext *ctx, struct pipe_texture *pt)
    uint *dest;
    uint i, j;
 
-   transfer = screen->get_tex_transfer(screen, pt, 0, 0, 0, PIPE_TRANSFER_WRITE,
-                                       0, 0, texSize, texSize);
+   transfer = st_cond_flush_get_tex_transfer(st_context(ctx),
+					     pt, 0, 0, 0, PIPE_TRANSFER_WRITE,
+					     0, 0, texSize, texSize);
    dest = (uint *) screen->transfer_map(screen, transfer);
 
    /* Pack four 1D maps into a 2D texture:
diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c
index 61687fbc3e..5c7206409c 100644
--- a/src/mesa/state_tracker/st_atom_rasterizer.c
+++ b/src/mesa/state_tracker/st_atom_rasterizer.c
@@ -180,22 +180,7 @@ static void update_raster_state( struct st_context *st )
 
    if (ctx->Polygon.StippleFlag)
       raster->poly_stipple_enable = 1;
-
-
-   /* _NEW_BUFFERS, _NEW_POLYGON
-    */
-   if (raster->fill_cw != PIPE_POLYGON_MODE_FILL ||
-       raster->fill_ccw != PIPE_POLYGON_MODE_FILL)
-   {
-      GLfloat mrd = (ctx->DrawBuffer ? 
-		     ctx->DrawBuffer->_MRD : 
-		     1.0f);
-
-      raster->offset_units = ctx->Polygon.OffsetFactor * mrd;
-      raster->offset_scale = (ctx->Polygon.OffsetUnits * mrd *
-			    st->polygon_offset_scale);
-   }
-      
+     
    /* _NEW_POINT
     */
    raster->point_size = ctx->Point.Size;
@@ -207,7 +192,8 @@ static void update_raster_state( struct st_context *st )
    raster->point_sprite = ctx->Point.PointSprite;
    for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
       if (ctx->Point.CoordReplace[i]) {
-         if (ctx->Point.SpriteOrigin == GL_UPPER_LEFT)
+         if ((ctx->Point.SpriteOrigin == GL_UPPER_LEFT) ^
+             (st_fb_orientation(ctx->DrawBuffer) == Y_0_BOTTOM))
             raster->sprite_coord_mode[i] = PIPE_SPRITE_COORD_UPPER_LEFT;
          else 
             raster->sprite_coord_mode[i] = PIPE_SPRITE_COORD_LOWER_LEFT;
diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c
index fc1ff5be04..ee649be885 100644
--- a/src/mesa/state_tracker/st_atom_shader.c
+++ b/src/mesa/state_tracker/st_atom_shader.c
@@ -175,11 +175,12 @@ find_translated_vp(struct st_context *st,
 
    /* See if we need to translate vertex program to TGSI form */
    if (xvp->serialNo != stvp->serialNo) {
-      GLuint outAttr, dummySlot;
+      GLuint outAttr;
       const GLbitfield outputsWritten = stvp->Base.Base.OutputsWritten;
       GLuint numVpOuts = 0;
       GLboolean emitPntSize = GL_FALSE, emitBFC0 = GL_FALSE, emitBFC1 = GL_FALSE;
-      GLint maxGeneric;
+      GLbitfield usedGenerics = 0x0;
+      GLbitfield usedOutputSlots = 0x0;
 
       /* Compute mapping of vertex program outputs to slots, which depends
        * on the fragment program's input->slot mapping.
@@ -192,10 +193,12 @@ find_translated_vp(struct st_context *st,
 
          if (outAttr == VERT_RESULT_HPOS) {
             /* always put xformed position into slot zero */
-            xvp->output_to_slot[VERT_RESULT_HPOS] = 0;
+            GLuint slot = 0;
+            xvp->output_to_slot[VERT_RESULT_HPOS] = slot;
             xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_POSITION;
             xvp->output_to_semantic_index[outAttr] = 0;
             numVpOuts++;
+            usedOutputSlots |= (1 << slot);
          }
          else if (outputsWritten & (1 << outAttr)) {
             /* see if the frag prog wants this vert output */
@@ -209,6 +212,12 @@ find_translated_vp(struct st_context *st,
                   xvp->output_to_semantic_name[outAttr] = stfp->input_semantic_name[fpInSlot];
                   xvp->output_to_semantic_index[outAttr] = stfp->input_semantic_index[fpInSlot];
                   numVpOuts++;
+                  usedOutputSlots |= (1 << vpOutSlot);
+               }
+               else {
+#if 0 /*debug*/
+                  printf("VP output %d not used by FP\n", outAttr);
+#endif
                }
             }
             else if (outAttr == VERT_RESULT_PSIZ)
@@ -226,45 +235,49 @@ find_translated_vp(struct st_context *st,
 
       /* must do these last */
       if (emitPntSize) {
-         xvp->output_to_slot[VERT_RESULT_PSIZ] = numVpOuts++;
+         GLuint slot = numVpOuts++;
+         xvp->output_to_slot[VERT_RESULT_PSIZ] = slot;
          xvp->output_to_semantic_name[VERT_RESULT_PSIZ] = TGSI_SEMANTIC_PSIZE;
          xvp->output_to_semantic_index[VERT_RESULT_PSIZ] = 0;
+         usedOutputSlots |= (1 << slot);
       }
       if (emitBFC0) {
-         xvp->output_to_slot[VERT_RESULT_BFC0] = numVpOuts++;
+         GLuint slot = numVpOuts++;
+         xvp->output_to_slot[VERT_RESULT_BFC0] = slot;
          xvp->output_to_semantic_name[VERT_RESULT_BFC0] = TGSI_SEMANTIC_COLOR;
          xvp->output_to_semantic_index[VERT_RESULT_BFC0] = 0;
+         usedOutputSlots |= (1 << slot);
       }
       if (emitBFC1) {
-         xvp->output_to_slot[VERT_RESULT_BFC1] = numVpOuts++;
-         xvp->output_to_semantic_name[VERT_RESULT_BFC0] = TGSI_SEMANTIC_COLOR;
-         xvp->output_to_semantic_index[VERT_RESULT_BFC0] = 1;
+         GLuint slot = numVpOuts++;
+         xvp->output_to_slot[VERT_RESULT_BFC1] = slot;
+         xvp->output_to_semantic_name[VERT_RESULT_BFC1] = TGSI_SEMANTIC_COLOR;
+         xvp->output_to_semantic_index[VERT_RESULT_BFC1] = 1;
+         usedOutputSlots |= (1 << slot);
       }
 
-      /* Unneeded vertex program outputs will go to this slot.
-       * We could use this info to do dead code elimination in the
-       * vertex program.
-       */
-      dummySlot = numVpOuts;
-
-      /* find max GENERIC slot index */
-      maxGeneric = -1;
+      /* build usedGenerics mask */
+      usedGenerics = 0x0;
       for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) {
          if (xvp->output_to_semantic_name[outAttr] == TGSI_SEMANTIC_GENERIC) {
-            maxGeneric = MAX2(maxGeneric,
-                              xvp->output_to_semantic_index[outAttr]);
+            usedGenerics |= (1 << xvp->output_to_semantic_index[outAttr]);
          }
       }
 
-      /* Map vert program outputs that aren't used to the dummy slot
-       * (and an unused generic attribute slot).
+      /* For each vertex program output that doesn't match up to a fragment
+       * program input, map the vertex program output to a free slot and
+       * free generic attribute.
        */
       for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) {
          if (outputsWritten & (1 << outAttr)) {
             if (xvp->output_to_slot[outAttr] == UNUSED) {
-               xvp->output_to_slot[outAttr] = dummySlot;
+               GLint freeGeneric = _mesa_ffs(~usedGenerics) - 1;
+               GLint freeSlot = _mesa_ffs(~usedOutputSlots) - 1;
+               usedGenerics |= (1 << freeGeneric);
+               usedOutputSlots |= (1 << freeSlot);
+               xvp->output_to_slot[outAttr] = freeSlot;
                xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_GENERIC;
-               xvp->output_to_semantic_index[outAttr] = maxGeneric + 1;
+               xvp->output_to_semantic_index[outAttr] = freeGeneric;
             }
          }
 
diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c
index 3f9a825a15..95181578f6 100644
--- a/src/mesa/state_tracker/st_cb_accum.c
+++ b/src/mesa/state_tracker/st_cb_accum.c
@@ -40,16 +40,14 @@
 #include "st_draw.h"
 #include "st_public.h"
 #include "st_format.h"
+#include "st_texture.h"
+#include "st_inlines.h"
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
 #include "util/u_tile.h"
 
 
-#define UNCLAMPED_FLOAT_TO_SHORT(us, f)  \
-   us = ( (short) ( CLAMP((f), -1.0, 1.0) * 32767.0F) )
-
-
 /**
  * For hardware that supports deep color buffers, we could accelerate
  * most/all the accum operations with blending/texturing.
@@ -57,73 +55,20 @@
  */
 
 
-/**
- * Wrapper for pipe_get_tile_rgba().  Do format/cpp override to make the
- * tile util function think the surface is 16bit/channel, even if it's not.
- * See also: st_renderbuffer_alloc_storage()
- */
-static void
-acc_get_tile_rgba(struct pipe_context *pipe, struct pipe_transfer *acc_pt,
-                  uint x, uint y, uint w, uint h, float *p)
-{
-   const enum pipe_format f = acc_pt->format;
-   const struct pipe_format_block b = acc_pt->block;
-
-   acc_pt->format = DEFAULT_ACCUM_PIPE_FORMAT;
-   acc_pt->block.size = 8;
-   acc_pt->block.width = 1;
-   acc_pt->block.height = 1;
-
-   pipe_get_tile_rgba(acc_pt, x, y, w, h, p);
-
-   acc_pt->format = f;
-   acc_pt->block = b;
-}
-
-
-/**
- * Wrapper for pipe_put_tile_rgba().  Do format/cpp override to make the
- * tile util function think the surface is 16bit/channel, even if it's not.
- * See also: st_renderbuffer_alloc_storage()
- */
-static void
-acc_put_tile_rgba(struct pipe_context *pipe, struct pipe_transfer *acc_pt,
-                  uint x, uint y, uint w, uint h, const float *p)
-{
-   enum pipe_format f = acc_pt->format;
-   const struct pipe_format_block b = acc_pt->block;
-
-   acc_pt->format = DEFAULT_ACCUM_PIPE_FORMAT;
-   acc_pt->block.size = 8;
-   acc_pt->block.width = 1;
-   acc_pt->block.height = 1;
-
-   pipe_put_tile_rgba(acc_pt, x, y, w, h, p);
-
-   acc_pt->format = f;
-   acc_pt->block = b;
-}
-
-
-
 void
 st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
 {
    struct st_renderbuffer *acc_strb = st_renderbuffer(rb);
-   struct pipe_transfer *acc_pt;
-   struct pipe_screen *screen = ctx->st->pipe->screen;
    const GLint xpos = ctx->DrawBuffer->_Xmin;
    const GLint ypos = ctx->DrawBuffer->_Ymin;
    const GLint width = ctx->DrawBuffer->_Xmax - xpos;
    const GLint height = ctx->DrawBuffer->_Ymax - ypos;
-   GLubyte *map;
+   size_t stride = acc_strb->stride;
+   GLubyte *data = acc_strb->data;
 
-   acc_pt = screen->get_tex_transfer(screen, acc_strb->texture, 0, 0, 0,
-                                     PIPE_TRANSFER_WRITE, xpos, ypos,
-                                     width, height);
-   map = screen->transfer_map(screen, acc_pt);
-
-   /* note acc_strb->format might not equal acc_pt->format */
+   if(!data)
+      return;
+   
    switch (acc_strb->format) {
    case PIPE_FORMAT_R16G16B16A16_SNORM:
       {
@@ -133,7 +78,7 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
          GLshort a = FLOAT_TO_SHORT(ctx->Accum.ClearColor[3]);
          int i, j;
          for (i = 0; i < height; i++) {
-            GLshort *dst = (GLshort *) (map + i * acc_pt->stride + xpos * 8);
+            GLshort *dst = (GLshort *) (data + (ypos + i) * stride + xpos * 8);
             for (j = 0; j < width; j++) {
                dst[0] = r;
                dst[1] = g;
@@ -147,9 +92,6 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb)
    default:
       _mesa_problem(ctx, "unexpected format in st_clear_accum_buffer()");
    }
-
-   screen->transfer_unmap(screen, acc_pt);
-   screen->tex_transfer_destroy(acc_pt);
 }
 
 
@@ -159,25 +101,18 @@ accum_mad(GLcontext *ctx, GLfloat scale, GLfloat bias,
           GLint xpos, GLint ypos, GLint width, GLint height,
           struct st_renderbuffer *acc_strb)
 {
-   struct pipe_screen *screen = ctx->st->pipe->screen;
-   struct pipe_transfer *acc_pt;
-   GLubyte *map;
+   size_t stride = acc_strb->stride;
+   GLubyte *data = acc_strb->data;
 
-   acc_pt = screen->get_tex_transfer(screen, acc_strb->texture, 0, 0, 0,
-                                     PIPE_TRANSFER_READ_WRITE, xpos, ypos,
-                                     width, height);
-   map = screen->transfer_map(screen, acc_pt);
-
-   /* note acc_strb->format might not equal acc_pt->format */
    switch (acc_strb->format) {
    case PIPE_FORMAT_R16G16B16A16_SNORM:
       {
          int i, j;
          for (i = 0; i < height; i++) {
-            GLshort *acc = (GLshort *) (map + (ypos + i) * acc_pt->stride + xpos * 8);
+            GLshort *acc = (GLshort *) (data + (ypos + i) * stride + xpos * 8);
             for (j = 0; j < width * 4; j++) {
-               float val = SHORT_TO_FLOAT(acc[j]) * scale + bias;
-               acc[j] = FLOAT_TO_SHORT(val);
+               float val = SHORT_TO_FLOAT(*acc) * scale + bias;
+               *acc++ = FLOAT_TO_SHORT(val);
             }
          }
       }
@@ -185,86 +120,95 @@ accum_mad(GLcontext *ctx, GLfloat scale, GLfloat bias,
    default:
       _mesa_problem(NULL, "unexpected format in st_clear_accum_buffer()");
    }
-
-   screen->transfer_unmap(screen, acc_pt);
-   screen->tex_transfer_destroy(acc_pt);
 }
 
 
 static void
-accum_accum(struct pipe_context *pipe, GLfloat value,
+accum_accum(struct st_context *st, GLfloat value,
             GLint xpos, GLint ypos, GLint width, GLint height,
             struct st_renderbuffer *acc_strb,
             struct st_renderbuffer *color_strb)
 {
+   struct pipe_context *pipe = st->pipe;
    struct pipe_screen *screen = pipe->screen;
-   struct pipe_transfer *acc_trans, *color_trans;
-   GLfloat *colorBuf, *accBuf;
-   GLint i;
-
-   acc_trans = screen->get_tex_transfer(screen, acc_strb->texture, 0, 0, 0,
-                                        PIPE_TRANSFER_READ, xpos, ypos,
-                                        width, height);
+   struct pipe_transfer *color_trans;
+   size_t stride = acc_strb->stride;
+   GLubyte *data = acc_strb->data;
+   GLfloat *buf;
 
-   color_trans = screen->get_tex_transfer(screen, color_strb->texture, 0, 0, 0,
-                                          PIPE_TRANSFER_READ, xpos, ypos,
-                                          width, height);
+   color_trans = st_cond_flush_get_tex_transfer(st, color_strb->texture,
+						0, 0, 0,
+						PIPE_TRANSFER_READ, xpos, ypos,
+						width, height);
 
-   colorBuf = (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat));
-   accBuf = (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat));
+   buf = (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat));
 
-   pipe_get_tile_rgba(color_trans, 0, 0, width, height, colorBuf);
-   acc_get_tile_rgba(pipe, acc_trans, 0, 0, width, height, accBuf);
+   pipe_get_tile_rgba(color_trans, 0, 0, width, height, buf);
 
-   for (i = 0; i < 4 * width * height; i++) {
-      accBuf[i] = accBuf[i] + colorBuf[i] * value;
+   switch (acc_strb->format) {
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      {
+         const GLfloat *color = buf;
+         int i, j;
+         for (i = 0; i < height; i++) {
+            GLshort *acc = (GLshort *) (data + (ypos + i) * stride + xpos * 8);
+            for (j = 0; j < width * 4; j++) {
+               float val = *color++ * value;
+               *acc++ += FLOAT_TO_SHORT(val);
+            }
+         }
+      }
+      break;
+   default:
+      _mesa_problem(NULL, "unexpected format in st_clear_accum_buffer()");
    }
 
-   screen->tex_transfer_destroy(acc_trans);
-   acc_trans = screen->get_tex_transfer(screen, acc_strb->texture, 0, 0, 0,
-                                        PIPE_TRANSFER_WRITE, xpos, ypos,
-                                        width, height);
-
-   acc_put_tile_rgba(pipe, acc_trans, 0, 0, width, height, accBuf);
-
-   _mesa_free(colorBuf);
-   _mesa_free(accBuf);
-   screen->tex_transfer_destroy(acc_trans);
+   _mesa_free(buf);
    screen->tex_transfer_destroy(color_trans);
 }
 
 
 static void
-accum_load(struct pipe_context *pipe, GLfloat value,
+accum_load(struct st_context *st, GLfloat value,
            GLint xpos, GLint ypos, GLint width, GLint height,
            struct st_renderbuffer *acc_strb,
            struct st_renderbuffer *color_strb)
 {
+   struct pipe_context *pipe = st->pipe;
    struct pipe_screen *screen = pipe->screen;
-   struct pipe_transfer *acc_trans, *color_trans;
+   struct pipe_transfer *color_trans;
+   size_t stride = acc_strb->stride;
+   GLubyte *data = acc_strb->data;
    GLfloat *buf;
-   GLint i;
 
-   acc_trans = screen->get_tex_transfer(screen, acc_strb->texture, 0, 0, 0,
-                                        PIPE_TRANSFER_WRITE, xpos, ypos,
-                                        width, height);
-
-   color_trans = screen->get_tex_transfer(screen, color_strb->texture, 0, 0, 0,
-                                        PIPE_TRANSFER_READ, xpos, ypos,
-                                        width, height);
+   color_trans = st_cond_flush_get_tex_transfer(st, color_strb->texture,
+						0, 0, 0,
+						PIPE_TRANSFER_READ, xpos, ypos,
+						width, height);
 
    buf = (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat));
 
    pipe_get_tile_rgba(color_trans, 0, 0, width, height, buf);
 
-   for (i = 0; i < 4 * width * height; i++) {
-      buf[i] = buf[i] * value;
+   switch (acc_strb->format) {
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      {
+         const GLfloat *color = buf;
+         int i, j;
+         for (i = 0; i < height; i++) {
+            GLshort *acc = (GLshort *) (data + (ypos + i) * stride + xpos * 8);
+            for (j = 0; j < width * 4; j++) {
+               float val = *color++ * value;
+               *acc++ = FLOAT_TO_SHORT(val);
+            }
+         }
+      }
+      break;
+   default:
+      _mesa_problem(NULL, "unexpected format in st_clear_accum_buffer()");
    }
 
-   acc_put_tile_rgba(pipe, acc_trans, 0, 0, width, height, buf);
-
    _mesa_free(buf);
-   screen->tex_transfer_destroy(acc_trans);
    screen->tex_transfer_destroy(color_trans);
 }
 
@@ -278,45 +222,58 @@ accum_return(GLcontext *ctx, GLfloat value,
    struct pipe_context *pipe = ctx->st->pipe;
    struct pipe_screen *screen = pipe->screen;
    const GLubyte *colormask = ctx->Color.ColorMask;
-   struct pipe_transfer *acc_trans, *color_trans;
-   GLfloat *abuf, *cbuf = NULL;
-   GLint i, ch;
-
-   abuf = (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat));
-
-   acc_trans = screen->get_tex_transfer(screen, acc_strb->texture, 0, 0, 0,
-                                        PIPE_TRANSFER_READ, xpos, ypos,
-                                        width, height);
+   enum pipe_transfer_usage usage;
+   struct pipe_transfer *color_trans;
+   size_t stride = acc_strb->stride;
+   const GLubyte *data = acc_strb->data;
+   GLfloat *buf;
 
-   color_trans = screen->get_tex_transfer(screen, color_strb->texture, 0, 0, 0,
-                                          PIPE_TRANSFER_READ_WRITE, xpos, ypos,
-                                          width, height);
+   buf = (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat));
 
-   acc_get_tile_rgba(pipe, acc_trans, 0, 0, width, height, abuf);
+   if (!colormask[0] || !colormask[1] || !colormask[2] || !colormask[3])
+      usage = PIPE_TRANSFER_READ_WRITE;
+   else
+      usage = PIPE_TRANSFER_WRITE;
+   
+   color_trans = st_cond_flush_get_tex_transfer(st_context(ctx),
+						color_strb->texture, 0, 0, 0,
+						usage,
+						xpos, ypos,
+						width, height);
 
-   if (!colormask[0] || !colormask[1] || !colormask[2] || !colormask[3]) {
-      cbuf = (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat));
-      pipe_get_tile_rgba(color_trans, 0, 0, width, height, cbuf);
-   }
+   if (usage != PIPE_TRANSFER_WRITE)
+      pipe_get_tile_rgba(color_trans, 0, 0, width, height, buf);
 
-   for (i = 0; i < width * height; i++) {
-      for (ch = 0; ch < 4; ch++) {
-         if (colormask[ch]) {
-            GLfloat val = abuf[i * 4 + ch] * value;
-            abuf[i * 4 + ch] = CLAMP(val, 0.0f, 1.0f);
-         }
-         else {
-            abuf[i * 4 + ch] = cbuf[i * 4 + ch];
+   switch (acc_strb->format) {
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      {
+         GLfloat *color = buf;
+         int i, j, ch;
+         for (i = 0; i < height; i++) {
+            const GLshort *acc = (const GLshort *) (data + (ypos + i) * stride + xpos * 8);
+            for (j = 0; j < width; j++) {
+               for (ch = 0; ch < 4; ch++) {
+                  if (colormask[ch]) {
+                     GLfloat val = SHORT_TO_FLOAT(*acc * value);
+                     *color = CLAMP(val, 0.0f, 1.0f);
+                  }
+                  else {
+                     /* No change */
+                  }
+                  ++acc;
+                  ++color;
+               }
+            }
          }
       }
+      break;
+   default:
+      _mesa_problem(NULL, "unexpected format in st_clear_accum_buffer()");
    }
 
-   pipe_put_tile_rgba(color_trans, 0, 0, width, height, abuf);
+   pipe_put_tile_rgba(color_trans, 0, 0, width, height, buf);
 
-   _mesa_free(abuf);
-   if (cbuf)
-      _mesa_free(cbuf);
-   screen->tex_transfer_destroy(acc_trans);
+   _mesa_free(buf);
    screen->tex_transfer_destroy(color_trans);
 }
 
@@ -325,7 +282,6 @@ static void
 st_Accum(GLcontext *ctx, GLenum op, GLfloat value)
 {
    struct st_context *st = ctx->st;
-   struct pipe_context *pipe = st->pipe;
    struct st_renderbuffer *acc_strb
      = st_renderbuffer(ctx->DrawBuffer->Attachment[BUFFER_ACCUM].Renderbuffer);
    struct st_renderbuffer *color_strb
@@ -336,6 +292,9 @@ st_Accum(GLcontext *ctx, GLenum op, GLfloat value)
    const GLint width = ctx->DrawBuffer->_Xmax - xpos;
    const GLint height = ctx->DrawBuffer->_Ymax - ypos;
 
+   if(!acc_strb->data)
+      return;
+   
    /* make sure color bufs aren't cached */
    st_flush( st, PIPE_FLUSH_RENDER_CACHE, NULL );
 
@@ -352,11 +311,11 @@ st_Accum(GLcontext *ctx, GLenum op, GLfloat value)
       break;
    case GL_ACCUM:
       if (value != 0.0F) {
-         accum_accum(pipe, value, xpos, ypos, width, height, acc_strb, color_strb);
+         accum_accum(st, value, xpos, ypos, width, height, acc_strb, color_strb);
       }
       break;
    case GL_LOAD:
-      accum_load(pipe, value, xpos, ypos, width, height, acc_strb, color_strb);
+      accum_load(st, value, xpos, ypos, width, height, acc_strb, color_strb);
       break;
    case GL_RETURN:
       accum_return(ctx, value, xpos, ypos, width, height, acc_strb, color_strb);
diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c
index 2d547dd072..8709633557 100644
--- a/src/mesa/state_tracker/st_cb_bitmap.c
+++ b/src/mesa/state_tracker/st_cb_bitmap.c
@@ -47,6 +47,8 @@
 #include "st_cb_program.h"
 #include "st_mesa_to_tgsi.h"
 #include "st_texture.h"
+#include "st_inlines.h"
+
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
@@ -147,7 +149,7 @@ make_bitmap_fragment_program(GLcontext *ctx, GLuint samplerIndex)
       p->Instructions[ic].SrcReg[0].Swizzle = SWIZZLE_XXXX;
 
    p->Instructions[ic].SrcReg[0].Index = 0;
-   p->Instructions[ic].SrcReg[0].NegateBase = NEGATE_XYZW;
+   p->Instructions[ic].SrcReg[0].Negate = NEGATE_XYZW;
    ic++;
 
    /* END; */
@@ -330,15 +332,16 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height,
     * Create texture to hold bitmap pattern.
     */
    pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, ctx->st->bitmap.tex_format,
-                          0, width, height, 1, 0,
+                          0, width, height, 1,
                           PIPE_TEXTURE_USAGE_SAMPLER);
    if (!pt) {
       _mesa_unmap_bitmap_pbo(ctx, unpack);
       return NULL;
    }
 
-   transfer = screen->get_tex_transfer(screen, pt, 0, 0, 0, PIPE_TRANSFER_WRITE,
-                                       0, 0, width, height);
+   transfer = st_no_flush_get_tex_transfer(st_context(ctx), pt, 0, 0, 0,
+					   PIPE_TRANSFER_WRITE,
+					   0, 0, width, height);
 
    dest = screen->transfer_map(screen, transfer);
 
@@ -425,11 +428,11 @@ setup_bitmap_vertex_data(struct st_context *st,
    }
 
    /* put vertex data into vbuf */
-   pipe_buffer_write(pipe->screen, 
-                     st->bitmap.vbuf, 
-                     st->bitmap.vbuf_slot * sizeof st->bitmap.vertices,
-                     sizeof st->bitmap.vertices,
-                     st->bitmap.vertices);
+   st_no_flush_pipe_buffer_write(st,
+				 st->bitmap.vbuf,
+				 st->bitmap.vbuf_slot * sizeof st->bitmap.vertices,
+				 sizeof st->bitmap.vertices,
+				 st->bitmap.vertices);
 
    return st->bitmap.vbuf_slot++ * sizeof st->bitmap.vertices;
 }
@@ -570,8 +573,10 @@ reset_cache(struct st_context *st)
    cache->ymin = 1000000;
    cache->ymax = -1000000;
 
-   if (cache->trans)
+   if (cache->trans) {
       screen->tex_transfer_destroy(cache->trans);
+      cache->trans = NULL;
+   }
 
    assert(!cache->texture);
 
@@ -579,16 +584,27 @@ reset_cache(struct st_context *st)
    cache->texture = st_texture_create(st, PIPE_TEXTURE_2D,
                                       st->bitmap.tex_format, 0,
                                       BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT,
-                                      1, 0,
-                                      PIPE_TEXTURE_USAGE_SAMPLER);
+                                      1, PIPE_TEXTURE_USAGE_SAMPLER);
+
+}
+
+static void
+create_cache_trans(struct st_context *st)
+{
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct bitmap_cache *cache = st->bitmap.cache;
+
+   if (cache->trans)
+      return;
 
    /* Map the texture transfer.
     * Subsequent glBitmap calls will write into the texture image.
     */
-   cache->trans = screen->get_tex_transfer(screen, cache->texture, 0, 0, 0,
-                                           PIPE_TRANSFER_WRITE, 0, 0,
-                                           BITMAP_CACHE_WIDTH,
-                                           BITMAP_CACHE_HEIGHT);
+   cache->trans = st_no_flush_get_tex_transfer(st, cache->texture, 0, 0, 0,
+					       PIPE_TRANSFER_WRITE, 0, 0,
+					       BITMAP_CACHE_WIDTH,
+					       BITMAP_CACHE_HEIGHT);
    cache->buffer = screen->transfer_map(screen, cache->trans);
 
    /* init image to all 0xff */
@@ -620,11 +636,13 @@ st_flush_bitmap_cache(struct st_context *st)
          /* The texture transfer has been mapped until now.
           * So unmap and release the texture transfer before drawing.
           */
-         screen->transfer_unmap(screen, cache->trans);
-         cache->buffer = NULL;
+         if (cache->trans) {
+            screen->transfer_unmap(screen, cache->trans);
+            cache->buffer = NULL;
 
-         screen->tex_transfer_destroy(cache->trans);
-         cache->trans = NULL;
+            screen->tex_transfer_destroy(cache->trans);
+            cache->trans = NULL;
+         }
 
          draw_bitmap_quad(st->ctx,
                           cache->xpos,
@@ -709,6 +727,9 @@ accum_bitmap(struct st_context *st,
    if (y + height > cache->ymax)
       cache->ymax = y + height;
 
+   /* create the transfer if needed */
+   create_cache_trans(st);
+
    unpack_bitmap(st, px, py, width, height, unpack, bitmap,
                  cache->buffer, BITMAP_CACHE_WIDTH);
 
@@ -821,8 +842,7 @@ st_destroy_bitmap(struct st_context *st)
    struct pipe_screen *screen = pipe->screen;
    struct bitmap_cache *cache = st->bitmap.cache;
 
-   screen->transfer_unmap(screen, cache->trans);
-   screen->tex_transfer_destroy(cache->trans);
+
 
    if (st->bitmap.vs) {
       cso_delete_vertex_shader(st->cso_context, st->bitmap.vs);
@@ -834,7 +854,11 @@ st_destroy_bitmap(struct st_context *st)
       st->bitmap.vbuf = NULL;
    }
 
-   if (st->bitmap.cache) {
+   if (cache) {
+      if (cache->trans) {
+         screen->transfer_unmap(screen, cache->trans);
+         screen->tex_transfer_destroy(cache->trans);
+      }
       pipe_texture_reference(&st->bitmap.cache->texture, NULL);
       _mesa_free(st->bitmap.cache);
       st->bitmap.cache = NULL;
diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c
index 3651e4ae7d..7021d73208 100644
--- a/src/mesa/state_tracker/st_cb_bufferobjects.c
+++ b/src/mesa/state_tracker/st_cb_bufferobjects.c
@@ -26,10 +26,17 @@
  **************************************************************************/
 
 
+/**
+ * Functions for pixel buffer objects and vertex/element buffer objects.
+ */
+
+
 #include "main/imports.h"
 #include "main/mtypes.h"
+#include "main/arrayobj.h"
 #include "main/bufferobj.h"
 
+#include "st_inlines.h"
 #include "st_context.h"
 #include "st_cb_bufferobjects.h"
 
@@ -38,14 +45,6 @@
 #include "pipe/p_inlines.h"
 
 
-
-/* Pixel buffers and Vertex/index buffers are handled through these
- * mesa callbacks.  Framebuffer/Renderbuffer objects are
- * created/managed elsewhere.
- */
-
-
-
 /**
  * There is some duplication between mesa's bufferobjects and our
  * bufmgr buffers.  Both have an integer handle and a hashtable to
@@ -97,13 +96,13 @@ st_bufferobj_subdata(GLcontext *ctx,
 		     GLsizeiptrARB size,
 		     const GLvoid * data, struct gl_buffer_object *obj)
 {
-   struct pipe_context *pipe = st_context(ctx)->pipe;
    struct st_buffer_object *st_obj = st_buffer_object(obj);
 
    if (offset >= st_obj->size || size > (st_obj->size - offset))
       return;
 
-   pipe_buffer_write(pipe->screen, st_obj->buffer, offset, size, data);
+   st_cond_flush_pipe_buffer_write(st_context(ctx), st_obj->buffer,
+				   offset, size, data);
 }
 
 
@@ -117,13 +116,13 @@ st_bufferobj_get_subdata(GLcontext *ctx,
                          GLsizeiptrARB size,
                          GLvoid * data, struct gl_buffer_object *obj)
 {
-   struct pipe_context *pipe = st_context(ctx)->pipe;
    struct st_buffer_object *st_obj = st_buffer_object(obj);
 
    if (offset >= st_obj->size || size > (st_obj->size - offset))
       return;
 
-   pipe_buffer_read(pipe->screen, st_obj->buffer, offset, size, data);
+   st_cond_flush_pipe_buffer_read(st_context(ctx), st_obj->buffer,
+				  offset, size, data);
 }
 
 
@@ -168,10 +167,16 @@ st_bufferobj_data(GLcontext *ctx,
 
    st_obj->buffer = pipe_buffer_create( pipe->screen, 32, buffer_usage, size );
 
+   if (!st_obj->buffer) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glBufferDataARB");
+      return;
+   }
+
    st_obj->size = size;
 
    if (data)
-      st_bufferobj_subdata(ctx, target, 0, size, data, obj);
+      st_no_flush_pipe_buffer_write(st_context(ctx), st_obj->buffer, 0,
+				    size, data);
 }
 
 
@@ -182,9 +187,8 @@ static void *
 st_bufferobj_map(GLcontext *ctx, GLenum target, GLenum access,
                  struct gl_buffer_object *obj)
 {
-   struct pipe_context *pipe = st_context(ctx)->pipe;
    struct st_buffer_object *st_obj = st_buffer_object(obj);
-   GLuint flags;
+   uint flags;
 
    switch (access) {
    case GL_WRITE_ONLY:
@@ -200,8 +204,10 @@ st_bufferobj_map(GLcontext *ctx, GLenum target, GLenum access,
       break;      
    }
 
-   obj->Pointer = pipe_buffer_map(pipe->screen, st_obj->buffer, flags);
-   if(obj->Pointer) {
+   obj->Pointer = st_cond_flush_pipe_buffer_map(st_context(ctx),
+						st_obj->buffer,
+						flags);
+   if (obj->Pointer) {
       obj->Offset = 0;
       obj->Length = obj->Size;
    }
@@ -209,7 +215,6 @@ st_bufferobj_map(GLcontext *ctx, GLenum target, GLenum access,
 }
 
 
-
 /**
  * Called via glMapBufferRange().
  */
@@ -220,7 +225,7 @@ st_bufferobj_map_range(GLcontext *ctx, GLenum target,
 {
    struct pipe_context *pipe = st_context(ctx)->pipe;
    struct st_buffer_object *st_obj = st_buffer_object(obj);
-   GLuint flags = 0;
+   uint flags = 0x0;
    char *map;
 
    if (access & GL_MAP_WRITE_BIT)
@@ -242,8 +247,8 @@ st_bufferobj_map_range(GLcontext *ctx, GLenum target,
 
    map = obj->Pointer = pipe_buffer_map_range(pipe->screen, st_obj->buffer, offset, length, flags);
    if(obj->Pointer) {
-      obj->Offset = 0;
-      obj->Length = obj->Size;
+      obj->Offset = offset;
+      obj->Length = length;
       map += offset;
    }
    
@@ -262,7 +267,6 @@ st_bufferobj_flush_mapped_range(GLcontext *ctx, GLenum target,
    /* Subrange is relative to mapped range */
    assert(offset >= 0);
    assert(length >= 0);
-   assert(offset < obj->Length);
    assert(offset + length <= obj->Length);
    
    pipe_buffer_flush_mapped_range(pipe->screen, st_obj->buffer, 
@@ -287,6 +291,43 @@ st_bufferobj_unmap(GLcontext *ctx, GLenum target, struct gl_buffer_object *obj)
 }
 
 
+/**
+ * Called via glCopyBufferSubData().
+ */
+static void
+st_copy_buffer_subdata(GLcontext *ctx,
+                       struct gl_buffer_object *src,
+                       struct gl_buffer_object *dst,
+                       GLintptr readOffset, GLintptr writeOffset,
+                       GLsizeiptr size)
+{
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_buffer_object *srcObj = st_buffer_object(src);
+   struct st_buffer_object *dstObj = st_buffer_object(dst);
+   ubyte *srcPtr, *dstPtr;
+
+   /* buffer should not already be mapped */
+   assert(!src->Pointer);
+   assert(!dst->Pointer);
+
+   srcPtr = (ubyte *) pipe_buffer_map_range(pipe->screen,
+                                            srcObj->buffer,
+                                            readOffset, size,
+                                            PIPE_BUFFER_USAGE_CPU_READ);
+
+   dstPtr = (ubyte *) pipe_buffer_map_range(pipe->screen,
+                                            dstObj->buffer,
+                                            writeOffset, size,
+                                            PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   if (srcPtr && dstPtr)
+      _mesa_memcpy(dstPtr + writeOffset, srcPtr + readOffset, size);
+
+   pipe_buffer_unmap(pipe->screen, srcObj->buffer);
+   pipe_buffer_unmap(pipe->screen, dstObj->buffer);
+}
+
+
 void
 st_init_bufferobject_functions(struct dd_function_table *functions)
 {
@@ -299,4 +340,9 @@ st_init_bufferobject_functions(struct dd_function_table *functions)
    functions->MapBufferRange = st_bufferobj_map_range;
    functions->FlushMappedBufferRange = st_bufferobj_flush_mapped_range;
    functions->UnmapBuffer = st_bufferobj_unmap;
+   functions->CopyBufferSubData = st_copy_buffer_subdata;
+
+   /* For GL_APPLE_vertex_array_object */
+   functions->NewArrayObject = _mesa_new_array_object;
+   functions->DeleteArrayObject = _mesa_delete_array_object;
 }
diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c
index 5bdc6a1330..668acbccb8 100644
--- a/src/mesa/state_tracker/st_cb_clear.c
+++ b/src/mesa/state_tracker/st_cb_clear.c
@@ -45,6 +45,7 @@
 #include "st_program.h"
 #include "st_public.h"
 #include "st_mesa_to_tgsi.h"
+#include "st_inlines.h"
 
 #include "pipe/p_context.h"
 #include "pipe/p_inlines.h"
@@ -102,20 +103,6 @@ st_destroy_clear(struct st_context *st)
 }
 
 
-static GLboolean
-is_depth_stencil_format(enum pipe_format pipeFormat)
-{
-   switch (pipeFormat) {
-   case PIPE_FORMAT_S8Z24_UNORM:
-   case PIPE_FORMAT_Z24S8_UNORM:
-      return GL_TRUE;
-   default:
-      return GL_FALSE;
-   }
-}
-
-
-
 /**
  * Draw a screen-aligned quadrilateral.
  * Coords are window coords with y=0=bottom.  These will be passed
@@ -166,10 +153,10 @@ draw_quad(GLcontext *ctx,
    }
 
    /* put vertex data into vbuf */
-   pipe_buffer_write(pipe->screen, st->clear.vbuf, 
-                     st->clear.vbuf_slot * sizeof(st->clear.vertices),
-                     sizeof(st->clear.vertices),
-                     st->clear.vertices);
+   st_no_flush_pipe_buffer_write(st, st->clear.vbuf,
+				 st->clear.vbuf_slot * sizeof(st->clear.vertices),
+				 sizeof(st->clear.vertices),
+				 st->clear.vertices);
 
    /* draw */
    util_draw_vertex_buffer(pipe, 
@@ -330,7 +317,7 @@ static INLINE GLboolean
 check_clear_depth_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb)
 {
    const struct st_renderbuffer *strb = st_renderbuffer(rb);
-   const GLboolean isDS = is_depth_stencil_format(strb->surface->format);
+   const GLboolean isDS = pf_is_depth_and_stencil(strb->surface->format);
 
    if (ctx->Scissor.Enabled)
       return TRUE;
@@ -350,7 +337,7 @@ static INLINE GLboolean
 check_clear_stencil_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb)
 {
    const struct st_renderbuffer *strb = st_renderbuffer(rb);
-   const GLboolean isDS = is_depth_stencil_format(strb->surface->format);
+   const GLboolean isDS = pf_is_depth_and_stencil(strb->surface->format);
    const GLuint stencilMax = (1 << rb->StencilBits) - 1;
    const GLboolean maskStencil
       = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax;
diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c
index ebb1d1142a..2027b713ce 100644
--- a/src/mesa/state_tracker/st_cb_drawpixels.c
+++ b/src/mesa/state_tracker/st_cb_drawpixels.c
@@ -53,6 +53,8 @@
 #include "st_format.h"
 #include "st_mesa_to_tgsi.h"
 #include "st_texture.h"
+#include "st_inlines.h"
+
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
@@ -144,6 +146,8 @@ combined_drawpix_fragment_program(GLcontext *ctx)
       st->pixel_xfer.xfer_prog_sn = st->pixel_xfer.program->serialNo;
       st->pixel_xfer.user_prog_sn = st->fp->serialNo;
       st->pixel_xfer.combined_prog_sn = stfp->serialNo;
+      /* can't reference new program directly, already have a reference on it */
+      st_reference_fragprog(st, &st->pixel_xfer.combined_prog, NULL);
       st->pixel_xfer.combined_prog = stfp;
    }
 
@@ -351,8 +355,7 @@ make_texture(struct st_context *st,
    if (!pixels)
       return NULL;
 
-   pt = st_texture_create(st, PIPE_TEXTURE_2D, pipeFormat, 0, width, height,
-			  1, 0,
+   pt = st_texture_create(st, PIPE_TEXTURE_2D, pipeFormat, 0, width, height, 1,
                           PIPE_TEXTURE_USAGE_SAMPLER);
    if (!pt) {
       _mesa_unmap_drawpix_pbo(ctx, unpack);
@@ -369,9 +372,9 @@ make_texture(struct st_context *st,
       /* we'll do pixel transfer in a fragment shader */
       ctx->_ImageTransferState = 0x0;
 
-      transfer = screen->get_tex_transfer(screen, pt, 0, 0, 0,
-                                          PIPE_TRANSFER_WRITE, 0, 0,
-                                          width, height);
+      transfer = st_no_flush_get_tex_transfer(st, pt, 0, 0, 0,
+					      PIPE_TRANSFER_WRITE, 0, 0,
+					      width, height);
 
       /* map texture transfer */
       dest = screen->transfer_map(screen, transfer);
@@ -491,7 +494,7 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z,
       /* allocate/load buffer object with vertex data */
       buf = pipe_buffer_create(pipe->screen, 32, PIPE_BUFFER_USAGE_VERTEX,
                                sizeof(verts));
-      pipe_buffer_write(pipe->screen, buf, 0, sizeof(verts), verts);
+      st_no_flush_pipe_buffer_write(st, buf, 0, sizeof(verts), verts);
 
       util_draw_vertex_buffer(pipe, buf, 0,
                               PIPE_PRIM_QUADS,
@@ -627,13 +630,12 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
    struct pipe_context *pipe = st->pipe;
    struct pipe_screen *screen = pipe->screen;
    struct st_renderbuffer *strb;
+   enum pipe_transfer_usage usage;
    struct pipe_transfer *pt;
    const GLboolean zoom = ctx->Pixel.ZoomX != 1.0 || ctx->Pixel.ZoomY != 1.0;
    GLint skipPixels;
    ubyte *stmap;
 
-   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
-
    strb = st_renderbuffer(ctx->DrawBuffer->
                           Attachment[BUFFER_STENCIL].Renderbuffer);
 
@@ -641,9 +643,15 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
       y = ctx->DrawBuffer->Height - y - height;
    }
 
-   pt = screen->get_tex_transfer(screen, strb->texture, 0, 0, 0,
-                                 PIPE_TRANSFER_WRITE, x, y,
-                                 width, height);
+   if(format != GL_DEPTH_STENCIL && 
+      pf_get_component_bits( strb->format, PIPE_FORMAT_COMP_Z ) != 0)
+      usage = PIPE_TRANSFER_READ_WRITE;
+   else
+      usage = PIPE_TRANSFER_WRITE;
+   
+   pt = st_cond_flush_get_tex_transfer(st_context(ctx), strb->texture, 0, 0, 0,
+				       usage, x, y,
+				       width, height);
 
    stmap = screen->transfer_map(screen, pt);
 
@@ -693,6 +701,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
             case PIPE_FORMAT_S8_UNORM:
                {
                   ubyte *dest = stmap + spanY * pt->stride + spanX;
+                  assert(usage == PIPE_TRANSFER_WRITE);
                   memcpy(dest, sValues, spanWidth);
                }
                break;
@@ -700,6 +709,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
                if (format == GL_DEPTH_STENCIL) {
                   uint *dest = (uint *) (stmap + spanY * pt->stride + spanX*4);
                   GLint k;
+                  assert(usage == PIPE_TRANSFER_WRITE);
                   for (k = 0; k < spanWidth; k++) {
                      dest[k] = zValues[k] | (sValues[k] << 24);
                   }
@@ -707,6 +717,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
                else {
                   uint *dest = (uint *) (stmap + spanY * pt->stride + spanX*4);
                   GLint k;
+                  assert(usage == PIPE_TRANSFER_READ_WRITE);
                   for (k = 0; k < spanWidth; k++) {
                      dest[k] = (dest[k] & 0xffffff) | (sValues[k] << 24);
                   }
@@ -716,13 +727,15 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
                if (format == GL_DEPTH_STENCIL) {
                   uint *dest = (uint *) (stmap + spanY * pt->stride + spanX*4);
                   GLint k;
+                  assert(usage == PIPE_TRANSFER_WRITE);
                   for (k = 0; k < spanWidth; k++) {
-                     dest[k] = zValues[k] | (sValues[k] << 24);
+                     dest[k] = (zValues[k] << 8) | (sValues[k] & 0xff);
                   }
                }
                else {
                   uint *dest = (uint *) (stmap + spanY * pt->stride + spanX*4);
                   GLint k;
+                  assert(usage == PIPE_TRANSFER_READ_WRITE);
                   for (k = 0; k < spanWidth; k++) {
                      dest[k] = (dest[k] & 0xffffff00) | (sValues[k] & 0xff);
                   }
@@ -756,7 +769,6 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
    struct st_vertex_program *stvp;
    struct st_context *st = ctx->st;
    struct pipe_surface *ps;
-   GLuint bufferFormat;
    const GLfloat *color;
 
    if (format == GL_STENCIL_INDEX ||
@@ -784,8 +796,6 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
       color = NULL;
    }
 
-   bufferFormat = ps->format;
-
    /* draw with textured quad */
    {
       struct pipe_texture *pt
@@ -810,6 +820,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
 {
    struct st_renderbuffer *rbDraw = st_renderbuffer(ctx->DrawBuffer->_StencilBuffer);
    struct pipe_screen *screen = ctx->st->pipe->screen;
+   enum pipe_transfer_usage usage;
    struct pipe_transfer *ptDraw;
    ubyte *drawMap;
    ubyte *buffer;
@@ -826,9 +837,15 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
                           GL_STENCIL_INDEX, GL_UNSIGNED_BYTE,
                           &ctx->DefaultPacking, buffer);
 
-   ptDraw = screen->get_tex_transfer(screen, rbDraw->texture, 0, 0, 0,
-                                     PIPE_TRANSFER_WRITE, dstx, dsty,
-                                     width, height);
+   if(pf_get_component_bits( rbDraw->format, PIPE_FORMAT_COMP_Z ) != 0)
+      usage = PIPE_TRANSFER_READ_WRITE;
+   else
+      usage = PIPE_TRANSFER_WRITE;
+   
+   ptDraw = st_cond_flush_get_tex_transfer(st_context(ctx),
+					   rbDraw->texture, 0, 0, 0,
+					   usage, dstx, dsty,
+					   width, height);
 
    assert(ptDraw->block.width == 1);
    assert(ptDraw->block.height == 1);
@@ -857,6 +874,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
          {
             uint *dst4 = (uint *) dst;
             int j;
+            assert(usage == PIPE_TRANSFER_READ_WRITE);
             for (j = 0; j < width; j++) {
                *dst4 = (*dst4 & 0xffffff) | (src[j] << 24);
                dst4++;
@@ -867,6 +885,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
          {
             uint *dst4 = (uint *) dst;
             int j;
+            assert(usage == PIPE_TRANSFER_READ_WRITE);
             for (j = 0; j < width; j++) {
                *dst4 = (*dst4 & 0xffffff00) | (src[j] & 0xff);
                dst4++;
@@ -874,6 +893,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
          }
          break;
       case PIPE_FORMAT_S8_UNORM:
+         assert(usage == PIPE_TRANSFER_WRITE);
          memcpy(dst, src, width);
          break;
       default:
@@ -904,11 +924,38 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
    GLfloat *color;
    enum pipe_format srcFormat, texFormat;
 
-   /* make sure rendering has completed */
    pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
 
    st_validate_state(st);
 
+   if (srcx < 0) {
+      width -= -srcx;
+      dstx += -srcx;
+      srcx = 0;
+   }
+
+   if (srcy < 0) {
+      height -= -srcy;
+      dsty += -srcy;
+      srcy = 0;
+   }
+
+   if (dstx < 0) {
+      width -= -dstx;
+      srcx += -dstx;
+      dstx = 0;
+   }
+
+   if (dsty < 0) {
+      height -= -dsty;
+      srcy += -dsty;
+      dsty = 0;
+   }
+
+   if (width < 0 || height < 0)
+      return;
+
+
    if (type == GL_STENCIL) {
       /* can't use texturing to do stencil */
       copy_stencil_pixels(ctx, srcx, srcy, width, height, dstx, dsty);
@@ -950,15 +997,24 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
       }
    }
 
+   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+      srcy = ctx->DrawBuffer->Height - srcy - height;
+
+      if (srcy < 0) {
+         height -= -srcy;
+         srcy = 0;
+      }
+      
+      if (height < 0)
+         return;
+   }
+
    pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, texFormat, 0,
-                          width, height, 1, 0,
+                          width, height, 1,
                           PIPE_TEXTURE_USAGE_SAMPLER);
    if (!pt)
       return;
 
-   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
-      srcy = ctx->DrawBuffer->Height - srcy - height;
-   }
 
    if (srcFormat == texFormat) {
       /* copy source framebuffer surface into mipmap/texture */
@@ -978,13 +1034,13 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
    else {
       /* CPU-based fallback/conversion */
       struct pipe_transfer *ptRead =
-         screen->get_tex_transfer(screen, rbRead->texture, 0, 0, 0,
-                                  PIPE_TRANSFER_READ, srcx, srcy, width,
-                                  height);
+         st_cond_flush_get_tex_transfer(st, rbRead->texture, 0, 0, 0,
+					PIPE_TRANSFER_READ, srcx, srcy, width,
+					height);
 
       struct pipe_transfer *ptTex =
-         screen->get_tex_transfer(screen, pt, 0, 0, 0, PIPE_TRANSFER_WRITE,
-                                  0, 0, width, height);
+         st_cond_flush_get_tex_transfer(st, pt, 0, 0, 0, PIPE_TRANSFER_WRITE,
+					0, 0, width, height);
 
       if (type == GL_COLOR) {
          /* alternate path using get/put_tile() */
diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c
index f74d0d46d0..ecdb988033 100644
--- a/src/mesa/state_tracker/st_cb_fbo.c
+++ b/src/mesa/state_tracker/st_cb_fbo.c
@@ -88,94 +88,92 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
 {
    struct pipe_context *pipe = ctx->st->pipe;
    struct st_renderbuffer *strb = st_renderbuffer(rb);
-   struct pipe_texture template;
-   unsigned surface_usage;
-
-   /* Free the old surface and texture
-    */
-   pipe_surface_reference( &strb->surface, NULL );
-   pipe_texture_reference( &strb->texture, NULL );
-
-
-   memset(&template, 0, sizeof(template));
-
-   if (strb->format != PIPE_FORMAT_NONE) {
-      template.format = strb->format;
-   }
-   else {
-      template.format = st_choose_renderbuffer_format(pipe, internalFormat);
-   }
-
+   enum pipe_format format;
+
+   if (strb->format != PIPE_FORMAT_NONE)
+      format = strb->format;
+   else
+      format = st_choose_renderbuffer_format(pipe, internalFormat);
+      
+   /* init renderbuffer fields */
    strb->Base.Width  = width;
    strb->Base.Height = height;
-   init_renderbuffer_bits(strb, template.format);
-
-   template.target = PIPE_TEXTURE_2D;
-   template.compressed = 0;
-   pf_get_block(template.format, &template.block);
-   template.width[0] = width;
-   template.height[0] = height;
-   template.depth[0] = 1;
-   template.last_level = 0;
-   template.nr_samples = rb->NumSamples;
-
-   if (pf_is_depth_stencil(template.format)) {
-      template.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL;
+   init_renderbuffer_bits(strb, format);
+
+   strb->defined = GL_FALSE;  /* undefined contents now */
+
+   if(strb->software) {
+      struct pipe_format_block block;
+      size_t size;
+      
+      _mesa_free(strb->data);
+
+      assert(strb->format != PIPE_FORMAT_NONE);
+      pf_get_block(strb->format, &block);
+      
+      strb->stride = pf_get_stride(&block, width);
+      size = pf_get_2d_size(&block, strb->stride, height);
+      
+      strb->data = _mesa_malloc(size);
+      
+      return strb->data != NULL;
    }
    else {
-      template.tex_usage = (PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
-                            PIPE_TEXTURE_USAGE_RENDER_TARGET);
-   }
-
-
-   /* Probably need dedicated flags for surface usage too: 
-    */
-   surface_usage = (PIPE_BUFFER_USAGE_GPU_READ |
-                    PIPE_BUFFER_USAGE_GPU_WRITE);
-#if 0
-                    PIPE_BUFFER_USAGE_CPU_READ |
-                    PIPE_BUFFER_USAGE_CPU_WRITE);
-#endif
-
-   strb->texture = pipe->screen->texture_create( pipe->screen,
-                                                 &template );
+      struct pipe_texture template;
+      unsigned surface_usage;
+    
+      /* Free the old surface and texture
+       */
+      pipe_surface_reference( &strb->surface, NULL );
+      pipe_texture_reference( &strb->texture, NULL );
 
-   /* Special path for accum buffers.  
-    *
-    * Try a different surface format.  Since accum buffers are s/w
-    * only for now, the surface pixel format doesn't really matter,
-    * only that the buffer is large enough.
-    */
-   if (!strb->texture && template.format == DEFAULT_ACCUM_PIPE_FORMAT) 
-   {
-      /* Actually, just setting this usage value should be sufficient
-       * to tell the driver to go ahead and allocate the buffer, even
-       * if HW doesn't support the format.
+      /* Setup new texture template.
        */
-      template.tex_usage = 0;
-      surface_usage = (PIPE_BUFFER_USAGE_CPU_READ |
+      memset(&template, 0, sizeof(template));
+      template.target = PIPE_TEXTURE_2D;
+      template.format = format;
+      pf_get_block(format, &template.block);
+      template.width[0] = width;
+      template.height[0] = height;
+      template.depth[0] = 1;
+      template.last_level = 0;
+      template.nr_samples = rb->NumSamples;
+      if (pf_is_depth_stencil(format)) {
+         template.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL;
+      }
+      else {
+         template.tex_usage = (PIPE_TEXTURE_USAGE_DISPLAY_TARGET |
+                               PIPE_TEXTURE_USAGE_RENDER_TARGET);
+      }
+
+      /* Probably need dedicated flags for surface usage too: 
+       */
+      surface_usage = (PIPE_BUFFER_USAGE_GPU_READ |
+                       PIPE_BUFFER_USAGE_GPU_WRITE);
+#if 0
+                       PIPE_BUFFER_USAGE_CPU_READ |
                        PIPE_BUFFER_USAGE_CPU_WRITE);
+#endif
 
       strb->texture = pipe->screen->texture_create( pipe->screen,
                                                     &template );
 
-   }
-
-   if (!strb->texture) 
-      return FALSE;
+      if (!strb->texture) 
+         return FALSE;
 
-   strb->surface = pipe->screen->get_tex_surface( pipe->screen,
-                                                  strb->texture,
-                                                  0, 0, 0,
-                                                  surface_usage );
+      strb->surface = pipe->screen->get_tex_surface( pipe->screen,
+                                                     strb->texture,
+                                                     0, 0, 0,
+                                                     surface_usage );
 
-   assert(strb->surface->texture);
-   assert(strb->surface->format);
-   assert(strb->surface->width == width);
-   assert(strb->surface->height == height);
+      assert(strb->surface->texture);
+      assert(strb->surface->format);
+      assert(strb->surface->width == width);
+      assert(strb->surface->height == height);
 
 
-   return strb->surface != NULL;
+      return strb->surface != NULL;
+   }
 }
 
 
@@ -189,6 +187,7 @@ st_renderbuffer_delete(struct gl_renderbuffer *rb)
    ASSERT(strb);
    pipe_surface_reference(&strb->surface, NULL);
    pipe_texture_reference(&strb->texture, NULL);
+   _mesa_free(strb->data);
    _mesa_free(strb);
 }
 
@@ -245,7 +244,7 @@ st_new_renderbuffer(GLcontext *ctx, GLuint name)
  * renderbuffer).  The window system code determines the format.
  */
 struct gl_renderbuffer *
-st_new_renderbuffer_fb(enum pipe_format format, int samples)
+st_new_renderbuffer_fb(enum pipe_format format, int samples, boolean sw)
 {
    struct st_renderbuffer *strb;
 
@@ -259,7 +258,8 @@ st_new_renderbuffer_fb(enum pipe_format format, int samples)
    strb->Base.ClassID = 0x4242; /* just a unique value */
    strb->Base.NumSamples = samples;
    strb->format = format;
-
+   strb->software = sw;
+   
    switch (format) {
    case PIPE_FORMAT_A8R8G8B8_UNORM:
    case PIPE_FORMAT_B8G8R8A8_UNORM:
@@ -290,7 +290,7 @@ st_new_renderbuffer_fb(enum pipe_format format, int samples)
       strb->Base.InternalFormat = GL_STENCIL_INDEX8_EXT;
       strb->Base._BaseFormat = GL_STENCIL_INDEX;
       break;
-   case DEFAULT_ACCUM_PIPE_FORMAT: /*PIPE_FORMAT_R16G16B16A16_SNORM*/
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
       strb->Base.InternalFormat = GL_RGBA16;
       strb->Base._BaseFormat = GL_RGBA;
       break;
@@ -346,6 +346,7 @@ st_render_texture(GLcontext *ctx,
                   struct gl_framebuffer *fb,
                   struct gl_renderbuffer_attachment *att)
 {
+   struct pipe_screen *screen = ctx->st->pipe->screen;
    struct st_renderbuffer *strb;
    struct gl_renderbuffer *rb;
    struct pipe_texture *pt = st_get_texobj_texture(att->Texture);
@@ -368,6 +369,8 @@ st_render_texture(GLcontext *ctx,
    rb->AllocStorage = NULL; /* should not get called */
    strb = st_renderbuffer(rb);
 
+   assert(strb->Base.RefCount > 0);
+
    /* get the texture for the texture object */
    stObj = st_texture_object(att->Texture);
 
@@ -387,7 +390,14 @@ st_render_texture(GLcontext *ctx,
 
    pipe_surface_reference(&strb->surface, NULL);
 
-   /* the new surface will be created during framebuffer validation */
+   /* new surface for rendering into the texture */
+   strb->surface = screen->get_tex_surface(screen,
+                                           strb->texture,
+                                           strb->rtt_face,
+                                           strb->rtt_level,
+                                           strb->rtt_slice,
+                                           PIPE_BUFFER_USAGE_GPU_READ |
+                                           PIPE_BUFFER_USAGE_GPU_WRITE);
 
    init_renderbuffer_bits(strb, pt->format);
 
@@ -455,6 +465,134 @@ st_validate_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb)
 }
 
 
+/**
+ * Copy back color buffer to front color buffer.
+ */
+static void
+copy_back_to_front(struct st_context *st,
+                   struct gl_framebuffer *fb,
+                   gl_buffer_index frontIndex,
+                   gl_buffer_index backIndex)
+
+{
+   struct st_framebuffer *stfb = (struct st_framebuffer *) fb;
+   struct pipe_surface *surf_front, *surf_back;
+
+   (void) st_get_framebuffer_surface(stfb, frontIndex, &surf_front);
+   (void) st_get_framebuffer_surface(stfb, backIndex, &surf_back);
+
+   if (surf_front && surf_back) {
+      st->pipe->surface_copy(st->pipe,
+                             surf_front, 0, 0,  /* dest */
+                             surf_back, 0, 0,   /* src */
+                             fb->Width, fb->Height);
+   }
+}
+
+
+/**
+ * Check if we're drawing into, or read from, a front color buffer.  If the
+ * front buffer is missing, create it now.
+ *
+ * The back color buffer must exist since we'll use its format/samples info
+ * for creating the front buffer.
+ *
+ * \param frontIndex  either BUFFER_FRONT_LEFT or BUFFER_FRONT_RIGHT
+ * \param backIndex  either BUFFER_BACK_LEFT or BUFFER_BACK_RIGHT
+ */
+static void
+check_create_front_buffer(GLcontext *ctx, struct gl_framebuffer *fb,
+                          gl_buffer_index frontIndex,
+                          gl_buffer_index backIndex)
+{
+   if (fb->Attachment[frontIndex].Renderbuffer == NULL) {
+      GLboolean create = GL_FALSE;
+
+      /* check if drawing to or reading from front buffer */
+      if (fb->_ColorReadBufferIndex == frontIndex) {
+         create = GL_TRUE;
+      }
+      else {
+         GLuint b;
+         for (b = 0; b < fb->_NumColorDrawBuffers; b++) {
+            if (fb->_ColorDrawBufferIndexes[b] == frontIndex) {
+               create = GL_TRUE;
+               break;
+            }
+         }
+      }
+
+      if (create) {
+         struct st_renderbuffer *back;
+         struct gl_renderbuffer *front;
+         enum pipe_format colorFormat;
+         uint samples;
+
+         if (0)
+            _mesa_debug(ctx, "Allocate new front buffer\n");
+
+         /* get back renderbuffer info */
+         back = st_renderbuffer(fb->Attachment[backIndex].Renderbuffer);
+         colorFormat = back->format;
+         samples = back->Base.NumSamples;
+
+         /* create front renderbuffer */
+         front = st_new_renderbuffer_fb(colorFormat, samples, FALSE);
+         _mesa_add_renderbuffer(fb, frontIndex, front);
+
+         /* alloc texture/surface for new front buffer */
+         front->AllocStorage(ctx, front, front->InternalFormat,
+                             fb->Width, fb->Height);
+
+         /* initialize the front color buffer contents by copying
+          * the back buffer.
+          */
+         copy_back_to_front(ctx->st, fb, frontIndex, backIndex);
+      }
+   }
+}
+
+
+/**
+ * If front left/right color buffers are missing, create them now.
+ */
+static void
+check_create_front_buffers(GLcontext *ctx, struct gl_framebuffer *fb)
+{
+   /* check if we need to create the front left buffer now */
+   check_create_front_buffer(ctx, fb, BUFFER_FRONT_LEFT, BUFFER_BACK_LEFT);
+
+   if (fb->Visual.stereoMode) {
+      check_create_front_buffer(ctx, fb, BUFFER_FRONT_RIGHT, BUFFER_BACK_RIGHT);
+   }
+
+   st_invalidate_state(ctx, _NEW_BUFFERS);
+}
+
+
+/**
+ * Called via glDrawBuffer.
+ */
+static void
+st_DrawBuffers(GLcontext *ctx, GLsizei count, const GLenum *buffers)
+{
+   (void) count;
+   (void) buffers;
+   check_create_front_buffers(ctx, ctx->DrawBuffer);
+}
+
+
+/**
+ * Called via glReadBuffer.
+ */
+static void
+st_ReadBuffer(GLcontext *ctx, GLenum buffer)
+{
+   (void) buffer;
+   check_create_front_buffers(ctx, ctx->ReadBuffer);
+}
+
+
 void st_init_fbo_functions(struct dd_function_table *functions)
 {
    functions->NewFramebuffer = st_new_framebuffer;
@@ -467,4 +605,7 @@ void st_init_fbo_functions(struct dd_function_table *functions)
    /* no longer needed by core Mesa, drivers handle resizes...
    functions->ResizeBuffers = st_resize_buffers;
    */
+
+   functions->DrawBuffers = st_DrawBuffers;
+   functions->ReadBuffer = st_ReadBuffer;
 }
diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h
index 44fa9fe9a4..bea6eb89c3 100644
--- a/src/mesa/state_tracker/st_cb_fbo.h
+++ b/src/mesa/state_tracker/st_cb_fbo.h
@@ -30,10 +30,6 @@
 #define ST_CB_FBO_H
 
 
-#define DEFAULT_ACCUM_PIPE_FORMAT PIPE_FORMAT_R16G16B16A16_SNORM
-
-
-
 /**
  * Derived renderbuffer class.  Just need to add a pointer to the
  * pipe surface.
@@ -44,7 +40,15 @@ struct st_renderbuffer
    struct pipe_texture *texture;
    struct pipe_surface *surface; /* temporary view into texture */
    enum pipe_format format;  /** preferred format, or PIPE_FORMAT_NONE */
-
+   GLboolean defined;        /**< defined contents? */
+
+   /**
+    * Used only when hardware accumulation buffers are not supported.
+    */
+   boolean software;
+   size_t stride;
+   void *data;
+   
    struct st_texture_object *rtt;  /**< GL render to texture's texture */
    int rtt_level, rtt_face, rtt_slice;
 
@@ -62,7 +66,7 @@ st_renderbuffer(struct gl_renderbuffer *rb)
 
 
 extern struct gl_renderbuffer *
-st_new_renderbuffer_fb(enum pipe_format format, int samples);
+st_new_renderbuffer_fb(enum pipe_format format, int samples, boolean sw);
 
 extern void
 st_init_fbo_functions(struct dd_function_table *functions);
diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c
index 7d7d3823c9..8ceeeabcd3 100644
--- a/src/mesa/state_tracker/st_cb_flush.c
+++ b/src/mesa/state_tracker/st_cb_flush.c
@@ -47,10 +47,19 @@
 #include "util/u_blit.h"
 
 
+/** Check if we have a front color buffer and if it's been drawn to. */
 static INLINE GLboolean
 is_front_buffer_dirty(struct st_context *st)
 {
-   return st->frontbuffer_status == FRONT_STATUS_DIRTY;
+   if (st->frontbuffer_status == FRONT_STATUS_DIRTY) {
+      return GL_TRUE;
+   }
+   else {
+      GLframebuffer *fb = st->ctx->DrawBuffer;
+      struct st_renderbuffer *strb
+         = st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+      return strb && strb->defined;
+   }
 }
 
 
@@ -82,7 +91,7 @@ display_front_buffer(struct st_context *st)
 void st_flush( struct st_context *st, uint pipeFlushFlags,
                struct pipe_fence_handle **fence )
 {
-   FLUSH_VERTICES(st->ctx, 0);
+   FLUSH_CURRENT(st->ctx, 0);
 
    /* Release any vertex buffers that might potentially be accessed in
     * successive frames:
diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c
index 7dd2352739..3bcccd0df4 100644
--- a/src/mesa/state_tracker/st_cb_rasterpos.c
+++ b/src/mesa/state_tracker/st_cb_rasterpos.c
@@ -194,9 +194,10 @@ new_draw_rastpos_stage(GLcontext *ctx, struct draw_context *draw)
    rs->stage.destroy = rastpos_destroy;
    rs->ctx = ctx;
 
-   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+   for (i = 0; i < Elements(rs->array); i++) {
       rs->array[i].Size = 4;
       rs->array[i].Type = GL_FLOAT;
+      rs->array[i].Format = GL_RGBA;
       rs->array[i].Stride = 0;
       rs->array[i].StrideB = 0;
       rs->array[i].Ptr = (GLubyte *) ctx->Current.Attrib[i];
diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c
index 9ce5f3fe84..ccf1a0b563 100644
--- a/src/mesa/state_tracker/st_cb_readpixels.c
+++ b/src/mesa/state_tracker/st_cb_readpixels.c
@@ -42,13 +42,15 @@
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
 #include "util/u_tile.h"
+
 #include "st_context.h"
 #include "st_cb_bitmap.h"
 #include "st_cb_readpixels.h"
 #include "st_cb_fbo.h"
 #include "st_format.h"
 #include "st_public.h"
-
+#include "st_texture.h"
+#include "st_inlines.h"
 
 /**
  * Special case for reading stencil buffer.
@@ -73,8 +75,11 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
    }
 
    /* Create a read transfer from the renderbuffer's texture */
-   pt = screen->get_tex_transfer(screen, strb->texture,  0, 0, 0,
-                                 PIPE_TRANSFER_READ, x, y, width, height);
+
+   pt = st_cond_flush_get_tex_transfer(st_context(ctx), strb->texture,
+				       0, 0, 0,
+				       PIPE_TRANSFER_READ, x, y,
+				       width, height);
 
    /* map the stencil buffer */
    stmap = screen->transfer_map(screen, pt);
@@ -107,7 +112,7 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
       case PIPE_FORMAT_S8Z24_UNORM:
          if (format == GL_DEPTH_STENCIL) {
             const uint *src = (uint *) (stmap + srcY * pt->stride);
-            const GLfloat scale = 1.0 / (0xffffff);
+            const GLfloat scale = 1.0f / (0xffffff);
             GLint k;
             for (k = 0; k < width; k++) {
                sValues[k] = src[k] >> 24;
@@ -125,7 +130,7 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
       case PIPE_FORMAT_Z24S8_UNORM:
          if (format == GL_DEPTH_STENCIL) {
             const uint *src = (uint *) (stmap + srcY * pt->stride);
-            const GLfloat scale = 1.0 / (0xffffff);
+            const GLfloat scale = 1.0f / (0xffffff);
             GLint k;
             for (k = 0; k < width; k++) {
                sValues[k] = src[k] & 0xff;
@@ -240,8 +245,10 @@ st_fast_readpixels(GLcontext *ctx, struct st_renderbuffer *strb,
          y = strb->texture->height[0] - y - height;
       }
 
-      trans = screen->get_tex_transfer(screen, strb->texture, 0, 0, 0,
-                                       PIPE_TRANSFER_READ, x, y, width, height);
+      trans = st_cond_flush_get_tex_transfer(st_context(ctx), strb->texture,
+					     0, 0, 0,
+					     PIPE_TRANSFER_READ, x, y,
+					     width, height);
       if (!trans) {
          return GL_FALSE;
       }
@@ -350,7 +357,6 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
    if (!dest)
       return;
 
-   /* make sure rendering has completed */
    st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL);
 
    if (format == GL_STENCIL_INDEX ||
@@ -395,8 +401,10 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
    }
 
    /* Create a read transfer from the renderbuffer's texture */
-   trans = screen->get_tex_transfer(screen, strb->texture,  0, 0, 0,
-                                    PIPE_TRANSFER_READ, x, y, width, height);
+   trans = st_cond_flush_get_tex_transfer(st_context(ctx), strb->texture,
+					  0, 0, 0,
+					  PIPE_TRANSFER_READ, x, y,
+					  width, height);
 
    /* determine bottom-to-top vs. top-to-bottom order */
    if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
@@ -437,11 +445,16 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
             }
          }
          else {
-            /* untested, but simple: */
+            /* XXX: unreachable code -- should be before st_read_stencil_pixels */
             assert(format == GL_DEPTH_STENCIL_EXT);
             for (i = 0; i < height; i++) {
+               GLuint *zshort = (GLuint *)dst;
                pipe_get_tile_raw(trans, 0, y, width, 1, dst, 0);
                y += yStep;
+               /* Reverse into 24/8 */
+               for (j = 0; j < width; j++) {
+                  zshort[j] = (zshort[j] << 8) | (zshort[j] >> 24);
+               }
                dst += dstStride;
             }
          }
@@ -464,7 +477,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
             }
          }
          else {
-            /* untested, but simple: */
+            /* XXX: unreachable code -- should be before st_read_stencil_pixels */
             assert(format == GL_DEPTH_STENCIL_EXT);
             for (i = 0; i < height; i++) {
                pipe_get_tile_raw(trans, 0, y, width, 1, dst, 0);
diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c
index 8013e69e8e..14b78d1253 100644
--- a/src/mesa/state_tracker/st_cb_texture.c
+++ b/src/mesa/state_tracker/st_cb_texture.c
@@ -26,6 +26,7 @@
  **************************************************************************/
 
 #include "main/mfeatures.h"
+#include "main/bufferobj.h"
 #if FEATURE_convolve
 #include "main/convolve.h"
 #endif
@@ -49,12 +50,15 @@
 #include "state_tracker/st_public.h"
 #include "state_tracker/st_texture.h"
 #include "state_tracker/st_gen_mipmap.h"
+#include "state_tracker/st_inlines.h"
+#include "state_tracker/st_atom.h"
 
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
 #include "util/u_tile.h"
 #include "util/u_blit.h"
+#include "util/u_surface.h"
 
 
 #define DBG if (0) printf
@@ -88,7 +92,7 @@ gl_target_to_pipe(GLenum target)
  * Return nominal bytes per texel for a compressed format, 0 for non-compressed
  * format.
  */
-static int
+static GLuint
 compressed_num_bytes(GLuint mesaFormat)
 {
    switch(mesaFormat) {
@@ -110,6 +114,25 @@ compressed_num_bytes(GLuint mesaFormat)
 }
 
 
+static GLboolean
+is_compressed_mesa_format(const struct gl_texture_format *format)
+{
+   switch (format->MesaFormat) {
+   case MESA_FORMAT_RGB_DXT1:
+   case MESA_FORMAT_RGBA_DXT1:
+   case MESA_FORMAT_RGBA_DXT3:
+   case MESA_FORMAT_RGBA_DXT5:
+   case MESA_FORMAT_SRGB_DXT1:
+   case MESA_FORMAT_SRGBA_DXT1:
+   case MESA_FORMAT_SRGBA_DXT3:
+   case MESA_FORMAT_SRGBA_DXT5:
+      return GL_TRUE;
+   default:
+      return GL_FALSE;
+   }
+}
+
+
 /** called via ctx->Driver.NewTextureImage() */
 static struct gl_texture_image *
 st_NewTextureImage(GLcontext * ctx)
@@ -169,7 +192,7 @@ st_FreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage)
  * than COPY_DWORDS would:
  * XXX Put this in src/mesa/main/imports.h ???
  */
-#if defined(i386) || defined(__i386__)
+#if defined(PIPE_CC_GCC) && defined(PIPE_ARCH_X86)
 static INLINE void *
 __memcpy(void *to, const void *from, size_t n)
 {
@@ -227,6 +250,21 @@ logbase2(int n)
 
 
 /**
+ * Return default texture usage bitmask for the given texture format.
+ */
+static GLuint
+default_usage(enum pipe_format fmt)
+{
+   GLuint usage = PIPE_TEXTURE_USAGE_SAMPLER;
+   if (pf_is_depth_stencil(fmt))
+      usage |= PIPE_TEXTURE_USAGE_DEPTH_STENCIL;
+   else
+      usage |= PIPE_TEXTURE_USAGE_RENDER_TARGET;
+   return usage;
+}
+
+
+/**
  * Allocate a pipe_texture object for the given st_texture_object using
  * the given st_texture_image to guess the mipmap size/levels.
  *
@@ -250,7 +288,7 @@ guess_and_alloc_texture(struct st_context *st,
    GLuint width = stImage->base.Width2;  /* size w/out border */
    GLuint height = stImage->base.Height2;
    GLuint depth = stImage->base.Depth2;
-   GLuint i, comp_byte = 0;
+   GLuint i, usage;
    enum pipe_format fmt;
 
    DBG("%s\n", __FUNCTION__);
@@ -308,10 +346,10 @@ guess_and_alloc_texture(struct st_context *st,
       lastLevel = firstLevel + MAX2(MAX2(l2width, l2height), l2depth);
    }
 
-   if (stImage->base.IsCompressed)
-      comp_byte = compressed_num_bytes(stImage->base.TexFormat->MesaFormat);
-
    fmt = st_mesa_format_to_pipe_format(stImage->base.TexFormat->MesaFormat);
+
+   usage = default_usage(fmt);
+
    stObj->pt = st_texture_create(st,
                                  gl_target_to_pipe(stObj->base.Target),
                                  fmt,
@@ -319,11 +357,7 @@ guess_and_alloc_texture(struct st_context *st,
                                  width,
                                  height,
                                  depth,
-                                 comp_byte,
-                                 ( (pf_is_depth_stencil(fmt) ?
-                                   PIPE_TEXTURE_USAGE_DEPTH_STENCIL :
-                                   PIPE_TEXTURE_USAGE_RENDER_TARGET) |
-                                   PIPE_TEXTURE_USAGE_SAMPLER ));
+                                 usage);
 
    DBG("%s - success\n", __FUNCTION__);
 }
@@ -368,6 +402,109 @@ strip_texture_border(GLint border,
 
 
 /**
+ * Try to do texture compression via rendering.  If the Gallium driver
+ * can render into a compressed surface this will allow us to do texture
+ * compression.
+ * \return GL_TRUE for success, GL_FALSE for failure
+ */
+static GLboolean
+compress_with_blit(GLcontext * ctx,
+                   GLenum target, GLint level,
+                   GLint xoffset, GLint yoffset, GLint zoffset,
+                   GLint width, GLint height, GLint depth,
+                   GLenum format, GLenum type, const void *pixels,
+                   const struct gl_pixelstore_attrib *unpack,
+                   struct gl_texture_image *texImage)
+{
+   const GLuint dstImageOffsets[1] = {0};
+   struct st_texture_image *stImage = st_texture_image(texImage);
+   struct pipe_screen *screen = ctx->st->pipe->screen;
+   const struct gl_texture_format *mesa_format;
+   struct pipe_texture templ;
+   struct pipe_texture *src_tex;
+   struct pipe_surface *dst_surface;
+   struct pipe_transfer *tex_xfer;
+   void *map;
+
+
+   if (!stImage->pt) {
+      /* XXX: Can this happen? Should we assert? */
+      return GL_FALSE;
+   }
+
+   /* get destination surface (in the compressed texture) */
+   dst_surface = screen->get_tex_surface(screen, stImage->pt,
+                                         stImage->face, stImage->level, 0,
+                                         PIPE_BUFFER_USAGE_GPU_WRITE);
+   if (!dst_surface) {
+      /* can't render into this format (or other problem) */
+      return GL_FALSE;
+   }
+
+   /* Choose format for the temporary RGBA texture image.
+    */
+   mesa_format = st_ChooseTextureFormat(ctx, GL_RGBA, format, type);
+   assert(mesa_format);
+   if (!mesa_format)
+      return GL_FALSE;
+
+   /* Create the temporary source texture
+    */
+   memset(&templ, 0, sizeof(templ));
+   templ.target = PIPE_TEXTURE_2D;
+   templ.format = st_mesa_format_to_pipe_format(mesa_format->MesaFormat);
+   pf_get_block(templ.format, &templ.block);
+   templ.width[0] = width;
+   templ.height[0] = height;
+   templ.depth[0] = 1;
+   templ.last_level = 0;
+   templ.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER;
+   src_tex = screen->texture_create(screen, &templ);
+
+   if (!src_tex)
+      return GL_FALSE;
+
+   /* Put user's tex data into the temporary texture
+    */
+   tex_xfer = st_cond_flush_get_tex_transfer(st_context(ctx), src_tex,
+					     0, 0, 0, /* face, level are zero */
+					     PIPE_TRANSFER_WRITE,
+					     0, 0, width, height); /* x, y, w, h */
+   map = screen->transfer_map(screen, tex_xfer);
+
+   mesa_format->StoreImage(ctx, 2, GL_RGBA, mesa_format,
+                           map,              /* dest ptr */
+                           0, 0, 0,          /* dest x/y/z offset */
+                           tex_xfer->stride, /* dest row stride (bytes) */
+                           dstImageOffsets,  /* image offsets (for 3D only) */
+                           width, height, 1, /* size */
+                           format, type,     /* source format/type */
+                           pixels,           /* source data */
+                           unpack);          /* source data packing */
+
+   screen->transfer_unmap(screen, tex_xfer);
+   screen->tex_transfer_destroy(tex_xfer);
+
+   /* copy / compress image */
+   util_blit_pixels_tex(ctx->st->blit,
+                        src_tex,          /* pipe_texture (src) */
+                        0, 0,             /* src x0, y0 */
+                        width, height,    /* src x1, y1 */
+                        dst_surface,      /* pipe_surface (dst) */
+                        xoffset, yoffset, /* dst x0, y0 */
+                        xoffset + width,  /* dst x1 */
+                        yoffset + height, /* dst y1 */
+                        0.0,              /* z */
+                        PIPE_TEX_MIPFILTER_NEAREST);
+
+   pipe_surface_reference(&dst_surface, NULL);
+   pipe_texture_reference(&src_tex, NULL);
+
+   return GL_TRUE;
+}
+
+
+/**
  * Do glTexImage1/2/3D().
  */
 static void
@@ -381,8 +518,9 @@ st_TexImage(GLcontext * ctx,
             const struct gl_pixelstore_attrib *unpack,
             struct gl_texture_object *texObj,
             struct gl_texture_image *texImage,
-            GLsizei imageSize, int compressed)
+            GLsizei imageSize, GLboolean compressed_src)
 {
+   struct pipe_screen *screen = ctx->st->pipe->screen;
    struct st_texture_object *stObj = st_texture_object(texObj);
    struct st_texture_image *stImage = st_texture_image(texImage);
    GLint postConvWidth, postConvHeight;
@@ -395,8 +533,7 @@ st_TexImage(GLcontext * ctx,
 
    /* gallium does not support texture borders, strip it off */
    if (border) {
-      strip_texture_border(border, &width, &height, &depth,
-                           unpack, &unpackNB);
+      strip_texture_border(border, &width, &height, &depth, unpack, &unpackNB);
       unpack = &unpackNB;
       texImage->Width = width;
       texImage->Height = height;
@@ -512,11 +649,12 @@ st_TexImage(GLcontext * ctx,
     * the expectation that the texture will be set up but nothing
     * more will be done.  This is where those calls return:
     */
-   if (compressed) {
+   if (compressed_src) {
       pixels = _mesa_validate_pbo_compressed_teximage(ctx, imageSize, pixels,
 						      unpack,
 						      "glCompressedTexImage");
-   } else {
+   }
+   else {
       pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, 1,
 					   format, type,
 					   pixels, unpack, "glTexImage");
@@ -524,12 +662,28 @@ st_TexImage(GLcontext * ctx,
    if (!pixels)
       return;
 
+   /* See if we can do texture compression with a blit/render.
+    */
+   if (!compressed_src &&
+       !ctx->Mesa_DXTn &&
+       is_compressed_mesa_format(texImage->TexFormat) &&
+       screen->is_format_supported(screen,
+                                   stImage->pt->format,
+                                   stImage->pt->target,
+                                   PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)) {
+      if (compress_with_blit(ctx, target, level, 0, 0, 0, width, height, depth,
+                             format, type, pixels, unpack, texImage)) {
+         goto done;
+      }
+   }
+
    if (stImage->pt) {
       texImage->Data = st_texture_image_map(ctx->st, stImage, 0,
                                             PIPE_TRANSFER_WRITE, 0, 0,
                                             stImage->base.Width,
                                             stImage->base.Height);
-      dstRowStride = stImage->transfer->stride;
+      if(stImage->transfer)
+         dstRowStride = stImage->transfer->stride;
    }
    else {
       /* Allocate regular memory and store the image there temporarily.   */
@@ -559,16 +713,16 @@ st_TexImage(GLcontext * ctx,
     * the blitter to copy.  Or, use the hardware to do the format
     * conversion and copy:
     */
-   if (compressed) {
+   if (compressed_src) {
       memcpy(texImage->Data, pixels, imageSize);
    }
    else {
-      GLuint srcImageStride = _mesa_image_image_stride(unpack, width, height,
-						       format, type);
-      int i;
+      const GLuint srcImageStride =
+         _mesa_image_image_stride(unpack, width, height, format, type);
+      GLint i;
       const GLubyte *src = (const GLubyte *) pixels;
 
-      for (i = 0; i++ < depth;) {
+      for (i = 0; i < depth; i++) {
 	 if (!texImage->TexFormat->StoreImage(ctx, dims, 
 					      texImage->_BaseFormat, 
 					      texImage->TexFormat, 
@@ -581,9 +735,11 @@ st_TexImage(GLcontext * ctx,
 	    _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
 	 }
 
-	 if (stImage->pt && i < depth) {
+	 if (stImage->pt && i + 1 < depth) {
+            /* unmap this slice */
 	    st_texture_image_unmap(ctx->st, stImage);
-	    texImage->Data = st_texture_image_map(ctx->st, stImage, i,
+            /* map next slice of 3D texture */
+	    texImage->Data = st_texture_image_map(ctx->st, stImage, i + 1,
                                                   PIPE_TRANSFER_WRITE, 0, 0,
                                                   stImage->base.Width,
                                                   stImage->base.Height);
@@ -594,7 +750,8 @@ st_TexImage(GLcontext * ctx,
 
    _mesa_unmap_teximage_pbo(ctx, unpack);
 
-   if (stImage->pt) {
+done:
+   if (stImage->pt && texImage->Data) {
       st_texture_image_unmap(ctx->st, stImage);
       texImage->Data = NULL;
    }
@@ -616,9 +773,9 @@ st_TexImage3D(GLcontext * ctx,
               struct gl_texture_object *texObj,
               struct gl_texture_image *texImage)
 {
-   st_TexImage(ctx, 3, target, level,
-                 internalFormat, width, height, depth, border,
-                 format, type, pixels, unpack, texObj, texImage, 0, 0);
+   st_TexImage(ctx, 3, target, level, internalFormat, width, height, depth,
+               border, format, type, pixels, unpack, texObj, texImage,
+               0, GL_FALSE);
 }
 
 
@@ -632,9 +789,8 @@ st_TexImage2D(GLcontext * ctx,
               struct gl_texture_object *texObj,
               struct gl_texture_image *texImage)
 {
-   st_TexImage(ctx, 2, target, level,
-                 internalFormat, width, height, 1, border,
-                 format, type, pixels, unpack, texObj, texImage, 0, 0);
+   st_TexImage(ctx, 2, target, level, internalFormat, width, height, 1, border,
+               format, type, pixels, unpack, texObj, texImage, 0, GL_FALSE);
 }
 
 
@@ -648,9 +804,8 @@ st_TexImage1D(GLcontext * ctx,
               struct gl_texture_object *texObj,
               struct gl_texture_image *texImage)
 {
-   st_TexImage(ctx, 1, target, level,
-                 internalFormat, width, 1, 1, border,
-                 format, type, pixels, unpack, texObj, texImage, 0, 0);
+   st_TexImage(ctx, 1, target, level, internalFormat, width, 1, 1, border,
+               format, type, pixels, unpack, texObj, texImage, 0, GL_FALSE);
 }
 
 
@@ -662,12 +817,96 @@ st_CompressedTexImage2D(GLcontext *ctx, GLenum target, GLint level,
                         struct gl_texture_object *texObj,
                         struct gl_texture_image *texImage)
 {
-   st_TexImage(ctx, 2, target, level,
-		 internalFormat, width, height, 1, border,
-		 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, 1);
+   st_TexImage(ctx, 2, target, level, internalFormat, width, height, 1, border,
+               0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, GL_TRUE);
 }
 
 
+
+/**
+ * glGetTexImage() helper: decompress a compressed texture by rendering
+ * a textured quad.  Store the results in the user's buffer.
+ */
+static void
+decompress_with_blit(GLcontext * ctx, GLenum target, GLint level,
+                     GLenum format, GLenum type, GLvoid *pixels,
+                     struct gl_texture_object *texObj,
+                     struct gl_texture_image *texImage)
+{
+   struct pipe_screen *screen = ctx->st->pipe->screen;
+   struct st_texture_image *stImage = st_texture_image(texImage);
+   const GLuint width = texImage->Width;
+   const GLuint height = texImage->Height;
+   struct pipe_surface *dst_surface;
+   struct pipe_texture *dst_texture;
+   struct pipe_transfer *tex_xfer;
+
+   /* create temp / dest surface */
+   if (!util_create_rgba_surface(screen, width, height,
+                                 &dst_texture, &dst_surface)) {
+      _mesa_problem(ctx, "util_create_rgba_surface() failed "
+                    "in decompress_with_blit()");
+      return;
+   }
+
+   /* blit/render/decompress */
+   util_blit_pixels_tex(ctx->st->blit,
+                        stImage->pt,      /* pipe_texture (src) */
+                        0, 0,             /* src x0, y0 */
+                        width, height,    /* src x1, y1 */
+                        dst_surface,      /* pipe_surface (dst) */
+                        0, 0,             /* dst x0, y0 */
+                        width, height,    /* dst x1, y1 */
+                        0.0,              /* z */
+                        PIPE_TEX_MIPFILTER_NEAREST);
+
+   /* map the dst_surface so we can read from it */
+   tex_xfer = st_cond_flush_get_tex_transfer(st_context(ctx),
+					     dst_texture, 0, 0, 0,
+					     PIPE_TRANSFER_READ,
+					     0, 0, width, height);
+
+   pixels = _mesa_map_readpix_pbo(ctx, &ctx->Pack, pixels);
+
+   /* copy/pack data into user buffer */
+   if (st_equal_formats(stImage->pt->format, format, type)) {
+      /* memcpy */
+      const uint bytesPerRow = width * pf_get_size(stImage->pt->format);
+      ubyte *map = screen->transfer_map(screen, tex_xfer);
+      GLuint row;
+      for (row = 0; row < height; row++) {
+         GLvoid *dest = _mesa_image_address2d(&ctx->Pack, pixels, width,
+                                              height, format, type, row, 0);
+         memcpy(dest, map, bytesPerRow);
+         map += tex_xfer->stride;
+      }
+      screen->transfer_unmap(screen, tex_xfer);
+   }
+   else {
+      /* format translation via floats */
+      GLuint row;
+      for (row = 0; row < height; row++) {
+         const GLbitfield transferOps = 0x0; /* bypassed for glGetTexImage() */
+         GLfloat rgba[4 * MAX_WIDTH];
+         GLvoid *dest = _mesa_image_address2d(&ctx->Pack, pixels, width,
+                                              height, format, type, row, 0);
+
+         /* get float[4] rgba row from surface */
+         pipe_get_tile_rgba(tex_xfer, 0, row, width, 1, rgba);
+
+         _mesa_pack_rgba_span_float(ctx, width, (GLfloat (*)[4]) rgba, format,
+                                    type, dest, &ctx->Pack, transferOps);
+      }
+   }
+
+   _mesa_unmap_readpix_pbo(ctx, &ctx->Pack);
+
+   /* destroy the temp / dest surface */
+   util_destroy_rgba_surface(dst_texture, dst_surface);
+}
+
+
+
 /**
  * Need to map texture image into memory before copying image data,
  * then unmap it.
@@ -676,22 +915,36 @@ static void
 st_get_tex_image(GLcontext * ctx, GLenum target, GLint level,
                  GLenum format, GLenum type, GLvoid * pixels,
                  struct gl_texture_object *texObj,
-                 struct gl_texture_image *texImage, int compressed)
+                 struct gl_texture_image *texImage, GLboolean compressed_dst)
 {
    struct st_texture_image *stImage = st_texture_image(texImage);
-   GLuint dstImageStride = _mesa_image_image_stride(&ctx->Pack,
-                                                    texImage->Width,
-						    texImage->Height,
-                                                    format, type);
-   GLuint depth;
-   GLuint i;
+   const GLuint dstImageStride =
+      _mesa_image_image_stride(&ctx->Pack, texImage->Width, texImage->Height,
+                               format, type);
+   GLuint depth, i;
    GLubyte *dest;
 
+   if (stImage->pt &&
+       pf_is_compressed(stImage->pt->format) &&
+       !compressed_dst) {
+      /* Need to decompress the texture.
+       * We'll do this by rendering a textured quad.
+       * Note that we only expect RGBA formats (no Z/depth formats).
+       */
+      decompress_with_blit(ctx, target, level, format, type, pixels,
+                           texObj, texImage);
+      return;
+   }
+
    /* Map */
    if (stImage->pt) {
       /* Image is stored in hardware format in a buffer managed by the
        * kernel.  Need to explicitly map and unmap it.
        */
+
+      st_teximage_flush_before_map(ctx->st, stImage->pt, 0, level,
+				   PIPE_TRANSFER_READ);
+
       texImage->Data = st_texture_image_map(ctx->st, stImage, 0,
                                             PIPE_TRANSFER_READ, 0, 0,
                                             stImage->base.Width,
@@ -715,18 +968,21 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level,
 
    dest = (GLubyte *) pixels;
 
-   for (i = 0; i++ < depth;) {
-      if (compressed) {
+   for (i = 0; i < depth; i++) {
+      if (compressed_dst) {
 	 _mesa_get_compressed_teximage(ctx, target, level, dest,
 				       texObj, texImage);
-      } else {
+      }
+      else {
 	 _mesa_get_teximage(ctx, target, level, format, type, dest,
 			    texObj, texImage);
       }
 
-      if (stImage->pt && i < depth) {
+      if (stImage->pt && i + 1 < depth) {
+         /* unmap this slice */
 	 st_texture_image_unmap(ctx->st, stImage);
-	 texImage->Data = st_texture_image_map(ctx->st, stImage, i,
+         /* map next slice of 3D texture */
+	 texImage->Data = st_texture_image_map(ctx->st, stImage, i + 1,
                                                PIPE_TRANSFER_READ, 0, 0,
                                                stImage->base.Width,
                                                stImage->base.Height);
@@ -750,8 +1006,8 @@ st_GetTexImage(GLcontext * ctx, GLenum target, GLint level,
                struct gl_texture_object *texObj,
                struct gl_texture_image *texImage)
 {
-   st_get_tex_image(ctx, target, level, format, type, pixels,
-                    texObj, texImage, 0);
+   st_get_tex_image(ctx, target, level, format, type, pixels, texObj, texImage,
+                    GL_FALSE);
 }
 
 
@@ -761,17 +1017,14 @@ st_GetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level,
                          struct gl_texture_object *texObj,
                          struct gl_texture_image *texImage)
 {
-   st_get_tex_image(ctx, target, level, 0, 0, pixels,
-                    (struct gl_texture_object *) texObj,
-                    (struct gl_texture_image *) texImage, 1);
+   st_get_tex_image(ctx, target, level, 0, 0, pixels, texObj, texImage,
+                    GL_TRUE);
 }
 
 
 
 static void
-st_TexSubimage(GLcontext * ctx,
-               GLint dims,
-               GLenum target, GLint level,
+st_TexSubimage(GLcontext *ctx, GLint dims, GLenum target, GLint level,
                GLint xoffset, GLint yoffset, GLint zoffset,
                GLint width, GLint height, GLint depth,
                GLenum format, GLenum type, const void *pixels,
@@ -779,11 +1032,12 @@ st_TexSubimage(GLcontext * ctx,
                struct gl_texture_object *texObj,
                struct gl_texture_image *texImage)
 {
+   struct pipe_screen *screen = ctx->st->pipe->screen;
    struct st_texture_image *stImage = st_texture_image(texImage);
    GLuint dstRowStride;
-   GLuint srcImageStride = _mesa_image_image_stride(packing, width, height,
-						    format, type);
-   int i;
+   const GLuint srcImageStride =
+      _mesa_image_image_stride(packing, width, height, format, type);
+   GLint i;
    const GLubyte *src;
 
    DBG("%s target %s level %d offset %d,%d %dx%d\n", __FUNCTION__,
@@ -796,10 +1050,28 @@ st_TexSubimage(GLcontext * ctx,
    if (!pixels)
       return;
 
+   /* See if we can do texture compression with a blit/render.
+    */
+   if (!ctx->Mesa_DXTn &&
+       is_compressed_mesa_format(texImage->TexFormat) &&
+       screen->is_format_supported(screen,
+                                   stImage->pt->format,
+                                   stImage->pt->target,
+                                   PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)) {
+      if (compress_with_blit(ctx, target, level,
+                             xoffset, yoffset, zoffset,
+                             width, height, depth,
+                             format, type, pixels, packing, texImage)) {
+         goto done;
+      }
+   }
+
    /* Map buffer if necessary.  Need to lock to prevent other contexts
     * from uploading the buffer under us.
     */
    if (stImage->pt) {
+      st_teximage_flush_before_map(ctx->st, stImage->pt, 0, level,
+				   PIPE_TRANSFER_WRITE);
       texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset, 
                                             PIPE_TRANSFER_WRITE,
                                             xoffset, yoffset,
@@ -814,7 +1086,7 @@ st_TexSubimage(GLcontext * ctx,
    src = (const GLubyte *) pixels;
    dstRowStride = stImage->transfer->stride;
 
-   for (i = 0; i++ < depth;) {
+   for (i = 0; i < depth; i++) {
       if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat,
 					   texImage->TexFormat,
 					   texImage->Data,
@@ -826,10 +1098,12 @@ st_TexSubimage(GLcontext * ctx,
 	 _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage");
       }
 
-      if (stImage->pt && i < depth) {
-         /* map next slice of 3D texture */
+      if (stImage->pt && i + 1 < depth) {
+         /* unmap this slice */
 	 st_texture_image_unmap(ctx->st, stImage);
-	 texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset + i,
+         /* map next slice of 3D texture */
+	 texImage->Data = st_texture_image_map(ctx->st, stImage,
+                                               zoffset + i + 1,
                                                PIPE_TRANSFER_WRITE,
                                                xoffset, yoffset,
                                                width, height);
@@ -837,88 +1111,74 @@ st_TexSubimage(GLcontext * ctx,
       }
    }
 
-   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
-      ctx->Driver.GenerateMipmap(ctx, target, texObj);
-   }
-
    _mesa_unmap_teximage_pbo(ctx, packing);
 
+done:
    if (stImage->pt) {
       st_texture_image_unmap(ctx->st, stImage);
       texImage->Data = NULL;
    }
+
+   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+      ctx->Driver.GenerateMipmap(ctx, target, texObj);
+   }
 }
 
 
 
 static void
-st_TexSubImage3D(GLcontext * ctx,
-                   GLenum target,
-                   GLint level,
-                   GLint xoffset, GLint yoffset, GLint zoffset,
-                   GLsizei width, GLsizei height, GLsizei depth,
-                   GLenum format, GLenum type,
-                   const GLvoid * pixels,
-                   const struct gl_pixelstore_attrib *packing,
-                   struct gl_texture_object *texObj,
-                   struct gl_texture_image *texImage)
+st_TexSubImage3D(GLcontext *ctx, GLenum target, GLint level,
+                 GLint xoffset, GLint yoffset, GLint zoffset,
+                 GLsizei width, GLsizei height, GLsizei depth,
+                 GLenum format, GLenum type, const GLvoid *pixels,
+                 const struct gl_pixelstore_attrib *packing,
+                 struct gl_texture_object *texObj,
+                 struct gl_texture_image *texImage)
 {
-   st_TexSubimage(ctx, 3, target, level,
-                  xoffset, yoffset, zoffset,
-                  width, height, depth,
-                  format, type, pixels, packing, texObj, texImage);
+   st_TexSubimage(ctx, 3, target, level, xoffset, yoffset, zoffset,
+                  width, height, depth, format, type,
+                  pixels, packing, texObj, texImage);
 }
 
 
 static void
-st_TexSubImage2D(GLcontext * ctx,
-                   GLenum target,
-                   GLint level,
-                   GLint xoffset, GLint yoffset,
-                   GLsizei width, GLsizei height,
-                   GLenum format, GLenum type,
-                   const GLvoid * pixels,
-                   const struct gl_pixelstore_attrib *packing,
-                   struct gl_texture_object *texObj,
-                   struct gl_texture_image *texImage)
+st_TexSubImage2D(GLcontext *ctx, GLenum target, GLint level,
+                 GLint xoffset, GLint yoffset,
+                 GLsizei width, GLsizei height,
+                 GLenum format, GLenum type, const GLvoid * pixels,
+                 const struct gl_pixelstore_attrib *packing,
+                 struct gl_texture_object *texObj,
+                 struct gl_texture_image *texImage)
 {
-   st_TexSubimage(ctx, 2, target, level,
-                  xoffset, yoffset, 0,
-                  width, height, 1,
-                  format, type, pixels, packing, texObj, texImage);
+   st_TexSubimage(ctx, 2, target, level, xoffset, yoffset, 0,
+                  width, height, 1, format, type,
+                  pixels, packing, texObj, texImage);
 }
 
 
 static void
-st_TexSubImage1D(GLcontext * ctx,
-                   GLenum target,
-                   GLint level,
-                   GLint xoffset,
-                   GLsizei width,
-                   GLenum format, GLenum type,
-                   const GLvoid * pixels,
-                   const struct gl_pixelstore_attrib *packing,
-                   struct gl_texture_object *texObj,
-                   struct gl_texture_image *texImage)
+st_TexSubImage1D(GLcontext *ctx, GLenum target, GLint level,
+                 GLint xoffset, GLsizei width, GLenum format, GLenum type,
+                 const GLvoid * pixels,
+                 const struct gl_pixelstore_attrib *packing,
+                 struct gl_texture_object *texObj,
+                 struct gl_texture_image *texImage)
 {
-   st_TexSubimage(ctx, 1, target, level,
-                  xoffset, 0, 0,
-                  width, 1, 1,
+   st_TexSubimage(ctx, 1, target, level, xoffset, 0, 0, width, 1, 1,
                   format, type, pixels, packing, texObj, texImage);
 }
 
 
 
 /**
- * Do a CopyTexSubImage operation using a read transfer from the source, a write
- * transfer to the destination and get_tile()/put_tile() to access the pixels/texels.
+ * Do a CopyTexSubImage operation using a read transfer from the source,
+ * a write transfer to the destination and get_tile()/put_tile() to access
+ * the pixels/texels.
  *
  * Note: srcY=0=TOP of renderbuffer
  */
 static void
-fallback_copy_texsubimage(GLcontext *ctx,
-                          GLenum target,
-                          GLint level,
+fallback_copy_texsubimage(GLcontext *ctx, GLenum target, GLint level,
                           struct st_renderbuffer *strb,
                           struct st_texture_image *stImage,
                           GLenum baseFormat,
@@ -937,12 +1197,15 @@ fallback_copy_texsubimage(GLcontext *ctx,
       srcY = strb->Base.Height - srcY - height;
    }
 
-   src_trans = screen->get_tex_transfer( screen,
-                                         strb->texture,
-                                         0, 0, 0,
-                                         PIPE_TRANSFER_READ,
-                                         srcX, srcY,
-                                         width, height);
+   src_trans = st_cond_flush_get_tex_transfer( st_context(ctx),
+					       strb->texture,
+					       0, 0, 0,
+					       PIPE_TRANSFER_READ,
+					       srcX, srcY,
+					       width, height);
+
+   st_teximage_flush_before_map(ctx->st, stImage->pt, 0, 0,
+				PIPE_TRANSFER_WRITE);
 
    texDest = st_texture_image_map(ctx->st, stImage, 0, PIPE_TRANSFER_WRITE,
                                   destX, destY, width, height);
@@ -980,8 +1243,8 @@ fallback_copy_texsubimage(GLcontext *ctx,
 
       if (tempSrc && texDest) {
          const GLint dims = 2;
+         const GLint dstRowStride = stImage->transfer->stride;
          struct gl_texture_image *texImage = &stImage->base;
-         GLint dstRowStride = stImage->transfer->stride;
          struct gl_pixelstore_attrib unpack = ctx->DefaultPacking;
 
          if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
@@ -1054,8 +1317,12 @@ st_copy_texsubimage(GLcontext *ctx,
    GLboolean use_fallback = GL_TRUE;
    GLboolean matching_base_formats;
 
-   /* any rendering in progress must complete before we grab the fb image */
-   st_finish(ctx->st);
+   /* any rendering in progress must flushed before we grab the fb image */
+   st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   /* make sure finalize_textures has been called? 
+    */
+   if (0) st_validate_state(ctx->st);
 
    /* determine if copying depth or color data */
    if (texBaseFormat == GL_DEPTH_COMPONENT ||
@@ -1070,6 +1337,39 @@ st_copy_texsubimage(GLcontext *ctx,
       strb = st_renderbuffer(fb->_ColorReadBuffer);
    }
 
+   if (!strb || !strb->surface || !stImage->pt) {
+      debug_printf("%s: null strb or stImage\n", __FUNCTION__);
+      return;
+   }
+
+   if (srcX < 0) {
+      width -= -srcX;
+      destX += -srcX;
+      srcX = 0;
+   }
+
+   if (srcY < 0) {
+      height -= -srcY;
+      destY += -srcY;
+      srcY = 0;
+   }
+
+   if (destX < 0) {
+      width -= -destX;
+      srcX += -destX;
+      destX = 0;
+   }
+
+   if (destY < 0) {
+      height -= -destY;
+      srcY += -destY;
+      destY = 0;
+   }
+
+   if (width < 0 || height < 0)
+      return;
+
+
    assert(strb);
    assert(strb->surface);
    assert(stImage->pt);
@@ -1095,7 +1395,6 @@ st_copy_texsubimage(GLcontext *ctx,
       if (src_format == dest_format && !do_flip) {
          /* use surface_copy() / blit */
 
-
          dest_surface = screen->get_tex_surface(screen, stImage->pt,
                                                 stImage->face, stImage->level,
                                                 destZ,
@@ -1122,7 +1421,7 @@ st_copy_texsubimage(GLcontext *ctx,
                                            PIPE_TEXTURE_USAGE_RENDER_TARGET,
                                            0)) {
          /* draw textured quad to do the copy */
-         int srcY0, srcY1;
+         GLint srcY0, srcY1;
 
          dest_surface = screen->get_tex_surface(screen, stImage->pt,
                                                 stImage->face, stImage->level,
@@ -1179,11 +1478,6 @@ st_CopyTexImage1D(GLcontext * ctx, GLenum target, GLint level,
    struct gl_texture_image *texImage =
       _mesa_select_tex_image(ctx, texObj, target, level);
 
-#if 0
-   if (border)
-      goto fail;
-#endif
-
    /* Setup or redefine the texture object, texture and texture
     * image.  Don't populate yet.  
     */
@@ -1274,8 +1568,8 @@ calculate_first_last_level(struct st_texture_object *stObj)
     * and having firstLevel and lastLevel as signed prevents the need for
     * extra sign checks.
     */
-   int firstLevel;
-   int lastLevel;
+   GLint firstLevel;
+   GLint lastLevel;
 
    /* Yes, this looks overly complicated, but it's all needed.
     */
@@ -1329,16 +1623,21 @@ copy_image_data_to_texture(struct st_context *st,
 
       /* More straightforward upload.  
        */
-      st_texture_image_data(st->pipe,
-                               stObj->pt,
-                               stImage->face,
-                               dstLevel,
-                               stImage->base.Data,
-                               stImage->base.RowStride * 
-                               stObj->pt->block.size,
-                               stImage->base.RowStride *
-                               stImage->base.Height *
-                               stObj->pt->block.size);
+
+      st_teximage_flush_before_map(st, stObj->pt, stImage->face, dstLevel,
+				   PIPE_TRANSFER_WRITE);
+
+
+      st_texture_image_data(st,
+                            stObj->pt,
+                            stImage->face,
+                            dstLevel,
+                            stImage->base.Data,
+                            stImage->base.RowStride * 
+                            stObj->pt->block.size,
+                            stImage->base.RowStride *
+                            stImage->base.Height *
+                            stObj->pt->block.size);
       _mesa_align_free(stImage->base.Data);
       stImage->base.Data = NULL;
    }
@@ -1360,9 +1659,7 @@ st_finalize_texture(GLcontext *ctx,
 {
    struct st_texture_object *stObj = st_texture_object(tObj);
    const GLuint nr_faces = (stObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
-   int comp_byte = 0;
-   int cpp;
-   GLuint face;
+   GLuint cpp, face;
    struct st_texture_image *firstImage;
 
    *needFlush = GL_FALSE;
@@ -1384,14 +1681,12 @@ st_finalize_texture(GLcontext *ctx,
    if (firstImage->pt &&
        firstImage->pt != stObj->pt &&
        firstImage->pt->last_level >= stObj->lastLevel) {
-
       pipe_texture_reference(&stObj->pt, firstImage->pt);
    }
 
    /* FIXME: determine format block instead of cpp */
    if (firstImage->base.IsCompressed) {
-      comp_byte = compressed_num_bytes(firstImage->base.TexFormat->MesaFormat);
-      cpp = comp_byte;
+      cpp = compressed_num_bytes(firstImage->base.TexFormat->MesaFormat);
    }
    else {
       cpp = firstImage->base.TexFormat->TexelBytes;
@@ -1409,8 +1704,9 @@ st_finalize_texture(GLcontext *ctx,
           stObj->pt->width[0] != firstImage->base.Width2 ||
           stObj->pt->height[0] != firstImage->base.Height2 ||
           stObj->pt->depth[0] != firstImage->base.Depth2 ||
-          stObj->pt->block.size/stObj->pt->block.width != cpp || /* Nominal bytes per pixel */
-          stObj->pt->compressed != firstImage->base.IsCompressed) {
+          /* Nominal bytes per pixel: */
+          stObj->pt->block.size / stObj->pt->block.width != cpp)
+      {
          pipe_texture_reference(&stObj->pt, NULL);
          ctx->st->dirty.st |= ST_NEW_FRAMEBUFFER;
       }
@@ -1421,6 +1717,8 @@ st_finalize_texture(GLcontext *ctx,
    if (!stObj->pt) {
       const enum pipe_format fmt =
          st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat);
+      GLuint usage = default_usage(fmt);
+
       stObj->pt = st_texture_create(ctx->st,
                                     gl_target_to_pipe(stObj->base.Target),
                                     fmt,
@@ -1428,11 +1726,7 @@ st_finalize_texture(GLcontext *ctx,
                                     firstImage->base.Width2,
                                     firstImage->base.Height2,
                                     firstImage->base.Depth2,
-                                    comp_byte,
-                                    ( (pf_is_depth_stencil(fmt) ?
-                                      PIPE_TEXTURE_USAGE_DEPTH_STENCIL :
-                                      PIPE_TEXTURE_USAGE_RENDER_TARGET) |
-                                      PIPE_TEXTURE_USAGE_SAMPLER ));
+                                    usage);
 
       if (!stObj->pt) {
          _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c
index b27274725f..92ddffc014 100644
--- a/src/mesa/state_tracker/st_context.c
+++ b/src/mesa/state_tracker/st_context.c
@@ -177,6 +177,12 @@ struct st_context *st_create_context(struct pipe_context *pipe,
 
    ctx = _mesa_create_context(visual, shareCtx, &funcs, NULL);
 
+   /* XXX: need a capability bit in gallium to query if the pipe
+    * driver prefers DP4 or MUL/MAD for vertex transformation.
+    */
+   if (debug_get_bool_option("MESA_MVP_DP4", FALSE))
+      _mesa_set_mvp_with_dp4( ctx, GL_TRUE );
+
    return st_create_context_priv(ctx, pipe);
 }
 
@@ -227,6 +233,7 @@ void st_destroy_context( struct st_context *st )
    struct pipe_context *pipe = st->pipe;
    struct cso_context *cso = st->cso_context;
    GLcontext *ctx = st->ctx;
+   GLuint i;
 
    /* need to unbind and destroy CSO objects before anything else */
    cso_release_all(st->cso_context);
@@ -234,6 +241,12 @@ void st_destroy_context( struct st_context *st )
    st_reference_fragprog(st, &st->fp, NULL);
    st_reference_vertprog(st, &st->vp, NULL);
 
+   /* release framebuffer surfaces */
+   for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) {
+      pipe_surface_reference(&st->state.framebuffer.cbufs[i], NULL);
+   }
+   pipe_surface_reference(&st->state.framebuffer.zsbuf, NULL);
+
    _mesa_delete_program_cache(st->ctx, st->pixel_xfer.cache);
 
    _vbo_DestroyContext(st->ctx);
@@ -250,9 +263,10 @@ void st_destroy_context( struct st_context *st )
 }
 
 
-void st_make_current(struct st_context *st,
-                     struct st_framebuffer *draw,
-                     struct st_framebuffer *read)
+GLboolean
+st_make_current(struct st_context *st,
+                struct st_framebuffer *draw,
+                struct st_framebuffer *read)
 {
    /* Call this periodically to detect when the user has begun using
     * GL rendering from multiple threads.
@@ -261,7 +275,8 @@ void st_make_current(struct st_context *st,
 
    if (st) {
       GLboolean firstTime = st->ctx->FirstTimeCurrent;
-      _mesa_make_current(st->ctx, &draw->Base, &read->Base);
+      if(!_mesa_make_current(st->ctx, &draw->Base, &read->Base))
+         return GL_FALSE;
       /* Need to initialize viewport here since draw->Base->Width/Height
        * will still be zero at this point.
        * This could be improved, but would require rather extensive work
@@ -273,12 +288,19 @@ void st_make_current(struct st_context *st,
          _mesa_set_scissor(st->ctx, 0, 0, w, h);
 
       }
+      return GL_TRUE;
    }
    else {
-      _mesa_make_current(NULL, NULL, NULL);
+      return _mesa_make_current(NULL, NULL, NULL);
    }
 }
 
+struct st_context *st_get_current(void)
+{
+   GET_CURRENT_CONTEXT(ctx);
+
+   return (ctx == NULL) ? NULL : ctx->st;
+}
 
 void st_copy_context_state(struct st_context *dst,
                            struct st_context *src,
diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h
index ae8c2978bf..18adb35e87 100644
--- a/src/mesa/state_tracker/st_context.h
+++ b/src/mesa/state_tracker/st_context.h
@@ -45,6 +45,7 @@ struct blit_state;
 struct bitmap_cache;
 
 
+/** XXX we'd like to get rid of these */
 #define FRONT_STATUS_UNDEFINED    0
 #define FRONT_STATUS_DIRTY        1
 #define FRONT_STATUS_COPY_OF_BACK 2
@@ -111,22 +112,15 @@ struct st_context
       struct gl_fragment_program *fragment_program;
    } cb;
 
-   GLuint frontbuffer_status;  /**< one of FRONT_STATUS_ */
+   GLuint frontbuffer_status;  /**< one of FRONT_STATUS_ (XXX to be removed) */
 
    char vendor[100];
    char renderer[100];
 
-   /* State to be validated:
-    */
-   struct st_tracked_state **atoms;
-   GLuint nr_atoms;
-
    struct st_state_flags dirty;
 
    GLboolean missing_textures;
 
-   GLfloat polygon_offset_scale; /* ?? */
-
    /** Mapping from VERT_RESULT_x to post-transformed vertex slot */
    const GLuint *vertex_result_to_slot;
 
diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c
index 225541a30b..8e036223c6 100644
--- a/src/mesa/state_tracker/st_draw.c
+++ b/src/mesa/state_tracker/st_draw.c
@@ -159,12 +159,21 @@ static GLuint fixed_types[4] = {
  * Return a PIPE_FORMAT_x for the given GL datatype and size.
  */
 GLuint
-st_pipe_vertex_format(GLenum type, GLuint size, GLboolean normalized)
+st_pipe_vertex_format(GLenum type, GLuint size, GLenum format,
+                      GLboolean normalized)
 {
    assert((type >= GL_BYTE && type <= GL_DOUBLE) ||
           type == GL_FIXED);
    assert(size >= 1);
    assert(size <= 4);
+   assert(format == GL_RGBA || format == GL_BGRA);
+
+   if (format == GL_BGRA) {
+      /* this is an odd-ball case */
+      assert(type == GL_UNSIGNED_BYTE);
+      assert(normalized);
+      return PIPE_FORMAT_B8G8R8A8_UNORM;
+   }
 
    if (normalized) {
       switch (type) {
@@ -392,6 +401,7 @@ setup_interleaved_attribs(GLcontext *ctx,
       velements[attr].src_format =
          st_pipe_vertex_format(arrays[mesaAttr]->Type,
                                arrays[mesaAttr]->Size,
+                               arrays[mesaAttr]->Format,
                                arrays[mesaAttr]->Normalized);
       assert(velements[attr].src_format);
    }
@@ -479,6 +489,7 @@ setup_non_interleaved_attribs(GLcontext *ctx,
       velements[attr].src_format
          = st_pipe_vertex_format(arrays[mesaAttr]->Type,
                                  arrays[mesaAttr]->Size,
+                                 arrays[mesaAttr]->Format,
                                  arrays[mesaAttr]->Normalized);
       assert(velements[attr].src_format);
    }
diff --git a/src/mesa/state_tracker/st_draw.h b/src/mesa/state_tracker/st_draw.h
index da04fce8e2..dcfe7e1536 100644
--- a/src/mesa/state_tracker/st_draw.h
+++ b/src/mesa/state_tracker/st_draw.h
@@ -62,7 +62,8 @@ st_feedback_draw_vbo(GLcontext *ctx,
 /* Internal function:
  */
 extern GLuint
-st_pipe_vertex_format(GLenum type, GLuint size, GLboolean normalized);
+st_pipe_vertex_format(GLenum type, GLuint size, GLenum format,
+                      GLboolean normalized);
 
 
 /**
diff --git a/src/mesa/state_tracker/st_draw_feedback.c b/src/mesa/state_tracker/st_draw_feedback.c
index e533afd051..2712c131c0 100644
--- a/src/mesa/state_tracker/st_draw_feedback.c
+++ b/src/mesa/state_tracker/st_draw_feedback.c
@@ -178,6 +178,7 @@ st_feedback_draw_vbo(GLcontext *ctx,
       velements[attr].src_format = 
          st_pipe_vertex_format(arrays[mesaAttr]->Type,
                                arrays[mesaAttr]->Size,
+                               arrays[mesaAttr]->Format,
                                arrays[mesaAttr]->Normalized);
       assert(velements[attr].src_format);
 
@@ -196,13 +197,10 @@ st_feedback_draw_vbo(GLcontext *ctx,
    draw_set_vertex_elements(draw, vp->num_inputs, velements);
 
    if (ib) {
-      unsigned indexSize;
       struct gl_buffer_object *bufobj = ib->obj;
-      struct st_buffer_object *stobj = st_buffer_object(bufobj);
+      unsigned indexSize;
       void *map;
 
-      index_buffer_handle = stobj->buffer;
-
       switch (ib->type) {
       case GL_UNSIGNED_INT:
          indexSize = 4;
@@ -215,9 +213,19 @@ st_feedback_draw_vbo(GLcontext *ctx,
 	 return;
       }
 
-      map = pipe_buffer_map(pipe->screen, index_buffer_handle,
-                            PIPE_BUFFER_USAGE_CPU_READ);
-      draw_set_mapped_element_buffer(draw, indexSize, map);
+      if (bufobj && bufobj->Name) {
+         struct st_buffer_object *stobj = st_buffer_object(bufobj);
+
+         index_buffer_handle = stobj->buffer;
+
+         map = pipe_buffer_map(pipe->screen, index_buffer_handle,
+                               PIPE_BUFFER_USAGE_CPU_READ);
+
+         draw_set_mapped_element_buffer(draw, indexSize, map);
+      }
+      else {
+         draw_set_mapped_element_buffer(draw, indexSize, (void *) ib->ptr);
+      }
    }
    else {
       /* no index/element buffer */
@@ -252,7 +260,7 @@ st_feedback_draw_vbo(GLcontext *ctx,
          draw_set_mapped_vertex_buffer(draw, i, NULL);
       }
    }
-   if (ib) {
+   if (index_buffer_handle) {
       pipe_buffer_unmap(pipe->screen, index_buffer_handle);
       draw_set_mapped_element_buffer(draw, 0, NULL);
    }
diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c
index 8f6be50774..8ed1211db6 100644
--- a/src/mesa/state_tracker/st_extensions.c
+++ b/src/mesa/state_tracker/st_extensions.c
@@ -138,6 +138,7 @@ void st_init_extensions(struct st_context *st)
    /*
     * Extensions that are supported by all Gallium drivers:
     */
+   ctx->Extensions.ARB_copy_buffer = GL_TRUE;
    ctx->Extensions.ARB_multisample = GL_TRUE;
    ctx->Extensions.ARB_fragment_program = GL_TRUE;
    ctx->Extensions.ARB_texture_border_clamp = GL_TRUE; /* XXX temp */
@@ -167,6 +168,9 @@ void st_init_extensions(struct st_context *st)
    ctx->Extensions.EXT_texture_env_combine = GL_TRUE;
    ctx->Extensions.EXT_texture_env_dot3 = GL_TRUE;
    ctx->Extensions.EXT_texture_lod_bias = GL_TRUE;
+   ctx->Extensions.EXT_vertex_array_bgra = GL_TRUE;
+
+   ctx->Extensions.APPLE_vertex_array_object = GL_TRUE;
 
    ctx->Extensions.NV_blend_square = GL_TRUE;
    ctx->Extensions.NV_texgen_reflection = GL_TRUE;
diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c
index 9e2d60c926..b243c249e3 100644
--- a/src/mesa/state_tracker/st_format.c
+++ b/src/mesa/state_tracker/st_format.c
@@ -392,7 +392,7 @@ default_depth_format(struct pipe_screen *screen,
  *                   or PIPE_TEXTURE_USAGE_SAMPLER
  */
 enum pipe_format
-st_choose_format(struct pipe_context *pipe, GLint internalFormat,
+st_choose_format(struct pipe_context *pipe, GLenum internalFormat,
                  enum pipe_texture_target target, unsigned tex_usage)
 {
    struct pipe_screen *screen = pipe->screen;
@@ -594,9 +594,13 @@ st_choose_format(struct pipe_context *pipe, GLint internalFormat,
 
 
 static GLboolean
-is_stencil_format(GLenum format)
+is_depth_or_stencil_format(GLenum internalFormat)
 {
-   switch (format) {
+   switch (internalFormat) {
+   case GL_DEPTH_COMPONENT:
+   case GL_DEPTH_COMPONENT16:
+   case GL_DEPTH_COMPONENT24:
+   case GL_DEPTH_COMPONENT32:
    case GL_STENCIL_INDEX:
    case GL_STENCIL_INDEX1_EXT:
    case GL_STENCIL_INDEX4_EXT:
@@ -614,10 +618,10 @@ is_stencil_format(GLenum format)
  * Called by FBO code to choose a PIPE_FORMAT_ for drawing surfaces.
  */
 enum pipe_format
-st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat)
+st_choose_renderbuffer_format(struct pipe_context *pipe, GLenum internalFormat)
 {
    uint usage;
-   if (is_stencil_format(internalFormat))
+   if (is_depth_or_stencil_format(internalFormat))
       usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL;
    else
       usage = PIPE_TEXTURE_USAGE_RENDER_TARGET;
@@ -716,3 +720,23 @@ st_ChooseTextureFormat(GLcontext *ctx, GLint internalFormat,
 
    return translate_gallium_format_to_mesa_format(pFormat);
 }
+
+
+/**
+ * Test if a gallium format is equivalent to a GL format/type.
+ */
+GLboolean
+st_equal_formats(enum pipe_format pFormat, GLenum format, GLenum type)
+{
+   switch (pFormat) {
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      return format == GL_RGBA && type == GL_UNSIGNED_BYTE;
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      return format == GL_BGRA && type == GL_UNSIGNED_BYTE;
+   case PIPE_FORMAT_R5G6B5_UNORM:
+      return format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5;
+   /* XXX more combos... */
+   default:
+      return GL_FALSE;
+   }
+}
diff --git a/src/mesa/state_tracker/st_format.h b/src/mesa/state_tracker/st_format.h
index 3f5ac3201b..9d9e02fe9b 100644
--- a/src/mesa/state_tracker/st_format.h
+++ b/src/mesa/state_tracker/st_format.h
@@ -64,11 +64,11 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat);
 
 
 extern enum pipe_format
-st_choose_format(struct pipe_context *pipe, GLint internalFormat,
+st_choose_format(struct pipe_context *pipe, GLenum internalFormat,
                  enum pipe_texture_target target, unsigned tex_usage);
 
 extern enum pipe_format
-st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat);
+st_choose_renderbuffer_format(struct pipe_context *pipe, GLenum internalFormat);
 
 
 extern const struct gl_texture_format *
@@ -76,4 +76,8 @@ st_ChooseTextureFormat(GLcontext * ctx, GLint internalFormat,
                        GLenum format, GLenum type);
 
 
+extern GLboolean
+st_equal_formats(enum pipe_format pFormat, GLenum format, GLenum type);
+
+
 #endif /* ST_CB_TEXIMAGE_H */
diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c
index daaad65cca..7072cbe62c 100644
--- a/src/mesa/state_tracker/st_framebuffer.c
+++ b/src/mesa/state_tracker/st_framebuffer.c
@@ -58,25 +58,25 @@ st_create_framebuffer( const __GLcontextModes *visual,
 
       _mesa_initialize_framebuffer(&stfb->Base, visual);
 
-      {
-         /* fake frontbuffer */
-         /* XXX allocation should only happen in the unusual case
-            it's actually needed */
-         struct gl_renderbuffer *rb
-            = st_new_renderbuffer_fb(colorFormat, samples);
-         _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb);
-      }
-
       if (visual->doubleBufferMode) {
          struct gl_renderbuffer *rb
-            = st_new_renderbuffer_fb(colorFormat, samples);
+            = st_new_renderbuffer_fb(colorFormat, samples, FALSE);
          _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb);
       }
+      else {
+         /* Only allocate front buffer right now if we're single buffered.
+          * If double-buffered, allocate front buffer on demand later.
+          * See check_create_front_buffers().
+          */
+         struct gl_renderbuffer *rb
+            = st_new_renderbuffer_fb(colorFormat, samples, FALSE);
+         _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb);
+      }
 
       if (depthFormat == stencilFormat && depthFormat != PIPE_FORMAT_NONE) {
          /* combined depth/stencil buffer */
          struct gl_renderbuffer *depthStencilRb
-            = st_new_renderbuffer_fb(depthFormat, samples);
+            = st_new_renderbuffer_fb(depthFormat, samples, FALSE);
          /* note: bind RB to two attachment points */
          _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthStencilRb);
          _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, depthStencilRb);
@@ -87,34 +87,35 @@ st_create_framebuffer( const __GLcontextModes *visual,
          if (visual->depthBits == 32) {
             /* 32-bit depth buffer */
             struct gl_renderbuffer *depthRb
-               = st_new_renderbuffer_fb(depthFormat, samples);
+               = st_new_renderbuffer_fb(depthFormat, samples, FALSE);
             _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb);
          }
          else if (visual->depthBits == 24) {
             /* 24-bit depth buffer, ignore stencil bits */
             struct gl_renderbuffer *depthRb
-               = st_new_renderbuffer_fb(depthFormat, samples);
+               = st_new_renderbuffer_fb(depthFormat, samples, FALSE);
             _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb);
          }
          else if (visual->depthBits > 0) {
             /* 16-bit depth buffer */
             struct gl_renderbuffer *depthRb
-               = st_new_renderbuffer_fb(depthFormat, samples);
+               = st_new_renderbuffer_fb(depthFormat, samples, FALSE);
             _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb);
          }
 
          if (visual->stencilBits > 0) {
             /* 8-bit stencil */
             struct gl_renderbuffer *stencilRb
-               = st_new_renderbuffer_fb(stencilFormat, samples);
+               = st_new_renderbuffer_fb(stencilFormat, samples, FALSE);
             _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, stencilRb);
          }
       }
 
       if (visual->accumRedBits > 0) {
          /* 16-bit/channel accum */
+         /* TODO: query the pipe screen for accumulation buffer format support */
          struct gl_renderbuffer *accumRb
-            = st_new_renderbuffer_fb(DEFAULT_ACCUM_PIPE_FORMAT, 0); /* XXX accum isn't multisampled right? */
+            = st_new_renderbuffer_fb(PIPE_FORMAT_R16G16B16A16_SNORM, 0, TRUE);
          _mesa_add_renderbuffer(&stfb->Base, BUFFER_ACCUM, accumRb);
       }
 
@@ -293,6 +294,115 @@ st_notify_swapbuffers(struct st_framebuffer *stfb)
 }
 
 
+/**
+ * Swap the front/back color buffers.  Exchange the front/back pointers
+ * and update some derived state.
+ * No need to call st_notify_swapbuffers() first.
+ *
+ * For a single-buffered framebuffer, no swap occurs, but we still return
+ * the pointer(s) to the front color buffer(s).
+ *
+ * \param front_left  returns pointer to front-left renderbuffer after swap
+ * \param front_right  returns pointer to front-right renderbuffer after swap
+ */
+void
+st_swapbuffers(struct st_framebuffer *stfb,
+               struct pipe_surface **front_left,
+               struct pipe_surface **front_right)
+{
+   struct gl_framebuffer *fb = &stfb->Base;
+
+   GET_CURRENT_CONTEXT(ctx);
+
+   if (ctx && ctx->DrawBuffer == &stfb->Base) {
+      st_flush( ctx->st, 
+		PIPE_FLUSH_RENDER_CACHE | 
+		PIPE_FLUSH_SWAPBUFFERS |
+		PIPE_FLUSH_FRAME,
+                NULL );
+   }
+
+   if (!fb->Visual.doubleBufferMode) {
+      /* single buffer mode - return pointers to front surfaces */
+      if (front_left) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+         *front_left = strb->surface;
+      }
+      if (front_right) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer);
+         *front_right = strb ? strb->surface : NULL;
+      }
+      return;
+   }
+
+   /* swap left buffers */
+   if (fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer &&
+       fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer) {
+      struct gl_renderbuffer *rbTemp;
+      rbTemp = fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer;
+      fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer =
+         fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer;
+      fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer = rbTemp;
+      if (front_left) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+         *front_left = strb->surface;
+      }
+      /* mark back buffer contents as undefined */
+      {
+         struct st_renderbuffer *back =
+            st_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer);
+         back->defined = GL_FALSE;
+      }
+   }
+   else {
+      /* no front buffer, display the back buffer */
+      if (front_left) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer);
+         *front_left = strb->surface;
+      }
+   }
+
+   /* swap right buffers (for stereo) */
+   if (fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer &&
+       fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer) {
+      struct gl_renderbuffer *rbTemp;
+      rbTemp = fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer;
+      fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer =
+         fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer;
+      fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer = rbTemp;
+      if (front_right) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_FRONT_RIGHT].Renderbuffer);
+         *front_right = strb->surface;
+      }
+      /* mark back buffer contents as undefined */
+      {
+         struct st_renderbuffer *back =
+            st_renderbuffer(fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer);
+         back->defined = GL_FALSE;
+      }
+   }
+   else {
+      /* no front right buffer, display back right buffer (if exists) */
+      if (front_right) {
+         struct st_renderbuffer *strb =
+            st_renderbuffer(fb->Attachment[BUFFER_BACK_RIGHT].Renderbuffer);
+         *front_right = strb ? strb->surface : NULL;
+      }
+   }
+
+   /* Update the _ColorDrawBuffers[] array and _ColorReadBuffer pointer */
+   _mesa_update_framebuffer(ctx);
+
+   /* Make sure we draw into the new back surface */
+   st_invalidate_state(ctx, _NEW_BUFFERS);
+}
+
+
 void *st_framebuffer_private( struct st_framebuffer *stfb )
 {
    return stfb->Private;
diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c
index 9cc2176d5e..dc6d77825f 100644
--- a/src/mesa/state_tracker/st_gen_mipmap.c
+++ b/src/mesa/state_tracker/st_gen_mipmap.c
@@ -47,6 +47,7 @@
 #include "st_program.h"
 #include "st_texture.h"
 #include "st_cb_texture.h"
+#include "st_inlines.h"
 
 
 /**
@@ -86,7 +87,7 @@ st_render_mipmap(struct st_context *st,
    assert(target != GL_TEXTURE_3D); /* not done yet */
 
    /* check if we can render in the texture's format */
-   if (!screen->is_format_supported(screen, pt->format, target,
+   if (!screen->is_format_supported(screen, pt->format, pt->target,
                                     PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)) {
       return FALSE;
    }
@@ -122,27 +123,34 @@ fallback_generate_mipmap(GLcontext *ctx, GLenum target,
       struct pipe_transfer *srcTrans, *dstTrans;
       const ubyte *srcData;
       ubyte *dstData;
+      int srcStride, dstStride;
 
-      srcTrans = screen->get_tex_transfer(screen, pt, face, srcLevel, zslice,
-                                          PIPE_TRANSFER_READ, 0, 0,
-                                          pt->width[srcLevel],
-                                          pt->height[srcLevel]);
-      dstTrans = screen->get_tex_transfer(screen, pt, face, dstLevel, zslice,
-                                          PIPE_TRANSFER_WRITE, 0, 0,
-                                          pt->width[dstLevel],
-                                          pt->height[dstLevel]);
+      srcTrans = st_cond_flush_get_tex_transfer(st_context(ctx), pt, face,
+						srcLevel, zslice,
+						PIPE_TRANSFER_READ, 0, 0,
+						pt->width[srcLevel],
+						pt->height[srcLevel]);
+
+      dstTrans = st_cond_flush_get_tex_transfer(st_context(ctx), pt, face,
+						dstLevel, zslice,
+						PIPE_TRANSFER_WRITE, 0, 0,
+						pt->width[dstLevel],
+						pt->height[dstLevel]);
 
       srcData = (ubyte *) screen->transfer_map(screen, srcTrans);
       dstData = (ubyte *) screen->transfer_map(screen, dstTrans);
 
+      srcStride = srcTrans->stride / srcTrans->block.size;
+      dstStride = dstTrans->stride / dstTrans->block.size;
+
       _mesa_generate_mipmap_level(target, datatype, comps,
                    0 /*border*/,
                    pt->width[srcLevel], pt->height[srcLevel], pt->depth[srcLevel],
                    srcData,
-                   srcTrans->stride, /* stride in bytes */
+                   srcStride, /* stride in texels */
                    pt->width[dstLevel], pt->height[dstLevel], pt->depth[dstLevel],
                    dstData,
-                   dstTrans->stride); /* stride in bytes */
+                   dstStride); /* stride in texels */
 
       screen->transfer_unmap(screen, srcTrans);
       screen->transfer_unmap(screen, dstTrans);
diff --git a/src/mesa/state_tracker/st_inlines.h b/src/mesa/state_tracker/st_inlines.h
new file mode 100644
index 0000000000..a41cfeb96f
--- /dev/null
+++ b/src/mesa/state_tracker/st_inlines.h
@@ -0,0 +1,154 @@
+/**************************************************************************
+ * 
+ * Copyright 2009 VMware, Inc.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+/**
+ * Functions for checking if buffers/textures are referenced when we need
+ * to read/write from/to them.  Flush when needed.
+ */
+
+#ifndef ST_INLINES_H
+#define ST_INLINES_H
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_state.h"
+
+#include "st_context.h"
+#include "st_texture.h"
+#include "st_public.h"
+
+static INLINE struct pipe_transfer *
+st_cond_flush_get_tex_transfer(struct st_context *st,
+			       struct pipe_texture *pt,
+			       unsigned int face,
+			       unsigned int level,
+			       unsigned int zslice,
+			       enum pipe_transfer_usage usage,
+			       unsigned int x, unsigned int y,
+			       unsigned int w, unsigned int h)
+{
+   struct pipe_screen *screen = st->pipe->screen;
+
+   st_teximage_flush_before_map(st, pt, face, level, usage);
+   return screen->get_tex_transfer(screen, pt, face, level, zslice, usage,
+				   x, y, w, h);
+}
+
+static INLINE struct pipe_transfer *
+st_no_flush_get_tex_transfer(struct st_context *st,
+			     struct pipe_texture *pt,
+			     unsigned int face,
+			     unsigned int level,
+			     unsigned int zslice,
+			     enum pipe_transfer_usage usage,
+			     unsigned int x, unsigned int y,
+			     unsigned int w, unsigned int h)
+{
+   struct pipe_screen *screen = st->pipe->screen;
+
+   return screen->get_tex_transfer(screen, pt, face, level,
+				   zslice, usage, x, y, w, h);
+}
+
+static INLINE void *
+st_cond_flush_pipe_buffer_map(struct st_context *st,
+			      struct pipe_buffer *buf,
+			      unsigned int map_flags)
+{
+   struct pipe_context *pipe = st->pipe;
+   unsigned int referenced = pipe->is_buffer_referenced(pipe, buf);
+
+   if (referenced && ((referenced & PIPE_REFERENCED_FOR_WRITE) ||
+		      (map_flags & PIPE_BUFFER_USAGE_CPU_WRITE)))
+      st_flush(st, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   return pipe_buffer_map(pipe->screen, buf, map_flags);
+}
+
+static INLINE void *
+st_no_flush_pipe_buffer_map(struct st_context *st,
+			    struct pipe_buffer *buf,
+			    unsigned int map_flags)
+{
+   return pipe_buffer_map(st->pipe->screen, buf, map_flags);
+}
+
+
+static INLINE void
+st_cond_flush_pipe_buffer_write(struct st_context *st,
+				struct pipe_buffer *buf,
+				unsigned int offset,
+				unsigned int size,
+				const void * data)
+{
+   struct pipe_context *pipe = st->pipe;
+
+   if (pipe->is_buffer_referenced(pipe, buf))
+      st_flush(st, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   pipe_buffer_write(pipe->screen, buf, offset, size, data);
+}
+
+static INLINE void
+st_no_flush_pipe_buffer_write(struct st_context *st,
+			      struct pipe_buffer *buf,
+			      unsigned int offset,
+			      unsigned int size,
+			      const void * data)
+{
+   pipe_buffer_write(st->pipe->screen, buf, offset, size, data);
+}
+
+static INLINE void
+st_cond_flush_pipe_buffer_read(struct st_context *st,
+			       struct pipe_buffer *buf,
+			       unsigned int offset,
+			       unsigned int size,
+			       void * data)
+{
+   struct pipe_context *pipe = st->pipe;
+
+   if (pipe->is_buffer_referenced(pipe, buf) & PIPE_REFERENCED_FOR_WRITE)
+      st_flush(st, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   pipe_buffer_read(pipe->screen, buf, offset, size, data);
+}
+
+static INLINE void
+st_no_flush_pipe_buffer_read(struct st_context *st,
+			     struct pipe_buffer *buf,
+			     unsigned int offset,
+			     unsigned int size,
+			     void * data)
+{
+   pipe_buffer_read(st->pipe->screen, buf, offset, size, data);
+}
+
+#endif
+
diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c
index ffa607dd87..43c9afccc3 100644
--- a/src/mesa/state_tracker/st_mesa_to_tgsi.c
+++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c
@@ -275,8 +275,8 @@ compile_instruction(
       /* swizzle (ext swizzle also depends on negation) */
       {
          GLuint swz[4];
-         GLboolean extended = (inst->SrcReg[i].NegateBase != NEGATE_NONE &&
-                               inst->SrcReg[i].NegateBase != NEGATE_XYZW);
+         GLboolean extended = (inst->SrcReg[i].Negate != NEGATE_NONE &&
+                               inst->SrcReg[i].Negate != NEGATE_XYZW);
          for( j = 0; j < 4; j++ ) {
             swz[j] = GET_SWZ( inst->SrcReg[i].Swizzle, j );
             if (swz[j] > SWIZZLE_W)
@@ -296,20 +296,20 @@ compile_instruction(
          }
       }
 
-      if( inst->SrcReg[i].NegateBase == NEGATE_XYZW ) {
+      if( inst->SrcReg[i].Negate == NEGATE_XYZW ) {
          fullsrc->SrcRegister.Negate = 1;
       }
-      else if( inst->SrcReg[i].NegateBase != NEGATE_NONE ) {
-         if( inst->SrcReg[i].NegateBase & NEGATE_X ) {
+      else if( inst->SrcReg[i].Negate != NEGATE_NONE ) {
+         if( inst->SrcReg[i].Negate & NEGATE_X ) {
             fullsrc->SrcRegisterExtSwz.NegateX = 1;
          }
-         if( inst->SrcReg[i].NegateBase & NEGATE_Y ) {
+         if( inst->SrcReg[i].Negate & NEGATE_Y ) {
             fullsrc->SrcRegisterExtSwz.NegateY = 1;
          }
-         if( inst->SrcReg[i].NegateBase & NEGATE_Z ) {
+         if( inst->SrcReg[i].Negate & NEGATE_Z ) {
             fullsrc->SrcRegisterExtSwz.NegateZ = 1;
          }
-         if( inst->SrcReg[i].NegateBase & NEGATE_W ) {
+         if( inst->SrcReg[i].Negate & NEGATE_W ) {
             fullsrc->SrcRegisterExtSwz.NegateW = 1;
          }
       }
@@ -318,10 +318,6 @@ compile_instruction(
          fullsrc->SrcRegisterExtMod.Absolute = 1;
       }
 
-      if( inst->SrcReg[i].NegateAbs ) {
-         fullsrc->SrcRegisterExtMod.Negate = 1;
-      }
-
       if( inst->SrcReg[i].RelAddr ) {
          fullsrc->SrcRegister.Indirect = 1;
 
diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c
index 6348e83d8a..72ca852458 100644
--- a/src/mesa/state_tracker/st_program.c
+++ b/src/mesa/state_tracker/st_program.c
@@ -49,26 +49,12 @@
 #include "cso_cache/cso_context.h"
 
 
-#define ST_MAX_SHADER_TOKENS 4096
+#define ST_MAX_SHADER_TOKENS (8 * 1024)
 
 
 #define TGSI_DEBUG 0
 
 
-/** XXX we should use the version of this from u_memory.h but including
- * that header causes symbol collisions.
- */
-static INLINE void *
-mem_dup(const void *src, uint size)
-{
-   void *dup = _mesa_malloc(size);
-   if (dup)
-      _mesa_memcpy(dup, src, size);
-   return dup;
-}
-
-
-
 /**
  * Translate a Mesa vertex shader into a TGSI shader.
  * \param outputMapping  to map vertex program output registers (VERT_RESULT_x)
@@ -84,7 +70,7 @@ st_translate_vertex_program(struct st_context *st,
                             const ubyte *outputSemanticIndex)
 {
    struct pipe_context *pipe = st->pipe;
-   struct tgsi_token tokens[ST_MAX_SHADER_TOKENS];
+   struct tgsi_token *tokens;
    GLuint defaultOutputMapping[VERT_RESULT_MAX];
    struct pipe_shader_state vs;
    GLuint attr, i;
@@ -102,6 +88,13 @@ st_translate_vertex_program(struct st_context *st,
    GLbitfield input_flags[MAX_PROGRAM_INPUTS];
    GLbitfield output_flags[MAX_PROGRAM_OUTPUTS];
 
+   tokens =  (struct tgsi_token *)MALLOC(ST_MAX_SHADER_TOKENS * sizeof *tokens);
+   if(!tokens) {
+      /* FIXME: propagate error to the caller */
+      assert(0);
+      return;
+   }
+
    memset(&vs, 0, sizeof(vs));
    memset(input_flags, 0, sizeof(input_flags));
    memset(output_flags, 0, sizeof(output_flags));
@@ -169,6 +162,14 @@ st_translate_vertex_program(struct st_context *st,
          case VERT_ATTRIB_GENERIC5:
          case VERT_ATTRIB_GENERIC6:
          case VERT_ATTRIB_GENERIC7:
+         case VERT_ATTRIB_GENERIC8:
+         case VERT_ATTRIB_GENERIC9:
+         case VERT_ATTRIB_GENERIC10:
+         case VERT_ATTRIB_GENERIC11:
+         case VERT_ATTRIB_GENERIC12:
+         case VERT_ATTRIB_GENERIC13:
+         case VERT_ATTRIB_GENERIC14:
+         case VERT_ATTRIB_GENERIC15:
             assert(attr < VERT_ATTRIB_MAX);
             vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC;
             vs_input_semantic_index[slot] = num_generic++;
@@ -289,9 +290,6 @@ st_translate_vertex_program(struct st_context *st,
       }
    }
 
-   assert(vs_output_semantic_name[0] == TGSI_SEMANTIC_POSITION);
-
-
    if (outputMapping) {
       /* find max output slot referenced to compute vs_num_outputs */
       GLuint maxSlot = 0;
@@ -339,7 +337,9 @@ st_translate_vertex_program(struct st_context *st,
    assert(num_tokens < ST_MAX_SHADER_TOKENS);
 
    vs.tokens = (struct tgsi_token *)
-      mem_dup(tokens, num_tokens * sizeof(tokens[0]));
+      _mesa_realloc(tokens,
+                    ST_MAX_SHADER_TOKENS * sizeof *tokens,
+                    num_tokens * sizeof *tokens);
 
    stvp->num_inputs = vs_num_inputs;
    stvp->state = vs; /* struct copy */
@@ -367,7 +367,7 @@ st_translate_fragment_program(struct st_context *st,
                               const GLuint inputMapping[])
 {
    struct pipe_context *pipe = st->pipe;
-   struct tgsi_token tokens[ST_MAX_SHADER_TOKENS];
+   struct tgsi_token *tokens;
    GLuint outputMapping[FRAG_RESULT_MAX];
    GLuint defaultInputMapping[FRAG_ATTRIB_MAX];
    struct pipe_shader_state fs;
@@ -387,6 +387,13 @@ st_translate_fragment_program(struct st_context *st,
    GLbitfield input_flags[MAX_PROGRAM_INPUTS];
    GLbitfield output_flags[MAX_PROGRAM_OUTPUTS];
 
+   tokens =  (struct tgsi_token *)MALLOC(ST_MAX_SHADER_TOKENS * sizeof *tokens);
+   if(!tokens) {
+      /* FIXME: propagate error to the caller */
+      assert(0);
+      return;
+   }
+
    memset(&fs, 0, sizeof(fs));
    memset(input_flags, 0, sizeof(input_flags));
    memset(output_flags, 0, sizeof(output_flags));
@@ -484,14 +491,14 @@ st_translate_fragment_program(struct st_context *st,
                /* handled above */
                assert(0);
                break;
-            case FRAG_RESULT_COLOR:
+            default:
+               assert(attr == FRAG_RESULT_COLOR ||
+                      (FRAG_RESULT_DATA0 <= attr && attr < FRAG_RESULT_MAX));
                fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_COLOR;
                fs_output_semantic_index[fs_num_outputs] = numColors;
                outputMapping[attr] = fs_num_outputs;
                numColors++;
                break;
-            default:
-               assert(0);
             }
 
             output_flags[fs_num_outputs] = stfp->Base.Base.OutputFlags[attr];
@@ -528,7 +535,9 @@ st_translate_fragment_program(struct st_context *st,
    assert(num_tokens < ST_MAX_SHADER_TOKENS);
 
    fs.tokens = (struct tgsi_token *)
-      mem_dup(tokens, num_tokens * sizeof(tokens[0]));
+      _mesa_realloc(tokens,
+                    ST_MAX_SHADER_TOKENS * sizeof *tokens,
+                    num_tokens * sizeof *tokens);
 
    stfp->state = fs; /* struct copy */
    stfp->driver_shader = pipe->create_fs_state(pipe, &fs);
diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h
index 030314372f..04d3a3d7c2 100644
--- a/src/mesa/state_tracker/st_public.h
+++ b/src/mesa/state_tracker/st_public.h
@@ -91,9 +91,11 @@ void *st_framebuffer_private( struct st_framebuffer *stfb );
 
 void st_unreference_framebuffer( struct st_framebuffer *stfb );
 
-void st_make_current(struct st_context *st,
-                     struct st_framebuffer *draw,
-                     struct st_framebuffer *read);
+GLboolean st_make_current(struct st_context *st,
+                          struct st_framebuffer *draw,
+                          struct st_framebuffer *read);
+
+struct st_context *st_get_current(void);
 
 void st_flush( struct st_context *st, uint pipeFlushFlags,
                struct pipe_fence_handle **fence );
@@ -101,6 +103,10 @@ void st_finish( struct st_context *st );
 
 void st_notify_swapbuffers(struct st_framebuffer *stfb);
 
+void st_swapbuffers(struct st_framebuffer *stfb,
+                    struct pipe_surface **front_left,
+                    struct pipe_surface **front_right);
+
 int st_set_teximage(struct pipe_texture *pt, int target);
 
 /** Redirect rendering into stfb's surface to a texture image */
diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c
index 3f90ad502c..10faa633ea 100644
--- a/src/mesa/state_tracker/st_texture.c
+++ b/src/mesa/state_tracker/st_texture.c
@@ -30,6 +30,7 @@
 #include "st_public.h"
 #include "st_texture.h"
 #include "st_cb_fbo.h"
+#include "st_inlines.h"
 #include "main/enums.h"
 #include "main/teximage.h"
 #include "main/texstore.h"
@@ -78,7 +79,6 @@ st_texture_create(struct st_context *st,
 		  GLuint width0,
 		  GLuint height0,
 		  GLuint depth0,
-		  GLuint compress_byte,
                   GLuint usage )
 {
    struct pipe_texture pt, *newtex;
@@ -101,7 +101,6 @@ st_texture_create(struct st_context *st,
    pt.width[0] = width0;
    pt.height[0] = height0;
    pt.depth[0] = depth0;
-   pt.compressed = compress_byte ? 1 : 0;
    pf_get_block(format, &pt.block);
    pt.tex_usage = usage;
 
@@ -128,8 +127,7 @@ st_texture_match_image(const struct pipe_texture *pt,
 
    /* Check if this image's format matches the established texture's format.
     */
-   if (st_mesa_format_to_pipe_format(image->TexFormat->MesaFormat) != pt->format ||
-       image->IsCompressed != pt->compressed)
+   if (st_mesa_format_to_pipe_format(image->TexFormat->MesaFormat) != pt->format)
       return GL_FALSE;
 
    /* Test if this image's size matches what's expected in the
@@ -191,13 +189,15 @@ st_texture_image_map(struct st_context *st, struct st_texture_image *stImage,
 		     GLuint zoffset, enum pipe_transfer_usage usage,
                      GLuint x, GLuint y, GLuint w, GLuint h)
 {
-   struct pipe_screen *screen = st->pipe->screen;
+   struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
    struct pipe_texture *pt = stImage->pt;
+
    DBG("%s \n", __FUNCTION__);
 
-   stImage->transfer = screen->get_tex_transfer(screen, pt, stImage->face,
-                                                stImage->level, zoffset, 
-                                                usage, x, y, w, h);
+   stImage->transfer = st_no_flush_get_tex_transfer(st, pt, stImage->face,
+						    stImage->level, zoffset,
+						    usage, x, y, w, h);
 
    if (stImage->transfer)
       return screen->transfer_map(screen, stImage->transfer);
@@ -254,13 +254,14 @@ st_surface_data(struct pipe_context *pipe,
 /* Upload data for a particular image.
  */
 void
-st_texture_image_data(struct pipe_context *pipe,
+st_texture_image_data(struct st_context *st,
                       struct pipe_texture *dst,
                       GLuint face,
                       GLuint level,
                       void *src,
                       GLuint src_row_stride, GLuint src_image_stride)
 {
+   struct pipe_context *pipe = st->pipe;
    struct pipe_screen *screen = pipe->screen;
    GLuint depth = dst->depth[level];
    GLuint i;
@@ -268,11 +269,12 @@ st_texture_image_data(struct pipe_context *pipe,
    struct pipe_transfer *dst_transfer;
 
    DBG("%s\n", __FUNCTION__);
+
    for (i = 0; i < depth; i++) {
-      dst_transfer = screen->get_tex_transfer(screen, dst, face, level, i,
-                                              PIPE_TRANSFER_WRITE, 0, 0,
-                                              dst->width[level],
-                                              dst->height[level]);
+      dst_transfer = st_no_flush_get_tex_transfer(st, dst, face, level, i,
+						  PIPE_TRANSFER_WRITE, 0, 0,
+						  dst->width[level],
+						  dst->height[level]);
 
       st_surface_data(pipe, dst_transfer,
 		      0, 0,                             /* dstx, dsty */
@@ -484,3 +486,20 @@ st_release_teximage(struct st_framebuffer *stfb, uint surfIndex,
 
    return 1;
 }
+
+void
+st_teximage_flush_before_map(struct st_context *st,
+			     struct pipe_texture *pt,
+			     unsigned int face,
+			     unsigned int level,
+			     enum pipe_transfer_usage usage)
+{
+   struct pipe_context *pipe = st->pipe;
+   unsigned referenced =
+      pipe->is_texture_referenced(pipe, pt, face, level);
+
+   if (referenced && ((referenced & PIPE_REFERENCED_FOR_WRITE) ||
+		      usage == PIPE_TRANSFER_WRITE ||
+		      usage == PIPE_TRANSFER_READ_WRITE))
+      st_flush(st, PIPE_FLUSH_RENDER_CACHE, NULL);
+}
diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h
index 28c2f580f6..b9d447cb56 100644
--- a/src/mesa/state_tracker/st_texture.h
+++ b/src/mesa/state_tracker/st_texture.h
@@ -108,7 +108,6 @@ st_texture_create(struct st_context *st,
                   GLuint width0,
                   GLuint height0,
                   GLuint depth0,
-                  GLuint compress_byte,
                   GLuint tex_usage );
 
 
@@ -157,7 +156,7 @@ st_texture_texel_offset(const struct pipe_texture * pt,
 /* Upload an image into a texture
  */
 extern void
-st_texture_image_data(struct pipe_context *pipe,
+st_texture_image_data(struct st_context *st,
                       struct pipe_texture *dst,
                       GLuint face, GLuint level, void *src,
                       GLuint src_row_pitch, GLuint src_image_pitch);
@@ -171,5 +170,10 @@ st_texture_image_copy(struct pipe_context *pipe,
                       struct pipe_texture *src,
                       GLuint face);
 
-
+extern void
+st_teximage_flush_before_map(struct st_context *st,
+			     struct pipe_texture *pt,
+			     unsigned int face,
+			     unsigned int level,
+			     enum pipe_transfer_usage usage);
 #endif