Merge commit 'eee/mesa-es' into android

51 files changed, 4411 insertions, 2628 deletions

diff --git a/src/mesa/drivers/dri/intel/intel_batchbuffer.c b/src/mesa/drivers/dri/intel/intel_batchbuffer.c
index 9d9937289a..6aa36d10b1 100644
--- a/src/mesa/drivers/dri/intel/intel_batchbuffer.c
+++ b/src/mesa/drivers/dri/intel/intel_batchbuffer.c
@@ -195,7 +195,16 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file,
 {
    struct intel_context *intel = batch->intel;
    GLuint used = batch->ptr - batch->map;
-   GLboolean was_locked = intel->locked;
+
+   if (intel->first_post_swapbuffers_batch == NULL) {
+      intel->first_post_swapbuffers_batch = intel->batch->buf;
+      drm_intel_bo_reference(intel->first_post_swapbuffers_batch);
+   }
+
+   if (intel->first_post_swapbuffers_batch == NULL) {
+      intel->first_post_swapbuffers_batch = intel->batch->buf;
+      drm_intel_bo_reference(intel->first_post_swapbuffers_batch);
+   }
 
    if (used == 0) {
       batch->cliprect_mode = IGNORE_CLIPRECTS;
@@ -207,7 +216,7 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file,
 	      used);
 
    /* Emit a flush if the bufmgr doesn't do it for us. */
-   if (!intel->ttm) {
+   if (intel->always_flush_cache || !intel->ttm) {
       *(GLuint *) (batch->ptr) = intel->vtbl.flush_cmd();
       batch->ptr += 4;
       used = batch->ptr - batch->map;
@@ -243,13 +252,9 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file,
    /* TODO: Just pass the relocation list and dma buffer up to the
     * kernel.
     */
-   if (!was_locked)
-      LOCK_HARDWARE(intel);
-
+   LOCK_HARDWARE(intel);
    do_flush_locked(batch, used, GL_FALSE);
-
-   if (!was_locked)
-      UNLOCK_HARDWARE(intel);
+   UNLOCK_HARDWARE(intel);
 
    if (INTEL_DEBUG & DEBUG_SYNC) {
       fprintf(stderr, "waiting for idle\n");
diff --git a/src/mesa/drivers/dri/intel/intel_blit.c b/src/mesa/drivers/dri/intel/intel_blit.c
index 208f90c0ab..43141c509c 100644
--- a/src/mesa/drivers/dri/intel/intel_blit.c
+++ b/src/mesa/drivers/dri/intel/intel_blit.c
@@ -26,12 +26,11 @@
  **************************************************************************/
 
 
-#include <stdio.h>
-#include <errno.h>
-
 #include "main/mtypes.h"
 #include "main/context.h"
 #include "main/enums.h"
+#include "main/texformat.h"
+#include "main/colormac.h"
 
 #include "intel_blit.h"
 #include "intel_buffers.h"
@@ -98,14 +97,16 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv,
       ASSERT(src->cpp == dst->cpp);
 
       if (cpp == 2) {
-	 BR13 = (0xCC << 16) | (1 << 24);
+	 BR13 = (0xCC << 16) | BR13_565;
 	 CMD = XY_SRC_COPY_BLT_CMD;
       }
       else {
-	 BR13 = (0xCC << 16) | (1 << 24) | (1 << 25);
+	 BR13 = (0xCC << 16) | BR13_8888;
 	 CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
       }
 
+      assert(src->tiling != I915_TILING_Y);
+      assert(dst->tiling != I915_TILING_Y);
 #ifndef I915
       if (src->tiling != I915_TILING_NONE) {
 	 CMD |= XY_SRC_TILED;
@@ -173,64 +174,6 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv,
    UNLOCK_HARDWARE(intel);
 }
 
-
-
-
-void
-intelEmitFillBlit(struct intel_context *intel,
-		  GLuint cpp,
-		  GLshort dst_pitch,
-		  dri_bo *dst_buffer,
-		  GLuint dst_offset,
-		  uint32_t dst_tiling,
-		  GLshort x, GLshort y,
-		  GLshort w, GLshort h,
-		  GLuint color)
-{
-   GLuint BR13, CMD;
-   BATCH_LOCALS;
-
-   dst_pitch *= cpp;
-
-   switch (cpp) {
-   case 1:
-   case 2:
-   case 3:
-      BR13 = (0xF0 << 16) | (1 << 24);
-      CMD = XY_COLOR_BLT_CMD;
-      break;
-   case 4:
-      BR13 = (0xF0 << 16) | (1 << 24) | (1 << 25);
-      CMD = XY_COLOR_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
-      break;
-   default:
-      return;
-   }
-#ifndef I915
-   if (dst_tiling != I915_TILING_NONE) {
-      CMD |= XY_DST_TILED;
-      dst_pitch /= 4;
-   }
-#endif
-
-   DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
-       __FUNCTION__, dst_buffer, dst_pitch, dst_offset, x, y, w, h);
-
-   assert(w > 0);
-   assert(h > 0);
-
-   BEGIN_BATCH(6, NO_LOOP_CLIPRECTS);
-   OUT_BATCH(CMD);
-   OUT_BATCH(BR13 | dst_pitch);
-   OUT_BATCH((y << 16) | x);
-   OUT_BATCH(((y + h) << 16) | (x + w));
-   OUT_RELOC(dst_buffer,
-	     I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
-	     dst_offset);
-   OUT_BATCH(color);
-   ADVANCE_BATCH();
-}
-
 static GLuint translate_raster_op(GLenum logicop)
 {
    switch(logicop) {
@@ -257,7 +200,7 @@ static GLuint translate_raster_op(GLenum logicop)
 
 /* Copy BitBlt
  */
-void
+GLboolean
 intelEmitCopyBlit(struct intel_context *intel,
 		  GLuint cpp,
 		  GLshort src_pitch,
@@ -279,6 +222,19 @@ intelEmitCopyBlit(struct intel_context *intel,
    dri_bo *aper_array[3];
    BATCH_LOCALS;
 
+   if (dst_tiling != I915_TILING_NONE) {
+      if (dst_offset & 4095)
+	 return GL_FALSE;
+      if (dst_tiling == I915_TILING_Y)
+	 return GL_FALSE;
+   }
+   if (src_tiling != I915_TILING_NONE) {
+      if (src_offset & 4095)
+	 return GL_FALSE;
+      if (src_tiling == I915_TILING_Y)
+	 return GL_FALSE;
+   }
+
    /* do space/cliprects check before going any further */
    do {
        aper_array[0] = intel->batch->buf;
@@ -293,12 +249,7 @@ intelEmitCopyBlit(struct intel_context *intel,
    } while (pass < 2);
 
    if (pass >= 2) {
-       GLboolean locked = GL_FALSE;       
-       if (!intel->locked) {
-           LOCK_HARDWARE(intel);
-           locked = GL_TRUE;
-       }
-
+       LOCK_HARDWARE(intel);
        dri_bo_map(dst_buffer, GL_TRUE);
        dri_bo_map(src_buffer, GL_FALSE);
        _mesa_copy_rect((GLubyte *)dst_buffer->virtual + dst_offset,
@@ -312,11 +263,9 @@ intelEmitCopyBlit(struct intel_context *intel,
        
        dri_bo_unmap(src_buffer);
        dri_bo_unmap(dst_buffer);
-       
-       if (locked)
-           UNLOCK_HARDWARE(intel);
+       UNLOCK_HARDWARE(intel);
 
-       return;
+       return GL_TRUE;
    }
 
    intel_batchbuffer_require_space(intel->batch, 8 * 4, NO_LOOP_CLIPRECTS);
@@ -332,17 +281,18 @@ intelEmitCopyBlit(struct intel_context *intel,
 
    switch (cpp) {
    case 1:
+      CMD = XY_SRC_COPY_BLT_CMD;
+      break;
    case 2:
-   case 3:
-      BR13 |= (1 << 24);
+      BR13 |= BR13_565;
       CMD = XY_SRC_COPY_BLT_CMD;
       break;
    case 4:
-      BR13 |= (1 << 24) | (1 << 25);
+      BR13 |= BR13_8888;
       CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
       break;
    default:
-      return;
+      return GL_FALSE;
    }
 
 #ifndef I915
@@ -357,7 +307,7 @@ intelEmitCopyBlit(struct intel_context *intel,
 #endif
 
    if (dst_y2 <= dst_y || dst_x2 <= dst_x) {
-      return;
+      return GL_TRUE;
    }
 
    assert(dst_x < dst_x2);
@@ -379,6 +329,8 @@ intelEmitCopyBlit(struct intel_context *intel,
    ADVANCE_BATCH();
 
    intel_batchbuffer_emit_mi_flush(intel->batch);
+
+   return GL_TRUE;
 }
 
 
@@ -481,10 +433,9 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask)
             const GLbitfield bufBit = 1 << buf;
             if ((clearMask & bufBit) && !(bufBit & skipBuffers)) {
                /* OK, clear this renderbuffer */
-               struct intel_region *irb_region =
-		  intel_get_rb_region(fb, buf);
+	       struct intel_renderbuffer *irb = intel_get_renderbuffer(fb, buf);
                dri_bo *write_buffer =
-                  intel_region_buffer(intel, irb_region,
+                  intel_region_buffer(intel, irb->region,
                                       all ? INTEL_WRITE_FULL :
                                       INTEL_WRITE_PART);
 
@@ -492,15 +443,13 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask)
                GLint pitch, cpp;
                GLuint BR13, CMD;
 
-               ASSERT(irb_region);
-
-               pitch = irb_region->pitch;
-               cpp = irb_region->cpp;
+               pitch = irb->region->pitch;
+               cpp = irb->region->cpp;
 
                DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
                    __FUNCTION__,
-                   irb_region->buffer, (pitch * cpp),
-                   irb_region->draw_offset,
+                   irb->region->buffer, (pitch * cpp),
+                   irb->region->draw_offset,
                    b.x1, b.y1, b.x2 - b.x1, b.y2 - b.y1);
 
 	       BR13 = 0xf0 << 16;
@@ -508,7 +457,7 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask)
 
                /* Setup the blit command */
                if (cpp == 4) {
-                  BR13 |= (1 << 24) | (1 << 25);
+                  BR13 |= BR13_8888;
                   if (buf == BUFFER_DEPTH || buf == BUFFER_STENCIL) {
                      if (clearMask & BUFFER_BIT_DEPTH)
                         CMD |= XY_BLT_WRITE_RGB;
@@ -521,12 +470,14 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask)
                   }
                }
                else {
-                  ASSERT(cpp == 2 || cpp == 0);
-                  BR13 |= (1 << 24);
+                  ASSERT(cpp == 2);
+                  BR13 |= BR13_565;
                }
 
+	       assert(irb->region->tiling != I915_TILING_Y);
+
 #ifndef I915
-	       if (irb_region->tiling != I915_TILING_NONE) {
+	       if (irb->region->tiling != I915_TILING_NONE) {
 		  CMD |= XY_DST_TILED;
 		  pitch /= 4;
 	       }
@@ -537,9 +488,36 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask)
                   clearVal = clear_depth;
                }
                else {
-                  clearVal = (cpp == 4)
-                     ? intel->ClearColor8888 : intel->ClearColor565;
-               }
+		  uint8_t clear[4];
+		  GLclampf *color = ctx->Color.ClearColor;
+
+		  CLAMPED_FLOAT_TO_UBYTE(clear[0], color[0]);
+		  CLAMPED_FLOAT_TO_UBYTE(clear[1], color[1]);
+		  CLAMPED_FLOAT_TO_UBYTE(clear[2], color[2]);
+		  CLAMPED_FLOAT_TO_UBYTE(clear[3], color[3]);
+
+		  switch (irb->texformat->MesaFormat) {
+		  case MESA_FORMAT_ARGB8888:
+		     clearVal = intel->ClearColor8888;
+		     break;
+		  case MESA_FORMAT_RGB565:
+		     clearVal = intel->ClearColor565;
+		     break;
+		  case MESA_FORMAT_ARGB4444:
+		     clearVal = PACK_COLOR_4444(clear[3], clear[0],
+						clear[1], clear[2]);
+		     break;
+		  case MESA_FORMAT_ARGB1555:
+		     clearVal = PACK_COLOR_1555(clear[3], clear[0],
+						clear[1], clear[2]);
+		     break;
+		  default:
+		     _mesa_problem(ctx, "Unexpected renderbuffer format: %d\n",
+				   irb->texformat->MesaFormat);
+		     clearVal = 0;
+		  }
+	       }
+
                /*
                   _mesa_debug(ctx, "hardware blit clear buf %d rb id %d\n",
                   buf, irb->Base.Name);
@@ -555,20 +533,19 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask)
                OUT_BATCH((b.y2 << 16) | b.x2);
                OUT_RELOC(write_buffer,
 			 I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
-                         irb_region->draw_offset);
+                         irb->region->draw_offset);
                OUT_BATCH(clearVal);
                ADVANCE_BATCH();
                clearMask &= ~bufBit;    /* turn off bit, for faster loop exit */
             }
          }
       }
-      intel_batchbuffer_emit_mi_flush(intel->batch);
    }
 
    UNLOCK_HARDWARE(intel);
 }
 
-void
+GLboolean
 intelEmitImmediateColorExpandBlit(struct intel_context *intel,
 				  GLuint cpp,
 				  GLubyte *src_bits, GLuint src_size,
@@ -584,11 +561,19 @@ intelEmitImmediateColorExpandBlit(struct intel_context *intel,
    int dwords = ALIGN(src_size, 8) / 4;
    uint32_t opcode, br13, blit_cmd;
 
+   if (dst_tiling != I915_TILING_NONE) {
+      if (dst_offset & 4095)
+	 return GL_FALSE;
+      if (dst_tiling == I915_TILING_Y)
+	 return GL_FALSE;
+   }
+
    assert( logic_op - GL_CLEAR >= 0 );
    assert( logic_op - GL_CLEAR < 0x10 );
+   assert(dst_pitch > 0);
 
    if (w < 0 || h < 0)
-      return;
+      return GL_TRUE;
 
    dst_pitch *= cpp;
 
@@ -645,4 +630,46 @@ intelEmitImmediateColorExpandBlit(struct intel_context *intel,
 			   REFERENCES_CLIPRECTS );
 
    intel_batchbuffer_emit_mi_flush(intel->batch);
+
+   return GL_TRUE;
+}
+
+/* We don't have a memmove-type blit like some other hardware, so we'll do a
+ * rectangular blit covering a large space, then emit 1-scanline blit at the
+ * end to cover the last if we need.
+ */
+void
+intel_emit_linear_blit(struct intel_context *intel,
+		       drm_intel_bo *dst_bo,
+		       unsigned int dst_offset,
+		       drm_intel_bo *src_bo,
+		       unsigned int src_offset,
+		       unsigned int size)
+{
+   GLuint pitch, height;
+
+   /* The pitch is a signed value. */
+   pitch = MIN2(size, (1 << 15) - 1);
+   height = size / pitch;
+   intelEmitCopyBlit(intel, 1,
+		     pitch, src_bo, src_offset, I915_TILING_NONE,
+		     pitch, dst_bo, dst_offset, I915_TILING_NONE,
+		     0, 0, /* src x/y */
+		     0, 0, /* dst x/y */
+		     pitch, height, /* w, h */
+		     GL_COPY);
+
+   src_offset += pitch * height;
+   dst_offset += pitch * height;
+   size -= pitch * height;
+   assert (size < (1 << 15));
+   if (size != 0) {
+      intelEmitCopyBlit(intel, 1,
+			size, src_bo, src_offset, I915_TILING_NONE,
+			size, dst_bo, dst_offset, I915_TILING_NONE,
+			0, 0, /* src x/y */
+			0, 0, /* dst x/y */
+			size, 1, /* w, h */
+			GL_COPY);
+   }
 }
diff --git a/src/mesa/drivers/dri/intel/intel_blit.h b/src/mesa/drivers/dri/intel/intel_blit.h
index 52065b13ed..240cb7cd1b 100644
--- a/src/mesa/drivers/dri/intel/intel_blit.h
+++ b/src/mesa/drivers/dri/intel/intel_blit.h
@@ -35,7 +35,8 @@ extern void intelCopyBuffer(const __DRIdrawablePrivate * dpriv,
 
 extern void intelClearWithBlit(GLcontext * ctx, GLbitfield mask);
 
-extern void intelEmitCopyBlit(struct intel_context *intel,
+GLboolean
+intelEmitCopyBlit(struct intel_context *intel,
                               GLuint cpp,
                               GLshort src_pitch,
                               dri_bo *src_buffer,
@@ -50,16 +51,7 @@ extern void intelEmitCopyBlit(struct intel_context *intel,
                               GLshort w, GLshort h,
 			      GLenum logicop );
 
-extern void intelEmitFillBlit(struct intel_context *intel,
-                              GLuint cpp,
-                              GLshort dst_pitch,
-                              dri_bo *dst_buffer,
-                              GLuint dst_offset,
-			      uint32_t dst_tiling,
-                              GLshort x, GLshort y,
-                              GLshort w, GLshort h, GLuint color);
-
-void
+GLboolean
 intelEmitImmediateColorExpandBlit(struct intel_context *intel,
 				  GLuint cpp,
 				  GLubyte *src_bits, GLuint src_size,
@@ -71,5 +63,11 @@ intelEmitImmediateColorExpandBlit(struct intel_context *intel,
 				  GLshort x, GLshort y,
 				  GLshort w, GLshort h,
 				  GLenum logic_op);
+void intel_emit_linear_blit(struct intel_context *intel,
+			    drm_intel_bo *dst_bo,
+			    unsigned int dst_offset,
+			    drm_intel_bo *src_bo,
+			    unsigned int src_offset,
+			    unsigned int size);
 
 #endif
diff --git a/src/mesa/drivers/dri/intel/intel_buffer_objects.c b/src/mesa/drivers/dri/intel/intel_buffer_objects.c
index 60d7bb3770..c55c5c426e 100644
--- a/src/mesa/drivers/dri/intel/intel_buffer_objects.c
+++ b/src/mesa/drivers/dri/intel/intel_buffer_objects.c
@@ -28,16 +28,18 @@
 
 #include "main/imports.h"
 #include "main/mtypes.h"
+#include "main/macros.h"
 #include "main/bufferobj.h"
 
 #include "intel_context.h"
+#include "intel_blit.h"
 #include "intel_buffer_objects.h"
 #include "intel_batchbuffer.h"
 #include "intel_regions.h"
 
-static GLboolean intel_bufferobj_unmap(GLcontext * ctx,
-				       GLenum target,
-				       struct gl_buffer_object *obj);
+static GLboolean
+intel_bufferobj_unmap(GLcontext * ctx,
+                      GLenum target, struct gl_buffer_object *obj);
 
 /** Allocates a new dri_bo to store the data for the buffer object. */
 static void
@@ -105,11 +107,13 @@ intel_bufferobj_free(GLcontext * ctx, struct gl_buffer_object *obj)
    assert(intel_obj);
 
    /* Buffer objects are automatically unmapped when deleting according
-    * to the spec.
+    * to the spec, but Mesa doesn't do UnmapBuffer for us at context destroy
+    * (though it does if you call glDeleteBuffers)
     */
    if (obj->Pointer)
       intel_bufferobj_unmap(ctx, 0, obj);
 
+   _mesa_free(intel_obj->sys_buffer);
    if (intel_obj->region) {
       intel_bufferobj_release_region(intel, intel_obj);
    }
@@ -126,9 +130,10 @@ intel_bufferobj_free(GLcontext * ctx, struct gl_buffer_object *obj)
  * Allocate space for and store data in a buffer object.  Any data that was
  * previously stored in the buffer object is lost.  If data is NULL,
  * memory will be allocated, but no copy will occur.
- * Called via glBufferDataARB().
+ * Called via ctx->Driver.BufferData().
+ * \return GL_TRUE for success, GL_FALSE if out of memory
  */
-static void
+static GLboolean
 intel_bufferobj_data(GLcontext * ctx,
                      GLenum target,
                      GLsizeiptrARB size,
@@ -141,11 +146,7 @@ intel_bufferobj_data(GLcontext * ctx,
    intel_obj->Base.Size = size;
    intel_obj->Base.Usage = usage;
 
-   /* Buffer objects are automatically unmapped when creating new data buffers
-    * according to the spec.
-    */
-   if (obj->Pointer)
-      intel_bufferobj_unmap(ctx, 0, obj);
+   assert(!obj->Pointer); /* Mesa should have unmapped it */
 
    if (intel_obj->region)
       intel_bufferobj_release_region(intel, intel_obj);
@@ -154,12 +155,32 @@ intel_bufferobj_data(GLcontext * ctx,
       dri_bo_unreference(intel_obj->buffer);
       intel_obj->buffer = NULL;
    }
+   _mesa_free(intel_obj->sys_buffer);
+   intel_obj->sys_buffer = NULL;
+
    if (size != 0) {
+#ifdef I915
+      /* On pre-965, stick VBOs in system memory, as we're always doing swtnl
+       * with their contents anyway.
+       */
+      if (target == GL_ARRAY_BUFFER || target == GL_ELEMENT_ARRAY_BUFFER) {
+	 intel_obj->sys_buffer = _mesa_malloc(size);
+	 if (intel_obj->sys_buffer != NULL) {
+	    if (data != NULL)
+	       memcpy(intel_obj->sys_buffer, data, size);
+	    return GL_TRUE;
+	 }
+      }
+#endif
       intel_bufferobj_alloc_buffer(intel, intel_obj);
+      if (!intel_obj->buffer)
+         return GL_FALSE;
 
       if (data != NULL)
 	 dri_bo_subdata(intel_obj->buffer, 0, size, data);
    }
+
+   return GL_TRUE;
 }
 
 
@@ -184,7 +205,10 @@ intel_bufferobj_subdata(GLcontext * ctx,
    if (intel_obj->region)
       intel_bufferobj_cow(intel, intel_obj);
 
-   dri_bo_subdata(intel_obj->buffer, offset, size, data);
+   if (intel_obj->sys_buffer)
+      memcpy((char *)intel_obj->sys_buffer + offset, data, size);
+   else
+      dri_bo_subdata(intel_obj->buffer, offset, size, data);
 }
 
 
@@ -216,11 +240,21 @@ intel_bufferobj_map(GLcontext * ctx,
 {
    struct intel_context *intel = intel_context(ctx);
    struct intel_buffer_object *intel_obj = intel_buffer_object(obj);
+   GLboolean read_only = (access == GL_READ_ONLY_ARB);
+   GLboolean write_only = (access == GL_WRITE_ONLY_ARB);
 
-   /* XXX: Translate access to flags arg below:
-    */
    assert(intel_obj);
 
+   if (intel_obj->sys_buffer) {
+      obj->Pointer = intel_obj->sys_buffer;
+      return obj->Pointer;
+   }
+
+   /* Flush any existing batchbuffer that might have written to this
+    * buffer.
+    */
+   intelFlush(ctx);
+
    if (intel_obj->region)
       intel_bufferobj_cow(intel, intel_obj);
 
@@ -229,27 +263,214 @@ intel_bufferobj_map(GLcontext * ctx,
       return NULL;
    }
 
-   dri_bo_map(intel_obj->buffer, GL_TRUE);
+   if (write_only && intel->intelScreen->kernel_exec_fencing) {
+      drm_intel_gem_bo_map_gtt(intel_obj->buffer);
+      intel_obj->mapped_gtt = GL_TRUE;
+   } else {
+      drm_intel_bo_map(intel_obj->buffer, !read_only);
+      intel_obj->mapped_gtt = GL_FALSE;
+   }
+
    obj->Pointer = intel_obj->buffer->virtual;
+   obj->Length = obj->Size;
+   obj->Offset = 0;
+
+   return obj->Pointer;
+}
+
+/**
+ * Called via glMapBufferRange().
+ *
+ * The goal of this extension is to allow apps to accumulate their rendering
+ * at the same time as they accumulate their buffer object.  Without it,
+ * you'd end up blocking on execution of rendering every time you mapped
+ * the buffer to put new data in.
+ *
+ * We support it in 3 ways: If unsynchronized, then don't bother
+ * flushing the batchbuffer before mapping the buffer, which can save blocking
+ * in many cases.  If we would still block, and they allow the whole buffer
+ * to be invalidated, then just allocate a new buffer to replace the old one.
+ * If not, and we'd block, and they allow the subrange of the buffer to be
+ * invalidated, then we can make a new little BO, let them write into that,
+ * and blit it into the real BO at unmap time.
+ */
+static void *
+intel_bufferobj_map_range(GLcontext * ctx,
+			  GLenum target, GLintptr offset, GLsizeiptr length,
+			  GLbitfield access, struct gl_buffer_object *obj)
+{
+   struct intel_context *intel = intel_context(ctx);
+   struct intel_buffer_object *intel_obj = intel_buffer_object(obj);
+
+   assert(intel_obj);
+
+   /* _mesa_MapBufferRange (GL entrypoint) sets these, but the vbo module also
+    * internally uses our functions directly.
+    */
+   obj->Offset = offset;
+   obj->Length = length;
+   obj->AccessFlags = access;
+
+   if (intel_obj->sys_buffer) {
+      obj->Pointer = intel_obj->sys_buffer + offset;
+      return obj->Pointer;
+   }
+
+   if (intel_obj->region)
+      intel_bufferobj_cow(intel, intel_obj);
+
+   /* If the mapping is synchronized with other GL operations, flush
+    * the batchbuffer so that GEM knows about the buffer access for later
+    * syncing.
+    */
+   if (!(access & GL_MAP_UNSYNCHRONIZED_BIT))
+      intelFlush(ctx);
+
+   if (intel_obj->buffer == NULL) {
+      obj->Pointer = NULL;
+      return NULL;
+   }
+
+   /* If the user doesn't care about existing buffer contents and mapping
+    * would cause us to block, then throw out the old buffer.
+    */
+   if (!(access & GL_MAP_UNSYNCHRONIZED_BIT) &&
+       (access & GL_MAP_INVALIDATE_BUFFER_BIT) &&
+       drm_intel_bo_busy(intel_obj->buffer)) {
+      drm_intel_bo_unreference(intel_obj->buffer);
+      intel_obj->buffer = dri_bo_alloc(intel->bufmgr, "bufferobj",
+				       intel_obj->Base.Size, 64);
+   }
+
+   /* If the user is mapping a range of an active buffer object but
+    * doesn't require the current contents of that range, make a new
+    * BO, and we'll copy what they put in there out at unmap or
+    * FlushRange time.
+    */
+   if ((access & GL_MAP_INVALIDATE_RANGE_BIT) &&
+       drm_intel_bo_busy(intel_obj->buffer)) {
+      if (access & GL_MAP_FLUSH_EXPLICIT_BIT) {
+	 intel_obj->range_map_buffer = _mesa_malloc(length);
+	 obj->Pointer = intel_obj->range_map_buffer;
+      } else {
+	 intel_obj->range_map_bo = drm_intel_bo_alloc(intel->bufmgr,
+						      "range map",
+						      length, 64);
+	 if (!(access & GL_MAP_READ_BIT) &&
+	     intel->intelScreen->kernel_exec_fencing) {
+	    drm_intel_gem_bo_map_gtt(intel_obj->range_map_bo);
+	    intel_obj->mapped_gtt = GL_TRUE;
+	 } else {
+	    drm_intel_bo_map(intel_obj->range_map_bo,
+			     (access & GL_MAP_WRITE_BIT) != 0);
+	    intel_obj->mapped_gtt = GL_FALSE;
+	 }
+	 obj->Pointer = intel_obj->range_map_bo->virtual;
+      }
+      return obj->Pointer;
+   }
+
+   if (!(access & GL_MAP_READ_BIT) &&
+       intel->intelScreen->kernel_exec_fencing) {
+      drm_intel_gem_bo_map_gtt(intel_obj->buffer);
+      intel_obj->mapped_gtt = GL_TRUE;
+   } else {
+      drm_intel_bo_map(intel_obj->buffer, (access & GL_MAP_WRITE_BIT) != 0);
+      intel_obj->mapped_gtt = GL_FALSE;
+   }
+
+   obj->Pointer = intel_obj->buffer->virtual + offset;
    return obj->Pointer;
 }
 
+/* Ideally we'd use a BO to avoid taking up cache space for the temporary
+ * data, but FlushMappedBufferRange may be followed by further writes to
+ * the pointer, so we would have to re-map after emitting our blit, which
+ * would defeat the point.
+ */
+static void
+intel_bufferobj_flush_mapped_range(GLcontext *ctx, GLenum target,
+				   GLintptr offset, GLsizeiptr length,
+				   struct gl_buffer_object *obj)
+{
+   struct intel_context *intel = intel_context(ctx);
+   struct intel_buffer_object *intel_obj = intel_buffer_object(obj);
+   drm_intel_bo *temp_bo;
+
+   /* Unless we're in the range map using a temporary system buffer,
+    * there's no work to do.
+    */
+   if (intel_obj->range_map_buffer == NULL)
+      return;
+
+   temp_bo = drm_intel_bo_alloc(intel->bufmgr, "range map flush", length, 64);
+
+   drm_intel_bo_subdata(temp_bo, 0, length, intel_obj->range_map_buffer);
+
+   intel_emit_linear_blit(intel,
+			  intel_obj->buffer, obj->Offset + offset,
+			  temp_bo, 0,
+			  length);
+
+   drm_intel_bo_unreference(temp_bo);
+}
+
 
 /**
- * Called via glMapBufferARB().
+ * Called via glUnmapBuffer().
  */
 static GLboolean
 intel_bufferobj_unmap(GLcontext * ctx,
                       GLenum target, struct gl_buffer_object *obj)
 {
+   struct intel_context *intel = intel_context(ctx);
    struct intel_buffer_object *intel_obj = intel_buffer_object(obj);
 
    assert(intel_obj);
-   if (intel_obj->buffer != NULL) {
-      assert(obj->Pointer);
-      dri_bo_unmap(intel_obj->buffer);
-      obj->Pointer = NULL;
+   assert(obj->Pointer);
+   if (intel_obj->sys_buffer != NULL) {
+      /* always keep the mapping around. */
+   } else if (intel_obj->range_map_buffer != NULL) {
+      /* Since we've emitted some blits to buffers that will (likely) be used
+       * in rendering operations in other cache domains in this batch, emit a
+       * flush.  Once again, we wish for a domain tracker in libdrm to cover
+       * usage inside of a batchbuffer.
+       */
+      intel_batchbuffer_emit_mi_flush(intel->batch);
+      free(intel_obj->range_map_buffer);
+      intel_obj->range_map_buffer = NULL;
+   } else if (intel_obj->range_map_bo != NULL) {
+      if (intel_obj->mapped_gtt) {
+	 drm_intel_gem_bo_unmap_gtt(intel_obj->range_map_bo);
+      } else {
+	 drm_intel_bo_unmap(intel_obj->range_map_bo);
+      }
+
+      intel_emit_linear_blit(intel,
+			     intel_obj->buffer, obj->Offset,
+			     intel_obj->range_map_bo, 0,
+			     obj->Length);
+
+      /* Since we've emitted some blits to buffers that will (likely) be used
+       * in rendering operations in other cache domains in this batch, emit a
+       * flush.  Once again, we wish for a domain tracker in libdrm to cover
+       * usage inside of a batchbuffer.
+       */
+      intel_batchbuffer_emit_mi_flush(intel->batch);
+
+      drm_intel_bo_unreference(intel_obj->range_map_bo);
+      intel_obj->range_map_bo = NULL;
+   } else if (intel_obj->buffer != NULL) {
+      if (intel_obj->mapped_gtt) {
+	 drm_intel_gem_bo_unmap_gtt(intel_obj->buffer);
+      } else {
+	 drm_intel_bo_unmap(intel_obj->buffer);
+      }
    }
+   obj->Pointer = NULL;
+   obj->Offset = 0;
+   obj->Length = 0;
+
    return GL_TRUE;
 }
 
@@ -266,19 +487,95 @@ intel_bufferobj_buffer(struct intel_context *intel,
       }
    }
 
+   if (intel_obj->buffer == NULL) {
+      void *sys_buffer = intel_obj->sys_buffer;
+
+      /* only one of buffer and sys_buffer could be non-NULL */
+      intel_bufferobj_alloc_buffer(intel, intel_obj);
+      intel_obj->sys_buffer = NULL;
+
+      intel_bufferobj_subdata(&intel->ctx,
+			      GL_ARRAY_BUFFER_ARB,
+			      0,
+			      intel_obj->Base.Size,
+			      sys_buffer,
+			      &intel_obj->Base);
+      _mesa_free(sys_buffer);
+      intel_obj->sys_buffer = NULL;
+   }
+
    return intel_obj->buffer;
 }
 
+static void
+intel_bufferobj_copy_subdata(GLcontext *ctx,
+			     struct gl_buffer_object *src,
+			     struct gl_buffer_object *dst,
+			     GLintptr read_offset, GLintptr write_offset,
+			     GLsizeiptr size)
+{
+   struct intel_context *intel = intel_context(ctx);
+   struct intel_buffer_object *intel_src = intel_buffer_object(src);
+   struct intel_buffer_object *intel_dst = intel_buffer_object(dst);
+   drm_intel_bo *src_bo, *dst_bo;
+
+   if (size == 0)
+      return;
+
+   /* If we're in system memory, just map and memcpy. */
+   if (intel_src->sys_buffer || intel_dst->sys_buffer) {
+      /* The same buffer may be used, but note that regions copied may
+       * not overlap.
+       */
+      if (src == dst) {
+	 char *ptr = intel_bufferobj_map(ctx, GL_COPY_WRITE_BUFFER,
+					 GL_READ_WRITE, dst);
+	 memcpy(ptr + write_offset, ptr + read_offset, size);
+	 intel_bufferobj_unmap(ctx, GL_COPY_WRITE_BUFFER, dst);
+      } else {
+	 const char *src_ptr;
+	 char *dst_ptr;
+
+	 src_ptr =  intel_bufferobj_map(ctx, GL_COPY_READ_BUFFER,
+					GL_READ_ONLY, src);
+	 dst_ptr =  intel_bufferobj_map(ctx, GL_COPY_WRITE_BUFFER,
+					GL_WRITE_ONLY, dst);
+
+	 memcpy(dst_ptr + write_offset, src_ptr + read_offset, size);
+
+	 intel_bufferobj_unmap(ctx, GL_COPY_READ_BUFFER, src);
+	 intel_bufferobj_unmap(ctx, GL_COPY_WRITE_BUFFER, dst);
+      }
+   }
+
+   /* Otherwise, we have real BOs, so blit them. */
+
+   dst_bo = intel_bufferobj_buffer(intel, intel_dst, INTEL_WRITE_PART);
+   src_bo = intel_bufferobj_buffer(intel, intel_src, INTEL_READ);
+
+   intel_emit_linear_blit(intel,
+			  dst_bo, write_offset,
+			  src_bo, read_offset, size);
+
+   /* Since we've emitted some blits to buffers that will (likely) be used
+    * in rendering operations in other cache domains in this batch, emit a
+    * flush.  Once again, we wish for a domain tracker in libdrm to cover
+    * usage inside of a batchbuffer.
+    */
+   intel_batchbuffer_emit_mi_flush(intel->batch);
+}
+
 void
-intel_bufferobj_init(struct intel_context *intel)
+intelInitBufferObjectFuncs(struct dd_function_table *functions)
 {
-   GLcontext *ctx = &intel->ctx;
-
-   ctx->Driver.NewBufferObject = intel_bufferobj_alloc;
-   ctx->Driver.DeleteBuffer = intel_bufferobj_free;
-   ctx->Driver.BufferData = intel_bufferobj_data;
-   ctx->Driver.BufferSubData = intel_bufferobj_subdata;
-   ctx->Driver.GetBufferSubData = intel_bufferobj_get_subdata;
-   ctx->Driver.MapBuffer = intel_bufferobj_map;
-   ctx->Driver.UnmapBuffer = intel_bufferobj_unmap;
+   functions->NewBufferObject = intel_bufferobj_alloc;
+   functions->DeleteBuffer = intel_bufferobj_free;
+   functions->BufferData = intel_bufferobj_data;
+   functions->BufferSubData = intel_bufferobj_subdata;
+   functions->GetBufferSubData = intel_bufferobj_get_subdata;
+   functions->MapBuffer = intel_bufferobj_map;
+   functions->MapBufferRange = intel_bufferobj_map_range;
+   functions->FlushMappedBufferRange = intel_bufferobj_flush_mapped_range;
+   functions->UnmapBuffer = intel_bufferobj_unmap;
+   functions->CopyBufferSubData = intel_bufferobj_copy_subdata;
 }
diff --git a/src/mesa/drivers/dri/intel/intel_buffer_objects.h b/src/mesa/drivers/dri/intel/intel_buffer_objects.h
index bf6dbd58f2..bf3e08a320 100644
--- a/src/mesa/drivers/dri/intel/intel_buffer_objects.h
+++ b/src/mesa/drivers/dri/intel/intel_buffer_objects.h
@@ -42,10 +42,19 @@ struct intel_buffer_object
 {
    struct gl_buffer_object Base;
    dri_bo *buffer;     /* the low-level buffer manager's buffer handle */
+   /** System memory buffer data, if not using a BO to store the data. */
+   void *sys_buffer;
 
    struct intel_region *region; /* Is there a zero-copy texture
                                    associated with this (pixel)
                                    buffer object? */
+
+   drm_intel_bo *range_map_bo;
+   void *range_map_buffer;
+   unsigned int range_map_offset;
+   GLsizei range_map_size;
+
+   GLboolean mapped_gtt;
 };
 
 
@@ -57,7 +66,7 @@ dri_bo *intel_bufferobj_buffer(struct intel_context *intel,
 
 /* Hook the bufferobject implementation into mesa: 
  */
-void intel_bufferobj_init(struct intel_context *intel);
+void intelInitBufferObjectFuncs(struct dd_function_table *functions);
 
 
 
@@ -69,10 +78,7 @@ void intel_bufferobj_init(struct intel_context *intel);
 static INLINE struct intel_buffer_object *
 intel_buffer_object(struct gl_buffer_object *obj)
 {
-   if (obj->Name)
-      return (struct intel_buffer_object *) obj;
-   else
-      return NULL;
+   return (struct intel_buffer_object *) obj;
 }
 
 /* Helpers for zerocopy image uploads.  See also intel_regions.h:
diff --git a/src/mesa/drivers/dri/intel/intel_buffers.c b/src/mesa/drivers/dri/intel/intel_buffers.c
index 0fd2f16a8f..e7357e78c5 100644
--- a/src/mesa/drivers/dri/intel/intel_buffers.c
+++ b/src/mesa/drivers/dri/intel/intel_buffers.c
@@ -25,25 +25,14 @@
  * 
  **************************************************************************/
 
-#include "intel_screen.h"
 #include "intel_context.h"
-#include "intel_blit.h"
 #include "intel_buffers.h"
-#include "intel_chipset.h"
-#include "intel_depthstencil.h"
 #include "intel_fbo.h"
 #include "intel_regions.h"
 #include "intel_batchbuffer.h"
-#include "intel_reg.h"
-#include "main/context.h"
 #include "main/framebuffer.h"
-#include "swrast/swrast.h"
-#include "utils.h"
 #include "drirenderbuffer.h"
-#include "vblank.h"
-#include "i915_drm.h"
 
-#define FILE_DEBUG_FLAG DEBUG_BLIT
 
 /**
  * XXX move this into a new dri/common/cliprects.c file.
@@ -114,7 +103,6 @@ intel_get_cliprects(struct intel_context *intel,
 		    int *x_off, int *y_off)
 {
    __DRIdrawablePrivate *dPriv = intel->driDrawable;
-   struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
 
    if (intel->constant_cliprect) {
       /* FBO or DRI2 rendering, which can just use the fb's size. */
@@ -143,399 +131,6 @@ intel_get_cliprects(struct intel_context *intel,
    }
 }
 
-/**
- * This will be called whenever the currently bound window is moved/resized.
- * XXX: actually, it seems to NOT be called when the window is only moved (BP).
- */
-void
-intelWindowMoved(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-   __DRIdrawablePrivate *dPriv = intel->driDrawable;
-   struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
-
-   if (!intel->intelScreen->driScrnPriv->dri2.enabled &&
-       intel->intelScreen->driScrnPriv->ddx_version.minor >= 7) {
-      volatile drm_i915_sarea_t *sarea = intel->sarea;
-      drm_clip_rect_t drw_rect = { .x1 = dPriv->x, .x2 = dPriv->x + dPriv->w,
-				   .y1 = dPriv->y, .y2 = dPriv->y + dPriv->h };
-      drm_clip_rect_t planeA_rect = { .x1 = sarea->planeA_x, .y1 = sarea->planeA_y,
-				     .x2 = sarea->planeA_x + sarea->planeA_w,
-				     .y2 = sarea->planeA_y + sarea->planeA_h };
-      drm_clip_rect_t planeB_rect = { .x1 = sarea->planeB_x, .y1 = sarea->planeB_y,
-				     .x2 = sarea->planeB_x + sarea->planeB_w,
-				     .y2 = sarea->planeB_y + sarea->planeB_h };
-      GLint areaA = driIntersectArea( drw_rect, planeA_rect );
-      GLint areaB = driIntersectArea( drw_rect, planeB_rect );
-      GLuint flags = dPriv->vblFlags;
-
-      /* Update vblank info
-       */
-      if (areaB > areaA || (areaA == areaB && areaB > 0)) {
-	 flags = dPriv->vblFlags | VBLANK_FLAG_SECONDARY;
-      } else {
-	 flags = dPriv->vblFlags & ~VBLANK_FLAG_SECONDARY;
-      }
-
-      /* Check to see if we changed pipes */
-      if (flags != dPriv->vblFlags && dPriv->vblFlags &&
-	  !(dPriv->vblFlags & VBLANK_FLAG_NO_IRQ)) {
-	 int64_t count;
-	 drmVBlank vbl;
-	 int i;
-
-	 /*
-	  * Deal with page flipping
-	  */
-	 vbl.request.type = DRM_VBLANK_ABSOLUTE;
-
-	 if ( dPriv->vblFlags & VBLANK_FLAG_SECONDARY ) {
-	    vbl.request.type |= DRM_VBLANK_SECONDARY;
-	 }
-
-	 for (i = 0; i < 2; i++) {
-	    if (!intel_fb->color_rb[i] ||
-		(intel_fb->vbl_waited - intel_fb->color_rb[i]->vbl_pending) <=
-		(1<<23))
-	       continue;
-
-	    vbl.request.sequence = intel_fb->color_rb[i]->vbl_pending;
-	    drmWaitVBlank(intel->driFd, &vbl);
-	 }
-
-	 /*
-	  * Update msc_base from old pipe
-	  */
-	 driDrawableGetMSC32(dPriv->driScreenPriv, dPriv, &count);
-	 dPriv->msc_base = count;
-	 /*
-	  * Then get new vblank_base and vblSeq values
-	  */
-	 dPriv->vblFlags = flags;
-	 driGetCurrentVBlank(dPriv);
-	 dPriv->vblank_base = dPriv->vblSeq;
-
-	 intel_fb->vbl_waited = dPriv->vblSeq;
-
-	 for (i = 0; i < 2; i++) {
-	    if (intel_fb->color_rb[i])
-	       intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_waited;
-	 }
-      }
-   } else {
-      dPriv->vblFlags &= ~VBLANK_FLAG_SECONDARY;
-   }
-
-   /* Update Mesa's notion of window size */
-   driUpdateFramebufferSize(ctx, dPriv);
-   intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */
-
-   /* Update hardware scissor */
-   if (ctx->Driver.Scissor != NULL) {
-      ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y,
-			  ctx->Scissor.Width, ctx->Scissor.Height);
-   }
-
-   /* Re-calculate viewport related state */
-   if (ctx->Driver.DepthRange != NULL)
-      ctx->Driver.DepthRange( ctx, ctx->Viewport.Near, ctx->Viewport.Far );
-}
-
-
-
-/* A true meta version of this would be very simple and additionally
- * machine independent.  Maybe we'll get there one day.
- */
-static void
-intelClearWithTris(struct intel_context *intel, GLbitfield mask)
-{
-   GLcontext *ctx = &intel->ctx;
-   struct gl_framebuffer *fb = ctx->DrawBuffer;
-   GLuint buf;
-
-   intel->vtbl.install_meta_state(intel);
-
-   /* Back and stencil cliprects are the same.  Try and do both
-    * buffers at once:
-    */
-   if (mask & (BUFFER_BIT_BACK_LEFT | BUFFER_BIT_STENCIL | BUFFER_BIT_DEPTH)) {
-      struct intel_region *backRegion =
-	 intel_get_rb_region(fb, BUFFER_BACK_LEFT);
-      struct intel_region *depthRegion =
-	 intel_get_rb_region(fb, BUFFER_DEPTH);
-
-      intel->vtbl.meta_draw_region(intel, backRegion, depthRegion);
-
-      if (mask & BUFFER_BIT_BACK_LEFT)
-	 intel->vtbl.meta_color_mask(intel, GL_TRUE);
-      else
-	 intel->vtbl.meta_color_mask(intel, GL_FALSE);
-
-      if (mask & BUFFER_BIT_STENCIL)
-	 intel->vtbl.meta_stencil_replace(intel,
-					  intel->ctx.Stencil.WriteMask[0],
-					  intel->ctx.Stencil.Clear);
-      else
-	 intel->vtbl.meta_no_stencil_write(intel);
-
-      if (mask & BUFFER_BIT_DEPTH)
-	 intel->vtbl.meta_depth_replace(intel);
-      else
-	 intel->vtbl.meta_no_depth_write(intel);
-
-      intel->vtbl.meta_draw_quad(intel,
-				 fb->_Xmin,
-				 fb->_Xmax,
-				 fb->_Ymin,
-				 fb->_Ymax,
-				 intel->ctx.Depth.Clear,
-				 intel->ClearColor8888,
-				 0, 0, 0, 0);   /* texcoords */
-
-      mask &= ~(BUFFER_BIT_BACK_LEFT | BUFFER_BIT_STENCIL | BUFFER_BIT_DEPTH);
-   }
-
-   /* clear the remaining (color) renderbuffers */
-   for (buf = 0; buf < BUFFER_COUNT && mask; buf++) {
-      const GLuint bufBit = 1 << buf;
-      if (mask & bufBit) {
-	 struct intel_renderbuffer *irbColor =
-	    intel_renderbuffer(fb->Attachment[buf].Renderbuffer);
-
-	 ASSERT(irbColor);
-
-	 intel->vtbl.meta_no_depth_write(intel);
-	 intel->vtbl.meta_no_stencil_write(intel);
-	 intel->vtbl.meta_color_mask(intel, GL_TRUE);
-	 intel->vtbl.meta_draw_region(intel, irbColor->region, NULL);
-
-	 intel->vtbl.meta_draw_quad(intel,
-				    fb->_Xmin,
-				    fb->_Xmax,
-				    fb->_Ymin,
-				    fb->_Ymax,
-				    0, intel->ClearColor8888,
-				    0, 0, 0, 0);   /* texcoords */
-
-	 mask &= ~bufBit;
-      }
-   }
-
-   intel->vtbl.leave_meta_state(intel);
-}
-
-static const char *buffer_names[] = {
-   [BUFFER_FRONT_LEFT] = "front",
-   [BUFFER_BACK_LEFT] = "back",
-   [BUFFER_FRONT_RIGHT] = "front right",
-   [BUFFER_BACK_RIGHT] = "back right",
-   [BUFFER_AUX0] = "aux0",
-   [BUFFER_AUX1] = "aux1",
-   [BUFFER_AUX2] = "aux2",
-   [BUFFER_AUX3] = "aux3",
-   [BUFFER_DEPTH] = "depth",
-   [BUFFER_STENCIL] = "stencil",
-   [BUFFER_ACCUM] = "accum",
-   [BUFFER_COLOR0] = "color0",
-   [BUFFER_COLOR1] = "color1",
-   [BUFFER_COLOR2] = "color2",
-   [BUFFER_COLOR3] = "color3",
-   [BUFFER_COLOR4] = "color4",
-   [BUFFER_COLOR5] = "color5",
-   [BUFFER_COLOR6] = "color6",
-   [BUFFER_COLOR7] = "color7",
-};
-
-/**
- * Called by ctx->Driver.Clear.
- */
-static void
-intelClear(GLcontext *ctx, GLbitfield mask)
-{
-   struct intel_context *intel = intel_context(ctx);
-   const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask);
-   GLbitfield tri_mask = 0;
-   GLbitfield blit_mask = 0;
-   GLbitfield swrast_mask = 0;
-   struct gl_framebuffer *fb = ctx->DrawBuffer;
-   GLuint i;
-
-   if (0)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   /* HW color buffers (front, back, aux, generic FBO, etc) */
-   if (colorMask == ~0) {
-      /* clear all R,G,B,A */
-      /* XXX FBO: need to check if colorbuffers are software RBOs! */
-      blit_mask |= (mask & BUFFER_BITS_COLOR);
-   }
-   else {
-      /* glColorMask in effect */
-      tri_mask |= (mask & BUFFER_BITS_COLOR);
-   }
-
-   /* HW stencil */
-   if (mask & BUFFER_BIT_STENCIL) {
-      const struct intel_region *stencilRegion
-         = intel_get_rb_region(fb, BUFFER_STENCIL);
-      if (stencilRegion) {
-         /* have hw stencil */
-         if (IS_965(intel->intelScreen->deviceID) ||
-	     (ctx->Stencil.WriteMask[0] & 0xff) != 0xff) {
-	    /* We have to use the 3D engine if we're clearing a partial mask
-	     * of the stencil buffer, or if we're on a 965 which has a tiled
-	     * depth/stencil buffer in a layout we can't blit to.
-	     */
-            tri_mask |= BUFFER_BIT_STENCIL;
-         }
-         else {
-            /* clearing all stencil bits, use blitting */
-            blit_mask |= BUFFER_BIT_STENCIL;
-         }
-      }
-   }
-
-   /* HW depth */
-   if (mask & BUFFER_BIT_DEPTH) {
-      /* clear depth with whatever method is used for stencil (see above) */
-      if (IS_965(intel->intelScreen->deviceID) ||
-	  tri_mask & BUFFER_BIT_STENCIL)
-         tri_mask |= BUFFER_BIT_DEPTH;
-      else
-         blit_mask |= BUFFER_BIT_DEPTH;
-   }
-
-   /* SW fallback clearing */
-   swrast_mask = mask & ~tri_mask & ~blit_mask;
-
-   for (i = 0; i < BUFFER_COUNT; i++) {
-      GLuint bufBit = 1 << i;
-      if ((blit_mask | tri_mask) & bufBit) {
-         if (!fb->Attachment[i].Renderbuffer->ClassID) {
-            blit_mask &= ~bufBit;
-            tri_mask &= ~bufBit;
-            swrast_mask |= bufBit;
-         }
-      }
-   }
-
-   if (blit_mask) {
-      if (INTEL_DEBUG & DEBUG_BLIT) {
-	 DBG("blit clear:");
-	 for (i = 0; i < BUFFER_COUNT; i++) {
-	    if (blit_mask & (1 << i))
-	       DBG(" %s", buffer_names[i]);
-	 }
-	 DBG("\n");
-      }
-      intelClearWithBlit(ctx, blit_mask);
-   }
-
-   if (tri_mask) {
-      if (INTEL_DEBUG & DEBUG_BLIT) {
-	 DBG("tri clear:");
-	 for (i = 0; i < BUFFER_COUNT; i++) {
-	    if (tri_mask & (1 << i))
-	       DBG(" %s", buffer_names[i]);
-	 }
-	 DBG("\n");
-      }
-      intelClearWithTris(intel, tri_mask);
-   }
-
-   if (swrast_mask) {
-      if (INTEL_DEBUG & DEBUG_BLIT) {
-	 DBG("swrast clear:");
-	 for (i = 0; i < BUFFER_COUNT; i++) {
-	    if (swrast_mask & (1 << i))
-	       DBG(" %s", buffer_names[i]);
-	 }
-	 DBG("\n");
-      }
-      _swrast_Clear(ctx, swrast_mask);
-   }
-}
-
-void
-intelSwapBuffers(__DRIdrawablePrivate * dPriv)
-{
-   __DRIscreenPrivate *psp = dPriv->driScreenPriv;
-
-   if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
-      GET_CURRENT_CONTEXT(ctx);
-      struct intel_context *intel;
-
-      if (ctx == NULL)
-	 return;
-
-      intel = intel_context(ctx);
-
-      if (ctx->Visual.doubleBufferMode) {
-	 GLboolean missed_target;
-	 struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
-	 int64_t ust;
-         
-	 _mesa_notifySwapBuffers(ctx);  /* flush pending rendering comands */
-
-	/*
-	 * The old swapping ioctl was incredibly racy, just wait for vblank
-	 * and do the swap ourselves.
-	 */
-	 driWaitForVBlank(dPriv, &missed_target);
-
-	 /*
-	  * Update each buffer's vbl_pending so we don't get too out of
-	  * sync
-	  */
-	 intel_get_renderbuffer(&intel_fb->Base,
-		   		BUFFER_BACK_LEFT)->vbl_pending = dPriv->vblSeq;
-         intel_get_renderbuffer(&intel_fb->Base,
-		   		BUFFER_FRONT_LEFT)->vbl_pending = dPriv->vblSeq;
-
-	 intelCopyBuffer(dPriv, NULL);
-
-	 intel_fb->swap_count++;
-	 (*psp->systemTime->getUST) (&ust);
-	 if (missed_target) {
-	    intel_fb->swap_missed_count++;
-	    intel_fb->swap_missed_ust = ust - intel_fb->swap_ust;
-	 }
-
-	 intel_fb->swap_ust = ust;
-      }
-      drmCommandNone(intel->driFd, DRM_I915_GEM_THROTTLE);
-
-   }
-   else {
-      /* XXX this shouldn't be an error but we can't handle it for now */
-      fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__);
-   }
-}
-
-void
-intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h)
-{
-   if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
-      struct intel_context *intel =
-         (struct intel_context *) dPriv->driContextPriv->driverPrivate;
-      GLcontext *ctx = &intel->ctx;
-
-      if (ctx->Visual.doubleBufferMode) {
-         drm_clip_rect_t rect;
-         rect.x1 = x + dPriv->x;
-         rect.y1 = (dPriv->h - y - h) + dPriv->y;
-         rect.x2 = rect.x1 + w;
-         rect.y2 = rect.y1 + h;
-         _mesa_notifySwapBuffers(ctx);  /* flush pending rendering comands */
-         intelCopyBuffer(dPriv, &rect);
-      }
-   }
-   else {
-      /* XXX this shouldn't be an error but we can't handle it for now */
-      fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__);
-   }
-}
-
 
 /**
  * Update the hardware state for drawing into a window or framebuffer object.
@@ -559,10 +154,10 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
       return;
    }
 
-   /* Do this here, note core Mesa, since this function is called from
+   /* Do this here, not core Mesa, since this function is called from
     * many places within the driver.
     */
-   if (ctx->NewState & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) {
+   if (ctx->NewState & _NEW_BUFFERS) {
       /* this updates the DrawBuffer->_NumColorDrawBuffers fields, etc */
       _mesa_update_framebuffer(ctx);
       /* this updates the DrawBuffer's Width/Height if it's a FBO */
@@ -577,9 +172,6 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
       return;
    }
 
-   if (fb->Name)
-      intel_validate_paired_depth_stencil(ctx, fb);
-
    /*
     * How many color buffers are we drawing into?
     */
@@ -587,7 +179,8 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
       /* writing to 0  */
       colorRegions[0] = NULL;
       intel->constant_cliprect = GL_TRUE;
-   } else if (fb->_NumColorDrawBuffers > 1) {
+   }
+   else if (fb->_NumColorDrawBuffers > 1) {
        int i;
        struct intel_renderbuffer *irb;
 
@@ -609,6 +202,8 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
 	       intel_batchbuffer_flush(intel->batch);
 	    intel->front_cliprects = GL_TRUE;
 	    colorRegions[0] = intel_get_rb_region(fb, BUFFER_FRONT_LEFT);
+
+	    intel->front_buffer_dirty = GL_TRUE;
 	 }
 	 else {
 	    if (!intel->constant_cliprect && intel->front_cliprects)
@@ -626,14 +221,6 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
       }
    }
 
-   /* Update culling direction which changes depending on the
-    * orientation of the buffer:
-    */
-   if (ctx->Driver.FrontFace)
-      ctx->Driver.FrontFace(ctx, ctx->Polygon.FrontFace);
-   else
-      ctx->NewState |= _NEW_POLYGON;
-
    if (!colorRegions[0]) {
       FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE);
    }
@@ -665,59 +252,50 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
    /***
     *** Stencil buffer
     *** This can only be hardware accelerated if we're using a
-    *** combined DEPTH_STENCIL buffer (for now anyway).
+    *** combined DEPTH_STENCIL buffer.
     ***/
    if (fb->_StencilBuffer && fb->_StencilBuffer->Wrapped) {
       irbStencil = intel_renderbuffer(fb->_StencilBuffer->Wrapped);
       if (irbStencil && irbStencil->region) {
          ASSERT(irbStencil->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT);
          FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE);
-         /* need to re-compute stencil hw state */
-	 if (ctx->Driver.Enable != NULL)
-	    ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled);
-	 else
-	    ctx->NewState |= _NEW_STENCIL;
-         if (!depthRegion)
-            depthRegion = irbStencil->region;
       }
       else {
          FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_TRUE);
       }
    }
    else {
-      /* XXX FBO: instead of FALSE, pass ctx->Stencil.Enabled ??? */
+      /* XXX FBO: instead of FALSE, pass ctx->Stencil._Enabled ??? */
       FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE);
-      /* need to re-compute stencil hw state */
-      if (ctx->Driver.Enable != NULL)
-	 ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled);
-      else
-	 ctx->NewState |= _NEW_STENCIL;
    }
 
    /*
-    * Update depth test state
+    * Update depth and stencil test state
     */
    if (ctx->Driver.Enable) {
-      if (ctx->Depth.Test && fb->Visual.depthBits > 0) {
-	 ctx->Driver.Enable(ctx, GL_DEPTH_TEST, GL_TRUE);
-      } else {
-	 ctx->Driver.Enable(ctx, GL_DEPTH_TEST, GL_FALSE);
-      }
-   } else {
-      ctx->NewState |= _NEW_DEPTH;
+      ctx->Driver.Enable(ctx, GL_DEPTH_TEST,
+                         (ctx->Depth.Test && fb->Visual.depthBits > 0));
+      ctx->Driver.Enable(ctx, GL_STENCIL_TEST,
+                         (ctx->Stencil.Enabled && fb->Visual.stencilBits > 0));
+   }
+   else {
+      /* Mesa's Stencil._Enabled field is updated when
+       * _NEW_BUFFERS | _NEW_STENCIL, but i965 code assumes that the value
+       * only changes with _NEW_STENCIL (which seems sensible).  So flag it
+       * here since this is the _NEW_BUFFERS path.
+       */
+      ctx->NewState |= (_NEW_DEPTH | _NEW_STENCIL);
    }
 
    intel->vtbl.set_draw_region(intel, colorRegions, depthRegion, 
-	fb->_NumColorDrawBuffers);
+                               fb->_NumColorDrawBuffers);
 
    /* update viewport since it depends on window size */
-   if (ctx->Driver.Viewport) {
-      ctx->Driver.Viewport(ctx, ctx->Viewport.X, ctx->Viewport.Y,
-			   ctx->Viewport.Width, ctx->Viewport.Height);
-   } else {
-      ctx->NewState |= _NEW_VIEWPORT;
-   }
-
+#ifdef I915
+   intelCalcViewport(ctx);
+#else
+   ctx->NewState |= _NEW_VIEWPORT;
+#endif
    /* Set state we know depends on drawable parameters:
     */
    if (ctx->Driver.Scissor)
@@ -729,12 +307,37 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
       ctx->Driver.DepthRange(ctx,
 			     ctx->Viewport.Near,
 			     ctx->Viewport.Far);
+
+   /* Update culling direction which changes depending on the
+    * orientation of the buffer:
+    */
+   if (ctx->Driver.FrontFace)
+      ctx->Driver.FrontFace(ctx, ctx->Polygon.FrontFace);
+   else
+      ctx->NewState |= _NEW_POLYGON;
 }
 
 
 static void
 intelDrawBuffer(GLcontext * ctx, GLenum mode)
 {
+   if ((ctx->DrawBuffer != NULL) && (ctx->DrawBuffer->Name == 0)) {
+      struct intel_context *const intel = intel_context(ctx);
+      const GLboolean was_front_buffer_rendering =
+	intel->is_front_buffer_rendering;
+
+      intel->is_front_buffer_rendering = (mode == GL_FRONT_LEFT)
+	|| (mode == GL_FRONT);
+
+      /* If we weren't front-buffer rendering before but we are now, make sure
+       * that the front-buffer has actually been allocated.
+       */
+      if (!was_front_buffer_rendering && intel->is_front_buffer_rendering) {
+	 intel_update_renderbuffers(intel->driContext,
+				    intel->driContext->driDrawablePriv);
+      }
+   }
+
    intel_draw_buffer(ctx, ctx->DrawBuffer);
 }
 
@@ -742,6 +345,23 @@ intelDrawBuffer(GLcontext * ctx, GLenum mode)
 static void
 intelReadBuffer(GLcontext * ctx, GLenum mode)
 {
+   if ((ctx->DrawBuffer != NULL) && (ctx->DrawBuffer->Name == 0)) {
+      struct intel_context *const intel = intel_context(ctx);
+      const GLboolean was_front_buffer_reading =
+	intel->is_front_buffer_reading;
+
+      intel->is_front_buffer_reading = (mode == GL_FRONT_LEFT)
+	|| (mode == GL_FRONT);
+
+      /* If we weren't front-buffer reading before but we are now, make sure
+       * that the front-buffer has actually been allocated.
+       */
+      if (!was_front_buffer_reading && intel->is_front_buffer_reading) {
+	 intel_update_renderbuffers(intel->driContext,
+				    intel->driContext->driDrawablePriv);
+      }
+   }
+
    if (ctx->ReadBuffer == ctx->DrawBuffer) {
       /* This will update FBO completeness status.
        * A framebuffer will be incomplete if the GL_READ_BUFFER setting
@@ -759,7 +379,6 @@ intelReadBuffer(GLcontext * ctx, GLenum mode)
 void
 intelInitBufferFuncs(struct dd_function_table *functions)
 {
-   functions->Clear = intelClear;
    functions->DrawBuffer = intelDrawBuffer;
    functions->ReadBuffer = intelReadBuffer;
 }
diff --git a/src/mesa/drivers/dri/intel/intel_buffers.h b/src/mesa/drivers/dri/intel/intel_buffers.h
index 0be1cee091..6069d38e9e 100644
--- a/src/mesa/drivers/dri/intel/intel_buffers.h
+++ b/src/mesa/drivers/dri/intel/intel_buffers.h
@@ -45,10 +45,6 @@ extern struct intel_region *intel_readbuf_region(struct intel_context *intel);
 
 extern struct intel_region *intel_drawbuf_region(struct intel_context *intel);
 
-extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv);
-
-extern void intelWindowMoved(struct intel_context *intel);
-
 extern void intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb);
 
 extern void intelInitBufferFuncs(struct dd_function_table *functions);
@@ -57,5 +53,8 @@ void intel_get_cliprects(struct intel_context *intel,
 			 struct drm_clip_rect **cliprects,
 			 unsigned int *num_cliprects,
 			 int *x_off, int *y_off);
+#ifdef I915
+void intelCalcViewport(GLcontext * ctx);
+#endif
 
 #endif /* INTEL_BUFFERS_H */
diff --git a/src/mesa/drivers/dri/intel/intel_chipset.h b/src/mesa/drivers/dri/intel/intel_chipset.h
index d1b4941601..3dc8653a73 100644
--- a/src/mesa/drivers/dri/intel/intel_chipset.h
+++ b/src/mesa/drivers/dri/intel/intel_chipset.h
@@ -46,6 +46,13 @@
 #define PCI_CHIP_G33_G			0x29C2
 #define PCI_CHIP_Q33_G			0x29D2
 
+#define PCI_CHIP_IGD_GM			0xA011
+#define PCI_CHIP_IGD_G			0xA001
+
+#define IS_IGDGM(devid)	(devid == PCI_CHIP_IGD_GM)
+#define IS_IGDG(devid)	(devid == PCI_CHIP_IGD_G)
+#define IS_IGD(devid) (IS_IGDG(devid) || IS_IGDGM(devid))
+
 #define PCI_CHIP_I965_G			0x29A2
 #define PCI_CHIP_I965_Q			0x2992
 #define PCI_CHIP_I965_G_1		0x2982
@@ -59,6 +66,10 @@
 #define PCI_CHIP_Q45_G                  0x2E12
 #define PCI_CHIP_G45_G                  0x2E22
 #define PCI_CHIP_G41_G                  0x2E32
+#define PCI_CHIP_B43_G                  0x2E42
+
+#define PCI_CHIP_ILD_G                  0x0042
+#define PCI_CHIP_ILM_G                  0x0046
 
 #define IS_MOBILE(devid)	(devid == PCI_CHIP_I855_GM || \
 				 devid == PCI_CHIP_I915_GM || \
@@ -66,15 +77,22 @@
 				 devid == PCI_CHIP_I945_GME || \
 				 devid == PCI_CHIP_I965_GM || \
 				 devid == PCI_CHIP_I965_GME || \
-				 devid == PCI_CHIP_GM45_GM)
+				 devid == PCI_CHIP_GM45_GM || \
+				 IS_IGD(devid) || \
+				 devid == PCI_CHIP_ILM_G)
 
 #define IS_G45(devid)           (devid == PCI_CHIP_IGD_E_G || \
                                  devid == PCI_CHIP_Q45_G || \
                                  devid == PCI_CHIP_G45_G || \
-                                 devid == PCI_CHIP_G41_G)
+                                 devid == PCI_CHIP_G41_G || \
+                                 devid == PCI_CHIP_B43_G)
 #define IS_GM45(devid)          (devid == PCI_CHIP_GM45_GM)
 #define IS_G4X(devid)		(IS_G45(devid) || IS_GM45(devid))
 
+#define IS_ILD(devid)           (devid == PCI_CHIP_ILD_G)
+#define IS_ILM(devid)           (devid == PCI_CHIP_ILM_G)
+#define IS_IGDNG(devid)           (IS_ILD(devid) || IS_ILM(devid))
+
 #define IS_915(devid)		(devid == PCI_CHIP_I915_G || \
 				 devid == PCI_CHIP_E7221_G || \
 				 devid == PCI_CHIP_I915_GM)
@@ -84,7 +102,7 @@
 				 devid == PCI_CHIP_I945_GME || \
 				 devid == PCI_CHIP_G33_G || \
 				 devid == PCI_CHIP_Q33_G || \
-				 devid == PCI_CHIP_Q35_G)
+				 devid == PCI_CHIP_Q35_G || IS_IGD(devid))
 
 #define IS_965(devid)		(devid == PCI_CHIP_I965_G || \
 				 devid == PCI_CHIP_I965_Q || \
@@ -92,7 +110,8 @@
 				 devid == PCI_CHIP_I965_GM || \
 				 devid == PCI_CHIP_I965_GME || \
 				 devid == PCI_CHIP_I946_GZ || \
-				 IS_G4X(devid))
+				 IS_G4X(devid) || \
+				 IS_IGDNG(devid))
 
 #define IS_9XX(devid)		(IS_915(devid) || \
 				 IS_945(devid) || \
diff --git a/src/mesa/drivers/dri/intel/intel_clear.c b/src/mesa/drivers/dri/intel/intel_clear.c
new file mode 100644
index 0000000000..bce23724b3
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_clear.c
@@ -0,0 +1,194 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2009 Intel Corporation.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "main/glheader.h"
+#include "main/mtypes.h"
+#include "swrast/swrast.h"
+#include "drivers/common/meta.h"
+
+#include "intel_context.h"
+#include "intel_blit.h"
+#include "intel_chipset.h"
+#include "intel_clear.h"
+#include "intel_fbo.h"
+#include "intel_pixel.h"
+#include "intel_regions.h"
+
+#define FILE_DEBUG_FLAG DEBUG_BLIT
+
+static const char *buffer_names[] = {
+   [BUFFER_FRONT_LEFT] = "front",
+   [BUFFER_BACK_LEFT] = "back",
+   [BUFFER_FRONT_RIGHT] = "front right",
+   [BUFFER_BACK_RIGHT] = "back right",
+   [BUFFER_DEPTH] = "depth",
+   [BUFFER_STENCIL] = "stencil",
+   [BUFFER_ACCUM] = "accum",
+   [BUFFER_AUX0] = "aux0",
+   [BUFFER_COLOR0] = "color0",
+   [BUFFER_COLOR1] = "color1",
+   [BUFFER_COLOR2] = "color2",
+   [BUFFER_COLOR3] = "color3",
+   [BUFFER_COLOR4] = "color4",
+   [BUFFER_COLOR5] = "color5",
+   [BUFFER_COLOR6] = "color6",
+   [BUFFER_COLOR7] = "color7",
+};
+
+/**
+ * Called by ctx->Driver.Clear.
+ */
+static void
+intelClear(GLcontext *ctx, GLbitfield mask)
+{
+   struct intel_context *intel = intel_context(ctx);
+   const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask);
+   GLbitfield tri_mask = 0;
+   GLbitfield blit_mask = 0;
+   GLbitfield swrast_mask = 0;
+   struct gl_framebuffer *fb = ctx->DrawBuffer;
+   GLuint i;
+
+   if (0)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   /* HW color buffers (front, back, aux, generic FBO, etc) */
+   if (colorMask == ~0) {
+      /* clear all R,G,B,A */
+      /* XXX FBO: need to check if colorbuffers are software RBOs! */
+      blit_mask |= (mask & BUFFER_BITS_COLOR);
+   }
+   else {
+      /* glColorMask in effect */
+      tri_mask |= (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT));
+   }
+
+   /* HW stencil */
+   if (mask & BUFFER_BIT_STENCIL) {
+      const struct intel_region *stencilRegion
+         = intel_get_rb_region(fb, BUFFER_STENCIL);
+      if (stencilRegion) {
+         /* have hw stencil */
+         if (stencilRegion->tiling == I915_TILING_Y ||
+	     (ctx->Stencil.WriteMask[0] & 0xff) != 0xff) {
+	    /* We have to use the 3D engine if we're clearing a partial mask
+	     * of the stencil buffer, or if we're on a 965 which has a tiled
+	     * depth/stencil buffer in a layout we can't blit to.
+	     */
+            tri_mask |= BUFFER_BIT_STENCIL;
+         }
+         else {
+            /* clearing all stencil bits, use blitting */
+            blit_mask |= BUFFER_BIT_STENCIL;
+         }
+      }
+   }
+
+   /* HW depth */
+   if (mask & BUFFER_BIT_DEPTH) {
+      const struct intel_region *irb = intel_get_rb_region(fb, BUFFER_DEPTH);
+
+      /* clear depth with whatever method is used for stencil (see above) */
+      if (irb->tiling == I915_TILING_Y || tri_mask & BUFFER_BIT_STENCIL)
+         tri_mask |= BUFFER_BIT_DEPTH;
+      else
+         blit_mask |= BUFFER_BIT_DEPTH;
+   }
+
+   /* If we're doing a tri pass for depth/stencil, include a likely color
+    * buffer with it.
+    */
+   if (mask & (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL)) {
+      int color_bit = _mesa_ffs(mask & BUFFER_BITS_COLOR);
+      if (color_bit != 0) {
+	 tri_mask |= blit_mask & (1 << (color_bit - 1));
+	 blit_mask &= ~(1 << (color_bit - 1));
+      }
+   }
+
+   /* SW fallback clearing */
+   swrast_mask = mask & ~tri_mask & ~blit_mask;
+
+   {
+      /* look for non-Intel renderbuffers (clear them with swrast) */
+      GLbitfield blit_or_tri = blit_mask | tri_mask;
+      while (blit_or_tri) {
+         GLuint i = _mesa_ffs(blit_or_tri) - 1;
+         GLbitfield bufBit = 1 << i;
+         if (!fb->Attachment[i].Renderbuffer->ClassID) {
+            blit_mask &= ~bufBit;
+            tri_mask &= ~bufBit;
+            swrast_mask |= bufBit;
+         }
+         blit_or_tri ^= bufBit;
+      }
+   }
+
+   if (blit_mask) {
+      if (INTEL_DEBUG & DEBUG_BLIT) {
+	 DBG("blit clear:");
+	 for (i = 0; i < BUFFER_COUNT; i++) {
+	    if (blit_mask & (1 << i))
+	       DBG(" %s", buffer_names[i]);
+	 }
+	 DBG("\n");
+      }
+      intelClearWithBlit(ctx, blit_mask);
+   }
+
+   if (tri_mask) {
+      if (INTEL_DEBUG & DEBUG_BLIT) {
+	 DBG("tri clear:");
+	 for (i = 0; i < BUFFER_COUNT; i++) {
+	    if (tri_mask & (1 << i))
+	       DBG(" %s", buffer_names[i]);
+	 }
+	 DBG("\n");
+      }
+      _mesa_meta_clear(&intel->ctx, tri_mask);
+   }
+
+   if (swrast_mask) {
+      if (INTEL_DEBUG & DEBUG_BLIT) {
+	 DBG("swrast clear:");
+	 for (i = 0; i < BUFFER_COUNT; i++) {
+	    if (swrast_mask & (1 << i))
+	       DBG(" %s", buffer_names[i]);
+	 }
+	 DBG("\n");
+      }
+      _swrast_Clear(ctx, swrast_mask);
+   }
+}
+
+
+void
+intelInitClearFuncs(struct dd_function_table *functions)
+{
+   functions->Clear = intelClear;
+}
diff --git a/src/mesa/drivers/dri/intel/intel_clear.h b/src/mesa/drivers/dri/intel/intel_clear.h
new file mode 100644
index 0000000000..7fd6b310a9
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_clear.h
@@ -0,0 +1,38 @@
+
+/**************************************************************************
+ * 
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef INTEL_CLEAR_H
+#define INTEL_CLEAR_H
+
+struct dd_function_table;
+
+extern void
+intelInitClearFuncs(struct dd_function_table *functions);
+
+
+#endif /* INTEL_CLEAR_H */
diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c
index c4a24d7397..7a2e7617d0 100644
--- a/src/mesa/drivers/dri/intel/intel_context.c
+++ b/src/mesa/drivers/dri/intel/intel_context.c
@@ -27,9 +27,10 @@
 
 
 #include "main/glheader.h"
+#include "main/colortab.h"
 #include "main/context.h"
-#include "main/matrix.h"
-#include "main/simple_list.h"
+#include "main/convolve.h"
+#include "main/arrayobj.h"
 #include "main/extensions.h"
 #include "main/framebuffer.h"
 #include "main/imports.h"
@@ -38,13 +39,8 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-
-#include "tnl/t_pipeline.h"
-#include "tnl/t_vertex.h"
-
 #include "drivers/common/driverfuncs.h"
-
-#include "intel_screen.h"
+#include "drivers/common/meta.h"
 
 #include "i830_dri.h"
 
@@ -52,52 +48,35 @@
 #include "intel_buffers.h"
 #include "intel_tex.h"
 #include "intel_batchbuffer.h"
-#include "intel_blit.h"
+#include "intel_clear.h"
+#include "intel_extensions.h"
 #include "intel_pixel.h"
 #include "intel_regions.h"
 #include "intel_buffer_objects.h"
 #include "intel_fbo.h"
+#include "intel_eglimage.h"
 #include "intel_decode.h"
 #include "intel_bufmgr.h"
+#include "intel_screen.h"
+#include "intel_swapbuffers.h"
 
 #include "drirenderbuffer.h"
 #include "vblank.h"
 #include "utils.h"
 #include "xmlpool.h"            /* for symbolic values of enum-type options */
+
+
 #ifndef INTEL_DEBUG
 int INTEL_DEBUG = (0);
 #endif
 
-#define need_GL_ARB_multisample
-#define need_GL_ARB_occlusion_query
-#define need_GL_ARB_point_parameters
-#define need_GL_ARB_shader_objects
-#define need_GL_ARB_texture_compression
-#define need_GL_ARB_vertex_buffer_object
-#define need_GL_ARB_vertex_program
-#define need_GL_ARB_vertex_shader
-#define need_GL_ARB_window_pos
-#define need_GL_EXT_blend_color
-#define need_GL_EXT_blend_equation_separate
-#define need_GL_EXT_blend_func_separate
-#define need_GL_EXT_blend_minmax
-#define need_GL_EXT_cull_vertex
-#define need_GL_EXT_fog_coord
-#define need_GL_EXT_framebuffer_object
-#define need_GL_EXT_multi_draw_arrays
-#define need_GL_EXT_point_parameters
-#define need_GL_EXT_secondary_color
-#define need_GL_ATI_separate_stencil
-#define need_GL_NV_point_sprite
-#define need_GL_NV_vertex_program
-#define need_GL_VERSION_2_0
-#define need_GL_VERSION_2_1
-
-#include "extension_helper.h"
-
-#define DRIVER_DATE                     "20090114"
+
+#define DRIVER_DATE                     "20090712 2009Q2 RC3"
 #define DRIVER_DATE_GEM                 "GEM " DRIVER_DATE
 
+
+static void intel_flush(GLcontext *ctx, GLboolean needs_mi_flush);
+
 static const GLubyte *
 intelGetString(GLcontext * ctx, GLenum name)
 {
@@ -151,6 +130,10 @@ intelGetString(GLcontext * ctx, GLenum name)
       case PCI_CHIP_Q33_G:
 	 chipset = "Intel(R) Q33";
 	 break;
+      case PCI_CHIP_IGD_GM:
+      case PCI_CHIP_IGD_G:
+	 chipset = "Intel(R) IGD";
+	 break;
       case PCI_CHIP_I965_Q:
 	 chipset = "Intel(R) 965Q";
 	 break;
@@ -182,6 +165,15 @@ intelGetString(GLcontext * ctx, GLenum name)
       case PCI_CHIP_G41_G:
          chipset = "Intel(R) G41";
          break;
+      case PCI_CHIP_B43_G:
+         chipset = "Intel(R) B43";
+         break;
+      case PCI_CHIP_ILD_G:
+         chipset = "Intel(R) IGDNG_D";
+         break;
+      case PCI_CHIP_ILM_G:
+         chipset = "Intel(R) IGDNG_M";
+         break;
       default:
          chipset = "Unknown Intel Chipset";
          break;
@@ -197,6 +189,24 @@ intelGetString(GLcontext * ctx, GLenum name)
    }
 }
 
+static unsigned
+intel_bits_per_pixel(const struct intel_renderbuffer *rb)
+{
+   switch (rb->Base._ActualFormat) {
+   case GL_RGB5:
+   case GL_DEPTH_COMPONENT16:
+      return 16;
+   case GL_RGB8:
+   case GL_RGBA8:
+   case GL_DEPTH_COMPONENT24:
+   case GL_DEPTH24_STENCIL8_EXT:
+   case GL_STENCIL_INDEX8_EXT:
+      return 32;
+   default:
+      return 0;
+   }
+}
+
 void
 intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
 {
@@ -204,7 +214,7 @@ intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
    struct intel_renderbuffer *rb;
    struct intel_region *region, *depth_region;
    struct intel_context *intel = context->driverPrivate;
-   __DRIbuffer *buffers;
+   __DRIbuffer *buffers = NULL;
    __DRIscreen *screen;
    int i, count;
    unsigned int attachments[10];
@@ -216,22 +226,65 @@ intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
 
    screen = intel->intelScreen->driScrnPriv;
 
-   i = 0;
-   if (intel_fb->color_rb[0])
-      attachments[i++] = __DRI_BUFFER_FRONT_LEFT;
-   if (intel_fb->color_rb[1])
-      attachments[i++] = __DRI_BUFFER_BACK_LEFT;
-   if (intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH))
-      attachments[i++] = __DRI_BUFFER_DEPTH;
-   if (intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL))
-      attachments[i++] = __DRI_BUFFER_STENCIL;
-
-   buffers = (*screen->dri2.loader->getBuffers)(drawable,
-						&drawable->w,
-						&drawable->h,
-						attachments, i,
-						&count,
-						drawable->loaderPrivate);
+   if (screen->dri2.loader
+       && (screen->dri2.loader->base.version > 2)
+       && (screen->dri2.loader->getBuffersWithFormat != NULL)) {
+      struct intel_renderbuffer *depth_rb;
+      struct intel_renderbuffer *stencil_rb;
+
+      i = 0;
+      if ((intel->is_front_buffer_rendering ||
+	   intel->is_front_buffer_reading ||
+	   !intel_fb->color_rb[1])
+	   && intel_fb->color_rb[0]) {
+	 attachments[i++] = __DRI_BUFFER_FRONT_LEFT;
+	 attachments[i++] = intel_bits_per_pixel(intel_fb->color_rb[0]);
+      }
+
+      if (intel_fb->color_rb[1]) {
+	 attachments[i++] = __DRI_BUFFER_BACK_LEFT;
+	 attachments[i++] = intel_bits_per_pixel(intel_fb->color_rb[1]);
+      }
+
+      depth_rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH);
+      stencil_rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL);
+
+      if ((depth_rb != NULL) && (stencil_rb != NULL)) {
+	 attachments[i++] = __DRI_BUFFER_DEPTH_STENCIL;
+	 attachments[i++] = intel_bits_per_pixel(depth_rb);
+      } else if (depth_rb != NULL) {
+	 attachments[i++] = __DRI_BUFFER_DEPTH;
+	 attachments[i++] = intel_bits_per_pixel(depth_rb);
+      } else if (stencil_rb != NULL) {
+	 attachments[i++] = __DRI_BUFFER_STENCIL;
+	 attachments[i++] = intel_bits_per_pixel(stencil_rb);
+      }
+
+      buffers =
+	 (*screen->dri2.loader->getBuffersWithFormat)(drawable,
+						      &drawable->w,
+						      &drawable->h,
+						      attachments, i / 2,
+						      &count,
+						      drawable->loaderPrivate);
+   } else if (screen->dri2.loader) {
+      i = 0;
+      if (intel_fb->color_rb[0])
+	 attachments[i++] = __DRI_BUFFER_FRONT_LEFT;
+      if (intel_fb->color_rb[1])
+	 attachments[i++] = __DRI_BUFFER_BACK_LEFT;
+      if (intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH))
+	 attachments[i++] = __DRI_BUFFER_DEPTH;
+      if (intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL))
+	 attachments[i++] = __DRI_BUFFER_STENCIL;
+
+      buffers = (*screen->dri2.loader->getBuffers)(drawable,
+						   &drawable->w,
+						   &drawable->h,
+						   attachments, i,
+						   &count,
+						   drawable->loaderPrivate);
+   }
 
    if (buffers == NULL)
       return;
@@ -259,6 +312,11 @@ intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
 	   region_name = "dri2 front buffer";
 	   break;
 
+       case __DRI_BUFFER_FAKE_FRONT_LEFT:
+	   rb = intel_fb->color_rb[0];
+	   region_name = "dri2 fake front buffer";
+	   break;
+
        case __DRI_BUFFER_BACK_LEFT:
 	   rb = intel_fb->color_rb[1];
 	   region_name = "dri2 back buffer";
@@ -269,6 +327,11 @@ intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
 	   region_name = "dri2 depth buffer";
 	   break;
 
+       case __DRI_BUFFER_DEPTH_STENCIL:
+	   rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH);
+	   region_name = "dri2 depth / stencil buffer";
+	   break;
+
        case __DRI_BUFFER_STENCIL:
 	   rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL);
 	   region_name = "dri2 stencil buffer";
@@ -282,6 +345,9 @@ intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
 	   return;
        }
 
+       if (rb == NULL)
+	  continue;
+
        if (rb->region) {
 	  dri_bo_flink(rb->region->buffer, &name);
 	  if (name == buffers[i].name)
@@ -312,6 +378,23 @@ intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
 
        intel_renderbuffer_set_region(rb, region);
        intel_region_release(&region);
+
+       if (buffers[i].attachment == __DRI_BUFFER_DEPTH_STENCIL) {
+	  rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL);
+	  if (rb != NULL) {
+	     struct intel_region *stencil_region = NULL;
+
+	     if (rb->region) {
+		dri_bo_flink(rb->region->buffer, &name);
+		if (name == buffers[i].name)
+		   continue;
+	     }
+
+	     intel_region_reference(&stencil_region, region);
+	     intel_renderbuffer_set_region(rb, stencil_region);
+	     intel_region_release(&stencil_region);
+	  }
+       }
    }
 
    driUpdateFramebufferSize(&intel->ctx, drawable);
@@ -328,9 +411,20 @@ intel_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h)
     if (!driContext->driScreenPriv->dri2.enabled)
 	return;
 
-    intel_update_renderbuffers(driContext, driContext->driDrawablePriv);
-    if (driContext->driDrawablePriv != driContext->driReadablePriv)
-	intel_update_renderbuffers(driContext, driContext->driReadablePriv);
+    if (!intel->meta.internal_viewport_call && ctx->DrawBuffer->Name == 0) {
+       /* If we're rendering to the fake front buffer, make sure all the pending
+	* drawing has landed on the real front buffer.  Otherwise when we
+	* eventually get to DRI2GetBuffersWithFormat the stale real front
+	* buffer contents will get copied to the new fake front buffer.
+	*/
+       if (intel->is_front_buffer_rendering) {
+	  intel_flush(ctx, GL_FALSE);
+       }
+
+       intel_update_renderbuffers(driContext, driContext->driDrawablePriv);
+       if (driContext->driDrawablePriv != driContext->driReadablePriv)
+	  intel_update_renderbuffers(driContext, driContext->driReadablePriv);
+    }
 
     old_viewport = ctx->Driver.Viewport;
     ctx->Driver.Viewport = NULL;
@@ -340,112 +434,6 @@ intel_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h)
     ctx->Driver.Viewport = old_viewport;
 }
 
-/**
- * Extension strings exported by the intel driver.
- *
- * Extensions supported by all chips supported by i830_dri, i915_dri, or
- * i965_dri.
- */
-static const struct dri_extension card_extensions[] = {
-   { "GL_ARB_multisample",                GL_ARB_multisample_functions },
-   { "GL_ARB_multitexture",               NULL },
-   { "GL_ARB_point_parameters",           GL_ARB_point_parameters_functions },
-   { "GL_ARB_texture_border_clamp",       NULL },
-   { "GL_ARB_texture_compression",        GL_ARB_texture_compression_functions },
-   { "GL_ARB_texture_cube_map",           NULL },
-   { "GL_ARB_texture_env_add",            NULL },
-   { "GL_ARB_texture_env_combine",        NULL },
-   { "GL_ARB_texture_env_crossbar",       NULL },
-   { "GL_ARB_texture_env_dot3",           NULL },
-   { "GL_ARB_texture_mirrored_repeat",    NULL },
-   { "GL_ARB_texture_rectangle",          NULL },
-   { "GL_ARB_vertex_buffer_object",       GL_ARB_vertex_buffer_object_functions },
-   { "GL_ARB_vertex_program",             GL_ARB_vertex_program_functions },
-   { "GL_ARB_window_pos",                 GL_ARB_window_pos_functions },
-   { "GL_EXT_blend_color",                GL_EXT_blend_color_functions },
-   { "GL_EXT_blend_equation_separate",    GL_EXT_blend_equation_separate_functions },
-   { "GL_EXT_blend_func_separate",        GL_EXT_blend_func_separate_functions },
-   { "GL_EXT_blend_minmax",               GL_EXT_blend_minmax_functions },
-   { "GL_EXT_blend_logic_op",             NULL },
-   { "GL_EXT_blend_subtract",             NULL },
-   { "GL_EXT_cull_vertex",                GL_EXT_cull_vertex_functions },
-   { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
-   { "GL_EXT_multi_draw_arrays",          GL_EXT_multi_draw_arrays_functions },
-   { "GL_EXT_packed_depth_stencil",       NULL },
-   { "GL_EXT_secondary_color",            GL_EXT_secondary_color_functions },
-   { "GL_EXT_stencil_wrap",               NULL },
-   { "GL_EXT_texture_edge_clamp",         NULL },
-   { "GL_EXT_texture_env_combine",        NULL },
-   { "GL_EXT_texture_env_dot3",           NULL },
-   { "GL_EXT_texture_filter_anisotropic", NULL },
-   { "GL_EXT_texture_lod_bias",           NULL },
-   { "GL_3DFX_texture_compression_FXT1",  NULL },
-   { "GL_APPLE_client_storage",           NULL },
-   { "GL_MESA_pack_invert",               NULL },
-   { "GL_MESA_ycbcr_texture",             NULL },
-   { "GL_NV_blend_square",                NULL },
-   { "GL_NV_point_sprite",                GL_NV_point_sprite_functions },
-   { "GL_NV_vertex_program",              GL_NV_vertex_program_functions },
-   { "GL_NV_vertex_program1_1",           NULL },
-   { "GL_SGIS_generate_mipmap",           NULL },
-   { NULL, NULL }
-};
-
-static const struct dri_extension brw_extensions[] = {
-   { "GL_ARB_depth_texture",              NULL },
-   { "GL_ARB_draw_buffers",               NULL },
-   { "GL_ARB_fragment_program",           NULL },
-   { "GL_ARB_fragment_program_shadow",    NULL },
-   { "GL_ARB_fragment_shader",            NULL },
-   { "GL_ARB_occlusion_query",            GL_ARB_occlusion_query_functions },
-   { "GL_ARB_point_sprite", 		  NULL },
-   { "GL_ARB_shader_objects",             GL_ARB_shader_objects_functions },
-   { "GL_ARB_shading_language_100",       GL_VERSION_2_0_functions },
-#if 0
-   /* Support for GLSL 1.20 is currently broken in core Mesa.
-    */
-   { "GL_ARB_shading_language_120",       GL_VERSION_2_1_functions },
-#endif
-   { "GL_ARB_shadow",                     NULL },
-   { "GL_ARB_texture_non_power_of_two",   NULL },
-   { "GL_ARB_vertex_shader",              GL_ARB_vertex_shader_functions },
-   { "GL_EXT_shadow_funcs",               NULL },
-   { "GL_EXT_texture_sRGB",		  NULL },
-   { "GL_ATI_separate_stencil",           GL_ATI_separate_stencil_functions },
-   { "GL_ATI_texture_env_combine3",       NULL },
-   { NULL,                                NULL }
-};
-
-static const struct dri_extension arb_oq_extensions[] = {
-   { NULL, NULL }
-};
-
-static const struct dri_extension ttm_extensions[] = {
-   { "GL_ARB_pixel_buffer_object",        NULL },
-   { "GL_EXT_framebuffer_object",         GL_EXT_framebuffer_object_functions },
-   { NULL, NULL }
-};
-
-/**
- * Initializes potential list of extensions if ctx == NULL, or actually enables
- * extensions for a context.
- */
-void intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging)
-{
-   struct intel_context *intel = ctx?intel_context(ctx):NULL;
-
-   /* Disable imaging extension until convolution is working in teximage paths.
-    */
-   enable_imaging = GL_FALSE;
-
-   driInitExtensions(ctx, card_extensions, enable_imaging);
-
-   if (intel == NULL || intel->ttm)
-      driInitExtensions(ctx, ttm_extensions, GL_FALSE);
-
-   if (intel == NULL || IS_965(intel->intelScreen->deviceID))
-      driInitExtensions(ctx, brw_extensions, GL_FALSE);
-}
 
 static const struct dri_debug_control debug_control[] = {
    { "tex",   DEBUG_TEXTURE},
@@ -496,9 +484,8 @@ intelInvalidateState(GLcontext * ctx, GLuint new_state)
       intel->vtbl.invalidate_state( intel, new_state );
 }
 
-
-void
-intelFlush(GLcontext * ctx)
+static void
+intel_flush(GLcontext *ctx, GLboolean needs_mi_flush)
 {
    struct intel_context *intel = intel_context(ctx);
 
@@ -512,10 +499,64 @@ intelFlush(GLcontext * ctx)
     * lands onscreen in a timely manner, even if the X Server doesn't trigger
     * a flush for us.
     */
-   intel_batchbuffer_emit_mi_flush(intel->batch);
+   if (needs_mi_flush)
+      intel_batchbuffer_emit_mi_flush(intel->batch);
 
    if (intel->batch->map != intel->batch->ptr)
       intel_batchbuffer_flush(intel->batch);
+
+   if ((ctx->DrawBuffer->Name == 0) && intel->front_buffer_dirty) {
+      __DRIscreen *const screen = intel->intelScreen->driScrnPriv;
+
+      if (screen->dri2.loader &&
+          (screen->dri2.loader->base.version >= 2)
+	  && (screen->dri2.loader->flushFrontBuffer != NULL)) {
+	 (*screen->dri2.loader->flushFrontBuffer)(intel->driDrawable,
+						  intel->driDrawable->loaderPrivate);
+
+	 /* Only clear the dirty bit if front-buffer rendering is no longer
+	  * enabled.  This is done so that the dirty bit can only be set in
+	  * glDrawBuffer.  Otherwise the dirty bit would have to be set at
+	  * each of N places that do rendering.  This has worse performances,
+	  * but it is much easier to get correct.
+	  */
+	 if (!intel->is_front_buffer_rendering) {
+	    intel->front_buffer_dirty = GL_FALSE;
+	 }
+      }
+   }
+}
+
+void
+intelFlush(GLcontext * ctx)
+{
+   intel_flush(ctx, GL_FALSE);
+}
+
+static void
+intel_glFlush(GLcontext *ctx)
+{
+   struct intel_context *intel = intel_context(ctx);
+
+   intel_flush(ctx, GL_TRUE);
+
+   /* We're using glFlush as an indicator that a frame is done, which is
+    * what DRI2 does before calling SwapBuffers (and means we should catch
+    * people doing front-buffer rendering, as well)..
+    *
+    * Wait for the swapbuffers before the one we just emitted, so we don't
+    * get too many swaps outstanding for apps that are GPU-heavy but not
+    * CPU-heavy.
+    *
+    * Unfortunately, we don't have a handle to the batch containing the swap,
+    * and getting our hands on that doesn't seem worth it, so we just us the
+    * first batch we emitted after the last swap.
+    */
+   if (intel->first_post_swapbuffers_batch != NULL) {
+      drm_intel_bo_wait_rendering(intel->first_post_swapbuffers_batch);
+      drm_intel_bo_unreference(intel->first_post_swapbuffers_batch);
+      intel->first_post_swapbuffers_batch = NULL;
+   }
 }
 
 void
@@ -531,7 +572,7 @@ intelFinish(GLcontext * ctx)
 
        irb = intel_renderbuffer(fb->_ColorDrawBuffers[i]);
 
-       if (irb->region)
+       if (irb && irb->region)
 	  dri_bo_wait_rendering(irb->region->buffer);
    }
    if (fb->_DepthBuffer) {
@@ -544,20 +585,26 @@ intelInitDriverFunctions(struct dd_function_table *functions)
 {
    _mesa_init_driver_functions(functions);
 
-   functions->Flush = intelFlush;
+   functions->Flush = intel_glFlush;
    functions->Finish = intelFinish;
    functions->GetString = intelGetString;
    functions->UpdateState = intelInvalidateState;
 
-   functions->CopyColorTable = _swrast_CopyColorTable;
-   functions->CopyColorSubTable = _swrast_CopyColorSubTable;
-   functions->CopyConvolutionFilter1D = _swrast_CopyConvolutionFilter1D;
-   functions->CopyConvolutionFilter2D = _swrast_CopyConvolutionFilter2D;
+   _MESA_INIT_COLORTABLE_FUNCTIONS(functions, _swrast_);
+   _MESA_INIT_CONVOLVE_FUNCTIONS(functions, _swrast_);
 
    intelInitTextureFuncs(functions);
+   intelInitTextureImageFuncs(functions);
+   intelInitTextureSubImageFuncs(functions);
+   intelInitTextureCopyImageFuncs(functions);
    intelInitStateFuncs(functions);
+   intelInitClearFuncs(functions);
    intelInitBufferFuncs(functions);
    intelInitPixelFuncs(functions);
+   intelInitBufferObjectFuncs(functions);
+   intel_init_syncobj_functions(functions);
+
+   intelInitEGLImageFuncs(functions);
 }
 
 
@@ -614,16 +661,20 @@ intelInitContext(struct intel_context *intel,
       }
    }
 
-   ctx->Const.MaxTextureMaxAnisotropy = 2.0;
-
    /* This doesn't yet catch all non-conformant rendering, but it's a
     * start.
     */
    if (getenv("INTEL_STRICT_CONFORMANCE")) {
-      intel->strict_conformance = 1;
+      unsigned int value = atoi(getenv("INTEL_STRICT_CONFORMANCE"));
+      if (value > 0) {
+         intel->conformance_mode = value;
+      }
+      else {
+         intel->conformance_mode = 1;
+      }
    }
 
-   if (intel->strict_conformance) {
+   if (intel->conformance_mode > 0) {
       ctx->Const.MinLineWidth = 1.0;
       ctx->Const.MinLineWidthAA = 1.0;
       ctx->Const.MaxLineWidth = 1.0;
@@ -649,7 +700,15 @@ intelInitContext(struct intel_context *intel,
     */
    _mesa_init_point(ctx);
 
+   meta_init_metaops(ctx, &intel->meta);
    ctx->Const.MaxColorAttachments = 4;  /* XXX FBO: review this */
+   if (IS_965(intelScreen->deviceID)) {
+      if (MAX_WIDTH > 8192)
+	 ctx->Const.MaxRenderbufferSize = 8192;
+   } else {
+      if (MAX_WIDTH > 2048)
+	 ctx->Const.MaxRenderbufferSize = 2048;
+   }
 
    /* Initialize the software rasterizer and helper modules. */
    _swrast_CreateContext(ctx);
@@ -661,6 +720,8 @@ intelInitContext(struct intel_context *intel,
    _swrast_allow_pixel_fog(ctx, GL_FALSE);
    _swrast_allow_vertex_fog(ctx, GL_TRUE);
 
+   _mesa_meta_init(ctx);
+
    intel->hw_stencil = mesaVis->stencilBits && mesaVis->depthBits == 24;
    intel->hw_stipple = 1;
 
@@ -691,8 +752,6 @@ intelInitContext(struct intel_context *intel,
 
    intel->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS);
 
-   _math_matrix_ctr(&intel->ViewportMatrix);
-
    if (IS_965(intelScreen->deviceID) && !intel->intelScreen->irq_active) {
       _mesa_printf("IRQs not active.  Exiting\n");
       exit(1);
@@ -709,7 +768,6 @@ intelInitContext(struct intel_context *intel,
 
    intel->batch = intel_batchbuffer_alloc(intel);
 
-   intel_bufferobj_init(intel);
    intel_fbo_init(intel);
 
    if (intel->ctx.Mesa_DXTn) {
@@ -719,6 +777,15 @@ intelInitContext(struct intel_context *intel,
    else if (driQueryOptionb(&intel->optionCache, "force_s3tc_enable")) {
       _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
    }
+   intel->use_texture_tiling = driQueryOptionb(&intel->optionCache,
+					       "texture_tiling");
+   if (intel->use_texture_tiling &&
+       !intel->intelScreen->kernel_exec_fencing) {
+      fprintf(stderr, "No kernel support for execution fencing, "
+	      "disabling texture tiling\n");
+      intel->use_texture_tiling = GL_FALSE;
+   }
+   intel->use_early_z = driQueryOptionb(&intel->optionCache, "early_z");
 
    intel->prim.primitive = ~0;
 
@@ -728,6 +795,16 @@ intelInitContext(struct intel_context *intel,
       intel->no_rast = 1;
    }
 
+   if (driQueryOptionb(&intel->optionCache, "always_flush_batch")) {
+      fprintf(stderr, "flushing batchbuffer before/after each draw call\n");
+      intel->always_flush_batch = 1;
+   }
+
+   if (driQueryOptionb(&intel->optionCache, "always_flush_cache")) {
+      fprintf(stderr, "flushing GPU caches before/after each draw call\n");
+      intel->always_flush_cache = 1;
+   }
+
    /* Disable all hardware rendering (skip emitting batches and fences/waits
     * to the kernel)
     */
@@ -748,6 +825,10 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv)
 
       INTEL_FIREVERTICES(intel);
 
+      _mesa_meta_free(&intel->ctx);
+
+      meta_destroy_metaops(&intel->meta);
+
       intel->vtbl.destroy(intel);
 
       release_texture_heaps = (intel->ctx.Shared->RefCount == 1);
@@ -765,15 +846,68 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv)
       intel->prim.vb = NULL;
       dri_bo_unreference(intel->prim.vb_bo);
       intel->prim.vb_bo = NULL;
+      dri_bo_unreference(intel->first_post_swapbuffers_batch);
+      intel->first_post_swapbuffers_batch = NULL;
 
       if (release_texture_heaps) {
-         /* This share group is about to go away, free our private
-          * texture object data.
+         /* Nothing is currently done here to free texture heaps;
+          * but we're not using the texture heap utilities, so I
+          * rather think we shouldn't.  I've taken a look, and can't
+          * find any private texture data hanging around anywhere, but
+          * I'm not yet certain there isn't any at all...
           */
-         if (INTEL_DEBUG & DEBUG_TEXTURE)
+         /* if (INTEL_DEBUG & DEBUG_TEXTURE)
             fprintf(stderr, "do something to free texture heaps\n");
+          */
       }
 
+      /* XXX In intelMakeCurrent() below, the context's static regions are 
+       * referenced inside the frame buffer; it's listed as a hack,
+       * with a comment of "XXX FBO temporary fix-ups!", but
+       * as long as it's there, we should release the regions here.
+       * The do/while loop around the block is used to allow the
+       * "continue" statements inside the block to exit the block,
+       * to avoid many layers of "if" constructs.
+       */
+      do {
+         __DRIdrawablePrivate * driDrawPriv = intel->driDrawable;
+         struct intel_framebuffer *intel_fb;
+         struct intel_renderbuffer *irbDepth, *irbStencil;
+         if (!driDrawPriv) {
+            /* We're already detached from the drawable; exit this block. */
+            continue;
+         }
+         intel_fb = (struct intel_framebuffer *) driDrawPriv->driverPrivate;
+         if (!intel_fb) {
+            /* The frame buffer is already gone; exit this block. */
+            continue;
+         }
+         irbDepth = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH);
+         irbStencil = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL);
+
+         /* If the regions of the frame buffer still match the regions
+          * of the context, release them.  If they've changed somehow,
+          * leave them alone.
+          */
+         if (intel_fb->color_rb[0] && intel_fb->color_rb[0]->region == intel->front_region) {
+	    intel_renderbuffer_set_region(intel_fb->color_rb[0], NULL);
+         }
+         if (intel_fb->color_rb[1] && intel_fb->color_rb[1]->region == intel->back_region) {
+	    intel_renderbuffer_set_region(intel_fb->color_rb[1], NULL);
+         }
+
+         if (irbDepth && irbDepth->region == intel->depth_region) {
+	    intel_renderbuffer_set_region(irbDepth, NULL);
+         }
+         /* Usually, the stencil buffer is the same as the depth buffer;
+          * but they're handled separately in MakeCurrent, so we'll
+          * handle them separately here.
+          */
+         if (irbStencil && irbStencil->region == intel->depth_region) {
+	    intel_renderbuffer_set_region(irbStencil, NULL);
+         }
+      } while (0);
+
       intel_region_release(&intel->front_region);
       intel_region_release(&intel->back_region);
       intel_region_release(&intel->depth_region);
@@ -782,6 +916,9 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv)
 
       /* free the Mesa context */
       _mesa_free_context_data(&intel->ctx);
+
+      FREE(intel);
+      driContextPriv->driverPrivate = NULL;
    }
 }
 
@@ -810,7 +947,10 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv,
           if (driDrawPriv != driReadPriv)
               intel_update_renderbuffers(driContextPriv, driReadPriv);
       } else {
-          /* XXX FBO temporary fix-ups! */
+          /* XXX FBO temporary fix-ups!  These are released in 
+           * intelDextroyContext(), above.  Changes here should be
+           * reflected there.
+           */
           /* if the renderbuffers don't have regions, init them from the context */
          struct intel_renderbuffer *irbDepth
             = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH);
@@ -858,6 +998,11 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv,
 		  ? driGetDefaultVBlankFlags(&intel->optionCache)
 		 : VBLANK_FLAG_NO_IRQ;
 
+	       /* Prevent error printf if one crtc is disabled, this will
+		* be properly calculated in intelWindowMoved() next.
+		*/
+		driDrawPriv->vblFlags = intelFixupVblank(intel, driDrawPriv);
+
 	       (*psp->systemTime->getUST) (&intel_fb->swap_ust);
 	       driDrawableInitVBlank(driDrawPriv);
 	       intel_fb->vbl_waited = driDrawPriv->vblSeq;
@@ -890,7 +1035,6 @@ intelContendedLock(struct intel_context *intel, GLuint flags)
    int me = intel->hHWContext;
 
    drmGetLock(intel->driFd, intel->hHWContext, flags);
-   intel->locked = 1;
 
    if (INTEL_DEBUG & DEBUG_LOCK)
       _mesa_printf("%s - got contended lock\n", __progname);
@@ -947,9 +1091,12 @@ void LOCK_HARDWARE( struct intel_context *intel )
     struct intel_framebuffer *intel_fb = NULL;
     struct intel_renderbuffer *intel_rb = NULL;
 
-    _glthread_LOCK_MUTEX(lockMutex);
-    assert(!intel->locked);
-    intel->locked = 1;
+    intel->locked++;
+    if (intel->locked >= 2)
+       return;
+
+    if (!sPriv->dri2.enabled)
+       _glthread_LOCK_MUTEX(lockMutex);
 
     if (intel->driDrawable) {
        intel_fb = intel->driDrawable->driverPrivate;
@@ -996,13 +1143,16 @@ void UNLOCK_HARDWARE( struct intel_context *intel )
 {
     __DRIscreen *sPriv = intel->driScreen;
 
-   intel->vtbl.note_unlock( intel );
-   intel->locked = 0;
+   intel->locked--;
+   if (intel->locked > 0)
+      return;
 
-   if (!sPriv->dri2.enabled)
-      DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext);
+   assert(intel->locked == 0);
 
-   _glthread_UNLOCK_MUTEX(lockMutex);
+   if (!sPriv->dri2.enabled) {
+      DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext);
+      _glthread_UNLOCK_MUTEX(lockMutex);
+   }
 
    if (INTEL_DEBUG & DEBUG_LOCK)
       _mesa_printf("%s - unlocked\n", __progname);
diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h
index 048286c196..03e7cf39d6 100644
--- a/src/mesa/drivers/dri/intel/intel_context.h
+++ b/src/mesa/drivers/dri/intel/intel_context.h
@@ -33,6 +33,7 @@
 #include "main/mtypes.h"
 #include "main/mm.h"
 #include "texmem.h"
+#include "dri_metaops.h"
 #include "drm.h"
 #include "intel_bufmgr.h"
 
@@ -48,6 +49,8 @@
 #define DV_PF_555  (1<<8)
 #define DV_PF_565  (2<<8)
 #define DV_PF_8888 (3<<8)
+#define DV_PF_4444 (8<<8)
+#define DV_PF_1555 (9<<8)
 
 struct intel_region;
 struct intel_context;
@@ -77,9 +80,19 @@ extern void intelFallback(struct intel_context *intel, GLuint bit,
 
 #define INTEL_MAX_FIXUP 64
 
+struct intel_sync_object {
+   struct gl_sync_object Base;
+
+   /** Batch associated with this sync object */
+   drm_intel_bo *bo;
+};
+
+/**
+ * intel_context is derived from Mesa's context class: GLcontext.
+ */
 struct intel_context
 {
-   GLcontext ctx;               /* the parent class */
+   GLcontext ctx;  /**< base class, must be first field */
 
    struct
    {
@@ -89,7 +102,6 @@ struct intel_context
       void (*new_batch) (struct intel_context * intel);
       void (*emit_invarient_state) (struct intel_context * intel);
       void (*note_fence) (struct intel_context *intel, GLuint fence);
-      void (*note_unlock) (struct intel_context *intel);
       void (*update_texture_state) (struct intel_context * intel);
 
       void (*render_start) (struct intel_context * intel);
@@ -100,7 +112,6 @@ struct intel_context
 			       GLuint num_regions);
 
       GLuint (*flush_cmd) (void);
-      void (*emit_flush) (struct intel_context *intel, GLuint unused);
 
       void (*reduced_primitive_state) (struct intel_context * intel,
                                        GLenum rprim);
@@ -157,18 +168,7 @@ struct intel_context
       void (*debug_batch)(struct intel_context *intel);
    } vtbl;
 
-   struct {
-      struct gl_fragment_program *bitmap_fp;
-      struct gl_vertex_program *passthrough_vp;
-
-      struct gl_fragment_program *saved_fp;
-      GLboolean saved_fp_enable;
-      struct gl_vertex_program *saved_vp;
-      GLboolean saved_vp_enable;
-
-      GLint saved_vp_x, saved_vp_y;
-      GLsizei saved_vp_width, saved_vp_height;
-   } meta;
+   struct dri_metaops meta;
 
    GLint refcount;
    GLuint Fallback;
@@ -188,6 +188,7 @@ struct intel_context
    GLboolean ttm;
 
    struct intel_batchbuffer *batch;
+   drm_intel_bo *first_post_swapbuffers_batch;
    GLboolean no_batch_wrap;
    unsigned batch_id;
 
@@ -209,10 +210,10 @@ struct intel_context
    char *prevLockFile;
    int prevLockLine;
 
-   GLubyte clear_chan[4];
    GLuint ClearColor565;
    GLuint ClearColor8888;
 
+
    /* Offsets of fields within the current vertex:
     */
    GLuint coloroffset;
@@ -229,7 +230,14 @@ struct intel_context
    GLboolean hw_stipple;
    GLboolean depth_buffer_is_float;
    GLboolean no_rast;
-   GLboolean strict_conformance;
+   GLboolean always_flush_batch;
+   GLboolean always_flush_cache;
+
+   /* 0 - nonconformant, best performance;
+    * 1 - fallback to sw for known conformance bugs
+    * 2 - always fallback to sw
+    */
+   GLuint conformance_mode;
 
    /* State for intelvb.c and inteltris.c.
     */
@@ -255,11 +263,40 @@ struct intel_context
     * flush time while the lock is held.
     */
    GLboolean constant_cliprect;
+
    /**
     * In !constant_cliprect mode, set to true if the front cliprects should be
     * used instead of back.
     */
    GLboolean front_cliprects;
+
+   /**
+    * Set if rendering has occured to the drawable's front buffer.
+    *
+    * This is used in the DRI2 case to detect that glFlush should also copy
+    * the contents of the fake front buffer to the real front buffer.
+    */
+   GLboolean front_buffer_dirty;
+
+   /**
+    * Track whether front-buffer rendering is currently enabled
+    *
+    * A separate flag is used to track this in order to support MRT more
+    * easily.
+    */
+   GLboolean is_front_buffer_rendering;
+   /**
+    * Track whether front-buffer is the current read target.
+    *
+    * This is closely associated with is_front_buffer_rendering, but may
+    * be set separately.  The DRI2 fake front buffer must be referenced
+    * either way.
+    */
+   GLboolean is_front_buffer_reading;
+
+   GLboolean use_texture_tiling;
+   GLboolean use_early_z;
+
    drm_clip_rect_t fboRect;     /**< cliprect for FBO rendering */
 
    int perf_boxes;
@@ -281,7 +318,7 @@ struct intel_context
    __DRIdrawablePrivate *driReadDrawable;
    __DRIscreenPrivate *driScreen;
    intelScreenPrivate *intelScreen;
-   volatile struct drm_i915_sarea *sarea;
+   volatile drm_i915_sarea_t *sarea;
 
    GLuint lastStamp;
 
@@ -312,6 +349,7 @@ extern char *__progname;
 
 #define ARRAY_SIZE(x) (sizeof(x) / sizeof(x[0]))
 #define ALIGN(value, alignment)  ((value + alignment - 1) & ~(alignment - 1))
+#define IS_POWER_OF_TWO(val) (((val) & (val - 1)) == 0)
 
 #define INTEL_FIREVERTICES(intel)		\
 do {						\
@@ -437,7 +475,8 @@ extern void intelFinish(GLcontext * ctx);
 extern void intelFlush(GLcontext * ctx);
 
 extern void intelInitDriverFunctions(struct dd_function_table *functions);
-extern void intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging);
+
+void intel_init_syncobj_functions(struct dd_function_table *functions);
 
 
 /* ================================================================
@@ -514,6 +553,9 @@ void intel_viewport(GLcontext * ctx, GLint x, GLint y,
 void intel_update_renderbuffers(__DRIcontext *context,
 				__DRIdrawable *drawable);
 
+void i915_set_buf_info_for_region(uint32_t *state, struct intel_region *region,
+				  uint32_t buffer_id);
+
 /*======================================================================
  * Inline conversion functions.  
  * These are better-typed than the macros used previously:
@@ -524,4 +566,10 @@ intel_context(GLcontext * ctx)
    return (struct intel_context *) ctx;
 }
 
+static INLINE GLboolean
+is_power_of_two(uint32_t value)
+{
+   return (value & (value - 1)) == 0;
+}
+
 #endif
diff --git a/src/mesa/drivers/dri/intel/intel_decode.c b/src/mesa/drivers/dri/intel/intel_decode.c
index 5f90ca22ec..a9dfe281cb 100644
--- a/src/mesa/drivers/dri/intel/intel_decode.c
+++ b/src/mesa/drivers/dri/intel/intel_decode.c
@@ -87,27 +87,28 @@ decode_mi(uint32_t *data, int count, uint32_t hw_offset, int *failures)
 
     struct {
 	uint32_t opcode;
+	int len_mask;
 	int min_len;
 	int max_len;
 	char *name;
     } opcodes_mi[] = {
-	{ 0x08, 1, 1, "MI_ARB_ON_OFF" },
-	{ 0x0a, 1, 1, "MI_BATCH_BUFFER_END" },
-	{ 0x31, 2, 2, "MI_BATCH_BUFFER_START" },
-	{ 0x14, 3, 3, "MI_DISPLAY_BUFFER_INFO" },
-	{ 0x04, 1, 1, "MI_FLUSH" },
-	{ 0x22, 3, 3, "MI_LOAD_REGISTER_IMM" },
-	{ 0x13, 2, 2, "MI_LOAD_SCAN_LINES_EXCL" },
-	{ 0x12, 2, 2, "MI_LOAD_SCAN_LINES_INCL" },
-	{ 0x00, 1, 1, "MI_NOOP" },
-	{ 0x11, 2, 2, "MI_OVERLAY_FLIP" },
-	{ 0x07, 1, 1, "MI_REPORT_HEAD" },
-	{ 0x18, 2, 2, "MI_SET_CONTEXT" },
-	{ 0x20, 3, 4, "MI_STORE_DATA_IMM" },
-	{ 0x21, 3, 4, "MI_STORE_DATA_INDEX" },
-	{ 0x24, 3, 3, "MI_STORE_REGISTER_MEM" },
-	{ 0x02, 1, 1, "MI_USER_INTERRUPT" },
-	{ 0x03, 1, 1, "MI_WAIT_FOR_EVENT" },
+	{ 0x08, 0, 1, 1, "MI_ARB_ON_OFF" },
+	{ 0x0a, 0, 1, 1, "MI_BATCH_BUFFER_END" },
+	{ 0x31, 0x3f, 2, 2, "MI_BATCH_BUFFER_START" },
+	{ 0x14, 0x3f, 3, 3, "MI_DISPLAY_BUFFER_INFO" },
+	{ 0x04, 0, 1, 1, "MI_FLUSH" },
+	{ 0x22, 0, 3, 3, "MI_LOAD_REGISTER_IMM" },
+	{ 0x13, 0x3f, 2, 2, "MI_LOAD_SCAN_LINES_EXCL" },
+	{ 0x12, 0x3f, 2, 2, "MI_LOAD_SCAN_LINES_INCL" },
+	{ 0x00, 0, 1, 1, "MI_NOOP" },
+	{ 0x11, 0x3f, 2, 2, "MI_OVERLAY_FLIP" },
+	{ 0x07, 0, 1, 1, "MI_REPORT_HEAD" },
+	{ 0x18, 0x3f, 2, 2, "MI_SET_CONTEXT" },
+	{ 0x20, 0x3f, 3, 4, "MI_STORE_DATA_IMM" },
+	{ 0x21, 0x3f, 3, 4, "MI_STORE_DATA_INDEX" },
+	{ 0x24, 0x3f, 3, 3, "MI_STORE_REGISTER_MEM" },
+	{ 0x02, 0, 1, 1, "MI_USER_INTERRUPT" },
+	{ 0x03, 0, 1, 1, "MI_WAIT_FOR_EVENT" },
     };
 
 
@@ -118,12 +119,14 @@ decode_mi(uint32_t *data, int count, uint32_t hw_offset, int *failures)
 
 	    instr_out(data, hw_offset, 0, "%s\n", opcodes_mi[opcode].name);
 	    if (opcodes_mi[opcode].max_len > 1) {
-		len = (data[0] & 0x000000ff) + 2;
+		len = (data[0] & opcodes_mi[opcode].len_mask) + 2;
 		if (len < opcodes_mi[opcode].min_len ||
 		    len > opcodes_mi[opcode].max_len)
 		{
-		    fprintf(out, "Bad length in %s\n",
-			    opcodes_mi[opcode].name);
+		    fprintf(out, "Bad length (%d) in %s, [%d, %d]\n",
+			    len, opcodes_mi[opcode].name,
+			    opcodes_mi[opcode].min_len,
+			    opcodes_mi[opcode].max_len);
 		}
 	    }
 
@@ -797,6 +800,7 @@ static int
 decode_3d_1d(uint32_t *data, int count, uint32_t hw_offset, int *failures, int i830)
 {
     unsigned int len, i, c, opcode, word, map, sampler, instr;
+    char *format;
 
     struct {
 	uint32_t opcode;
@@ -932,7 +936,7 @@ decode_3d_1d(uint32_t *data, int count, uint32_t hw_offset, int *failures, int i
 	instr_out(data, hw_offset, 0, "3DSTATE_PIXEL_SHADER_CONSTANTS\n");
 	len = (data[0] & 0x000000ff) + 2;
 
-	i = 1;
+	i = 2;
 	for (c = 0; c <= 31; c++) {
 	    if (data[1] & (1 << c)) {
 		if (i + 4 >= count)
@@ -952,7 +956,7 @@ decode_3d_1d(uint32_t *data, int count, uint32_t hw_offset, int *failures, int i
 	    }
 	}
 	if (len != i) {
-	    fprintf(out, "Bad count in 3DSTATE_MAP_STATE\n");
+	    fprintf(out, "Bad count in 3DSTATE_PIXEL_SHADER_CONSTANTS\n");
 	    (*failures)++;
 	}
 	return len;
@@ -998,6 +1002,35 @@ decode_3d_1d(uint32_t *data, int count, uint32_t hw_offset, int *failures, int i
 	    (*failures)++;
 	}
 	return len;
+    case 0x85:
+	len = (data[0] & 0x0000000f) + 2;
+
+	if (len != 2)
+	    fprintf(out, "Bad count in 3DSTATE_DEST_BUFFER_VARIABLES\n");
+	if (count < 2)
+	    BUFFER_FAIL(count, len, "3DSTATE_DEST_BUFFER_VARIABLES");
+
+	instr_out(data, hw_offset, 0,
+		  "3DSTATE_DEST_BUFFER_VARIABLES\n");
+
+	switch ((data[1] >> 8) & 0xf) {
+	case 0x0: format = "g8"; break;
+	case 0x1: format = "x1r5g5b5"; break;
+	case 0x2: format = "r5g6b5"; break;
+	case 0x3: format = "a8r8g8b8"; break;
+	case 0x4: format = "ycrcb_swapy"; break;
+	case 0x5: format = "ycrcb_normal"; break;
+	case 0x6: format = "ycrcb_swapuv"; break;
+	case 0x7: format = "ycrcb_swapuvy"; break;
+	case 0x8: format = "a4r4g4b4"; break;
+	case 0x9: format = "a1r5g5b5"; break;
+	case 0xa: format = "a2r10g10b10"; break;
+	default: format = "BAD"; break;
+	}
+	instr_out(data, hw_offset, 1, "%s format, early Z %sabled\n",
+		  format,
+		  (data[1] & (1 << 31)) ? "en" : "dis");
+	return len;
     }
 
     for (opcode = 0; opcode < sizeof(opcodes_3d_1d) / sizeof(opcodes_3d_1d[0]);
@@ -1510,7 +1543,7 @@ decode_3d_965(uint32_t *data, int count, uint32_t hw_offset, int *failures)
 
 	for (i = 1; i < len;) {
 	    instr_out(data, hw_offset, i, "buffer %d: %svalid, type 0x%04x, "
-		      "src offset 0x%04xd bytes\n",
+		      "src offset 0x%04x bytes\n",
 		      data[i] >> 27,
 		      data[i] & (1 << 26) ? "" : "in",
 		      (data[i] >> 16) & 0x1ff,
@@ -1592,7 +1625,7 @@ decode_3d_965(uint32_t *data, int count, uint32_t hw_offset, int *failures)
 		  "3DPRIMITIVE: %s %s\n",
 		  get_965_prim_type(data[0]),
 		  (data[0] & (1 << 15)) ? "random" : "sequential");
-	instr_out(data, hw_offset, 1, "primitive count\n");
+	instr_out(data, hw_offset, 1, "vertex count\n");
 	instr_out(data, hw_offset, 2, "start vertex\n");
 	instr_out(data, hw_offset, 3, "instance count\n");
 	instr_out(data, hw_offset, 4, "start instance\n");
diff --git a/src/mesa/drivers/dri/intel/intel_depthstencil.c b/src/mesa/drivers/dri/intel/intel_depthstencil.c
deleted file mode 100644
index 354b3bf0d7..0000000000
--- a/src/mesa/drivers/dri/intel/intel_depthstencil.c
+++ /dev/null
@@ -1,261 +0,0 @@
-/**************************************************************************
- * 
- * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
- * All Rights Reserved.
- * 
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- * 
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- * 
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- * 
- **************************************************************************/
-
-#include "main/glheader.h"
-#include "main/imports.h"
-#include "main/context.h"
-#include "main/depthstencil.h"
-#include "main/fbobject.h"
-#include "main/framebuffer.h"
-#include "main/hash.h"
-#include "main/mtypes.h"
-#include "main/renderbuffer.h"
-
-#include "intel_context.h"
-#include "intel_fbo.h"
-#include "intel_depthstencil.h"
-#include "intel_regions.h"
-#include "intel_span.h"
-
-/**
- * The GL_EXT_framebuffer_object allows the user to create their own
- * framebuffer objects consisting of color renderbuffers (0 or more),
- * depth renderbuffers (0 or 1) and stencil renderbuffers (0 or 1).
- *
- * The spec considers depth and stencil renderbuffers to be totally independent
- * buffers.  In reality, most graphics hardware today uses a combined
- * depth+stencil buffer (one 32-bit pixel = 24 bits of Z + 8 bits of stencil).
- *
- * This causes difficulty because the user may create some number of depth
- * renderbuffers and some number of stencil renderbuffers and bind them
- * together in framebuffers in any combination.
- *
- * This code manages all that.
- *
- * 1. Depth renderbuffers are always allocated in hardware as 32bpp
- *    GL_DEPTH24_STENCIL8 buffers.
- *
- * 2. Stencil renderbuffers are initially allocated in software as 8bpp
- *    GL_STENCIL_INDEX8 buffers.
- *
- * 3. Depth and Stencil renderbuffers use the PairedStencil and PairedDepth
- *    fields (respectively) to indicate if the buffer's currently paired
- *    with another stencil or depth buffer (respectively).
- *
- * 4. When a depth and stencil buffer are initially both attached to the
- *    current framebuffer, we merge the stencil buffer values into the
- *    depth buffer (really a depth+stencil buffer).  The then hardware uses
- *    the combined buffer.
- *
- * 5. Whenever a depth or stencil buffer is reallocated (with
- *    glRenderbufferStorage) we undo the pairing and copy the stencil values
- *    from the combined depth/stencil buffer back to the stencil-only buffer.
- *
- * 6. We also undo the pairing when we find a change in buffer bindings.
- *
- * 7. If a framebuffer is only using a depth renderbuffer (no stencil), we
- *    just use the combined depth/stencil buffer and ignore the stencil values.
- *
- * 8. If a framebuffer is only using a stencil renderbuffer (no depth) we have
- *    to promote the 8bpp software stencil buffer to a 32bpp hardware
- *    depth+stencil buffer.
- *
- */
-
-/**
- * Undo the pairing/interleaving between depth and stencil buffers.
- * irb should be a depth/stencil or stencil renderbuffer.
- */
-void
-intel_unpair_depth_stencil(GLcontext *ctx, struct intel_renderbuffer *irb)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct gl_renderbuffer *rb = &irb->Base;
-
-   if (irb->PairedStencil) {
-      /* irb is a depth/stencil buffer */
-      struct gl_renderbuffer *stencilRb;
-      struct intel_renderbuffer *stencilIrb;
-
-      ASSERT(rb->_ActualFormat == GL_DEPTH24_STENCIL8_EXT);
-
-      stencilRb = _mesa_lookup_renderbuffer(ctx, irb->PairedStencil);
-      stencilIrb = intel_renderbuffer(stencilRb);
-      if (stencilIrb) {
-         /* need to extract stencil values from the depth buffer */
-	 ASSERT(stencilIrb->PairedDepth == rb->Name);
-	 intel_renderbuffer_map(intel, rb);
-	 intel_renderbuffer_map(intel, stencilRb);
-#if 0
-         /* disable for now */
-	 _mesa_extract_stencil(ctx, rb, stencilRb);
-#endif
-	 intel_renderbuffer_unmap(intel, stencilRb);
-	 intel_renderbuffer_unmap(intel, rb);
-         stencilIrb->PairedDepth = 0;
-      }
-      irb->PairedStencil = 0;
-   }
-   else if (irb->PairedDepth) {
-      /* irb is a stencil buffer */
-      struct gl_renderbuffer *depthRb;
-      struct intel_renderbuffer *depthIrb;
-
-      ASSERT(rb->_ActualFormat == GL_STENCIL_INDEX8_EXT ||
-             rb->_ActualFormat == GL_DEPTH24_STENCIL8_EXT);
-
-      depthRb = _mesa_lookup_renderbuffer(ctx, irb->PairedDepth);
-      depthIrb = intel_renderbuffer(depthRb);
-      if (depthIrb) {
-         /* need to extract stencil values from the depth buffer */
-	 ASSERT(depthIrb->PairedStencil == rb->Name);
-	 intel_renderbuffer_map(intel, rb);
-	 intel_renderbuffer_map(intel, depthRb);
-#if 0
-         /* disable for now */
-	 _mesa_extract_stencil(ctx, depthRb, rb);
-#endif
-	 intel_renderbuffer_unmap(intel, depthRb);
-	 intel_renderbuffer_unmap(intel, rb);
-         depthIrb->PairedStencil = 0;
-      }
-      irb->PairedDepth = 0;
-   }
-   else {
-      _mesa_problem(ctx, "Problem in undo_depth_stencil_pairing");
-   }
-
-   ASSERT(irb->PairedStencil == 0);
-   ASSERT(irb->PairedDepth == 0);
-}
-
-
-/**
- * Examine the depth and stencil renderbuffers which are attached to the
- * framebuffer.  If both depth and stencil are attached, make sure that the
- * renderbuffers are 'paired' (combined).  If only depth or only stencil is
- * attached, undo any previous pairing.
- *
- * Must be called if NewState & _NEW_BUFFER (when renderbuffer attachments
- * change, for example).
- */
-void
-intel_validate_paired_depth_stencil(GLcontext * ctx,
-                                    struct gl_framebuffer *fb)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct intel_renderbuffer *depthRb, *stencilRb;
-
-   depthRb = intel_get_renderbuffer(fb, BUFFER_DEPTH);
-   stencilRb = intel_get_renderbuffer(fb, BUFFER_STENCIL);
-
-   if (depthRb && stencilRb) {
-      if (depthRb == stencilRb) {
-         /* Using a user-created combined depth/stencil buffer.
-          * Nothing to do.
-          */
-         ASSERT(depthRb->Base._BaseFormat == GL_DEPTH_STENCIL_EXT);
-         ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT);
-      }
-      else {
-         /* Separate depth/stencil buffers, need to interleave now */
-         ASSERT(depthRb->Base._BaseFormat == GL_DEPTH_COMPONENT ||
-                depthRb->Base._BaseFormat == GL_DEPTH_STENCIL);
-         ASSERT(stencilRb->Base._BaseFormat == GL_STENCIL_INDEX ||
-                stencilRb->Base._BaseFormat == GL_DEPTH_STENCIL);
-
-         /* may need to interleave depth/stencil now */
-         if (depthRb->PairedStencil == stencilRb->Base.Name) {
-            /* OK, the depth and stencil buffers are already interleaved */
-            ASSERT(stencilRb->PairedDepth == depthRb->Base.Name);
-         }
-         else {
-            /* need to setup new pairing/interleaving */
-            if (depthRb->PairedStencil) {
-               intel_unpair_depth_stencil(ctx, depthRb);
-            }
-            if (stencilRb->PairedDepth) {
-               intel_unpair_depth_stencil(ctx, stencilRb);
-            }
-
-            ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT);
-            ASSERT(stencilRb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT ||
-                   stencilRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT);
-
-            /* establish new pairing: interleave stencil into depth buffer */
-	    intel_renderbuffer_map(intel, &depthRb->Base);
-	    intel_renderbuffer_map(intel, &stencilRb->Base);
-            _mesa_insert_stencil(ctx, &depthRb->Base, &stencilRb->Base);
-	    intel_renderbuffer_unmap(intel, &stencilRb->Base);
-	    intel_renderbuffer_unmap(intel, &depthRb->Base);
-            depthRb->PairedStencil = stencilRb->Base.Name;
-            stencilRb->PairedDepth = depthRb->Base.Name;
-         }
-
-      }
-   }
-   else if (depthRb) {
-      /* Depth buffer but no stencil buffer.
-       * We'll use a GL_DEPTH24_STENCIL8 buffer and ignore the stencil bits.
-       */
-      /* can't assert this until storage is allocated:
-         ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT);
-       */
-      /* intel_undo any previous pairing */
-      if (depthRb->PairedStencil) {
-         intel_unpair_depth_stencil(ctx, depthRb);
-      }
-   }
-   else if (stencilRb) {
-      /* Stencil buffer but no depth buffer.
-       * Since h/w doesn't typically support just 8bpp stencil w/out Z,
-       * we'll use a GL_DEPTH24_STENCIL8 buffer and ignore the depth bits.
-       */
-      /* undo any previous pairing */
-      if (stencilRb->PairedDepth) {
-         intel_unpair_depth_stencil(ctx, stencilRb);
-      }
-      if (stencilRb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT) {
-         /* promote buffer to GL_DEPTH24_STENCIL8 for hw rendering */
-         _mesa_promote_stencil(ctx, &stencilRb->Base);
-         ASSERT(stencilRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT);
-      }
-   }
-
-   /* Finally, update the fb->_DepthBuffer and fb->_StencilBuffer fields */
-   _mesa_update_depth_buffer(ctx, fb, BUFFER_DEPTH);
-   if (depthRb && depthRb->PairedStencil)
-      _mesa_update_stencil_buffer(ctx, fb, BUFFER_DEPTH);
-   else
-      _mesa_update_stencil_buffer(ctx, fb, BUFFER_STENCIL);
-
-
-   /* The hardware should use fb->Attachment[BUFFER_DEPTH].Renderbuffer
-    * first, if present, then fb->Attachment[BUFFER_STENCIL].Renderbuffer
-    * if present.
-    */
-}
diff --git a/src/mesa/drivers/dri/intel/intel_depthstencil.h b/src/mesa/drivers/dri/intel/intel_depthstencil.h
deleted file mode 100644
index 740eb0d989..0000000000
--- a/src/mesa/drivers/dri/intel/intel_depthstencil.h
+++ /dev/null
@@ -1,15 +0,0 @@
-
-#ifndef INTEL_DEPTH_STENCIL_H
-#define INTEL_DEPTH_STENCIL_H
-
-#include "intel_fbo.h"
-
-extern void
-intel_unpair_depth_stencil(GLcontext * ctx, struct intel_renderbuffer *irb);
-
-extern void
-intel_validate_paired_depth_stencil(GLcontext * ctx,
-                                    struct gl_framebuffer *fb);
-
-
-#endif /* INTEL_DEPTH_STENCIL_H */
diff --git a/src/mesa/drivers/dri/intel/intel_depthtmp.h b/src/mesa/drivers/dri/intel/intel_depthtmp.h
new file mode 100644
index 0000000000..16d7708453
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_depthtmp.h
@@ -0,0 +1,54 @@
+/*
+ * Copyright © 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Eric Anholt <eric@anholt.net>
+ *
+ */
+
+/**
+ * Wrapper around the depthtmp.h macrofest to generate spans code for
+ * all the tiling styles.
+ */
+
+#define VALUE_TYPE INTEL_VALUE_TYPE
+#define WRITE_DEPTH(_x, _y, d) INTEL_WRITE_DEPTH(NO_TILE(_x, _y), d)
+#define READ_DEPTH(d, _x, _y) d = INTEL_READ_DEPTH(NO_TILE(_x, _y))
+#define TAG(x) INTEL_TAG(intel##x)
+#include "depthtmp.h"
+
+#define VALUE_TYPE INTEL_VALUE_TYPE
+#define WRITE_DEPTH(_x, _y, d) INTEL_WRITE_DEPTH(X_TILE(_x, _y), d)
+#define READ_DEPTH(d, _x, _y) d = INTEL_READ_DEPTH(X_TILE(_x, _y))
+#define TAG(x) INTEL_TAG(intel_XTile_##x)
+#include "depthtmp.h"
+
+#define VALUE_TYPE INTEL_VALUE_TYPE
+#define WRITE_DEPTH(_x, _y, d) INTEL_WRITE_DEPTH(Y_TILE(_x, _y), d)
+#define READ_DEPTH(d, _x, _y) d = INTEL_READ_DEPTH(Y_TILE(_x, _y))
+#define TAG(x) INTEL_TAG(intel_YTile_##x)
+#include "depthtmp.h"
+
+#undef INTEL_VALUE_TYPE
+#undef INTEL_WRITE_DEPTH
+#undef INTEL_READ_DEPTH
+#undef INTEL_TAG
diff --git a/src/mesa/drivers/dri/intel/intel_eglimage.c b/src/mesa/drivers/dri/intel/intel_eglimage.c
new file mode 100644
index 0000000000..fed8675c26
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_eglimage.c
@@ -0,0 +1,139 @@
+/**************************************************************************
+ * 
+ * Copyright (C) 2009 Chia-I Wu <olvaffe@gmail.com>
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "main/mtypes.h"
+#include "main/teximage.h"
+
+#include "intel_context.h"
+#include "intel_fbo.h"
+#include "intel_tex.h"
+
+#include "intel_eglimage.h"
+#include "main/eglimage.h"
+
+#include "EGL/internal/eglimage_dri.h"
+
+
+#if FEATURE_OES_EGL_image
+
+
+/* move to intel_fbo.c */
+static void
+copy_renderbuffer(struct intel_renderbuffer *dst,
+                  struct intel_renderbuffer *src)
+{
+   dst->Base.Width          = src->Base.Width;
+   dst->Base.Height         = src->Base.Height;
+   dst->Base.InternalFormat = src->Base.InternalFormat;
+   dst->Base._ActualFormat  = src->Base._ActualFormat;
+   dst->Base._BaseFormat    = src->Base._BaseFormat;
+
+   dst->Base.ColorEncoding  = src->Base.ColorEncoding;
+   dst->Base.ComponentType  = src->Base.ComponentType;
+
+   dst->Base.RedBits        = src->Base.RedBits;
+   dst->Base.GreenBits      = src->Base.GreenBits;
+   dst->Base.BlueBits       = src->Base.BlueBits;
+   dst->Base.AlphaBits      = src->Base.AlphaBits;
+   dst->Base.IndexBits      = src->Base.IndexBits;
+   dst->Base.DepthBits      = src->Base.DepthBits;
+   dst->Base.StencilBits    = src->Base.StencilBits;
+   dst->Base.NumSamples     = src->Base.NumSamples;
+
+   dst->Base.DataType       = src->Base.DataType;
+
+   intel_renderbuffer_set_region(dst, src->region);
+}
+
+static __DRIEGLImage *
+get_image(GLcontext *ctx, GLeglImageOES image)
+{
+   __DRIEGLImage *driImage;
+
+   driImage = _eglClientGetImageData((__DRIEGLImageHandle) image);
+   if (!driImage || driImage->magic != __DRI_EGL_IMAGE_MAGIC) {
+      _mesa_error(ctx, GL_INVALID_VALUE,
+                  "glEGLImageTargetTexture2DOES(image=0x%x)", image);
+      return NULL;
+   }
+
+   return driImage;
+}
+
+void
+intelEGLImageTargetRenderbufferStorage(GLcontext *ctx,
+                                          struct gl_renderbuffer *rb,
+                                          GLeglImageOES image)
+{
+   struct intel_renderbuffer *irb = intel_renderbuffer(rb);
+   __DRIEGLImage *driImage;
+   struct intel_framebuffer *image_fb;
+   struct intel_renderbuffer *image_rb;
+
+   driImage = get_image(ctx, image);
+   if (!driImage)
+      return;
+
+   image_fb = driImage->drawable->driverPrivate;
+   image_rb = image_fb->color_rb[0];
+
+   copy_renderbuffer(irb, image_rb);
+}
+
+
+void
+intelEGLImageTargetTexture2D(GLcontext *ctx,
+                                struct gl_texture_object *texObj,
+                                GLeglImageOES image)
+{
+   struct intel_context *intel = intel_context(ctx);
+   __DRIEGLImage *driImage;
+   GLint glx_texture_format;
+
+   driImage = get_image(ctx, image);
+   if (!driImage)
+      return;
+
+   /* only level == 0 is supported */
+   if (driImage->level) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glEGLImageTargetTexture2DOES(level=%d)", driImage->level);
+      return;
+   }
+
+   glx_texture_format = (driImage->texture_format_rgba)
+      ? GLX_TEXTURE_FORMAT_RGBA_EXT : GLX_TEXTURE_FORMAT_RGB_EXT;
+
+   /* TODO refactor intelSetTexBuffer2 to avoid lock... */
+   _mesa_unlock_texture(ctx, texObj);
+   intelSetTexBuffer2(intel->driContext, texObj->Target, glx_texture_format,
+                      driImage->drawable);
+   _mesa_lock_texture(ctx, texObj);
+}
+
+
+#endif /* FEATURE_OES_EGL_image */
diff --git a/src/mesa/drivers/dri/intel/intel_eglimage.h b/src/mesa/drivers/dri/intel/intel_eglimage.h
new file mode 100644
index 0000000000..ef1bf2eacd
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_eglimage.h
@@ -0,0 +1,61 @@
+/**************************************************************************
+ * 
+ * Copyright (C) 2009 Chia-I Wu <olvaffe@gmail.com>
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish, * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef INTEL_EGLIMAGE_H
+#define INTEL_EGLIMAGE_H
+
+#include "main/mtypes.h"
+
+/* main/eglimage.h might not exist */
+#if FEATURE_OES_EGL_image
+
+#ifndef GL_OES_EGL_image
+typedef void *GLeglImageOES;
+#endif
+
+#include "main/eglimage.h"
+
+extern void
+intelEGLImageTargetRenderbufferStorage(GLcontext *ctx,
+                                       struct gl_renderbuffer *rb,
+                                       GLeglImageOES image);
+
+extern void
+intelEGLImageTargetTexture2D(GLcontext *ctx,
+                             struct gl_texture_object *texObj,
+                             GLeglImageOES image);
+
+#else
+#define _MESA_INIT_EGLIMAGE_FUNCTIONS(driver, impl) do { } while (0)
+#endif /* FEATURE_OES_EGL_image */
+
+static INLINE void
+intelInitEGLImageFuncs(struct dd_function_table *functions)
+{
+   _MESA_INIT_EGLIMAGE_FUNCTIONS(functions, intel);
+}
+
+#endif /* INTEL_EGLIMAGE_H */
diff --git a/src/mesa/drivers/dri/intel/intel_extensions.c b/src/mesa/drivers/dri/intel/intel_extensions.c
new file mode 100644
index 0000000000..21af22fbfb
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_extensions.c
@@ -0,0 +1,213 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "intel_chipset.h"
+#include "intel_context.h"
+#include "intel_extensions.h"
+
+
+#define need_GL_ARB_copy_buffer
+#define need_GL_ARB_draw_elements_base_vertex
+#define need_GL_ARB_framebuffer_object
+#define need_GL_ARB_map_buffer_range
+#define need_GL_ARB_occlusion_query
+#define need_GL_ARB_point_parameters
+#define need_GL_ARB_shader_objects
+#define need_GL_ARB_sync
+#define need_GL_ARB_vertex_array_object
+#define need_GL_ARB_vertex_program
+#define need_GL_ARB_vertex_shader
+#define need_GL_ARB_window_pos
+#define need_GL_EXT_blend_color
+#define need_GL_EXT_blend_equation_separate
+#define need_GL_EXT_blend_func_separate
+#define need_GL_EXT_blend_minmax
+#define need_GL_EXT_cull_vertex
+#define need_GL_EXT_fog_coord
+#define need_GL_EXT_framebuffer_object
+#define need_GL_EXT_framebuffer_blit
+#define need_GL_EXT_gpu_program_parameters
+#define need_GL_EXT_point_parameters
+#define need_GL_EXT_provoking_vertex
+#define need_GL_EXT_secondary_color
+#define need_GL_EXT_stencil_two_side
+#define need_GL_APPLE_vertex_array_object
+#define need_GL_ATI_separate_stencil
+#define need_GL_ATI_envmap_bumpmap
+#define need_GL_NV_point_sprite
+#define need_GL_NV_vertex_program
+#define need_GL_VERSION_2_0
+#define need_GL_VERSION_2_1
+
+#include "extension_helper.h"
+
+
+/**
+ * Extension strings exported by the intel driver.
+ *
+ * Extensions supported by all chips supported by i830_dri, i915_dri, or
+ * i965_dri.
+ */
+static const struct dri_extension card_extensions[] = {
+   { "GL_ARB_copy_buffer",                GL_ARB_copy_buffer_functions },
+   { "GL_ARB_draw_elements_base_vertex",  GL_ARB_draw_elements_base_vertex_functions },
+   { "GL_ARB_half_float_pixel",           NULL },
+   { "GL_ARB_map_buffer_range",           GL_ARB_map_buffer_range_functions },
+   { "GL_ARB_multitexture",               NULL },
+   { "GL_ARB_point_parameters",           GL_ARB_point_parameters_functions },
+   { "GL_ARB_point_sprite",               NULL },
+   { "GL_ARB_sync",                       GL_ARB_sync_functions },
+   { "GL_ARB_texture_border_clamp",       NULL },
+   { "GL_ARB_texture_cube_map",           NULL },
+   { "GL_ARB_texture_env_add",            NULL },
+   { "GL_ARB_texture_env_combine",        NULL },
+   { "GL_ARB_texture_env_crossbar",       NULL },
+   { "GL_ARB_texture_env_dot3",           NULL },
+   { "GL_ARB_texture_mirrored_repeat",    NULL },
+   { "GL_ARB_texture_rectangle",          NULL },
+   { "GL_ARB_vertex_array_object",        GL_ARB_vertex_array_object_functions},
+   { "GL_ARB_vertex_program",             GL_ARB_vertex_program_functions },
+   { "GL_ARB_window_pos",                 GL_ARB_window_pos_functions },
+   { "GL_EXT_blend_color",                GL_EXT_blend_color_functions },
+   { "GL_EXT_blend_equation_separate",    GL_EXT_blend_equation_separate_functions },
+   { "GL_EXT_blend_func_separate",        GL_EXT_blend_func_separate_functions },
+   { "GL_EXT_blend_minmax",               GL_EXT_blend_minmax_functions },
+   { "GL_EXT_blend_logic_op",             NULL },
+   { "GL_EXT_blend_subtract",             NULL },
+   { "GL_EXT_cull_vertex",                GL_EXT_cull_vertex_functions },
+   { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
+   { "GL_EXT_gpu_program_parameters",     GL_EXT_gpu_program_parameters_functions },
+   { "GL_EXT_packed_depth_stencil",       NULL },
+   { "GL_EXT_provoking_vertex",           GL_EXT_provoking_vertex_functions },
+   { "GL_EXT_secondary_color",            GL_EXT_secondary_color_functions },
+   { "GL_EXT_stencil_wrap",               NULL },
+   { "GL_EXT_texture_edge_clamp",         NULL },
+   { "GL_EXT_texture_env_combine",        NULL },
+   { "GL_EXT_texture_env_dot3",           NULL },
+   { "GL_EXT_texture_filter_anisotropic", NULL },
+   { "GL_EXT_texture_lod_bias",           NULL },
+   { "GL_3DFX_texture_compression_FXT1",  NULL },
+   { "GL_APPLE_client_storage",           NULL },
+   { "GL_APPLE_vertex_array_object",      GL_APPLE_vertex_array_object_functions},
+   { "GL_MESA_pack_invert",               NULL },
+   { "GL_MESA_ycbcr_texture",             NULL },
+   { "GL_NV_blend_square",                NULL },
+   { "GL_NV_point_sprite",                GL_NV_point_sprite_functions },
+   { "GL_NV_vertex_program",              GL_NV_vertex_program_functions },
+   { "GL_NV_vertex_program1_1",           NULL },
+   { "GL_SGIS_generate_mipmap",           NULL },
+   { NULL, NULL }
+};
+
+
+/** i915 / i945-only extensions */
+static const struct dri_extension i915_extensions[] = {
+   { "GL_ARB_depth_texture",              NULL },
+   { "GL_ARB_fragment_program",           NULL },
+   { "GL_ARB_shadow",                     NULL },
+   { "GL_ARB_texture_non_power_of_two",   NULL },
+   { "GL_ATI_separate_stencil",           GL_ATI_separate_stencil_functions },
+   { "GL_ATI_texture_env_combine3",       NULL },
+   { "GL_EXT_shadow_funcs",               NULL },
+   { "GL_EXT_stencil_two_side",           GL_EXT_stencil_two_side_functions },
+   { "GL_NV_texture_env_combine4",        NULL },
+   { NULL,                                NULL }
+};
+
+
+/** i965-only extensions */
+static const struct dri_extension brw_extensions[] = {
+   { "GL_ARB_depth_clamp",                NULL },
+   { "GL_ARB_depth_texture",              NULL },
+   { "GL_ARB_fragment_program",           NULL },
+   { "GL_ARB_fragment_program_shadow",    NULL },
+   { "GL_ARB_fragment_shader",            NULL },
+   { "GL_ARB_framebuffer_object",         GL_ARB_framebuffer_object_functions},
+   { "GL_ARB_occlusion_query",            GL_ARB_occlusion_query_functions },
+   { "GL_ARB_point_sprite", 		  NULL },
+   { "GL_ARB_seamless_cube_map",          NULL },
+   { "GL_ARB_shader_objects",             GL_ARB_shader_objects_functions },
+   { "GL_ARB_shading_language_100",       GL_VERSION_2_0_functions },
+   { "GL_ARB_shading_language_120",       GL_VERSION_2_1_functions },
+   { "GL_ARB_shadow",                     NULL },
+   { "GL_MESA_texture_signed_rgba",       NULL },
+   { "GL_ARB_texture_non_power_of_two",   NULL },
+   { "GL_ARB_vertex_shader",              GL_ARB_vertex_shader_functions },
+   { "GL_EXT_shadow_funcs",               NULL },
+   { "GL_EXT_stencil_two_side",           GL_EXT_stencil_two_side_functions },
+   { "GL_EXT_texture_sRGB",		  NULL },
+   { "GL_EXT_texture_swizzle",		  NULL },
+   { "GL_EXT_vertex_array_bgra",	  NULL },
+   { "GL_ATI_envmap_bumpmap",             GL_ATI_envmap_bumpmap_functions },
+   { "GL_ATI_separate_stencil",           GL_ATI_separate_stencil_functions },
+   { "GL_ATI_texture_env_combine3",       NULL },
+   { "GL_NV_texture_env_combine4",        NULL },
+   { NULL,                                NULL }
+};
+
+
+static const struct dri_extension arb_oq_extensions[] = {
+   { NULL, NULL }
+};
+
+
+static const struct dri_extension ttm_extensions[] = {
+   { "GL_ARB_pixel_buffer_object",      NULL },
+   { "GL_EXT_framebuffer_blit",         GL_EXT_framebuffer_blit_functions },
+   { "GL_EXT_framebuffer_object",       GL_EXT_framebuffer_object_functions },
+#if FEATURE_OES_EGL_image
+   { "GL_OES_EGL_image",                NULL },
+#endif
+   { NULL, NULL }
+};
+
+
+/**
+ * Initializes potential list of extensions if ctx == NULL, or actually enables
+ * extensions for a context.
+ */
+void
+intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging)
+{
+   struct intel_context *intel = ctx?intel_context(ctx):NULL;
+
+   /* Disable imaging extension until convolution is working in teximage paths.
+    */
+   enable_imaging = GL_FALSE;
+
+   driInitExtensions(ctx, card_extensions, enable_imaging);
+
+   if (intel == NULL || intel->ttm)
+      driInitExtensions(ctx, ttm_extensions, GL_FALSE);
+
+   if (intel == NULL || IS_965(intel->intelScreen->deviceID))
+      driInitExtensions(ctx, brw_extensions, GL_FALSE);
+
+   if (intel == NULL || IS_915(intel->intelScreen->deviceID)
+       || IS_945(intel->intelScreen->deviceID))
+      driInitExtensions(ctx, i915_extensions, GL_FALSE);
+}
diff --git a/src/mesa/drivers/dri/intel/intel_extensions.h b/src/mesa/drivers/dri/intel/intel_extensions.h
new file mode 100644
index 0000000000..97147ecdb0
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_extensions.h
@@ -0,0 +1,36 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef INTEL_EXTENSIONS_H
+#define INTEL_EXTENSIONS_H
+
+
+extern void
+intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging);
+
+
+#endif
diff --git a/src/mesa/drivers/dri/intel/intel_fbo.c b/src/mesa/drivers/dri/intel/intel_fbo.c
index 7cf12619d6..084095fc96 100644
--- a/src/mesa/drivers/dri/intel/intel_fbo.c
+++ b/src/mesa/drivers/dri/intel/intel_fbo.c
@@ -27,6 +27,7 @@
 
 
 #include "main/imports.h"
+#include "main/macros.h"
 #include "main/mtypes.h"
 #include "main/fbobject.h"
 #include "main/framebuffer.h"
@@ -34,61 +35,17 @@
 #include "main/context.h"
 #include "main/texformat.h"
 #include "main/texrender.h"
+#include "drivers/common/meta.h"
 
 #include "intel_context.h"
 #include "intel_buffers.h"
-#include "intel_depthstencil.h"
 #include "intel_fbo.h"
 #include "intel_mipmap_tree.h"
 #include "intel_regions.h"
-#include "intel_span.h"
 
 
 #define FILE_DEBUG_FLAG DEBUG_FBO
 
-#define INTEL_RB_CLASS 0x12345678
-
-
-/* XXX FBO: move this to intel_context.h (inlined) */
-/**
- * Return a gl_renderbuffer ptr casted to intel_renderbuffer.
- * NULL will be returned if the rb isn't really an intel_renderbuffer.
- * This is determiend by checking the ClassID.
- */
-struct intel_renderbuffer *
-intel_renderbuffer(struct gl_renderbuffer *rb)
-{
-   struct intel_renderbuffer *irb = (struct intel_renderbuffer *) rb;
-   if (irb && irb->Base.ClassID == INTEL_RB_CLASS) {
-      /*_mesa_warning(NULL, "Returning non-intel Rb\n");*/
-      return irb;
-   }
-   else
-      return NULL;
-}
-
-
-struct intel_renderbuffer *
-intel_get_renderbuffer(struct gl_framebuffer *fb, int attIndex)
-{
-   if (attIndex >= 0)
-      return intel_renderbuffer(fb->Attachment[attIndex].Renderbuffer);
-   else
-      return NULL;
-}
-
-struct intel_region *
-intel_get_rb_region(struct gl_framebuffer *fb, GLuint attIndex)
-{
-   struct intel_renderbuffer *irb = intel_get_renderbuffer(fb, attIndex);
-
-   if (irb)
-      return irb->region;
-   else
-      return NULL;
-}
-
-
 
 /**
  * Create a new framebuffer object.
@@ -103,6 +60,7 @@ intel_new_framebuffer(GLcontext * ctx, GLuint name)
 }
 
 
+/** Called by gl_renderbuffer::Delete() */
 static void
 intel_delete_renderbuffer(struct gl_renderbuffer *rb)
 {
@@ -112,10 +70,6 @@ intel_delete_renderbuffer(struct gl_renderbuffer *rb)
 
    ASSERT(irb);
 
-   if (irb->PairedStencil || irb->PairedDepth) {
-      intel_unpair_depth_stencil(ctx, irb);
-   }
-
    if (irb->span_cache != NULL)
       _mesa_free(irb->span_cache);
 
@@ -127,7 +81,6 @@ intel_delete_renderbuffer(struct gl_renderbuffer *rb)
 }
 
 
-
 /**
  * Return a pointer to a specific pixel in a renderbuffer.
  */
@@ -142,7 +95,6 @@ intel_get_pointer(GLcontext * ctx, struct gl_renderbuffer *rb,
 }
 
 
-
 /**
  * Called via glRenderbufferStorageEXT() to set the format and allocate
  * storage for a user-created renderbuffer.
@@ -168,6 +120,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
       rb->RedBits = 5;
       rb->GreenBits = 6;
       rb->BlueBits = 5;
+      irb->texformat = &_mesa_texformat_rgb565;
       cpp = 2;
       break;
    case GL_RGB:
@@ -181,6 +134,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
       rb->GreenBits = 8;
       rb->BlueBits = 8;
       rb->AlphaBits = 0;
+      irb->texformat = &_mesa_texformat_argb8888; /* XXX: Need xrgb8888 */
       cpp = 4;
       break;
    case GL_RGBA:
@@ -197,6 +151,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
       rb->GreenBits = 8;
       rb->BlueBits = 8;
       rb->AlphaBits = 8;
+      irb->texformat = &_mesa_texformat_argb8888;
       cpp = 4;
       break;
    case GL_STENCIL_INDEX:
@@ -209,20 +164,15 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
       rb->DataType = GL_UNSIGNED_INT_24_8_EXT;
       rb->StencilBits = 8;
       cpp = 4;
+      irb->texformat = &_mesa_texformat_s8_z24;
       break;
    case GL_DEPTH_COMPONENT16:
-#if 0
       rb->_ActualFormat = GL_DEPTH_COMPONENT16;
       rb->DataType = GL_UNSIGNED_SHORT;
       rb->DepthBits = 16;
       cpp = 2;
+      irb->texformat = &_mesa_texformat_z16;
       break;
-#else
-      /* fall-through.
-       * 16bpp depth renderbuffer can't be paired with a stencil buffer so
-       * always used combined depth/stencil format.
-       */
-#endif
    case GL_DEPTH_COMPONENT:
    case GL_DEPTH_COMPONENT24:
    case GL_DEPTH_COMPONENT32:
@@ -230,6 +180,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
       rb->DataType = GL_UNSIGNED_INT_24_8_EXT;
       rb->DepthBits = 24;
       cpp = 4;
+      irb->texformat = &_mesa_texformat_s8_z24;
       break;
    case GL_DEPTH_STENCIL_EXT:
    case GL_DEPTH24_STENCIL8_EXT:
@@ -238,6 +189,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
       rb->DepthBits = 24;
       rb->StencilBits = 8;
       cpp = 4;
+      irb->texformat = &_mesa_texformat_s8_z24;
       break;
    default:
       _mesa_problem(ctx,
@@ -266,7 +218,9 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
       DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width,
 	  height, pitch);
 
-      irb->region = intel_region_alloc(intel, cpp, width, height, pitch);
+      irb->region = intel_region_alloc(intel, I915_TILING_NONE,
+				       cpp, width, height, pitch,
+				       GL_TRUE);
       if (!irb->region)
          return GL_FALSE;       /* out of memory? */
 
@@ -280,7 +234,6 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
 }
 
 
-
 /**
  * Called for each hardware renderbuffer when a _window_ is resized.
  * Just update fields.
@@ -298,6 +251,7 @@ intel_alloc_window_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
    return GL_TRUE;
 }
 
+
 static void
 intel_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb,
 		     GLuint width, GLuint height)
@@ -324,6 +278,8 @@ intel_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb,
    }
 }
 
+
+/** Dummy function for gl_renderbuffer::AllocStorage() */
 static GLboolean
 intel_nop_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
                         GLenum internalFormat, GLuint width, GLuint height)
@@ -343,10 +299,9 @@ intel_renderbuffer_set_region(struct intel_renderbuffer *rb,
    rb->region = NULL;
    intel_region_reference(&rb->region, region);
    intel_region_release(&old);
-
-   rb->pfPitch = region->pitch;
 }
 
+
 /**
  * Create a new intel_renderbuffer which corresponds to an on-screen window,
  * not a user-created renderbuffer.
@@ -376,6 +331,17 @@ intel_create_renderbuffer(GLenum intFormat)
       irb->Base.GreenBits = 6;
       irb->Base.BlueBits = 5;
       irb->Base.DataType = GL_UNSIGNED_BYTE;
+      irb->texformat = &_mesa_texformat_rgb565;
+      break;
+   case GL_RGB8:
+      irb->Base._ActualFormat = GL_RGB8;
+      irb->Base._BaseFormat = GL_RGB;
+      irb->Base.RedBits = 8;
+      irb->Base.GreenBits = 8;
+      irb->Base.BlueBits = 8;
+      irb->Base.AlphaBits = 0;
+      irb->Base.DataType = GL_UNSIGNED_BYTE;
+      irb->texformat = &_mesa_texformat_argb8888; /* XXX: Need xrgb8888 */
       break;
    case GL_RGBA8:
       irb->Base._ActualFormat = GL_RGBA8;
@@ -385,24 +351,28 @@ intel_create_renderbuffer(GLenum intFormat)
       irb->Base.BlueBits = 8;
       irb->Base.AlphaBits = 8;
       irb->Base.DataType = GL_UNSIGNED_BYTE;
+      irb->texformat = &_mesa_texformat_argb8888;
       break;
    case GL_STENCIL_INDEX8_EXT:
       irb->Base._ActualFormat = GL_STENCIL_INDEX8_EXT;
       irb->Base._BaseFormat = GL_STENCIL_INDEX;
       irb->Base.StencilBits = 8;
       irb->Base.DataType = GL_UNSIGNED_BYTE;
+      irb->texformat = &_mesa_texformat_s8_z24;
       break;
    case GL_DEPTH_COMPONENT16:
       irb->Base._ActualFormat = GL_DEPTH_COMPONENT16;
       irb->Base._BaseFormat = GL_DEPTH_COMPONENT;
       irb->Base.DepthBits = 16;
       irb->Base.DataType = GL_UNSIGNED_SHORT;
+      irb->texformat = &_mesa_texformat_z16;
       break;
    case GL_DEPTH_COMPONENT24:
       irb->Base._ActualFormat = GL_DEPTH24_STENCIL8_EXT;
       irb->Base._BaseFormat = GL_DEPTH_COMPONENT;
       irb->Base.DepthBits = 24;
       irb->Base.DataType = GL_UNSIGNED_INT;
+      irb->texformat = &_mesa_texformat_s8_z24;
       break;
    case GL_DEPTH24_STENCIL8_EXT:
       irb->Base._ActualFormat = GL_DEPTH24_STENCIL8_EXT;
@@ -410,6 +380,7 @@ intel_create_renderbuffer(GLenum intFormat)
       irb->Base.DepthBits = 24;
       irb->Base.StencilBits = 8;
       irb->Base.DataType = GL_UNSIGNED_INT_24_8_EXT;
+      irb->texformat = &_mesa_texformat_s8_z24;
       break;
    default:
       _mesa_problem(NULL,
@@ -466,9 +437,6 @@ intel_bind_framebuffer(GLcontext * ctx, GLenum target,
 {
    if (target == GL_FRAMEBUFFER_EXT || target == GL_DRAW_FRAMEBUFFER_EXT) {
       intel_draw_buffer(ctx, fb);
-      /* Integer depth range depends on depth buffer bits */
-      if (ctx->Driver.DepthRange != NULL)
-	 ctx->Driver.DepthRange(ctx, ctx->Viewport.Near, ctx->Viewport.Far);
    }
    else {
       /* don't need to do anything if target == GL_READ_FRAMEBUFFER_EXT */
@@ -492,10 +460,13 @@ intel_framebuffer_renderbuffer(GLcontext * ctx,
    intel_draw_buffer(ctx, fb);
 }
 
+
 static GLboolean
 intel_update_wrapper(GLcontext *ctx, struct intel_renderbuffer *irb, 
 		     struct gl_texture_image *texImage)
 {
+   irb->texformat = texImage->TexFormat;
+
    if (texImage->TexFormat == &_mesa_texformat_argb8888) {
       irb->Base._ActualFormat = GL_RGBA8;
       irb->Base._BaseFormat = GL_RGBA;
@@ -505,9 +476,21 @@ intel_update_wrapper(GLcontext *ctx, struct intel_renderbuffer *irb,
    else if (texImage->TexFormat == &_mesa_texformat_rgb565) {
       irb->Base._ActualFormat = GL_RGB5;
       irb->Base._BaseFormat = GL_RGB;
-      irb->Base.DataType = GL_UNSIGNED_SHORT;
+      irb->Base.DataType = GL_UNSIGNED_BYTE;
       DBG("Render to RGB5 texture OK\n");
    }
+   else if (texImage->TexFormat == &_mesa_texformat_argb1555) {
+      irb->Base._ActualFormat = GL_RGB5_A1;
+      irb->Base._BaseFormat = GL_RGBA;
+      irb->Base.DataType = GL_UNSIGNED_BYTE;
+      DBG("Render to ARGB1555 texture OK\n");
+   }
+   else if (texImage->TexFormat == &_mesa_texformat_argb4444) {
+      irb->Base._ActualFormat = GL_RGBA4;
+      irb->Base._BaseFormat = GL_RGBA;
+      irb->Base.DataType = GL_UNSIGNED_BYTE;
+      DBG("Render to ARGB4444 texture OK\n");
+   }
    else if (texImage->TexFormat == &_mesa_texformat_z16) {
       irb->Base._ActualFormat = GL_DEPTH_COMPONENT16;
       irb->Base._BaseFormat = GL_DEPTH_COMPONENT;
@@ -534,15 +517,15 @@ intel_update_wrapper(GLcontext *ctx, struct intel_renderbuffer *irb,
    irb->Base.BlueBits = texImage->TexFormat->BlueBits;
    irb->Base.AlphaBits = texImage->TexFormat->AlphaBits;
    irb->Base.DepthBits = texImage->TexFormat->DepthBits;
+   irb->Base.StencilBits = texImage->TexFormat->StencilBits;
 
    irb->Base.Delete = intel_delete_renderbuffer;
    irb->Base.AllocStorage = intel_nop_alloc_storage;
 
-   irb->RenderToTexture = GL_TRUE;
-
    return GL_TRUE;
 }
 
+
 /**
  * When glFramebufferTexture[123]D is called this function sets up the
  * gl_renderbuffer wrapper around the texture image.
@@ -551,7 +534,7 @@ intel_update_wrapper(GLcontext *ctx, struct intel_renderbuffer *irb,
 static struct intel_renderbuffer *
 intel_wrap_texture(GLcontext * ctx, struct gl_texture_image *texImage)
 {
-   const GLuint name = ~0;      /* not significant, but distinct for debugging */
+   const GLuint name = ~0;   /* not significant, but distinct for debugging */
    struct intel_renderbuffer *irb;
 
    /* make an intel_renderbuffer to wrap the texture image */
@@ -594,14 +577,16 @@ intel_render_texture(GLcontext * ctx,
 
    ASSERT(newImage);
 
-   if (newImage->Border != 0) {
-      /* Fallback on drawing to a texture with a border, which won't have a
-       * miptree.
+   intel_image = intel_texture_image(newImage);
+   if (!intel_image->mt) {
+      /* Fallback on drawing to a texture that doesn't have a miptree
+       * (has a border, width/height 0, etc.)
        */
-       _mesa_reference_renderbuffer(&att->Renderbuffer, NULL);
-       _mesa_render_texture(ctx, fb, att);
-       return;
-   } else if (!irb) {
+      _mesa_reference_renderbuffer(&att->Renderbuffer, NULL);
+      _mesa_render_texture(ctx, fb, att);
+      return;
+   }
+   else if (!irb) {
       irb = intel_wrap_texture(ctx, newImage);
       if (irb) {
          /* bind the wrapper to the attachment point */
@@ -612,7 +597,9 @@ intel_render_texture(GLcontext * ctx,
          _mesa_render_texture(ctx, fb, att);
          return;
       }
-   } if (!intel_update_wrapper(ctx, irb, newImage)) {
+   }
+
+   if (!intel_update_wrapper(ctx, irb, newImage)) {
        _mesa_reference_renderbuffer(&att->Renderbuffer, NULL);
        _mesa_render_texture(ctx, fb, att);
        return;
@@ -624,7 +611,6 @@ intel_render_texture(GLcontext * ctx,
        irb->Base.RefCount);
 
    /* point the renderbufer's region to the texture image region */
-   intel_image = intel_texture_image(newImage);
    if (irb->region != intel_image->mt->region) {
       if (irb->region)
 	 intel_region_release(&irb->region);
@@ -657,18 +643,59 @@ static void
 intel_finish_render_texture(GLcontext * ctx,
                             struct gl_renderbuffer_attachment *att)
 {
-   struct intel_renderbuffer *irb = intel_renderbuffer(att->Renderbuffer);
+   /* no-op
+    * Previously we released the renderbuffer's intel_region but
+    * that's not necessary and actually caused problems when trying
+    * to do a glRead/CopyPixels from the renderbuffer later.
+    * The region will be released later if the texture is replaced
+    * or the renderbuffer deleted.
+    *
+    * The intention of this driver hook is more of a "done rendering
+    * to texture, please re-twiddle/etc if necessary".
+    */
+}
 
-   DBG("End render texture (tid %x) tex %u\n", _glthread_GetID(), att->Texture->Name);
 
-   if (irb) {
-      /* just release the region */
-      intel_region_release(&irb->region);
+/**
+ * Do additional "completeness" testing of a framebuffer object.
+ */
+static void
+intel_validate_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb)
+{
+   const struct intel_renderbuffer *depthRb =
+      intel_get_renderbuffer(fb, BUFFER_DEPTH);
+   const struct intel_renderbuffer *stencilRb =
+      intel_get_renderbuffer(fb, BUFFER_STENCIL);
+   int i;
+
+   if (stencilRb && stencilRb != depthRb) {
+      /* we only support combined depth/stencil buffers, not separate
+       * stencil buffers.
+       */
+      fb->_Status = GL_FRAMEBUFFER_UNSUPPORTED_EXT;
    }
-   else if (att->Renderbuffer) {
-      /* software fallback */
-      _mesa_finish_render_texture(ctx, att);
-      /* XXX FBO: Need to unmap the buffer (or in intelSpanRenderStart???) */
+
+   for (i = 0; i < ctx->Const.MaxDrawBuffers; i++) {
+      struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[i];
+      struct intel_renderbuffer *irb = intel_renderbuffer(rb);
+
+      if (rb == NULL)
+	 continue;
+
+      if (irb == NULL) {
+	 fb->_Status = GL_FRAMEBUFFER_UNSUPPORTED_EXT;
+	 continue;
+      }
+
+      switch (irb->texformat->MesaFormat) {
+      case MESA_FORMAT_ARGB8888:
+      case MESA_FORMAT_RGB565:
+      case MESA_FORMAT_ARGB1555:
+      case MESA_FORMAT_ARGB4444:
+	 break;
+      default:
+	 fb->_Status = GL_FRAMEBUFFER_UNSUPPORTED_EXT;
+      }
    }
 }
 
@@ -687,4 +714,8 @@ intel_fbo_init(struct intel_context *intel)
    intel->ctx.Driver.RenderTexture = intel_render_texture;
    intel->ctx.Driver.FinishRenderTexture = intel_finish_render_texture;
    intel->ctx.Driver.ResizeBuffers = intel_resize_buffers;
+   intel->ctx.Driver.ValidateFramebuffer = intel_validate_framebuffer;
+#if FEATURE_EXT_framebuffer_blit
+   intel->ctx.Driver.BlitFramebuffer = _mesa_meta_blit_framebuffer;
+#endif
 }
diff --git a/src/mesa/drivers/dri/intel/intel_fbo.h b/src/mesa/drivers/dri/intel/intel_fbo.h
index b7e9280e8c..f0665af482 100644
--- a/src/mesa/drivers/dri/intel/intel_fbo.h
+++ b/src/mesa/drivers/dri/intel/intel_fbo.h
@@ -55,19 +55,13 @@ struct intel_framebuffer
 
 /**
  * Intel renderbuffer, derived from gl_renderbuffer.
- * Note: The PairedDepth and PairedStencil fields use renderbuffer IDs,
- * not pointers because in some circumstances a deleted renderbuffer could
- * result in a dangling pointer here.
  */
 struct intel_renderbuffer
 {
    struct gl_renderbuffer Base;
    struct intel_region *region;
-   GLuint pfPitch;              /* possibly paged flipped pitch */
-   GLboolean RenderToTexture;   /* RTT? */
 
-   GLuint PairedDepth;   /**< only used if this is a depth renderbuffer */
-   GLuint PairedStencil; /**< only used if this is a stencil renderbuffer */
+   const struct gl_texture_format *texformat;
 
    GLuint vbl_pending;   /**< vblank sequence number of pending flip */
 
@@ -75,48 +69,70 @@ struct intel_renderbuffer
    unsigned long span_cache_offset;
 };
 
-extern struct intel_renderbuffer *intel_renderbuffer(struct gl_renderbuffer
-                                                     *rb);
+
+/**
+ * gl_renderbuffer is a base class which we subclass.  The Class field
+ * is used for simple run-time type checking.
+ */
+#define INTEL_RB_CLASS 0x12345678
+
+
+/**
+ * Return a gl_renderbuffer ptr casted to intel_renderbuffer.
+ * NULL will be returned if the rb isn't really an intel_renderbuffer.
+ * This is determined by checking the ClassID.
+ */
+static INLINE struct intel_renderbuffer *
+intel_renderbuffer(struct gl_renderbuffer *rb)
+{
+   struct intel_renderbuffer *irb = (struct intel_renderbuffer *) rb;
+   if (irb && irb->Base.ClassID == INTEL_RB_CLASS) {
+      /*_mesa_warning(NULL, "Returning non-intel Rb\n");*/
+      return irb;
+   }
+   else
+      return NULL;
+}
+
+
+/**
+ * Return a framebuffer's renderbuffer, named by a BUFFER_x index.
+ */
+static INLINE struct intel_renderbuffer *
+intel_get_renderbuffer(struct gl_framebuffer *fb, int attIndex)
+{
+   if (attIndex >= 0)
+      return intel_renderbuffer(fb->Attachment[attIndex].Renderbuffer);
+   else
+      return NULL;
+}
+
 
 extern void
 intel_renderbuffer_set_region(struct intel_renderbuffer *irb,
 			      struct intel_region *region);
 
+
 extern struct intel_renderbuffer *
 intel_create_renderbuffer(GLenum intFormat);
 
-extern void intel_fbo_init(struct intel_context *intel);
-
-
-/* XXX make inline or macro */
-extern struct intel_renderbuffer *intel_get_renderbuffer(struct gl_framebuffer
-                                                         *fb,
-                                                         int attIndex);
-
-extern void intel_flip_renderbuffers(struct intel_framebuffer *intel_fb);
 
+extern void
+intel_fbo_init(struct intel_context *intel);
 
-/* XXX make inline or macro */
-extern struct intel_region *intel_get_rb_region(struct gl_framebuffer *fb,
-                                                GLuint attIndex);
 
+extern void
+intel_flip_renderbuffers(struct intel_framebuffer *intel_fb);
 
 
-/**
- * Are we currently rendering into a texture?
- */
-static INLINE GLboolean
-intel_rendering_to_texture(const GLcontext *ctx)
+static INLINE struct intel_region *
+intel_get_rb_region(struct gl_framebuffer *fb, GLuint attIndex)
 {
-   if (ctx->DrawBuffer->Name) {
-      /* User-created FBO */
-      const struct intel_renderbuffer *irb =
-         intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0]);
-      return irb && irb->RenderToTexture;
-   }
-   else {
-      return GL_FALSE;
-   }
+   struct intel_renderbuffer *irb = intel_get_renderbuffer(fb, attIndex);
+   if (irb)
+      return irb->region;
+   else
+      return NULL;
 }
 
 
diff --git a/src/mesa/drivers/dri/intel/intel_generatemipmap.c b/src/mesa/drivers/dri/intel/intel_generatemipmap.c
new file mode 100644
index 0000000000..0052abb42d
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_generatemipmap.c
@@ -0,0 +1,286 @@
+/*
+ * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright © 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Eric Anholt <eric@anholt.net>
+ *
+ */
+
+#include "main/glheader.h"
+#include "main/enums.h"
+#include "main/image.h"
+#include "main/mtypes.h"
+#include "main/macros.h"
+#include "main/bufferobj.h"
+#include "main/teximage.h"
+#include "main/texenv.h"
+#include "main/texobj.h"
+#include "main/texstate.h"
+#include "main/texparam.h"
+#include "main/varray.h"
+#include "main/attrib.h"
+#include "main/enable.h"
+#include "main/buffers.h"
+#include "main/fbobject.h"
+#include "main/framebuffer.h"
+#include "main/renderbuffer.h"
+#include "main/depth.h"
+#include "main/hash.h"
+#include "main/mipmap.h"
+#include "main/blend.h"
+#include "swrast/swrast.h"
+
+#include "intel_screen.h"
+#include "intel_context.h"
+#include "intel_batchbuffer.h"
+#include "intel_pixel.h"
+#include "intel_tex.h"
+#include "intel_mipmap_tree.h"
+
+#if FEATURE_attrib_stack
+static const char *intel_fp_tex2d =
+      "!!ARBfp1.0\n"
+      "TEX result.color, fragment.texcoord[0], texture[0], 2D;\n"
+      "END\n";
+
+static GLboolean
+intel_generate_mipmap_level(GLcontext *ctx, GLuint tex_name,
+			    int level, int width, int height)
+{
+   struct intel_context *intel = intel_context(ctx);
+   GLfloat vertices[4][2];
+   GLint status;
+
+   /* Set to source from the previous level */
+   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_BASE_LEVEL, level - 1);
+   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, level - 1);
+
+   /* Set to draw into the current level */
+   _mesa_FramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT,
+				 GL_COLOR_ATTACHMENT0_EXT,
+				 GL_TEXTURE_2D,
+				 tex_name,
+				 level);
+   /* Choose to render to the color attachment. */
+   _mesa_DrawBuffer(GL_COLOR_ATTACHMENT0_EXT);
+
+   status = _mesa_CheckFramebufferStatusEXT (GL_FRAMEBUFFER_EXT);
+   if (status != GL_FRAMEBUFFER_COMPLETE_EXT)
+      return GL_FALSE;
+
+   meta_set_passthrough_transform(&intel->meta);
+
+   /* XXX: Doing it right would involve setting up the transformation to do
+    * 0-1 mapping or something, and not changing the vertex data.
+    */
+   vertices[0][0] = 0;
+   vertices[0][1] = 0;
+   vertices[1][0] = width;
+   vertices[1][1] = 0;
+   vertices[2][0] = width;
+   vertices[2][1] = height;
+   vertices[3][0] = 0;
+   vertices[3][1] = height;
+
+   _mesa_VertexPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &vertices);
+   _mesa_Enable(GL_VERTEX_ARRAY);
+   meta_set_default_texrect(&intel->meta);
+
+   _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
+
+   meta_restore_texcoords(&intel->meta);
+   meta_restore_transform(&intel->meta);
+
+   return GL_TRUE;
+}
+
+static GLboolean
+intel_generate_mipmap_2d(GLcontext *ctx,
+			 GLenum target,
+			 struct gl_texture_object *texObj)
+{
+   struct intel_context *intel = intel_context(ctx);
+   GLint old_active_texture;
+   int level, max_levels, start_level, end_level;
+   GLuint fb_name;
+   GLboolean success = GL_FALSE;
+   struct gl_framebuffer *saved_fbo = NULL;
+
+   _mesa_PushAttrib(GL_ENABLE_BIT | GL_TEXTURE_BIT |
+		    GL_CURRENT_BIT | GL_COLOR_BUFFER_BIT |
+		    GL_DEPTH_BUFFER_BIT);
+   _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT);
+   old_active_texture = ctx->Texture.CurrentUnit;
+   _mesa_reference_framebuffer(&saved_fbo, ctx->DrawBuffer);
+
+   _mesa_Disable(GL_POLYGON_STIPPLE);
+   _mesa_Disable(GL_DEPTH_TEST);
+   _mesa_Disable(GL_STENCIL_TEST);
+   _mesa_ColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
+   _mesa_DepthMask(GL_FALSE);
+
+   /* Bind the given texture to GL_TEXTURE_2D with linear filtering for our
+    * minification.
+    */
+   _mesa_ActiveTextureARB(GL_TEXTURE0_ARB);
+   _mesa_Enable(GL_TEXTURE_2D);
+   _mesa_BindTexture(GL_TEXTURE_2D, texObj->Name);
+   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER,
+		       GL_LINEAR_MIPMAP_NEAREST);
+   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
+
+   /* Bind the new renderbuffer to the color attachment point. */
+   _mesa_GenFramebuffersEXT(1, &fb_name);
+   _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, fb_name);
+
+   meta_set_fragment_program(&intel->meta, &intel->meta.tex2d_fp,
+			     intel_fp_tex2d);
+   meta_set_passthrough_vertex_program(&intel->meta);
+
+   max_levels = _mesa_max_texture_levels(ctx, texObj->Target);
+   start_level = texObj->BaseLevel;
+   end_level = texObj->MaxLevel;
+
+   /* Loop generating level+1 from level. */
+   for (level = start_level; level < end_level && level < max_levels - 1; level++) {
+      const struct gl_texture_image *srcImage;
+      int width, height;
+
+      srcImage = _mesa_select_tex_image(ctx, texObj, target, level);
+      if (srcImage->Border != 0)
+	 goto fail;
+
+      width = srcImage->Width / 2;
+      if (width < 1)
+	 width = 1;
+      height = srcImage->Height / 2;
+      if (height < 1)
+	 height = 1;
+
+      if (width == srcImage->Width &&
+	  height == srcImage->Height) {
+	 /* Neither _mesa_max_texture_levels nor texObj->MaxLevel are the
+	  * maximum texture level for the object, so break out when we've gone
+	  * over the edge.
+	  */
+	 break;
+      }
+
+      /* Make sure that there's space allocated for the target level.
+       * We could skip this if there's already space allocated and save some
+       * time.
+       */
+      _mesa_TexImage2D(GL_TEXTURE_2D, level + 1, srcImage->InternalFormat,
+		       width, height, 0,
+		       GL_RGBA, GL_UNSIGNED_INT, NULL);
+
+      if (!intel_generate_mipmap_level(ctx, texObj->Name, level + 1,
+				       width, height))
+	 goto fail;
+   }
+
+   success = GL_TRUE;
+
+fail:
+   meta_restore_fragment_program(&intel->meta);
+   meta_restore_vertex_program(&intel->meta);
+
+   _mesa_DeleteFramebuffersEXT(1, &fb_name);
+   _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture);
+   if (saved_fbo)
+      _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, saved_fbo->Name);
+   _mesa_reference_framebuffer(&saved_fbo, NULL);
+   _mesa_PopClientAttrib();
+   _mesa_PopAttrib();
+
+   return success;
+}
+#endif /* FEATURE_attrib_stack */
+
+
+/**
+ * Generate new mipmap data from BASE+1 to BASE+p (the minimally-sized mipmap
+ * level).
+ *
+ * The texture object's miptree must be mapped.
+ *
+ * It would be really nice if this was just called by Mesa whenever mipmaps
+ * needed to be regenerated, rather than us having to remember to do so in
+ * each texture image modification path.
+ *
+ * This function should also include an accelerated path.
+ */
+void
+intel_generate_mipmap(GLcontext *ctx, GLenum target,
+                      struct gl_texture_object *texObj)
+{
+   struct intel_context *intel = intel_context(ctx);
+   struct intel_texture_object *intelObj = intel_texture_object(texObj);
+   GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
+   int face, i;
+
+#if FEATURE_attrib_stack
+   /* HW path */
+   if (target == GL_TEXTURE_2D &&
+       ctx->Extensions.EXT_framebuffer_object &&
+       ctx->Extensions.ARB_fragment_program &&
+       ctx->Extensions.ARB_vertex_program) {
+      GLboolean success;
+
+      /* We'll be accessing this texture using GL entrypoints, which should
+       * be resilient against other access to this texture.
+       */
+      _mesa_unlock_texture(ctx, texObj);
+      success = intel_generate_mipmap_2d(ctx, target, texObj);
+      _mesa_lock_texture(ctx, texObj);
+
+      if (success)
+	 return;
+   }
+#endif /* FEATURE_attrib_stack */
+
+   /* SW path */
+   intel_tex_map_level_images(intel, intelObj, texObj->BaseLevel);
+   _mesa_generate_mipmap(ctx, target, texObj);
+   intel_tex_unmap_level_images(intel, intelObj, texObj->BaseLevel);
+
+   /* Update the level information in our private data in the new images, since
+    * it didn't get set as part of a normal TexImage path.
+    */
+   for (face = 0; face < nr_faces; face++) {
+      for (i = texObj->BaseLevel + 1; i < texObj->MaxLevel; i++) {
+         struct intel_texture_image *intelImage;
+
+	 intelImage = intel_texture_image(texObj->Image[face][i]);
+	 if (intelImage == NULL)
+	    break;
+
+	 intelImage->level = i;
+	 intelImage->face = face;
+	 /* Unreference the miptree to signal that the new Data is a bare
+	  * pointer from mesa.
+	  */
+	 intel_miptree_release(intel, &intelImage->mt);
+      }
+   }
+}
diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
index bf1c3f03f0..c985da5aa2 100644
--- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
+++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
@@ -57,14 +57,16 @@ intel_miptree_create_internal(struct intel_context *intel,
 			      GLuint last_level,
 			      GLuint width0,
 			      GLuint height0,
-			      GLuint depth0, GLuint cpp, GLuint compress_byte)
+			      GLuint depth0, GLuint cpp, GLuint compress_byte,
+			      uint32_t tiling)
 {
    GLboolean ok;
    struct intel_mipmap_tree *mt = calloc(sizeof(*mt), 1);
 
-   DBG("%s target %s format %s level %d..%d\n", __FUNCTION__,
+   DBG("%s target %s format %s level %d..%d <-- %p\n", __FUNCTION__,
        _mesa_lookup_enum_by_nr(target),
-       _mesa_lookup_enum_by_nr(internal_format), first_level, last_level);
+       _mesa_lookup_enum_by_nr(internal_format), 
+       first_level, last_level, mt);
 
    mt->target = target_to_target(target);
    mt->internal_format = internal_format;
@@ -80,15 +82,16 @@ intel_miptree_create_internal(struct intel_context *intel,
 
 #ifdef I915
    if (IS_945(intel->intelScreen->deviceID))
-      ok = i945_miptree_layout(intel, mt);
+      ok = i945_miptree_layout(intel, mt, tiling);
    else
-      ok = i915_miptree_layout(intel, mt);
+      ok = i915_miptree_layout(intel, mt, tiling);
 #else
-   ok = brw_miptree_layout(intel, mt);
+   ok = brw_miptree_layout(intel, mt, tiling);
 #endif
 
    if (!ok) {
       free(mt);
+      DBG("%s not okay - returning NULL\n", __FUNCTION__);
       return NULL;
    }
 
@@ -98,18 +101,33 @@ intel_miptree_create_internal(struct intel_context *intel,
 struct intel_mipmap_tree *
 intel_miptree_create(struct intel_context *intel,
 		     GLenum target,
+		     GLenum base_format,
 		     GLenum internal_format,
 		     GLuint first_level,
 		     GLuint last_level,
 		     GLuint width0,
 		     GLuint height0,
-		     GLuint depth0, GLuint cpp, GLuint compress_byte)
+		     GLuint depth0, GLuint cpp, GLuint compress_byte,
+		     GLboolean expect_accelerated_upload)
 {
    struct intel_mipmap_tree *mt;
+   uint32_t tiling;
+
+   if (intel->use_texture_tiling && compress_byte == 0 &&
+       intel->intelScreen->kernel_exec_fencing) {
+      if (IS_965(intel->intelScreen->deviceID) &&
+	  (base_format == GL_DEPTH_COMPONENT ||
+	   base_format == GL_DEPTH_STENCIL_EXT))
+	 tiling = I915_TILING_Y;
+      else
+	 tiling = I915_TILING_X;
+   } else
+      tiling = I915_TILING_NONE;
 
    mt = intel_miptree_create_internal(intel, target, internal_format,
 				      first_level, last_level, width0,
-				      height0, depth0, cpp, compress_byte);
+				      height0, depth0, cpp, compress_byte,
+				      tiling);
    /*
     * pitch == 0 || height == 0  indicates the null texture
     */
@@ -117,10 +135,12 @@ intel_miptree_create(struct intel_context *intel,
       return NULL;
 
    mt->region = intel_region_alloc(intel,
+				   tiling,
 				   mt->cpp,
 				   mt->pitch,
 				   mt->total_height,
-				   mt->pitch);
+				   mt->pitch,
+				   expect_accelerated_upload);
 
    if (!mt->region) {
        free(mt);
@@ -145,7 +165,8 @@ intel_miptree_create_for_region(struct intel_context *intel,
    mt = intel_miptree_create_internal(intel, target, internal_format,
 				      first_level, last_level,
 				      region->width, region->height, 1,
-				      region->cpp, compress_byte);
+				      region->cpp, compress_byte,
+				      I915_TILING_NONE);
    if (!mt)
       return mt;
 #if 0
@@ -183,6 +204,7 @@ intel_miptree_create_for_region(struct intel_context *intel,
 
 int intel_miptree_pitch_align (struct intel_context *intel,
 			       struct intel_mipmap_tree *mt,
+			       uint32_t tiling,
 			       int pitch)
 {
 #ifdef I915
@@ -203,6 +225,11 @@ int intel_miptree_pitch_align (struct intel_context *intel,
 	 pitch_align = 4;
       }
 
+      if (tiling == I915_TILING_X)
+	 pitch_align = 512;
+      else if (tiling == I915_TILING_Y)
+	 pitch_align = 128;
+
       pitch = ALIGN(pitch * mt->cpp, pitch_align);
 
 #ifdef I915
@@ -326,23 +353,31 @@ intel_miptree_set_level_info(struct intel_mipmap_tree *mt,
 }
 
 
-
 void
-intel_miptree_set_image_offset(struct intel_mipmap_tree *mt,
-			       GLuint level, GLuint img,
-			       GLuint x, GLuint y)
+intel_miptree_set_image_offset_ex(struct intel_mipmap_tree *mt,
+                                  GLuint level, GLuint img,
+                                  GLuint x, GLuint y, 
+                                  GLuint offset)
 {
    if (img == 0 && level == 0)
       assert(x == 0 && y == 0);
 
    assert(img < mt->level[level].nr_images);
 
-   mt->level[level].image_offset[img] = (x + y * mt->pitch) * mt->cpp;
+   mt->level[level].image_offset[img] = (x + y * mt->pitch) * mt->cpp + offset;
 
    DBG("%s level %d img %d pos %d,%d image_offset %x\n",
        __FUNCTION__, level, img, x, y, mt->level[level].image_offset[img]);
 }
 
+void
+intel_miptree_set_image_offset(struct intel_mipmap_tree *mt,
+			       GLuint level, GLuint img,
+			       GLuint x, GLuint y)
+{
+    intel_miptree_set_image_offset_ex(mt, level, img, x, y, 0);
+}
+
 
 /* Although we use the image_offset[] array to store relative offsets
  * to cube faces, Mesa doesn't know anything about this and expects
@@ -452,11 +487,11 @@ intel_miptree_image_data(struct intel_context *intel,
 			0, 0,                             /* source x, y */
 			dst->level[level].width, height); /* width, height */
 
-      src += src_image_pitch * dst->cpp;
+      src = (char *)src + src_image_pitch * dst->cpp;
    }
 }
 
-extern GLuint intel_compressed_alignment(GLenum);
+extern void intel_get_texture_alignment_unit(GLenum, GLuint *, GLuint *);
 /* Copy mipmap image between trees
  */
 void
@@ -473,20 +508,37 @@ intel_miptree_image_copy(struct intel_context *intel,
    const GLuint *dst_depth_offset = intel_miptree_depth_offsets(dst, level);
    const GLuint *src_depth_offset = intel_miptree_depth_offsets(src, level);
    GLuint i;
+   GLboolean success;
 
    if (dst->compressed) {
-       GLuint alignment = intel_compressed_alignment(dst->internal_format);
+       GLuint align_w, align_h;
+
+       intel_get_texture_alignment_unit(dst->internal_format, &align_w, &align_h);
        height = (height + 3) / 4;
-       width = ((width + alignment - 1) & ~(alignment - 1));
+       width = ALIGN(width, align_w);
    }
 
    for (i = 0; i < depth; i++) {
-      intel_region_copy(intel,
-                        dst->region, dst_offset + dst_depth_offset[i],
-                        0,
-                        0,
-                        src->region, src_offset + src_depth_offset[i],
-                        0, 0, width, height);
+      success = intel_region_copy(intel,
+				  dst->region, dst_offset + dst_depth_offset[i],
+				  0, 0,
+				  src->region, src_offset + src_depth_offset[i],
+				  0, 0, width, height, GL_COPY);
+      if (!success) {
+	 GLubyte *src_ptr, *dst_ptr;
+
+	 src_ptr = intel_region_map(intel, src->region);
+	 dst_ptr = intel_region_map(intel, dst->region);
+
+	 _mesa_copy_rect(dst_ptr + dst_offset + dst_depth_offset[i],
+			 dst->cpp,
+			 dst->pitch,
+			 0, 0, width, height,
+			 src_ptr + src_offset + src_depth_offset[i],
+			 src->pitch,
+			 0, 0);
+	 intel_region_unmap(intel, src->region);
+	 intel_region_unmap(intel, dst->region);
+      }
    }
-
 }
diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.h b/src/mesa/drivers/dri/intel/intel_mipmap_tree.h
index c9537dbb9a..c890b2a0d0 100644
--- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.h
+++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.h
@@ -126,6 +126,7 @@ struct intel_mipmap_tree
 
 struct intel_mipmap_tree *intel_miptree_create(struct intel_context *intel,
                                                GLenum target,
+                                               GLenum base_format,
                                                GLenum internal_format,
                                                GLuint first_level,
                                                GLuint last_level,
@@ -133,7 +134,8 @@ struct intel_mipmap_tree *intel_miptree_create(struct intel_context *intel,
                                                GLuint height0,
                                                GLuint depth0,
                                                GLuint cpp,
-                                               GLuint compress_byte);
+                                               GLuint compress_byte,
+					       GLboolean expect_accelerated_upload);
 
 struct intel_mipmap_tree *
 intel_miptree_create_for_region(struct intel_context *intel,
@@ -147,6 +149,7 @@ intel_miptree_create_for_region(struct intel_context *intel,
 
 int intel_miptree_pitch_align (struct intel_context *intel,
 			       struct intel_mipmap_tree *mt,
+			       uint32_t tiling,
 			       int pitch);
 
 void intel_miptree_reference(struct intel_mipmap_tree **dst,
@@ -193,6 +196,11 @@ void intel_miptree_set_level_info(struct intel_mipmap_tree *mt,
                                   GLuint x, GLuint y,
                                   GLuint w, GLuint h, GLuint d);
 
+void intel_miptree_set_image_offset_ex(struct intel_mipmap_tree *mt,
+                                       GLuint level,
+                                       GLuint img, GLuint x, GLuint y,
+                                       GLuint offset);
+
 void intel_miptree_set_image_offset(struct intel_mipmap_tree *mt,
                                     GLuint level,
                                     GLuint img, GLuint x, GLuint y);
@@ -217,10 +225,13 @@ void intel_miptree_image_copy(struct intel_context *intel,
 /* i915_mipmap_tree.c:
  */
 GLboolean i915_miptree_layout(struct intel_context *intel,
-			      struct intel_mipmap_tree *mt);
+			      struct intel_mipmap_tree *mt,
+			      uint32_t tiling);
 GLboolean i945_miptree_layout(struct intel_context *intel,
-			      struct intel_mipmap_tree *mt);
+			      struct intel_mipmap_tree *mt,
+			      uint32_t tiling);
 GLboolean brw_miptree_layout(struct intel_context *intel,
-			     struct intel_mipmap_tree *mt);
+			     struct intel_mipmap_tree *mt,
+			     uint32_t tiling);
 
 #endif
diff --git a/src/mesa/drivers/dri/intel/intel_pixel.c b/src/mesa/drivers/dri/intel/intel_pixel.c
index cf2f32d384..b5850df9ff 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel.c
@@ -25,11 +25,17 @@
  * 
  **************************************************************************/
 
+#include "main/accum.h"
+#include "main/drawpix.h"
 #include "main/enums.h"
 #include "main/state.h"
+#include "main/bufferobj.h"
 #include "main/context.h"
 #include "main/enable.h"
 #include "main/matrix.h"
+#include "main/texstate.h"
+#include "main/varray.h"
+#include "main/viewport.h"
 #include "swrast/swrast.h"
 #include "shader/arbprogram.h"
 #include "shader/program.h"
@@ -112,7 +118,7 @@ intel_check_blit_fragment_ops(GLcontext * ctx, GLboolean src_alpha_is_one)
       return GL_FALSE;
    }
 
-   if (ctx->Stencil.Enabled) {
+   if (ctx->Stencil._Enabled) {
       DBG("fallback due to image stencil\n");
       return GL_FALSE;
    }
@@ -173,179 +179,12 @@ intel_check_blit_format(struct intel_region * region,
 }
 
 void
-intel_meta_set_passthrough_transform(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-
-   intel->meta.saved_vp_x = ctx->Viewport.X;
-   intel->meta.saved_vp_y = ctx->Viewport.Y;
-   intel->meta.saved_vp_width = ctx->Viewport.Width;
-   intel->meta.saved_vp_height = ctx->Viewport.Height;
-
-   _mesa_Viewport(0, 0, ctx->DrawBuffer->Width, ctx->DrawBuffer->Height);
-
-   _mesa_MatrixMode(GL_PROJECTION);
-   _mesa_PushMatrix();
-   _mesa_LoadIdentity();
-   _mesa_Ortho(0, ctx->DrawBuffer->Width, 0, ctx->DrawBuffer->Height, 1, -1);
-
-   _mesa_MatrixMode(GL_MODELVIEW);
-   _mesa_PushMatrix();
-   _mesa_LoadIdentity();
-}
-
-void
-intel_meta_restore_transform(struct intel_context *intel)
-{
-   _mesa_MatrixMode(GL_PROJECTION);
-   _mesa_PopMatrix();
-   _mesa_MatrixMode(GL_MODELVIEW);
-   _mesa_PopMatrix();
-
-   _mesa_Viewport(intel->meta.saved_vp_x, intel->meta.saved_vp_y,
-		  intel->meta.saved_vp_width, intel->meta.saved_vp_height);
-}
-
-/**
- * Set up a vertex program to pass through the position and first texcoord
- * for pixel path.
- */
-void
-intel_meta_set_passthrough_vertex_program(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-   static const char *vp =
-      "!!ARBvp1.0\n"
-      "TEMP vertexClip;\n"
-      "DP4 vertexClip.x, state.matrix.mvp.row[0], vertex.position;\n"
-      "DP4 vertexClip.y, state.matrix.mvp.row[1], vertex.position;\n"
-      "DP4 vertexClip.z, state.matrix.mvp.row[2], vertex.position;\n"
-      "DP4 vertexClip.w, state.matrix.mvp.row[3], vertex.position;\n"
-      "MOV result.position, vertexClip;\n"
-      "MOV result.texcoord[0], vertex.texcoord[0];\n"
-      "MOV result.color, vertex.color;\n"
-      "END\n";
-
-   assert(intel->meta.saved_vp == NULL);
-
-   _mesa_reference_vertprog(ctx, &intel->meta.saved_vp,
-			    ctx->VertexProgram.Current);
-   if (intel->meta.passthrough_vp == NULL) {
-      GLuint prog_name;
-      _mesa_GenPrograms(1, &prog_name);
-      _mesa_BindProgram(GL_VERTEX_PROGRAM_ARB, prog_name);
-      _mesa_ProgramStringARB(GL_VERTEX_PROGRAM_ARB,
-			     GL_PROGRAM_FORMAT_ASCII_ARB,
-			     strlen(vp), (const GLubyte *)vp);
-      _mesa_reference_vertprog(ctx, &intel->meta.passthrough_vp,
-			       ctx->VertexProgram.Current);
-      _mesa_DeletePrograms(1, &prog_name);
-   }
-
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
-			    intel->meta.passthrough_vp);
-   ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB,
-			   &intel->meta.passthrough_vp->Base);
-
-   intel->meta.saved_vp_enable = ctx->VertexProgram.Enabled;
-   _mesa_Enable(GL_VERTEX_PROGRAM_ARB);
-}
-
-/**
- * Restores the previous vertex program after
- * intel_meta_set_passthrough_vertex_program()
- */
-void
-intel_meta_restore_vertex_program(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-   _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current,
-			    intel->meta.saved_vp);
-   _mesa_reference_vertprog(ctx, &intel->meta.saved_vp, NULL);
-   ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB,
-			   &ctx->VertexProgram.Current->Base);
-
-   if (!intel->meta.saved_vp_enable)
-      _mesa_Disable(GL_VERTEX_PROGRAM_ARB);
-}
-
-/**
- * Binds the given program string to GL_FRAGMENT_PROGRAM_ARB, caching the
- * program object.
- */
-void
-intel_meta_set_fragment_program(struct intel_context *intel,
-				struct gl_fragment_program **prog,
-				const char *prog_string)
-{
-   GLcontext *ctx = &intel->ctx;
-   assert(intel->meta.saved_fp == NULL);
-
-   _mesa_reference_fragprog(ctx, &intel->meta.saved_fp,
-			    ctx->FragmentProgram.Current);
-   if (*prog == NULL) {
-      GLuint prog_name;
-      _mesa_GenPrograms(1, &prog_name);
-      _mesa_BindProgram(GL_FRAGMENT_PROGRAM_ARB, prog_name);
-      _mesa_ProgramStringARB(GL_FRAGMENT_PROGRAM_ARB,
-			     GL_PROGRAM_FORMAT_ASCII_ARB,
-			     strlen(prog_string), (const GLubyte *)prog_string);
-      _mesa_reference_fragprog(ctx, prog, ctx->FragmentProgram.Current);
-      /* Note that DeletePrograms unbinds the program on us */
-      _mesa_DeletePrograms(1, &prog_name);
-   }
-
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, *prog);
-   ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, &((*prog)->Base));
-
-   intel->meta.saved_fp_enable = ctx->FragmentProgram.Enabled;
-   _mesa_Enable(GL_FRAGMENT_PROGRAM_ARB);
-}
-
-/**
- * Restores the previous fragment program after
- * intel_meta_set_fragment_program()
- */
-void
-intel_meta_restore_fragment_program(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-
-   FLUSH_VERTICES(ctx, _NEW_PROGRAM);
-   _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current,
-			    intel->meta.saved_fp);
-   _mesa_reference_fragprog(ctx, &intel->meta.saved_fp, NULL);
-   ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB,
-			   &ctx->FragmentProgram.Current->Base);
-
-   if (!intel->meta.saved_fp_enable)
-      _mesa_Disable(GL_FRAGMENT_PROGRAM_ARB);
-}
-
-void
 intelInitPixelFuncs(struct dd_function_table *functions)
 {
-   functions->Accum = _swrast_Accum;
+   _MESA_INIT_ACCUM_FUNCTIONS(functions, _swrast_);
    if (!getenv("INTEL_NO_BLIT")) {
-      functions->Bitmap = intelBitmap;
-      functions->CopyPixels = intelCopyPixels;
-      functions->DrawPixels = intelDrawPixels;
-#ifdef I915
-      functions->ReadPixels = intelReadPixels;
-#endif
+      _MESA_INIT_DRAWPIX_FUNCTIONS(functions, intel);
    }
-}
-
-void
-intel_free_pixel_state(struct intel_context *intel)
-{
-   GLcontext *ctx = &intel->ctx;
-
-   _mesa_reference_vertprog(ctx, &intel->meta.passthrough_vp, NULL);
-   _mesa_reference_fragprog(ctx, &intel->meta.bitmap_fp, NULL);
+   functions->ReadPixels = intelReadPixels;
 }
 
diff --git a/src/mesa/drivers/dri/intel/intel_pixel.h b/src/mesa/drivers/dri/intel/intel_pixel.h
index 76b8781316..96a6dd17b2 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel.h
+++ b/src/mesa/drivers/dri/intel/intel_pixel.h
@@ -31,16 +31,6 @@
 #include "main/mtypes.h"
 
 void intelInitPixelFuncs(struct dd_function_table *functions);
-void intel_meta_set_passthrough_transform(struct intel_context *intel);
-void intel_meta_restore_transform(struct intel_context *intel);
-void intel_meta_set_passthrough_vertex_program(struct intel_context *intel);
-void intel_meta_restore_vertex_program(struct intel_context *intel);
-void intel_meta_set_fragment_program(struct intel_context *intel,
-				     struct gl_fragment_program **prog,
-				     const char *prog_string);
-void intel_meta_restore_fragment_program(struct intel_context *intel);
-void intel_free_pixel_state(struct intel_context *intel);
-
 GLboolean intel_check_blit_fragment_ops(GLcontext * ctx,
 					GLboolean src_alpha_is_one);
 
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
index 3a01f63dc7..9a0bcc07a5 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
@@ -42,6 +42,7 @@
 #include "main/varray.h"
 #include "main/attrib.h"
 #include "main/enable.h"
+#include "main/viewport.h"
 #include "shader/arbprogram.h"
 #include "glapi/dispatch.h"
 #include "swrast/swrast.h"
@@ -92,19 +93,12 @@ static const GLubyte *map_pbo( GLcontext *ctx,
    return ADD_POINTERS(buf, bitmap);
 }
 
-static GLboolean test_bit( const GLubyte *src,
-			    GLuint bit )
+static GLboolean test_bit( const GLubyte *src, GLuint bit )
 {
    return (src[bit/8] & (1<<(bit % 8))) ? 1 : 0;
 }
 
-static GLboolean test_msb_bit(const GLubyte *src, GLuint bit)
-{
-   return (src[bit/8] & (1<<(7 - (bit % 8)))) ? 1 : 0;
-}
-
-static void set_bit( GLubyte *dest,
-			  GLuint bit )
+static void set_bit( GLubyte *dest, GLuint bit )
 {
    dest[bit/8] |= 1 << (bit % 8);
 }
@@ -194,7 +188,7 @@ do_blit_bitmap( GLcontext *ctx,
    struct gl_framebuffer *fb = ctx->DrawBuffer;
    GLfloat tmpColor[4];
    GLubyte ubcolor[4];
-   GLuint color8888, color565;
+   GLuint color;
    unsigned int num_cliprects;
    drm_clip_rect_t *cliprects;
    int x_off, y_off;
@@ -215,7 +209,7 @@ do_blit_bitmap( GLcontext *ctx,
    if (!dst)
        return GL_FALSE;
 
-   if (unpack->BufferObj->Name) {
+   if (_mesa_is_bufferobj(unpack->BufferObj)) {
       bitmap = map_pbo(ctx, width, height, unpack, bitmap);
       if (bitmap == NULL)
 	 return GL_TRUE;	/* even though this is an error, we're done */
@@ -232,8 +226,11 @@ do_blit_bitmap( GLcontext *ctx,
    UNCLAMPED_FLOAT_TO_UBYTE(ubcolor[2], tmpColor[2]);
    UNCLAMPED_FLOAT_TO_UBYTE(ubcolor[3], tmpColor[3]);
 
-   color8888 = INTEL_PACKCOLOR8888(ubcolor[0], ubcolor[1], ubcolor[2], ubcolor[3]);
-   color565 = INTEL_PACKCOLOR565(ubcolor[0], ubcolor[1], ubcolor[2]);
+   if (dst->cpp == 2)
+      color = INTEL_PACKCOLOR565(ubcolor[0], ubcolor[1], ubcolor[2]);
+   else
+      color = INTEL_PACKCOLOR8888(ubcolor[0], ubcolor[1],
+				  ubcolor[2], ubcolor[3]);
 
    if (!intel_check_blit_fragment_ops(ctx, tmpColor[3] == 1.0F))
       return GL_FALSE;
@@ -307,21 +304,21 @@ do_blit_bitmap( GLcontext *ctx,
 				   fb->Name == 0 ? GL_TRUE : GL_FALSE) == 0)
 		  continue;
 
-	       /* 
-		*/
-	       intelEmitImmediateColorExpandBlit( intel,
-						  dst->cpp,
-						  (GLubyte *)stipple, 
-						  sz,
-						  (dst->cpp == 2) ? color565 : color8888,
-						  dst->pitch,
-						  dst->buffer,
-						  0,
-						  dst->tiling,
-						  box_x + px,
-						  box_y + py,
-						  w, h,
-						  logic_op);
+	       if (!intelEmitImmediateColorExpandBlit(intel,
+						      dst->cpp,
+						      (GLubyte *)stipple,
+						      sz,
+						      color,
+						      dst->pitch,
+						      dst->buffer,
+						      0,
+						      dst->tiling,
+						      box_x + px,
+						      box_y + py,
+						      w, h,
+						      logic_op)) {
+		  return GL_FALSE;
+	       }
 	    } 
 	 } 
       }
@@ -332,7 +329,7 @@ out:
    if (INTEL_DEBUG & DEBUG_SYNC)
       intel_batchbuffer_flush(intel->batch);
 
-   if (unpack->BufferObj->Name) {
+   if (_mesa_is_bufferobj(unpack->BufferObj)) {
       /* done with PBO so unmap it now */
       ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
                               unpack->BufferObj);
@@ -360,11 +357,8 @@ intel_texture_bitmap(GLcontext * ctx,
       "END\n";
    GLuint texname;
    GLfloat vertices[4][4];
-   GLfloat texcoords[4][2];
    GLint old_active_texture;
-   GLubyte *unpacked_bitmap;
    GLubyte *a8_bitmap;
-   int x, y;
    GLfloat dst_z;
 
    /* We need a fragment program for the KIL effect */
@@ -401,6 +395,20 @@ intel_texture_bitmap(GLcontext * ctx,
       return GL_FALSE;
    }
 
+   if (!ctx->Extensions.ARB_texture_non_power_of_two &&
+       (!is_power_of_two(width) || !is_power_of_two(height))) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr,
+		 "glBitmap() fallback: NPOT texture\n");
+      return GL_FALSE;
+   }
+
+   if (ctx->Fog.Enabled) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "glBitmap() fallback: fog\n");
+      return GL_FALSE;
+   }
+
    /* Check that we can load in a texture this big. */
    if (width > (1 << (ctx->Const.MaxTextureLevels - 1)) ||
        height > (1 << (ctx->Const.MaxTextureLevels - 1))) {
@@ -410,23 +418,17 @@ intel_texture_bitmap(GLcontext * ctx,
       return GL_FALSE;
    }
 
-   /* Convert the A1 bitmap to an A8 format suitable for glTexImage */
-   if (unpack->BufferObj->Name) {
+   if (_mesa_is_bufferobj(unpack->BufferObj)) {
       bitmap = map_pbo(ctx, width, height, unpack, bitmap);
       if (bitmap == NULL)
 	 return GL_TRUE;	/* even though this is an error, we're done */
    }
-   unpacked_bitmap = _mesa_unpack_bitmap(width, height, bitmap,
-					 unpack);
+
+   /* Convert the A1 bitmap to an A8 format suitable for glTexImage */
    a8_bitmap = _mesa_calloc(width * height);
-   for (y = 0; y < height; y++) {
-      for (x = 0; x < width; x++) {
-	 if (test_msb_bit(unpacked_bitmap, ALIGN(width, 8) * y + x))
-	    a8_bitmap[y * width + x] = 0xff;
-      }
-   }
-   _mesa_free(unpacked_bitmap);
-   if (unpack->BufferObj->Name) {
+   _mesa_expand_bitmap(width, height, unpack, bitmap, a8_bitmap, width, 0xff);
+
+   if (_mesa_is_bufferobj(unpack->BufferObj)) {
       /* done with PBO so unmap it now */
       ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
                               unpack->BufferObj);
@@ -459,15 +461,18 @@ intel_texture_bitmap(GLcontext * ctx,
 		    GL_ALPHA, GL_UNSIGNED_BYTE, a8_bitmap);
    _mesa_free(a8_bitmap);
 
-   intel_meta_set_fragment_program(intel, &intel->meta.bitmap_fp, fp);
+   meta_set_fragment_program(&intel->meta, &intel->meta.bitmap_fp, fp);
    _mesa_ProgramLocalParameter4fvARB(GL_FRAGMENT_PROGRAM_ARB, 0,
 				     ctx->Current.RasterColor);
-   intel_meta_set_passthrough_vertex_program(intel);
-   intel_meta_set_passthrough_transform(intel);
+   meta_set_passthrough_vertex_program(&intel->meta);
+   meta_set_passthrough_transform(&intel->meta);
 
    /* convert rasterpos Z from [0,1] to NDC coord in [-1,1] */
    dst_z = -1.0 + 2.0 * ctx->Current.RasterPos[2];
 
+   /* RasterPos[2] already takes into account the DepthRange mapping. */
+   _mesa_DepthRange(0.0, 1.0);
+
    vertices[0][0] = dst_x;
    vertices[0][1] = dst_y;
    vertices[0][2] = dst_z;
@@ -485,24 +490,15 @@ intel_texture_bitmap(GLcontext * ctx,
    vertices[3][2] = dst_z;
    vertices[3][3] = 1.0;
 
-   texcoords[0][0] = 0.0;
-   texcoords[0][1] = 0.0;
-   texcoords[1][0] = 1.0;
-   texcoords[1][1] = 0.0;
-   texcoords[2][0] = 1.0;
-   texcoords[2][1] = 1.0;
-   texcoords[3][0] = 0.0;
-   texcoords[3][1] = 1.0;
-
    _mesa_VertexPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &vertices);
-   _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &texcoords);
    _mesa_Enable(GL_VERTEX_ARRAY);
-   _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
-   CALL_DrawArrays(ctx->Exec, (GL_TRIANGLE_FAN, 0, 4));
+   meta_set_default_texrect(&intel->meta);
+   _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
 
-   intel_meta_restore_transform(intel);
-   intel_meta_restore_fragment_program(intel);
-   intel_meta_restore_vertex_program(intel);
+   meta_restore_texcoords(&intel->meta);
+   meta_restore_transform(&intel->meta);
+   meta_restore_fragment_program(&intel->meta);
+   meta_restore_vertex_program(&intel->meta);
 
    _mesa_PopClientAttrib();
    _mesa_Disable(GL_TEXTURE_2D); /* asserted that it was disabled at entry */
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_copy.c b/src/mesa/drivers/dri/intel/intel_pixel_copy.c
index 7c7aa6097c..07ca8f7ddb 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_copy.c
@@ -26,18 +26,13 @@
  **************************************************************************/
 
 #include "main/glheader.h"
-#include "main/enums.h"
 #include "main/image.h"
 #include "main/state.h"
 #include "main/mtypes.h"
-#include "main/macros.h"
-#include "swrast/swrast.h"
+#include "drivers/common/meta.h"
 
-#include "intel_screen.h"
 #include "intel_context.h"
-#include "intel_batchbuffer.h"
 #include "intel_buffers.h"
-#include "intel_blit.h"
 #include "intel_regions.h"
 #include "intel_pixel.h"
 
@@ -87,7 +82,7 @@ intel_check_copypixel_blit_fragment_ops(GLcontext * ctx)
             ctx->Color.AlphaEnabled ||
             ctx->Depth.Test ||
             ctx->Fog.Enabled ||
-            ctx->Stencil.Enabled ||
+            ctx->Stencil._Enabled ||
             !ctx->Color.ColorMask[0] ||
             !ctx->Color.ColorMask[1] ||
             !ctx->Color.ColorMask[2] ||
@@ -97,162 +92,6 @@ intel_check_copypixel_blit_fragment_ops(GLcontext * ctx)
 	    ctx->Color.BlendEnabled);
 }
 
-#ifdef I915
-/* Doesn't work for overlapping regions.  Could do a double copy or
- * just fallback.
- */
-static GLboolean
-do_texture_copypixels(GLcontext * ctx,
-                      GLint srcx, GLint srcy,
-                      GLsizei width, GLsizei height,
-                      GLint dstx, GLint dsty, GLenum type)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct intel_region *dst = intel_drawbuf_region(intel);
-   struct intel_region *src = copypix_src_region(intel, type);
-   GLenum src_format;
-   GLenum src_type;
-
-   DBG("%s %d,%d %dx%d --> %d,%d\n", __FUNCTION__, 
-       srcx, srcy, width, height, dstx, dsty);
-
-   if (!src || !dst || type != GL_COLOR)
-      return GL_FALSE;
-
-   if (ctx->_ImageTransferState) {
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-         fprintf(stderr, "%s: check_color failed\n", __FUNCTION__);
-      return GL_FALSE;
-   }
-
-   /* Can't handle overlapping regions.  Don't have sufficient control
-    * over rasterization to pull it off in-place.  Punt on these for
-    * now.
-    * 
-    * XXX: do a copy to a temporary. 
-    */
-   if (src->buffer == dst->buffer) {
-      drm_clip_rect_t srcbox;
-      drm_clip_rect_t dstbox;
-      drm_clip_rect_t tmp;
-
-      srcbox.x1 = srcx;
-      srcbox.y1 = srcy;
-      srcbox.x2 = srcx + width;
-      srcbox.y2 = srcy + height;
-
-      if (ctx->Pixel.ZoomX > 0) {
-	 dstbox.x1 = dstx;
-	 dstbox.x2 = dstx + width * ctx->Pixel.ZoomX;
-      } else {
-	 dstbox.x1 = dstx + width * ctx->Pixel.ZoomX;
-	 dstbox.x2 = dstx;
-      }
-      if (ctx->Pixel.ZoomY > 0) {
-	 dstbox.y1 = dsty;
-	 dstbox.y2 = dsty + height * ctx->Pixel.ZoomY;
-      } else {
-	 dstbox.y1 = dsty + height * ctx->Pixel.ZoomY;
-	 dstbox.y2 = dsty;
-      }
-
-      DBG("src %d,%d %d,%d\n", srcbox.x1, srcbox.y1, srcbox.x2, srcbox.y2);
-      DBG("dst %d,%d %d,%d (%dx%d) (%f,%f)\n", dstbox.x1, dstbox.y1, dstbox.x2, dstbox.y2,
-	  width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY);
-
-      if (intel_intersect_cliprects(&tmp, &srcbox, &dstbox)) {
-         DBG("%s: regions overlap\n", __FUNCTION__);
-         return GL_FALSE;
-      }
-   }
-
-   intelFlush(&intel->ctx);
-
-   intel->vtbl.install_meta_state(intel);
-
-   /* Is this true?  Also will need to turn depth testing on according
-    * to state:
-    */
-   intel->vtbl.meta_no_stencil_write(intel);
-   intel->vtbl.meta_no_depth_write(intel);
-
-   /* Set the 3d engine to draw into the destination region:
-    */
-   intel->vtbl.meta_draw_region(intel, dst, intel->depth_region);
-
-   intel->vtbl.meta_import_pixel_state(intel);
-
-   if (src->cpp == 2) {
-      src_format = GL_RGB;
-      src_type = GL_UNSIGNED_SHORT_5_6_5;
-   }
-   else {
-      src_format = GL_BGRA;
-      src_type = GL_UNSIGNED_BYTE;
-   }
-
-   /* Set the frontbuffer up as a large rectangular texture.
-    */
-   if (!intel->vtbl.meta_tex_rect_source(intel, src->buffer, 0,
-                                         src->pitch,
-                                         src->height, src_format, src_type)) {
-      intel->vtbl.leave_meta_state(intel);
-      return GL_FALSE;
-   }
-
-
-   intel->vtbl.meta_texture_blend_replace(intel);
-
-   LOCK_HARDWARE(intel);
-
-   if (intel->driDrawable->numClipRects) {
-      __DRIdrawablePrivate *dPriv = intel->driDrawable;
-
-
-      srcy = dPriv->h - srcy - height;  /* convert from gl to hardware coords */
-
-      srcx += dPriv->x;
-      srcy += dPriv->y;
-
-      /* Clip against the source region.  This is the only source
-       * clipping we do.  XXX: Just set the texcord wrap mode to clamp
-       * or similar.
-       *
-       */
-      if (0) {
-         GLint orig_x = srcx;
-         GLint orig_y = srcy;
-
-         if (!_mesa_clip_to_region(0, 0, src->pitch, src->height,
-                                   &srcx, &srcy, &width, &height))
-            goto out;
-
-         dstx += srcx - orig_x;
-         dsty += (srcy - orig_y) * ctx->Pixel.ZoomY;
-      }
-
-      /* Just use the regular cliprect mechanism...  Does this need to
-       * even hold the lock???
-       */
-      intel->vtbl.meta_draw_quad(intel,
-				 dstx,
-				 dstx + width * ctx->Pixel.ZoomX,
-				 dPriv->h - (dsty + height * ctx->Pixel.ZoomY),
-				 dPriv->h - (dsty), 0, /* XXX: what z value? */
-				 0x00ff00ff,
-				 srcx, srcx + width, srcy, srcy + height);
-
-    out:
-      intel->vtbl.leave_meta_state(intel);
-      intel_batchbuffer_emit_mi_flush(intel->batch);
-   }
-   UNLOCK_HARDWARE(intel);
-
-   DBG("%s: success\n", __FUNCTION__);
-   return GL_TRUE;
-}
-#endif /* I915 */
-
 
 /**
  * CopyPixels with the blitter.  Don't support zooming, pixel transfer, etc.
@@ -272,6 +111,12 @@ do_blit_copypixels(GLcontext * ctx,
    drm_clip_rect_t *cliprects;
    int x_off, y_off;
 
+   if (type == GL_DEPTH || type == GL_STENCIL) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "glCopyPixels() fallback: GL_DEPTH || GL_STENCIL\n");
+      return GL_FALSE;
+   }
+
    /* Update draw buffer bounds */
    _mesa_update_state(ctx);
 
@@ -362,14 +207,16 @@ do_blit_copypixels(GLcontext * ctx,
 				   &clip_x, &clip_y, &clip_w, &clip_h))
             continue;
 
-         intelEmitCopyBlit(intel, dst->cpp,
-			   src->pitch, src->buffer, 0, src->tiling,
-			   dst->pitch, dst->buffer, 0, dst->tiling,
-			   clip_x + delta_x, clip_y + delta_y, /* srcx, srcy */
-			   clip_x, clip_y, /* dstx, dsty */
-			   clip_w, clip_h,
-			   ctx->Color.ColorLogicOpEnabled ?
-			   ctx->Color.LogicOp : GL_COPY);
+	 if (!intel_region_copy(intel,
+				dst, 0, clip_x, clip_y,
+				src, 0, clip_x + delta_x, clip_y + delta_y,
+				clip_w, clip_h,
+				ctx->Color.ColorLogicOpEnabled ?
+				ctx->Color.LogicOp : GL_COPY)) {
+	    DBG("%s: blit failure\n", __FUNCTION__);
+	    UNLOCK_HARDWARE(intel);
+	    return GL_FALSE;
+	 }
       }
    }
 out:
@@ -392,12 +239,6 @@ intelCopyPixels(GLcontext * ctx,
    if (do_blit_copypixels(ctx, srcx, srcy, width, height, destx, desty, type))
       return;
 
-#ifdef I915
-   if (do_texture_copypixels(ctx, srcx, srcy, width, height, destx, desty, type))
-      return;
-#endif
-
-   DBG("fallback to _swrast_CopyPixels\n");
-
-   _swrast_CopyPixels(ctx, srcx, srcy, width, height, destx, desty, type);
+   /* this will use swrast if needed */
+   _mesa_meta_copy_pixels(ctx, srcx, srcy, width, height, destx, desty, type);
 }
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_draw.c b/src/mesa/drivers/dri/intel/intel_pixel_draw.c
index 0e83afa645..7fbb89fd6a 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_draw.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_draw.c
@@ -29,8 +29,6 @@
 #include "main/enums.h"
 #include "main/image.h"
 #include "main/mtypes.h"
-#include "main/macros.h"
-#include "main/bufferobj.h"
 #include "main/teximage.h"
 #include "main/texenv.h"
 #include "main/texobj.h"
@@ -41,155 +39,22 @@
 #include "main/enable.h"
 #include "main/buffers.h"
 #include "main/fbobject.h"
-#include "main/renderbuffer.h"
 #include "main/depth.h"
 #include "main/hash.h"
 #include "main/blend.h"
-#include "glapi/dispatch.h"
 #include "swrast/swrast.h"
+#include "drivers/common/meta.h"
 
-#include "intel_screen.h"
 #include "intel_context.h"
 #include "intel_batchbuffer.h"
 #include "intel_blit.h"
 #include "intel_buffers.h"
 #include "intel_regions.h"
 #include "intel_pixel.h"
-#include "intel_buffer_objects.h"
 #include "intel_fbo.h"
 
-static GLboolean
-intel_texture_drawpixels(GLcontext * ctx,
-			 GLint x, GLint y,
-			 GLsizei width, GLsizei height,
-			 GLenum format,
-			 GLenum type,
-			 const struct gl_pixelstore_attrib *unpack,
-			 const GLvoid *pixels)
-{
-   struct intel_context *intel = intel_context(ctx);
-   GLuint texname;
-   GLfloat vertices[4][4];
-   GLfloat texcoords[4][2];
-   GLfloat z;
-
-   /* We're going to mess with texturing with no regard to existing texture
-    * state, so if there is some set up we have to bail.
-    */
-   if (ctx->Texture._EnabledUnits != 0) {
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "glDrawPixels() fallback: texturing enabled\n");
-      return GL_FALSE;
-   }
-
-   /* Can't do textured DrawPixels with a fragment program, unless we were
-    * to generate a new program that sampled our texture and put the results
-    * in the fragment color before the user's program started.
-    */
-   if (ctx->FragmentProgram.Enabled) {
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "glDrawPixels() fallback: fragment program enabled\n");
-      return GL_FALSE;
-   }
-
-   /* We don't have a way to generate fragments with stencil values which *
-    * will set the resulting stencil value.
-    */
-   if (format == GL_STENCIL_INDEX)
-      return GL_FALSE;
-
-   /* Check that we can load in a texture this big. */
-   if (width > (1 << (ctx->Const.MaxTextureLevels - 1)) ||
-       height > (1 << (ctx->Const.MaxTextureLevels - 1))) {
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "glDrawPixels() fallback: bitmap too large (%dx%d)\n",
-		 width, height);
-      return GL_FALSE;
-   }
-
-   /* To do DEPTH_COMPONENT, we would need to change our setup to not draw to
-    * the color buffer, and sample the texture values into the fragment depth
-    * in a program.
-    */
-   if (format == GL_DEPTH_COMPONENT) {
-      if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr,
-		 "glDrawPixels() fallback: format == GL_DEPTH_COMPONENT\n");
-      return GL_FALSE;
-   }
-
-   _mesa_PushAttrib(GL_ENABLE_BIT | GL_TEXTURE_BIT |
-		    GL_CURRENT_BIT);
-   _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT);
-
-   /* XXX: pixel store stuff */
-   _mesa_Disable(GL_POLYGON_STIPPLE);
-
-   _mesa_ActiveTextureARB(GL_TEXTURE0_ARB);
-   _mesa_Enable(GL_TEXTURE_2D);
-   _mesa_GenTextures(1, &texname);
-   _mesa_BindTexture(GL_TEXTURE_2D, texname);
-   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
-   _mesa_TexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
-   _mesa_TexEnvf(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_REPLACE);
-   /*
-   _mesa_TexEnvf(GL_TEXTURE_ENV, GL_COMBINE_RGB, GL_REPLACE);
-   _mesa_TexEnvf(GL_TEXTURE_ENV, GL_COMBINE_ALPHA, GL_REPLACE);
-   */
-   _mesa_TexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, width, height, 0, format,
-		    type, pixels);
-
-   intel_meta_set_passthrough_transform(intel);
-
-   /* convert rasterpos Z from [0,1] to NDC coord in [-1,1] */
-   z = -1.0 + 2.0 * ctx->Current.RasterPos[2];
-
-   /* Create the vertex buffer based on the current raster pos.  The x and y
-    * we're handed are ctx->Current.RasterPos[0,1] rounded to integers.
-    * We also apply the depth.  However, the W component is already multiplied
-    * into ctx->Current.RasterPos[0,1,2] and we can ignore it at this point.
-    */
-   vertices[0][0] = x;
-   vertices[0][1] = y;
-   vertices[0][2] = z;
-   vertices[0][3] = 1.0;
-   vertices[1][0] = x + width * ctx->Pixel.ZoomX;
-   vertices[1][1] = y;
-   vertices[1][2] = z;
-   vertices[1][3] = 1.0;
-   vertices[2][0] = x + width * ctx->Pixel.ZoomX;
-   vertices[2][1] = y + height * ctx->Pixel.ZoomY;
-   vertices[2][2] = z;
-   vertices[2][3] = 1.0;
-   vertices[3][0] = x;
-   vertices[3][1] = y + height * ctx->Pixel.ZoomY;
-   vertices[3][2] = z;
-   vertices[3][3] = 1.0;
-
-   texcoords[0][0] = 0.0;
-   texcoords[0][1] = 0.0;
-   texcoords[1][0] = 1.0;
-   texcoords[1][1] = 0.0;
-   texcoords[2][0] = 1.0;
-   texcoords[2][1] = 1.0;
-   texcoords[3][0] = 0.0;
-   texcoords[3][1] = 1.0;
-
-   _mesa_VertexPointer(4, GL_FLOAT, 4 * sizeof(GLfloat), &vertices);
-   _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &texcoords);
-   _mesa_Enable(GL_VERTEX_ARRAY);
-   _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
-   CALL_DrawArrays(ctx->Exec, (GL_TRIANGLE_FAN, 0, 4));
-
-   intel_meta_restore_transform(intel);
-   _mesa_PopClientAttrib();
-   _mesa_PopAttrib();
-
-   _mesa_DeleteTextures(1, &texname);
-
-   return GL_TRUE;
-}
 
+/** XXX compare perf of this vs. _mesa_meta_draw_pixels(STENCIL) */
 static GLboolean
 intel_stencil_drawpixels(GLcontext * ctx,
 			 GLint x, GLint y,
@@ -202,13 +67,14 @@ intel_stencil_drawpixels(GLcontext * ctx,
    struct intel_context *intel = intel_context(ctx);
    GLuint texname, rb_name, fb_name, old_fb_name;
    GLfloat vertices[4][2];
-   GLfloat texcoords[4][2];
    struct intel_renderbuffer *irb;
    struct intel_renderbuffer *depth_irb;
    struct gl_renderbuffer *rb;
    struct gl_pixelstore_attrib old_unpack;
    GLstencil *stencil_pixels;
-   int row;
+   int row, y1, y2;
+   GLint old_active_texture;
+   GLboolean rendering_to_fbo = ctx->DrawBuffer->Name != 0;
 
    if (format != GL_STENCIL_INDEX)
       return GL_FALSE;
@@ -225,6 +91,10 @@ intel_stencil_drawpixels(GLcontext * ctx,
       return GL_FALSE;
    }
 
+   /* We don't support stencil testing/ops here */
+   if (ctx->Stencil._Enabled)
+      return GL_FALSE;
+
    /* We use FBOs for our wrapping of the depthbuffer into a color
     * destination.
     */
@@ -262,10 +132,19 @@ intel_stencil_drawpixels(GLcontext * ctx,
       return GL_FALSE;
    }
 
+   if (!ctx->Extensions.ARB_texture_non_power_of_two &&
+       (!is_power_of_two(width) || !is_power_of_two(height))) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr,
+		 "glDrawPixels(GL_STENCIL_INDEX) fallback: NPOT texture\n");
+      return GL_FALSE;
+   }
+
    _mesa_PushAttrib(GL_ENABLE_BIT | GL_TEXTURE_BIT |
 		    GL_CURRENT_BIT | GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
    _mesa_PushClientAttrib(GL_CLIENT_VERTEX_ARRAY_BIT);
    old_fb_name = ctx->DrawBuffer->Name;
+   old_active_texture = ctx->Texture.CurrentUnit;
 
    _mesa_Disable(GL_POLYGON_STIPPLE);
    _mesa_Disable(GL_DEPTH_TEST);
@@ -330,34 +209,37 @@ intel_stencil_drawpixels(GLcontext * ctx,
    ctx->Unpack = old_unpack;
    _mesa_free(stencil_pixels);
 
-   intel_meta_set_passthrough_transform(intel);
+   meta_set_passthrough_transform(&intel->meta);
 
+   /* Since we're rendering to the framebuffer as if it was an FBO,
+    * if it's the window system we have to flip the coordinates.
+    */
+   if (rendering_to_fbo) {
+      y1 = y;
+      y2 = y + height * ctx->Pixel.ZoomY;
+   } else {
+      y1 = irb->Base.Height - (y + height * ctx->Pixel.ZoomY);
+      y2 = irb->Base.Height - y;
+   }
    vertices[0][0] = x;
-   vertices[0][1] = y;
+   vertices[0][1] = y1;
    vertices[1][0] = x + width * ctx->Pixel.ZoomX;
-   vertices[1][1] = y;
+   vertices[1][1] = y1;
    vertices[2][0] = x + width * ctx->Pixel.ZoomX;
-   vertices[2][1] = y + height * ctx->Pixel.ZoomY;
+   vertices[2][1] = y2;
    vertices[3][0] = x;
-   vertices[3][1] = y + height * ctx->Pixel.ZoomY;
-
-   texcoords[0][0] = 0.0;
-   texcoords[0][1] = 0.0;
-   texcoords[1][0] = 1.0;
-   texcoords[1][1] = 0.0;
-   texcoords[2][0] = 1.0;
-   texcoords[2][1] = 1.0;
-   texcoords[3][0] = 0.0;
-   texcoords[3][1] = 1.0;
+   vertices[3][1] = y2;
 
    _mesa_VertexPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &vertices);
-   _mesa_TexCoordPointer(2, GL_FLOAT, 2 * sizeof(GLfloat), &texcoords);
    _mesa_Enable(GL_VERTEX_ARRAY);
-   _mesa_Enable(GL_TEXTURE_COORD_ARRAY);
-   CALL_DrawArrays(ctx->Exec, (GL_TRIANGLE_FAN, 0, 4));
+   meta_set_default_texrect(&intel->meta);
 
-   intel_meta_restore_transform(intel);
+   _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
 
+   meta_restore_texcoords(&intel->meta);
+   meta_restore_transform(&intel->meta);
+
+   _mesa_ActiveTextureARB(GL_TEXTURE0_ARB + old_active_texture);
    _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, old_fb_name);
 
    _mesa_PopClientAttrib();
@@ -379,17 +261,25 @@ intelDrawPixels(GLcontext * ctx,
                 const struct gl_pixelstore_attrib *unpack,
                 const GLvoid * pixels)
 {
-   if (intel_texture_drawpixels(ctx, x, y, width, height, format, type,
-				unpack, pixels))
-      return;
-
+#if 0
+   /* XXX this function doesn't seem to work reliably even when all
+    * the pre-requisite conditions are met.
+    * Note that this function is never hit with conform.
+    * Fall back to swrast because even the _mesa_meta_draw_pixels() approach
+    * isn't working because of an apparent stencil bug.
+    */
    if (intel_stencil_drawpixels(ctx, x, y, width, height, format, type,
 				unpack, pixels))
       return;
+#else
+   (void) intel_stencil_drawpixels; /* silence warning */
+   if (format == GL_STENCIL_INDEX) {
+      _swrast_DrawPixels(ctx, x, y, width, height, format, type,
+                         unpack, pixels);
+      return;
+   }
+#endif
 
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      _mesa_printf("%s: fallback to swrast\n", __FUNCTION__);
-
-   _swrast_DrawPixels(ctx, x, y, width, height, format, type,
-		      unpack, pixels);
+   _mesa_meta_draw_pixels(ctx, x, y, width, height, format, type,
+                          unpack, pixels);
 }
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_read.c b/src/mesa/drivers/dri/intel/intel_pixel_read.c
new file mode 100644
index 0000000000..bc67f6242a
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_pixel_read.c
@@ -0,0 +1,315 @@
+/**************************************************************************
+ *
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "main/glheader.h"
+#include "main/enums.h"
+#include "main/mtypes.h"
+#include "main/macros.h"
+#include "main/image.h"
+#include "main/bufferobj.h"
+#include "main/state.h"
+#include "swrast/swrast.h"
+
+#include "intel_screen.h"
+#include "intel_context.h"
+#include "intel_batchbuffer.h"
+#include "intel_blit.h"
+#include "intel_buffers.h"
+#include "intel_regions.h"
+#include "intel_pixel.h"
+#include "intel_buffer_objects.h"
+
+/* For many applications, the new ability to pull the source buffers
+ * back out of the GTT and then do the packing/conversion operations
+ * in software will be as much of an improvement as trying to get the
+ * blitter and/or texture engine to do the work.
+ *
+ * This step is gated on private backbuffers.
+ *
+ * Obviously the frontbuffer can't be pulled back, so that is either
+ * an argument for blit/texture readpixels, or for blitting to a
+ * temporary and then pulling that back.
+ *
+ * When the destination is a pbo, however, it's not clear if it is
+ * ever going to be pulled to main memory (though the access param
+ * will be a good hint).  So it sounds like we do want to be able to
+ * choose between blit/texture implementation on the gpu and pullback
+ * and cpu-based copying.
+ *
+ * Unless you can magically turn client memory into a PBO for the
+ * duration of this call, there will be a cpu-based copying step in
+ * any case.
+ */
+
+
+static GLboolean
+do_texture_readpixels(GLcontext * ctx,
+                      GLint x, GLint y, GLsizei width, GLsizei height,
+                      GLenum format, GLenum type,
+                      const struct gl_pixelstore_attrib *pack,
+                      struct intel_region *dest_region)
+{
+#if 0
+   struct intel_context *intel = intel_context(ctx);
+   intelScreenPrivate *screen = intel->intelScreen;
+   GLint pitch = pack->RowLength ? pack->RowLength : width;
+   __DRIdrawablePrivate *dPriv = intel->driDrawable;
+   int textureFormat;
+   GLenum glTextureFormat;
+   int destFormat, depthFormat, destPitch;
+   drm_clip_rect_t tmp;
+
+   if (INTEL_DEBUG & DEBUG_PIXEL)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+
+   if (ctx->_ImageTransferState ||
+       pack->SwapBytes || pack->LsbFirst || !pack->Invert) {
+      if (INTEL_DEBUG & DEBUG_PIXEL)
+         fprintf(stderr, "%s: check_color failed\n", __FUNCTION__);
+      return GL_FALSE;
+   }
+
+   intel->vtbl.meta_texrect_source(intel, intel_readbuf_region(intel));
+
+   if (!intel->vtbl.meta_render_dest(intel, dest_region, type, format)) {
+      if (INTEL_DEBUG & DEBUG_PIXEL)
+         fprintf(stderr, "%s: couldn't set dest %s/%s\n",
+                 __FUNCTION__,
+                 _mesa_lookup_enum_by_nr(type),
+                 _mesa_lookup_enum_by_nr(format));
+      return GL_FALSE;
+   }
+
+   LOCK_HARDWARE(intel);
+
+   if (intel->driDrawable->numClipRects) {
+      intel->vtbl.install_meta_state(intel);
+      intel->vtbl.meta_no_depth_write(intel);
+      intel->vtbl.meta_no_stencil_write(intel);
+
+      if (!driClipRectToFramebuffer(ctx->ReadBuffer, &x, &y, &width, &height)) {
+         UNLOCK_HARDWARE(intel);
+         SET_STATE(i830, state);
+         if (INTEL_DEBUG & DEBUG_PIXEL)
+            fprintf(stderr, "%s: cliprect failed\n", __FUNCTION__);
+         return GL_TRUE;
+      }
+
+      y = dPriv->h - y - height;
+      x += dPriv->x;
+      y += dPriv->y;
+
+
+      /* Set the frontbuffer up as a large rectangular texture.
+       */
+      intel->vtbl.meta_tex_rect_source(intel, src_region, textureFormat);
+
+
+      intel->vtbl.meta_texture_blend_replace(i830, glTextureFormat);
+
+
+      /* Set the 3d engine to draw into the destination region:
+       */
+
+      intel->vtbl.meta_draw_region(intel, dest_region);
+      intel->vtbl.meta_draw_format(intel, destFormat, depthFormat);     /* ?? */
+
+
+      /* Draw a single quad, no cliprects:
+       */
+      intel->vtbl.meta_disable_cliprects(intel);
+
+      intel->vtbl.draw_quad(intel,
+                            0, width, 0, height,
+                            0x00ff00ff, x, x + width, y, y + height);
+
+      intel->vtbl.leave_meta_state(intel);
+   }
+   UNLOCK_HARDWARE(intel);
+
+   intel_region_wait_fence(ctx, dest_region);   /* required by GL */
+   return GL_TRUE;
+#endif
+
+   return GL_FALSE;
+}
+
+
+
+
+static GLboolean
+do_blit_readpixels(GLcontext * ctx,
+                   GLint x, GLint y, GLsizei width, GLsizei height,
+                   GLenum format, GLenum type,
+                   const struct gl_pixelstore_attrib *pack, GLvoid * pixels)
+{
+   struct intel_context *intel = intel_context(ctx);
+   struct intel_region *src = intel_readbuf_region(intel);
+   struct intel_buffer_object *dst = intel_buffer_object(pack->BufferObj);
+   GLuint dst_offset;
+   GLuint rowLength;
+
+   if (INTEL_DEBUG & DEBUG_PIXEL)
+      _mesa_printf("%s\n", __FUNCTION__);
+
+   if (!src)
+      return GL_FALSE;
+
+   if (!_mesa_is_bufferobj(pack->BufferObj)) {
+      /* PBO only for now:
+       */
+      if (INTEL_DEBUG & DEBUG_PIXEL)
+         _mesa_printf("%s - not PBO\n", __FUNCTION__);
+      return GL_FALSE;
+   }
+
+
+   if (ctx->_ImageTransferState ||
+       !intel_check_blit_format(src, format, type)) {
+      if (INTEL_DEBUG & DEBUG_PIXEL)
+         _mesa_printf("%s - bad format for blit\n", __FUNCTION__);
+      return GL_FALSE;
+   }
+
+   if (pack->Alignment != 1 || pack->SwapBytes || pack->LsbFirst) {
+      if (INTEL_DEBUG & DEBUG_PIXEL)
+         _mesa_printf("%s: bad packing params\n", __FUNCTION__);
+      return GL_FALSE;
+   }
+
+   if (pack->RowLength > 0)
+      rowLength = pack->RowLength;
+   else
+      rowLength = width;
+
+   if (pack->Invert) {
+      if (INTEL_DEBUG & DEBUG_PIXEL)
+         _mesa_printf("%s: MESA_PACK_INVERT not done yet\n", __FUNCTION__);
+      return GL_FALSE;
+   }
+   else {
+      rowLength = -rowLength;
+   }
+
+   /* XXX 64-bit cast? */
+   dst_offset = (GLuint) _mesa_image_address(2, pack, pixels, width, height,
+                                             format, type, 0, 0, 0);
+
+
+   /* Although the blits go on the command buffer, need to do this and
+    * fire with lock held to guarentee cliprects are correct.
+    */
+   intelFlush(&intel->ctx);
+   LOCK_HARDWARE(intel);
+
+   if (intel->driDrawable->numClipRects) {
+      GLboolean all = (width * height * src->cpp == dst->Base.Size &&
+                       x == 0 && dst_offset == 0);
+
+      dri_bo *dst_buffer = intel_bufferobj_buffer(intel, dst,
+						  all ? INTEL_WRITE_FULL :
+						  INTEL_WRITE_PART);
+      __DRIdrawablePrivate *dPriv = intel->driDrawable;
+      int nbox = dPriv->numClipRects;
+      drm_clip_rect_t *box = dPriv->pClipRects;
+      drm_clip_rect_t rect;
+      drm_clip_rect_t src_rect;
+      int i;
+
+      src_rect.x1 = dPriv->x + x;
+      src_rect.y1 = dPriv->y + dPriv->h - (y + height);
+      src_rect.x2 = src_rect.x1 + width;
+      src_rect.y2 = src_rect.y1 + height;
+
+
+
+      for (i = 0; i < nbox; i++) {
+         if (!intel_intersect_cliprects(&rect, &src_rect, &box[i]))
+            continue;
+
+         if (!intelEmitCopyBlit(intel,
+				src->cpp,
+				src->pitch, src->buffer, 0, src->tiling,
+				rowLength, dst_buffer, dst_offset, GL_FALSE,
+				rect.x1,
+				rect.y1,
+				rect.x1 - src_rect.x1,
+				rect.y2 - src_rect.y2,
+				rect.x2 - rect.x1, rect.y2 - rect.y1,
+				GL_COPY)) {
+	    UNLOCK_HARDWARE(intel);
+	    return GL_FALSE;
+	 }
+      }
+   }
+   UNLOCK_HARDWARE(intel);
+
+   if (INTEL_DEBUG & DEBUG_PIXEL)
+      _mesa_printf("%s - DONE\n", __FUNCTION__);
+
+   return GL_TRUE;
+}
+
+void
+intelReadPixels(GLcontext * ctx,
+                GLint x, GLint y, GLsizei width, GLsizei height,
+                GLenum format, GLenum type,
+                const struct gl_pixelstore_attrib *pack, GLvoid * pixels)
+{
+   if (INTEL_DEBUG & DEBUG_PIXEL)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   intelFlush(ctx);
+
+#ifdef I915
+   if (do_blit_readpixels
+       (ctx, x, y, width, height, format, type, pack, pixels))
+      return;
+
+   if (do_texture_readpixels
+       (ctx, x, y, width, height, format, type, pack, pixels))
+      return;
+#else
+   (void)do_blit_readpixels;
+   (void)do_texture_readpixels;
+#endif
+
+   if (INTEL_DEBUG & DEBUG_PIXEL)
+      _mesa_printf("%s: fallback to swrast\n", __FUNCTION__);
+
+   /* Update Mesa state before calling down into _swrast_ReadPixels, as
+    * the spans code requires the computed buffer states to be up to date,
+    * but _swrast_ReadPixels only updates Mesa state after setting up
+    * the spans code.
+    */
+
+   if (ctx->NewState)
+      _mesa_update_state(ctx);
+
+   _swrast_ReadPixels(ctx, x, y, width, height, format, type, pack, pixels);
+}
diff --git a/src/mesa/drivers/dri/intel/intel_reg.h b/src/mesa/drivers/dri/intel/intel_reg.h
index 57ac8f0cc1..d19f1bae34 100644
--- a/src/mesa/drivers/dri/intel/intel_reg.h
+++ b/src/mesa/drivers/dri/intel/intel_reg.h
@@ -189,6 +189,19 @@
 
 #define S7_DEPTH_OFFSET_CONST_MASK     ~0
 
+/* p143 */
+#define _3DSTATE_BUF_INFO_CMD	(CMD_3D | (0x1d<<24) | (0x8e<<16) | 1)
+/* Dword 1 */
+#define BUF_3D_ID_COLOR_BACK	(0x3<<24)
+#define BUF_3D_ID_DEPTH 	(0x7<<24)
+#define BUF_3D_USE_FENCE	(1<<23)
+#define BUF_3D_TILED_SURFACE	(1<<22)
+#define BUF_3D_TILE_WALK_X	0
+#define BUF_3D_TILE_WALK_Y	(1<<21)
+#define BUF_3D_PITCH(x)         (((x)/4)<<2)
+/* Dword 2 */
+#define BUF_3D_ADDR(x)		((x) & ~0x3)
+
 /* Primitive dispatch on 830-945 */
 #define _3DPRIMITIVE			(CMD_3D | (0x1f << 24))
 #define PRIM_INDIRECT            (1<<23)
diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c
index 51ce32a967..a86c66a844 100644
--- a/src/mesa/drivers/dri/intel/intel_regions.c
+++ b/src/mesa/drivers/dri/intel/intel_regions.c
@@ -52,17 +52,77 @@
 
 #define FILE_DEBUG_FLAG DEBUG_REGION
 
+/* This should be set to the maximum backtrace size desired.
+ * Set it to 0 to disable backtrace debugging.
+ */
+#define DEBUG_BACKTRACE_SIZE 0
+
+#if DEBUG_BACKTRACE_SIZE == 0
+/* Use the standard debug output */
+#define _DBG(...) DBG(__VA_ARGS__)
+#else
+/* Use backtracing debug output */
+#define _DBG(...) {debug_backtrace(); DBG(__VA_ARGS__);}
+
+/* Backtracing debug support */
+#include <execinfo.h>
+
+static void
+debug_backtrace(void)
+{
+   void *trace[DEBUG_BACKTRACE_SIZE];
+   char **strings = NULL;
+   int traceSize;
+   register int i;
+
+   traceSize = backtrace(trace, DEBUG_BACKTRACE_SIZE);
+   strings = backtrace_symbols(trace, traceSize);
+   if (strings == NULL) {
+      DBG("no backtrace:");
+      return;
+   }
+
+   /* Spit out all the strings with a colon separator.  Ignore
+    * the first, since we don't really care about the call
+    * to debug_backtrace() itself.  Skip until the final "/" in
+    * the trace to avoid really long lines.
+    */
+   for (i = 1; i < traceSize; i++) {
+      char *p = strings[i], *slash = strings[i];
+      while (*p) {
+         if (*p++ == '/') {
+            slash = p;
+         }
+      }
+
+      DBG("%s:", slash);
+   }
+
+   /* Free up the memory, and we're done */
+   free(strings);
+}
+
+#endif
+
+
+
 /* XXX: Thread safety?
  */
 GLubyte *
 intel_region_map(struct intel_context *intel, struct intel_region *region)
 {
-   DBG("%s\n", __FUNCTION__);
+   intelFlush(&intel->ctx);
+
+   _DBG("%s %p\n", __FUNCTION__, region);
    if (!region->map_refcount++) {
       if (region->pbo)
          intel_region_cow(intel, region);
 
-      dri_bo_map(region->buffer, GL_TRUE);
+      if (region->tiling != I915_TILING_NONE &&
+	  intel->intelScreen->kernel_exec_fencing)
+	 drm_intel_gem_bo_map_gtt(region->buffer);
+      else
+	 dri_bo_map(region->buffer, GL_TRUE);
       region->map = region->buffer->virtual;
    }
 
@@ -72,9 +132,13 @@ intel_region_map(struct intel_context *intel, struct intel_region *region)
 void
 intel_region_unmap(struct intel_context *intel, struct intel_region *region)
 {
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s %p\n", __FUNCTION__, region);
    if (!--region->map_refcount) {
-      dri_bo_unmap(region->buffer);
+      if (region->tiling != I915_TILING_NONE &&
+	  intel->intelScreen->kernel_exec_fencing)
+	 drm_intel_gem_bo_unmap_gtt(region->buffer);
+      else
+	 dri_bo_unmap(region->buffer);
       region->map = NULL;
    }
 }
@@ -87,10 +151,10 @@ intel_region_alloc_internal(struct intel_context *intel,
 {
    struct intel_region *region;
 
-   DBG("%s\n", __FUNCTION__);
-
-   if (buffer == NULL)
+   if (buffer == NULL) {
+      _DBG("%s <-- NULL\n", __FUNCTION__);
       return NULL;
+   }
 
    region = calloc(sizeof(*region), 1);
    region->cpp = cpp;
@@ -104,19 +168,59 @@ intel_region_alloc_internal(struct intel_context *intel,
    region->tiling = I915_TILING_NONE;
    region->bit_6_swizzle = I915_BIT_6_SWIZZLE_NONE;
 
+   _DBG("%s <-- %p\n", __FUNCTION__, region);
    return region;
 }
 
 struct intel_region *
 intel_region_alloc(struct intel_context *intel,
-                   GLuint cpp, GLuint width, GLuint height, GLuint pitch)
+		   uint32_t tiling,
+                   GLuint cpp, GLuint width, GLuint height, GLuint pitch,
+		   GLboolean expect_accelerated_upload)
 {
    dri_bo *buffer;
+   struct intel_region *region;
+
+   /* If we're tiled, our allocations are in 8 or 32-row blocks, so
+    * failure to align our height means that we won't allocate enough pages.
+    *
+    * If we're untiled, we still have to align to 2 rows high because the
+    * data port accesses 2x2 blocks even if the bottom row isn't to be
+    * rendered, so failure to align means we could walk off the end of the
+    * GTT and fault.
+    */
+   if (tiling == I915_TILING_X)
+      height = ALIGN(height, 8);
+   else if (tiling == I915_TILING_Y)
+      height = ALIGN(height, 32);
+   else
+      height = ALIGN(height, 2);
 
-   buffer = dri_bo_alloc(intel->bufmgr, "region",
-			 pitch * cpp * height, 64);
+   /* If we're untiled, we have to align to 2 rows high because the
+    * data port accesses 2x2 blocks even if the bottom row isn't to be
+    * rendered, so failure to align means we could walk off the end of the
+    * GTT and fault.
+    */
+   height = ALIGN(height, 2);
 
-   return intel_region_alloc_internal(intel, cpp, width, height, pitch, buffer);
+   if (expect_accelerated_upload) {
+      buffer = drm_intel_bo_alloc_for_render(intel->bufmgr, "region",
+					     pitch * cpp * height, 64);
+   } else {
+      buffer = drm_intel_bo_alloc(intel->bufmgr, "region",
+				  pitch * cpp * height, 64);
+   }
+
+   region = intel_region_alloc_internal(intel, cpp, width, height,
+					pitch, buffer);
+
+   if (tiling != I915_TILING_NONE) {
+      assert(((pitch * cpp) & 127) == 0);
+      drm_intel_bo_set_tiling(buffer, &tiling, pitch * cpp);
+      drm_intel_bo_get_tiling(buffer, &region->tiling, &region->bit_6_swizzle);
+   }
+
+   return region;
 }
 
 struct intel_region *
@@ -152,7 +256,7 @@ void
 intel_region_reference(struct intel_region **dst, struct intel_region *src)
 {
    if (src)
-      DBG("%s %d\n", __FUNCTION__, src->refcount);
+      _DBG("%s %p %d\n", __FUNCTION__, src, src->refcount);
 
    assert(*dst == NULL);
    if (src) {
@@ -166,10 +270,12 @@ intel_region_release(struct intel_region **region_handle)
 {
    struct intel_region *region = *region_handle;
 
-   if (region == NULL)
+   if (region == NULL) {
+      _DBG("%s NULL\n", __FUNCTION__);
       return;
+   }
 
-   DBG("%s %d\n", __FUNCTION__, region->refcount - 1);
+   _DBG("%s %p %d\n", __FUNCTION__, region, region->refcount - 1);
 
    ASSERT(region->refcount > 0);
    region->refcount--;
@@ -243,9 +349,7 @@ intel_region_data(struct intel_context *intel,
                   const void *src, GLuint src_pitch,
                   GLuint srcx, GLuint srcy, GLuint width, GLuint height)
 {
-   GLboolean locked = GL_FALSE;
-
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s\n", __FUNCTION__);
 
    if (intel == NULL)
       return;
@@ -258,39 +362,33 @@ intel_region_data(struct intel_context *intel,
          intel_region_cow(intel, dst);
    }
 
-   if (!intel->locked) {
-      LOCK_HARDWARE(intel);
-      locked = GL_TRUE;
-   }
-
+   LOCK_HARDWARE(intel);
    _mesa_copy_rect(intel_region_map(intel, dst) + dst_offset,
                    dst->cpp,
                    dst->pitch,
                    dstx, dsty, width, height, src, src_pitch, srcx, srcy);
 
    intel_region_unmap(intel, dst);
-
-   if (locked)
-      UNLOCK_HARDWARE(intel);
-
+   UNLOCK_HARDWARE(intel);
 }
 
 /* Copy rectangular sub-regions. Need better logic about when to
  * push buffers into AGP - will currently do so whenever possible.
  */
-void
+GLboolean
 intel_region_copy(struct intel_context *intel,
                   struct intel_region *dst,
                   GLuint dst_offset,
                   GLuint dstx, GLuint dsty,
                   struct intel_region *src,
                   GLuint src_offset,
-                  GLuint srcx, GLuint srcy, GLuint width, GLuint height)
+                  GLuint srcx, GLuint srcy, GLuint width, GLuint height,
+		  GLenum logicop)
 {
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s\n", __FUNCTION__);
 
    if (intel == NULL)
-      return;
+      return GL_FALSE;
 
    if (dst->pbo) {
       if (dstx == 0 &&
@@ -302,41 +400,12 @@ intel_region_copy(struct intel_context *intel,
 
    assert(src->cpp == dst->cpp);
 
-   intelEmitCopyBlit(intel,
-                     dst->cpp,
-                     src->pitch, src->buffer, src_offset, src->tiling,
-                     dst->pitch, dst->buffer, dst_offset, dst->tiling,
-                     srcx, srcy, dstx, dsty, width, height,
-		     GL_COPY);
-}
-
-/* Fill a rectangular sub-region.  Need better logic about when to
- * push buffers into AGP - will currently do so whenever possible.
- */
-void
-intel_region_fill(struct intel_context *intel,
-                  struct intel_region *dst,
-                  GLuint dst_offset,
-                  GLuint dstx, GLuint dsty,
-                  GLuint width, GLuint height, GLuint color)
-{
-   DBG("%s\n", __FUNCTION__);
-
-   if (intel == NULL)
-      return;   
-
-   if (dst->pbo) {
-      if (dstx == 0 &&
-          dsty == 0 && width == dst->pitch && height == dst->height)
-         intel_region_release_pbo(intel, dst);
-      else
-         intel_region_cow(intel, dst);
-   }
-
-   intelEmitFillBlit(intel,
-                     dst->cpp,
-                     dst->pitch, dst->buffer, dst_offset, dst->tiling,
-                     dstx, dsty, width, height, color);
+   return intelEmitCopyBlit(intel,
+			    dst->cpp,
+			    src->pitch, src->buffer, src_offset, src->tiling,
+			    dst->pitch, dst->buffer, dst_offset, dst->tiling,
+			    srcx, srcy, dstx, dsty, width, height,
+			    logicop);
 }
 
 /* Attach to a pbo, discarding our data.  Effectively zero-copy upload
@@ -347,9 +416,13 @@ intel_region_attach_pbo(struct intel_context *intel,
                         struct intel_region *region,
                         struct intel_buffer_object *pbo)
 {
+   dri_bo *buffer;
+
    if (region->pbo == pbo)
       return;
 
+   _DBG("%s %p %p\n", __FUNCTION__, region, pbo);
+
    /* If there is already a pbo attached, break the cow tie now.
     * Don't call intel_region_release_pbo() as that would
     * unnecessarily allocate a new buffer we would have to immediately
@@ -365,10 +438,13 @@ intel_region_attach_pbo(struct intel_context *intel,
       region->buffer = NULL;
    }
 
+   /* make sure pbo has a buffer of its own */
+   buffer = intel_bufferobj_buffer(intel, pbo, INTEL_WRITE_FULL);
+
    region->pbo = pbo;
    region->pbo->region = region;
-   dri_bo_reference(pbo->buffer);
-   region->buffer = pbo->buffer;
+   dri_bo_reference(buffer);
+   region->buffer = buffer;
 }
 
 
@@ -379,6 +455,7 @@ void
 intel_region_release_pbo(struct intel_context *intel,
                          struct intel_region *region)
 {
+   _DBG("%s %p\n", __FUNCTION__, region);
    assert(region->buffer == region->pbo->buffer);
    region->pbo->region = NULL;
    region->pbo = NULL;
@@ -397,34 +474,27 @@ void
 intel_region_cow(struct intel_context *intel, struct intel_region *region)
 {
    struct intel_buffer_object *pbo = region->pbo;
-   GLboolean was_locked = intel->locked;
-
-   if (intel == NULL)
-      return;
+   GLboolean ok;
 
    intel_region_release_pbo(intel, region);
 
    assert(region->cpp * region->pitch * region->height == pbo->Base.Size);
 
-   DBG("%s (%d bytes)\n", __FUNCTION__, pbo->Base.Size);
+   _DBG("%s %p (%d bytes)\n", __FUNCTION__, region, pbo->Base.Size);
 
    /* Now blit from the texture buffer to the new buffer: 
     */
 
-   was_locked = intel->locked;
-   if (!was_locked)
-      LOCK_HARDWARE(intel);
-
-   intelEmitCopyBlit(intel,
-		     region->cpp,
-		     region->pitch, region->buffer, 0, region->tiling,
-		     region->pitch, pbo->buffer, 0, region->tiling,
-		     0, 0, 0, 0,
-		     region->pitch, region->height,
-		     GL_COPY);
-
-   if (!was_locked)
-      UNLOCK_HARDWARE(intel);
+   LOCK_HARDWARE(intel);
+   ok = intelEmitCopyBlit(intel,
+                          region->cpp,
+                          region->pitch, pbo->buffer, 0, region->tiling,
+                          region->pitch, region->buffer, 0, region->tiling,
+                          0, 0, 0, 0,
+                          region->pitch, region->height,
+                          GL_COPY);
+   assert(ok);
+   UNLOCK_HARDWARE(intel);
 }
 
 dri_bo *
@@ -453,6 +523,10 @@ intel_recreate_static(struct intel_context *intel,
    if (region == NULL) {
       region = calloc(sizeof(*region), 1);
       region->refcount = 1;
+      _DBG("%s creating new region %p\n", __FUNCTION__, region);
+   }
+   else {
+      _DBG("%s %p\n", __FUNCTION__, region);
    }
 
    if (intel->ctx.Visual.rgbBits == 24)
@@ -460,7 +534,8 @@ intel_recreate_static(struct intel_context *intel,
    else
       region->cpp = intel->ctx.Visual.rgbBits / 8;
    region->pitch = intelScreen->pitch;
-   region->height = intelScreen->height;     /* needed? */
+   region->width = intelScreen->width;
+   region->height = intelScreen->height;
 
    if (region->buffer != NULL) {
       dri_bo_unreference(region->buffer);
diff --git a/src/mesa/drivers/dri/intel/intel_regions.h b/src/mesa/drivers/dri/intel/intel_regions.h
index 4b120ba4ce..0d379bdc6e 100644
--- a/src/mesa/drivers/dri/intel/intel_regions.h
+++ b/src/mesa/drivers/dri/intel/intel_regions.h
@@ -73,8 +73,10 @@ struct intel_region
  * copied by calling intel_reference_region().
  */
 struct intel_region *intel_region_alloc(struct intel_context *intel,
-                                        GLuint cpp, GLuint width,
-                                        GLuint height, GLuint pitch);
+                                        uint32_t tiling,
+					GLuint cpp, GLuint width,
+                                        GLuint height, GLuint pitch,
+					GLboolean expect_accelerated_upload);
 
 struct intel_region *
 intel_region_alloc_for_handle(struct intel_context *intel,
@@ -108,21 +110,15 @@ void intel_region_data(struct intel_context *intel,
 
 /* Copy rectangular sub-regions
  */
-void intel_region_copy(struct intel_context *intel,
-                       struct intel_region *dest,
-                       GLuint dest_offset,
-                       GLuint destx, GLuint desty,
-                       struct intel_region *src,
-                       GLuint src_offset,
-                       GLuint srcx, GLuint srcy, GLuint width, GLuint height);
-
-/* Fill a rectangular sub-region
- */
-void intel_region_fill(struct intel_context *intel,
-                       struct intel_region *dest,
-                       GLuint dest_offset,
-                       GLuint destx, GLuint desty,
-                       GLuint width, GLuint height, GLuint color);
+GLboolean
+intel_region_copy(struct intel_context *intel,
+		  struct intel_region *dest,
+		  GLuint dest_offset,
+		  GLuint destx, GLuint desty,
+		  struct intel_region *src,
+		  GLuint src_offset,
+		  GLuint srcx, GLuint srcy, GLuint width, GLuint height,
+		  GLenum logicop);
 
 /* Helpers for zerocopy uploads, particularly texture image uploads:
  */
diff --git a/src/mesa/drivers/dri/intel/intel_screen.c b/src/mesa/drivers/dri/intel/intel_screen.c
index 2809fadd88..d547077929 100644
--- a/src/mesa/drivers/dri/intel/intel_screen.c
+++ b/src/mesa/drivers/dri/intel/intel_screen.c
@@ -29,27 +29,31 @@
 #include "main/glheader.h"
 #include "main/context.h"
 #include "main/framebuffer.h"
-#include "main/matrix.h"
 #include "main/renderbuffer.h"
-#include "main/simple_list.h"
+
 #include "utils.h"
 #include "vblank.h"
 #include "xmlpool.h"
 
-
-#include "intel_screen.h"
-
+#include "intel_batchbuffer.h"
 #include "intel_buffers.h"
-#include "intel_tex.h"
-#include "intel_span.h"
-#include "intel_fbo.h"
+#include "intel_bufmgr.h"
 #include "intel_chipset.h"
+#include "intel_extensions.h"
+#include "intel_fbo.h"
+#include "intel_regions.h"
+#include "intel_swapbuffers.h"
+#include "intel_screen.h"
+#include "intel_span.h"
+#include "intel_tex.h"
 
 #include "i915_drm.h"
 #include "i830_dri.h"
-#include "intel_regions.h"
-#include "intel_batchbuffer.h"
-#include "intel_bufmgr.h"
+
+#define DRI_CONF_TEXTURE_TILING(def) \
+	DRI_CONF_OPT_BEGIN(texture_tiling, bool, def)		\
+		DRI_CONF_DESC(en, "Enable texture tiling")	\
+	DRI_CONF_OPT_END					\
 
 PUBLIC const char __driConfigOptions[] =
    DRI_CONF_BEGIN
@@ -65,6 +69,17 @@ PUBLIC const char __driConfigOptions[] =
 	    DRI_CONF_ENUM(1, "Enable reuse of all sizes of buffer objects")
 	 DRI_CONF_DESC_END
       DRI_CONF_OPT_END
+
+#ifdef I915
+     DRI_CONF_TEXTURE_TILING(false)
+#else
+     DRI_CONF_TEXTURE_TILING(true)
+#endif
+
+      DRI_CONF_OPT_BEGIN(early_z, bool, false)
+	 DRI_CONF_DESC(en, "Enable early Z in classic mode (unstable, 945-only).")
+      DRI_CONF_OPT_END
+
    DRI_CONF_SECTION_END
    DRI_CONF_SECTION_QUALITY
       DRI_CONF_FORCE_S3TC_ENABLE(false)
@@ -72,10 +87,12 @@ PUBLIC const char __driConfigOptions[] =
    DRI_CONF_SECTION_END
    DRI_CONF_SECTION_DEBUG
      DRI_CONF_NO_RAST(false)
+     DRI_CONF_ALWAYS_FLUSH_BATCH(false)
+     DRI_CONF_ALWAYS_FLUSH_CACHE(false)
    DRI_CONF_SECTION_END
 DRI_CONF_END;
 
-const GLuint __driNConfigOptions = 6;
+const GLuint __driNConfigOptions = 10;
 
 #ifdef USE_NEW_INTERFACE
 static PFNGLXCREATECONTEXTMODES create_context_modes = NULL;
@@ -160,7 +177,7 @@ intelPrintSAREA(const drm_i915_sarea_t * sarea)
  * A number of the screen parameters are obtained/computed from
  * information in the SAREA.  This function updates those parameters.
  */
-void
+static void
 intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen,
                            drm_i915_sarea_t * sarea)
 {
@@ -210,6 +227,7 @@ static const __DRItexOffsetExtension intelTexOffsetExtension = {
 static const __DRItexBufferExtension intelTexBufferExtension = {
     { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION },
    intelSetTexBuffer,
+   intelSetTexBuffer2,
 };
 
 static int
@@ -279,7 +297,7 @@ intel_get_param(__DRIscreenPrivate *psp, int param, int *value)
 
    ret = drmCommandWriteRead(psp->fd, DRM_I915_GETPARAM, &gp, sizeof(gp));
    if (ret) {
-      fprintf(stderr, "drm_i915_getparam: %d\n", ret);
+      _mesa_warning(NULL, "drm_i915_getparam: %d", ret);
       return GL_FALSE;
    }
 
@@ -348,6 +366,7 @@ intelDestroyScreen(__DRIscreenPrivate * sPriv)
 
    dri_bufmgr_destroy(intelScreen->bufmgr);
    intelUnmapScreenRegions(intelScreen);
+   driDestroyOptionInfo(&intelScreen->optionCache);
 
    FREE(intelScreen);
    sPriv->private = NULL;
@@ -362,15 +381,13 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv,
                   __DRIdrawablePrivate * driDrawPriv,
                   const __GLcontextModes * mesaVis, GLboolean isPixmap)
 {
-   intelScreenPrivate *screen = (intelScreenPrivate *) driScrnPriv->private;
-
    if (isPixmap) {
       return GL_FALSE;          /* not implemented */
    }
    else {
       GLboolean swStencil = (mesaVis->stencilBits > 0 &&
                              mesaVis->depthBits != 24);
-      GLenum rgbFormat = (mesaVis->redBits == 5 ? GL_RGB5 : GL_RGBA8);
+      GLenum rgbFormat;
 
       struct intel_framebuffer *intel_fb = CALLOC_STRUCT(intel_framebuffer);
 
@@ -379,6 +396,13 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv,
 
       _mesa_initialize_framebuffer(&intel_fb->Base, mesaVis);
 
+      if (mesaVis->redBits == 5)
+	 rgbFormat = GL_RGB5;
+      else if (mesaVis->alphaBits == 0)
+	 rgbFormat = GL_RGB8;
+      else
+	 rgbFormat = GL_RGBA8;
+
       /* setup the hardware-based renderbuffers */
       intel_fb->color_rb[0] = intel_create_renderbuffer(rgbFormat);
       _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT,
@@ -432,7 +456,31 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv,
 static void
 intelDestroyBuffer(__DRIdrawablePrivate * driDrawPriv)
 {
-   _mesa_unreference_framebuffer((GLframebuffer **)(&(driDrawPriv->driverPrivate)));
+   struct intel_framebuffer *intel_fb = driDrawPriv->driverPrivate;
+   struct intel_renderbuffer *depth_rb;
+   struct intel_renderbuffer *stencil_rb;
+
+   if (intel_fb) {
+      if (intel_fb->color_rb[0]) {
+         intel_renderbuffer_set_region(intel_fb->color_rb[0], NULL);
+      }
+
+      if (intel_fb->color_rb[1]) {
+         intel_renderbuffer_set_region(intel_fb->color_rb[1], NULL);
+      }
+
+      depth_rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH);
+      if (depth_rb) {
+         intel_renderbuffer_set_region(depth_rb, NULL);
+      }
+
+      stencil_rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL);
+      if (stencil_rb) {
+         intel_renderbuffer_set_region(stencil_rb, NULL);
+      }
+   }
+
+   _mesa_reference_framebuffer((GLframebuffer **)(&(driDrawPriv->driverPrivate)), NULL);
 }
 
 
@@ -513,8 +561,6 @@ intelFillInModes(__DRIscreenPrivate *psp,
    __GLcontextModes *m;
    unsigned depth_buffer_factor;
    unsigned back_buffer_factor;
-   GLenum fb_format;
-   GLenum fb_type;
    int i;
 
    /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't
@@ -526,6 +572,7 @@ intelFillInModes(__DRIscreenPrivate *psp,
 
    uint8_t depth_bits_array[3];
    uint8_t stencil_bits_array[3];
+   uint8_t msaa_samples_array[1];
 
    depth_bits_array[0] = 0;
    depth_bits_array[1] = depth_bits;
@@ -542,22 +589,39 @@ intelFillInModes(__DRIscreenPrivate *psp,
 
    stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits;
 
+   msaa_samples_array[0] = 0;
+
    depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1;
    back_buffer_factor = (have_back_buffer) ? 3 : 1;
 
    if (pixel_bits == 16) {
-      fb_format = GL_RGB;
-      fb_type = GL_UNSIGNED_SHORT_5_6_5;
+      configs = driCreateConfigs(GL_RGB, GL_UNSIGNED_SHORT_5_6_5,
+				 depth_bits_array, stencil_bits_array,
+				 depth_buffer_factor, back_buffer_modes,
+				 back_buffer_factor,
+				 msaa_samples_array, 1);
    }
    else {
-      fb_format = GL_BGRA;
-      fb_type = GL_UNSIGNED_INT_8_8_8_8_REV;
+      __DRIconfig **configs_a8r8g8b8;
+      __DRIconfig **configs_x8r8g8b8;
+
+      configs_a8r8g8b8 = driCreateConfigs(GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV,
+					  depth_bits_array,
+					  stencil_bits_array,
+					  depth_buffer_factor,
+					  back_buffer_modes,
+					  back_buffer_factor,
+					  msaa_samples_array, 1);
+      configs_x8r8g8b8 = driCreateConfigs(GL_BGR, GL_UNSIGNED_INT_8_8_8_8_REV,
+					  depth_bits_array,
+					  stencil_bits_array,
+					  depth_buffer_factor,
+					  back_buffer_modes,
+					  back_buffer_factor,
+					  msaa_samples_array, 1);
+      configs = driConcatConfigs(configs_a8r8g8b8, configs_x8r8g8b8);
    }
 
-   configs = driCreateConfigs(fb_format, fb_type,
-			      depth_bits_array, stencil_bits_array,
-			      depth_buffer_factor, back_buffer_modes,
-			      back_buffer_factor);
    if (configs == NULL) {
     fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
               __LINE__);
@@ -584,6 +648,7 @@ intel_init_bufmgr(intelScreenPrivate *intelScreen)
    GLboolean gem_supported;
    struct drm_i915_getparam gp;
    __DRIscreenPrivate *spriv = intelScreen->driScrnPriv;
+   int num_fences = 0;
 
    intelScreen->no_hw = getenv("INTEL_NO_HW") != NULL;
 
@@ -613,10 +678,10 @@ intel_init_bufmgr(intelScreenPrivate *intelScreen)
    /* Otherwise, use the classic buffer manager. */
    if (intelScreen->bufmgr == NULL) {
       if (gem_disable) {
-	 fprintf(stderr, "GEM disabled.  Using classic.\n");
+	 _mesa_warning(NULL, "GEM disabled.  Using classic.");
       } else {
-	 fprintf(stderr, "Failed to initialize GEM.  "
-		 "Falling back to classic.\n");
+	 _mesa_warning(NULL,
+                       "Failed to initialize GEM.  Falling back to classic.");
       }
 
       if (intelScreen->tex.size == 0) {
@@ -634,8 +699,10 @@ intel_init_bufmgr(intelScreenPrivate *intelScreen)
 				&intelScreen->sarea->last_dispatch);
    }
 
-   /* XXX bufmgr should be per-screen, not per-context */
-   intelScreen->ttm = intelScreen->ttm;
+   if (intel_get_param(spriv, I915_PARAM_NUM_FENCES_AVAIL, &num_fences))
+      intelScreen->kernel_exec_fencing = !!num_fences;
+   else
+      intelScreen->kernel_exec_fencing = GL_FALSE;
 
    return GL_TRUE;
 }
@@ -719,6 +786,17 @@ static const
 __DRIconfig **intelInitScreen2(__DRIscreenPrivate *psp)
 {
    intelScreenPrivate *intelScreen;
+   GLenum fb_format[3];
+   GLenum fb_type[3];
+   /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't
+    * support pageflipping at all.
+    */
+   static const GLenum back_buffer_modes[] = {
+      GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML
+   };
+   uint8_t depth_bits[4], stencil_bits[4], msaa_samples_array[1];
+   int color;
+   __DRIconfig **configs = NULL;
 
    /* Calling driInitExtensions here, with a NULL context pointer,
     * does not actually enable the extensions.  It just makes sure
@@ -758,8 +836,71 @@ __DRIconfig **intelInitScreen2(__DRIscreenPrivate *psp)
    intelScreen->irq_active = 1;
    psp->extensions = intelScreenExtensions;
 
-   return driConcatConfigs(intelFillInModes(psp, 16, 16, 0, 1),
-			   intelFillInModes(psp, 32, 24, 8, 1));
+   depth_bits[0] = 0;
+   stencil_bits[0] = 0;
+   depth_bits[1] = 16;
+   stencil_bits[1] = 0;
+   depth_bits[2] = 24;
+   stencil_bits[2] = 0;
+   depth_bits[3] = 24;
+   stencil_bits[3] = 8;
+
+   msaa_samples_array[0] = 0;
+
+   fb_format[0] = GL_RGB;
+   fb_type[0] = GL_UNSIGNED_SHORT_5_6_5;
+
+   fb_format[1] = GL_BGR;
+   fb_type[1] = GL_UNSIGNED_INT_8_8_8_8_REV;
+
+   fb_format[2] = GL_BGRA;
+   fb_type[2] = GL_UNSIGNED_INT_8_8_8_8_REV;
+
+   depth_bits[0] = 0;
+   stencil_bits[0] = 0;
+
+   for (color = 0; color < ARRAY_SIZE(fb_format); color++) {
+      __DRIconfig **new_configs;
+      int depth_factor;
+
+      /* With DRI2 right now, GetBuffers always returns a depth/stencil buffer
+       * with the same cpp as the drawable.  So we can't support depth cpp !=
+       * color cpp currently.
+       */
+      if (fb_type[color] == GL_UNSIGNED_SHORT_5_6_5) {
+	 depth_bits[1] = 16;
+	 stencil_bits[1] = 0;
+
+	 depth_factor = 2;
+      } else {
+	 depth_bits[1] = 24;
+	 stencil_bits[1] = 0;
+	 depth_bits[2] = 24;
+	 stencil_bits[2] = 8;
+
+	 depth_factor = 3;
+      }
+      new_configs = driCreateConfigs(fb_format[color], fb_type[color],
+				     depth_bits,
+				     stencil_bits,
+				     depth_factor,
+				     back_buffer_modes,
+				     ARRAY_SIZE(back_buffer_modes),
+				     msaa_samples_array,
+				     ARRAY_SIZE(msaa_samples_array));
+      if (configs == NULL)
+	 configs = new_configs;
+      else
+	 configs = driConcatConfigs(configs, new_configs);
+   }
+
+   if (configs == NULL) {
+      fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__,
+              __LINE__);
+      return NULL;
+   }
+
+   return (const __DRIconfig **)configs;
 }
 
 const struct __DriverAPIRec driDriverAPI = {
diff --git a/src/mesa/drivers/dri/intel/intel_screen.h b/src/mesa/drivers/dri/intel/intel_screen.h
index fcd0d9c28c..a9b9e109a6 100644
--- a/src/mesa/drivers/dri/intel/intel_screen.h
+++ b/src/mesa/drivers/dri/intel/intel_screen.h
@@ -79,6 +79,7 @@ typedef struct
    GLboolean no_vbo;
    int ttm;
    dri_bufmgr *bufmgr;
+   GLboolean kernel_exec_fencing;
 
    /**
    * Configuration cache with default values for all contexts
@@ -92,10 +93,6 @@ extern GLboolean intelMapScreenRegions(__DRIscreenPrivate * sPriv);
 
 extern void intelUnmapScreenRegions(intelScreenPrivate * intelScreen);
 
-extern void
-intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen,
-                           drm_i915_sarea_t * sarea);
-
 extern void intelDestroyContext(__DRIcontextPrivate * driContextPriv);
 
 extern GLboolean intelUnbindContext(__DRIcontextPrivate * driContextPriv);
@@ -105,11 +102,6 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv,
                  __DRIdrawablePrivate * driDrawPriv,
                  __DRIdrawablePrivate * driReadPriv);
 
-extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv);
-
-extern void
-intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h);
-
 extern struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen);
 
 #endif
diff --git a/src/mesa/drivers/dri/intel/intel_span.c b/src/mesa/drivers/dri/intel/intel_span.c
index d9315043e6..1e73943457 100644
--- a/src/mesa/drivers/dri/intel/intel_span.c
+++ b/src/mesa/drivers/dri/intel/intel_span.c
@@ -29,6 +29,7 @@
 #include "main/macros.h"
 #include "main/mtypes.h"
 #include "main/colormac.h"
+#include "main/texformat.h"
 
 #include "intel_buffers.h"
 #include "intel_fbo.h"
@@ -131,6 +132,18 @@ pwrite_8(struct intel_renderbuffer *irb, uint32_t offset, uint8_t val)
    dri_bo_subdata(irb->region->buffer, offset, 1, &val);
 }
 
+static uint32_t
+z24s8_to_s8z24(uint32_t val)
+{
+   return (val << 24) | (val >> 8);
+}
+
+static uint32_t
+s8z24_to_z24s8(uint32_t val)
+{
+   return (val >> 24) | (val << 8);
+}
+
 static uint32_t no_tile_swizzle(struct intel_renderbuffer *irb,
 				int x, int y)
 {
@@ -150,7 +163,7 @@ static uint32_t x_tile_swizzle(struct intel_renderbuffer *irb,
 	int	x_tile_number, y_tile_number;
 	int	tile_off, tile_base;
 	
-	tile_stride = (irb->pfPitch * irb->region->cpp) << 3;
+	tile_stride = (irb->region->pitch * irb->region->cpp) << 3;
 
 	xbyte = x * irb->region->cpp;
 
@@ -190,7 +203,7 @@ static uint32_t x_tile_swizzle(struct intel_renderbuffer *irb,
 	printf("(%d,%d) -> %d + %d = %d (pitch = %d, tstride = %d)\n",
 	       x, y, tile_off, tile_base,
 	       tile_off + tile_base,
-	       irb->pfPitch, tile_stride);
+	       irb->region->pitch, tile_stride);
 #endif
 
 	return tile_base + tile_off;
@@ -205,7 +218,7 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 	int	x_tile_number, y_tile_number;
 	int	tile_off, tile_base;
 	
-	tile_stride = (irb->pfPitch * irb->region->cpp) << 5;
+	tile_stride = (irb->region->pitch * irb->region->cpp) << 5;
 
 	xbyte = x * irb->region->cpp;
 
@@ -255,8 +268,8 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 #define LOCAL_VARS							\
    struct intel_context *intel = intel_context(ctx);			\
    struct intel_renderbuffer *irb = intel_renderbuffer(rb);		\
-   const GLint yScale = irb->RenderToTexture ? 1 : -1;			\
-   const GLint yBias = irb->RenderToTexture ? 0 : irb->Base.Height - 1;	\
+   const GLint yScale = ctx->DrawBuffer->Name ? 1 : -1;			\
+   const GLint yBias = ctx->DrawBuffer->Name ? 0 : irb->Base.Height - 1;\
    unsigned int num_cliprects;						\
    struct drm_clip_rect *cliprects;					\
    int x_off, y_off;							\
@@ -293,107 +306,51 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 #define X_TILE(_X, _Y) x_tile_swizzle(irb, (_X) + x_off, (_Y) + y_off)
 #define Y_TILE(_X, _Y) y_tile_swizzle(irb, (_X) + x_off, (_Y) + y_off)
 
-/* 16 bit, RGB565 color spanline and pixel functions
- */
-#define SPANTMP_PIXEL_FMT GL_RGB
-#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5
-
-#define TAG(x)    intel##x##_RGB565
-#define TAG2(x,y) intel##x##_RGB565##y
-#define GET_VALUE(X, Y) pread_16(irb, NO_TILE(X, Y))
-#define PUT_VALUE(X, Y, V) pwrite_16(irb, NO_TILE(X, Y), V)
-#include "spantmp2.h"
-
-/* 32 bit, ARGB8888 color spanline and pixel functions
- */
-#define SPANTMP_PIXEL_FMT GL_BGRA
-#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
-
-#define TAG(x)    intel##x##_ARGB8888
-#define TAG2(x,y) intel##x##_ARGB8888##y
-#define GET_VALUE(X, Y) pread_32(irb, NO_TILE(X, Y))
-#define PUT_VALUE(X, Y, V) pwrite_32(irb, NO_TILE(X, Y), V)
-#include "spantmp2.h"
-
-/* 32 bit, xRGB8888 color spanline and pixel functions
- */
-#define SPANTMP_PIXEL_FMT GL_BGRA
-#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
-
-#define TAG(x)    intel##x##_xRGB8888
-#define TAG2(x,y) intel##x##_xRGB8888##y
-#define GET_VALUE(X, Y) pread_xrgb8888(irb, NO_TILE(X, Y))
-#define PUT_VALUE(X, Y, V) pwrite_xrgb8888(irb, NO_TILE(X, Y), V)
-#include "spantmp2.h"
-
-/* 16 bit RGB565 color tile spanline and pixel functions
- */
-
-#define SPANTMP_PIXEL_FMT GL_RGB
-#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5
-
-#define TAG(x)    intel_XTile_##x##_RGB565
-#define TAG2(x,y) intel_XTile_##x##_RGB565##y
-#define GET_VALUE(X, Y) pread_16(irb, X_TILE(X, Y))
-#define PUT_VALUE(X, Y, V) pwrite_16(irb, X_TILE(X, Y), V)
-#include "spantmp2.h"
-
-#define SPANTMP_PIXEL_FMT GL_RGB
-#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5
-
-#define TAG(x)    intel_YTile_##x##_RGB565
-#define TAG2(x,y) intel_YTile_##x##_RGB565##y
-#define GET_VALUE(X, Y) pread_16(irb, Y_TILE(X, Y))
-#define PUT_VALUE(X, Y, V) pwrite_16(irb, Y_TILE(X, Y), V)
-#include "spantmp2.h"
-
-/* 32 bit ARGB888 color tile spanline and pixel functions
- */
-
-#define SPANTMP_PIXEL_FMT GL_BGRA
-#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
-
-#define TAG(x)    intel_XTile_##x##_ARGB8888
-#define TAG2(x,y) intel_XTile_##x##_ARGB8888##y
-#define GET_VALUE(X, Y) pread_32(irb, X_TILE(X, Y))
-#define PUT_VALUE(X, Y, V) pwrite_32(irb, X_TILE(X, Y), V)
-#include "spantmp2.h"
-
-#define SPANTMP_PIXEL_FMT GL_BGRA
-#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
-
-#define TAG(x)    intel_YTile_##x##_ARGB8888
-#define TAG2(x,y) intel_YTile_##x##_ARGB8888##y
-#define GET_VALUE(X, Y) pread_32(irb, Y_TILE(X, Y))
-#define PUT_VALUE(X, Y, V) pwrite_32(irb, Y_TILE(X, Y), V)
-#include "spantmp2.h"
-
-/* 32 bit xRGB888 color tile spanline and pixel functions
- */
-
-#define SPANTMP_PIXEL_FMT GL_BGRA
-#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
-
-#define TAG(x)    intel_XTile_##x##_xRGB8888
-#define TAG2(x,y) intel_XTile_##x##_xRGB8888##y
-#define GET_VALUE(X, Y) pread_xrgb8888(irb, X_TILE(X, Y))
-#define PUT_VALUE(X, Y, V) pwrite_xrgb8888(irb, X_TILE(X, Y), V)
-#include "spantmp2.h"
-
-#define SPANTMP_PIXEL_FMT GL_BGRA
-#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
-
-#define TAG(x)    intel_YTile_##x##_xRGB8888
-#define TAG2(x,y) intel_YTile_##x##_xRGB8888##y
-#define GET_VALUE(X, Y) pread_xrgb8888(irb, Y_TILE(X, Y))
-#define PUT_VALUE(X, Y, V) pwrite_xrgb8888(irb, Y_TILE(X, Y), V)
-#include "spantmp2.h"
+/* r5g6b5 color span and pixel functions */
+#define INTEL_PIXEL_FMT GL_RGB
+#define INTEL_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5
+#define INTEL_READ_VALUE(offset) pread_16(irb, offset)
+#define INTEL_WRITE_VALUE(offset, v) pwrite_16(irb, offset, v)
+#define INTEL_TAG(x) x##_RGB565
+#include "intel_spantmp.h"
+
+/* a4r4g4b4 color span and pixel functions */
+#define INTEL_PIXEL_FMT GL_BGRA
+#define INTEL_PIXEL_TYPE GL_UNSIGNED_SHORT_4_4_4_4_REV
+#define INTEL_READ_VALUE(offset) pread_16(irb, offset)
+#define INTEL_WRITE_VALUE(offset, v) pwrite_16(irb, offset, v)
+#define INTEL_TAG(x) x##_ARGB4444
+#include "intel_spantmp.h"
+
+/* a1r5g5b5 color span and pixel functions */
+#define INTEL_PIXEL_FMT GL_BGRA
+#define INTEL_PIXEL_TYPE GL_UNSIGNED_SHORT_1_5_5_5_REV
+#define INTEL_READ_VALUE(offset) pread_16(irb, offset)
+#define INTEL_WRITE_VALUE(offset, v) pwrite_16(irb, offset, v)
+#define INTEL_TAG(x) x##_ARGB1555
+#include "intel_spantmp.h"
+
+/* a8r8g8b8 color span and pixel functions */
+#define INTEL_PIXEL_FMT GL_BGRA
+#define INTEL_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
+#define INTEL_READ_VALUE(offset) pread_32(irb, offset)
+#define INTEL_WRITE_VALUE(offset, v) pwrite_32(irb, offset, v)
+#define INTEL_TAG(x) x##_ARGB8888
+#include "intel_spantmp.h"
+
+/* x8r8g8b8 color span and pixel functions */
+#define INTEL_PIXEL_FMT GL_BGRA
+#define INTEL_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
+#define INTEL_READ_VALUE(offset) pread_xrgb8888(irb, offset)
+#define INTEL_WRITE_VALUE(offset, v) pwrite_xrgb8888(irb, offset, v)
+#define INTEL_TAG(x) x##_xRGB8888
+#include "intel_spantmp.h"
 
 #define LOCAL_DEPTH_VARS						\
    struct intel_context *intel = intel_context(ctx);			\
    struct intel_renderbuffer *irb = intel_renderbuffer(rb);		\
-   const GLint yScale = irb->RenderToTexture ? 1 : -1;			\
-   const GLint yBias = irb->RenderToTexture ? 0 : irb->Base.Height - 1; \
+   const GLint yScale = ctx->DrawBuffer->Name ? 1 : -1;			\
+   const GLint yBias = ctx->DrawBuffer->Name ? 0 : irb->Base.Height - 1;\
    unsigned int num_cliprects;						\
    struct drm_clip_rect *cliprects;					\
    int x_off, y_off;							\
@@ -402,98 +359,26 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 
 #define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS
 
-/**
- ** 16-bit depthbuffer functions.
- **/
-#define VALUE_TYPE GLushort
-#define WRITE_DEPTH(_x, _y, d) pwrite_16(irb, NO_TILE(_x, _y), d)
-#define READ_DEPTH(d, _x, _y) d = pread_16(irb, NO_TILE(_x, _y))
-#define TAG(x) intel##x##_z16
-#include "depthtmp.h"
-
+/* z16 depthbuffer functions. */
+#define INTEL_VALUE_TYPE GLushort
+#define INTEL_WRITE_DEPTH(offset, d) pwrite_16(irb, offset, d)
+#define INTEL_READ_DEPTH(offset) pread_16(irb, offset)
+#define INTEL_TAG(name) name##_z16
+#include "intel_depthtmp.h"
 
-/**
- ** 16-bit x tile depthbuffer functions.
- **/
-#define VALUE_TYPE GLushort
-#define WRITE_DEPTH(_x, _y, d) pwrite_16(irb, X_TILE(_x, _y), d)
-#define READ_DEPTH(d, _x, _y) d = pread_16(irb, X_TILE(_x, _y))
-#define TAG(x) intel_XTile_##x##_z16
-#include "depthtmp.h"
+/* z24 depthbuffer functions. */
+#define INTEL_VALUE_TYPE GLuint
+#define INTEL_WRITE_DEPTH(offset, d) pwrite_32(irb, offset, d)
+#define INTEL_READ_DEPTH(offset) pread_32(irb, offset)
+#define INTEL_TAG(name) name##_z24
+#include "intel_depthtmp.h"
 
-/**
- ** 16-bit y tile depthbuffer functions.
- **/
-#define VALUE_TYPE GLushort
-#define WRITE_DEPTH(_x, _y, d) pwrite_16(irb, Y_TILE(_x, _y), d)
-#define READ_DEPTH(d, _x, _y) d = pread_16(irb, Y_TILE(_x, _y))
-#define TAG(x) intel_YTile_##x##_z16
-#include "depthtmp.h"
-
-
-/**
- ** 24/8-bit interleaved depth/stencil functions
- ** Note: we're actually reading back combined depth+stencil values.
- ** The wrappers in main/depthstencil.c are used to extract the depth
- ** and stencil values.
- **/
-#define VALUE_TYPE GLuint
-
-/* Change ZZZS -> SZZZ */
-#define WRITE_DEPTH(_x, _y, d)					\
-   pwrite_32(irb, NO_TILE(_x, _y), ((d) >> 8) | ((d) << 24))
-
-/* Change SZZZ -> ZZZS */
-#define READ_DEPTH( d, _x, _y ) {				\
-   GLuint tmp = pread_32(irb, NO_TILE(_x, _y));			\
-   d = (tmp << 8) | (tmp >> 24);				\
-}
-
-#define TAG(x) intel##x##_z24_s8
-#include "depthtmp.h"
-
-
-/**
- ** 24/8-bit x-tile interleaved depth/stencil functions
- ** Note: we're actually reading back combined depth+stencil values.
- ** The wrappers in main/depthstencil.c are used to extract the depth
- ** and stencil values.
- **/
-#define VALUE_TYPE GLuint
-
-/* Change ZZZS -> SZZZ */
-#define WRITE_DEPTH(_x, _y, d)					\
-   pwrite_32(irb, X_TILE(_x, _y), ((d) >> 8) | ((d) << 24))
-
-/* Change SZZZ -> ZZZS */
-#define READ_DEPTH( d, _x, _y ) {				\
-   GLuint tmp = pread_32(irb, X_TILE(_x, _y));		\
-   d = (tmp << 8) | (tmp >> 24);				\
-}
-
-#define TAG(x) intel_XTile_##x##_z24_s8
-#include "depthtmp.h"
-
-/**
- ** 24/8-bit y-tile interleaved depth/stencil functions
- ** Note: we're actually reading back combined depth+stencil values.
- ** The wrappers in main/depthstencil.c are used to extract the depth
- ** and stencil values.
- **/
-#define VALUE_TYPE GLuint
-
-/* Change ZZZS -> SZZZ */
-#define WRITE_DEPTH(_x, _y, d)					\
-   pwrite_32(irb, Y_TILE(_x, _y), ((d) >> 8) | ((d) << 24))
-
-/* Change SZZZ -> ZZZS */
-#define READ_DEPTH( d, _x, _y ) {				\
-   GLuint tmp = pread_32(irb, Y_TILE(_x, _y));			\
-   d = (tmp << 8) | (tmp >> 24);				\
-}
-
-#define TAG(x) intel_YTile_##x##_z24_s8
-#include "depthtmp.h"
+/* z24s8 depthbuffer functions. */
+#define INTEL_VALUE_TYPE GLuint
+#define INTEL_WRITE_DEPTH(offset, d) pwrite_32(irb, offset, z24s8_to_s8z24(d))
+#define INTEL_READ_DEPTH(offset) s8z24_to_z24s8(pread_32(irb, offset))
+#define INTEL_TAG(name) name##_z24_s8
+#include "intel_depthtmp.h"
 
 
 /**
@@ -528,8 +413,6 @@ intel_renderbuffer_map(struct intel_context *intel, struct gl_renderbuffer *rb)
    if (irb == NULL || irb->region == NULL)
       return;
 
-   irb->pfPitch = irb->region->pitch;
-
    intel_set_span_functions(intel, rb);
 }
 
@@ -543,7 +426,6 @@ intel_renderbuffer_unmap(struct intel_context *intel,
       return;
 
    clear_span_cache(irb);
-   irb->pfPitch = 0;
 
    rb->GetRow = NULL;
    rb->PutRow = NULL;
@@ -562,23 +444,32 @@ intel_renderbuffer_unmap(struct intel_context *intel,
  * _ColorReadBuffer, _DepthBuffer or _StencilBuffer fields.
  */
 static void
-intel_map_unmap_buffers(struct intel_context *intel, GLboolean map)
+intel_map_unmap_framebuffer(struct intel_context *intel,
+			    struct gl_framebuffer *fb,
+			    GLboolean map)
 {
    GLcontext *ctx = &intel->ctx;
    GLuint i, j;
 
-   /* color draw buffers */
-   for (j = 0; j < ctx->DrawBuffer->_NumColorDrawBuffers; j++) {
+   /* color buffers */
+   if (fb == ctx->DrawBuffer) {
+      for (j = 0; j < fb->_NumColorDrawBuffers; j++) {
+	 if (map)
+	    intel_renderbuffer_map(intel, fb->_ColorDrawBuffers[j]);
+	 else
+	    intel_renderbuffer_unmap(intel, fb->_ColorDrawBuffers[j]);
+      }
+   } else {
       if (map)
-	 intel_renderbuffer_map(intel, ctx->DrawBuffer->_ColorDrawBuffers[j]);
+	 intel_renderbuffer_map(intel, fb->_ColorReadBuffer);
       else
-	 intel_renderbuffer_unmap(intel, ctx->DrawBuffer->_ColorDrawBuffers[j]);
+	 intel_renderbuffer_unmap(intel, fb->_ColorReadBuffer);
    }
 
    /* check for render to textures */
    for (i = 0; i < BUFFER_COUNT; i++) {
       struct gl_renderbuffer_attachment *att =
-         ctx->DrawBuffer->Attachment + i;
+         fb->Attachment + i;
       struct gl_texture_object *tex = att->Texture;
       if (tex) {
          /* render to texture */
@@ -590,36 +481,28 @@ intel_map_unmap_buffers(struct intel_context *intel, GLboolean map)
       }
    }
 
-   /* color read buffers */
-   if (map)
-      intel_renderbuffer_map(intel, ctx->ReadBuffer->_ColorReadBuffer);
-   else
-      intel_renderbuffer_unmap(intel, ctx->ReadBuffer->_ColorReadBuffer);
-
    /* depth buffer (Note wrapper!) */
-   if (ctx->DrawBuffer->_DepthBuffer) {
+   if (fb->_DepthBuffer) {
       if (map)
-	 intel_renderbuffer_map(intel, ctx->DrawBuffer->_DepthBuffer->Wrapped);
+	 intel_renderbuffer_map(intel, fb->_DepthBuffer->Wrapped);
       else
 	 intel_renderbuffer_unmap(intel,
-				  ctx->DrawBuffer->_DepthBuffer->Wrapped);
+				  fb->_DepthBuffer->Wrapped);
    }
 
    /* stencil buffer (Note wrapper!) */
-   if (ctx->DrawBuffer->_StencilBuffer) {
+   if (fb->_StencilBuffer) {
       if (map)
 	 intel_renderbuffer_map(intel,
-				ctx->DrawBuffer->_StencilBuffer->Wrapped);
+				fb->_StencilBuffer->Wrapped);
       else
 	 intel_renderbuffer_unmap(intel,
-				  ctx->DrawBuffer->_StencilBuffer->Wrapped);
+				  fb->_StencilBuffer->Wrapped);
    }
 }
 
-
-
 /**
- * Prepare for softare rendering.  Map current read/draw framebuffers'
+ * Prepare for software rendering.  Map current read/draw framebuffers'
  * renderbuffes and all currently bound texture objects.
  *
  * Old note: Moved locking out to get reasonable span performance.
@@ -640,11 +523,12 @@ intelSpanRenderStart(GLcontext * ctx)
       }
    }
 
-   intel_map_unmap_buffers(intel, GL_TRUE);
+   intel_map_unmap_framebuffer(intel, ctx->DrawBuffer, GL_TRUE);
+   intel_map_unmap_framebuffer(intel, ctx->ReadBuffer, GL_TRUE);
 }
 
 /**
- * Called when done softare rendering.  Unmap the buffers we mapped in
+ * Called when done software rendering.  Unmap the buffers we mapped in
  * the above function.
  */
 void
@@ -662,7 +546,8 @@ intelSpanRenderFinish(GLcontext * ctx)
       }
    }
 
-   intel_map_unmap_buffers(intel, GL_FALSE);
+   intel_map_unmap_framebuffer(intel, ctx->DrawBuffer, GL_FALSE);
+   intel_map_unmap_framebuffer(intel, ctx->ReadBuffer, GL_FALSE);
 
    UNLOCK_HARDWARE(intel);
 }
@@ -696,8 +581,8 @@ intel_set_span_functions(struct intel_context *intel,
    else
       tiling = I915_TILING_NONE;
 
-   if (rb->_ActualFormat == GL_RGB5) {
-      /* 565 RGB */
+   switch (irb->texformat->MesaFormat) {
+   case MESA_FORMAT_RGB565:
       switch (tiling) {
       case I915_TILING_NONE:
       default:
@@ -710,38 +595,67 @@ intel_set_span_functions(struct intel_context *intel,
 	 intel_YTile_InitPointers_RGB565(rb);
 	 break;
       }
-   }
-   else if (rb->_ActualFormat == GL_RGB8) {
-      /* 8888 RGBx */
+      break;
+   case MESA_FORMAT_ARGB4444:
       switch (tiling) {
       case I915_TILING_NONE:
       default:
-	 intelInitPointers_xRGB8888(rb);
+	 intelInitPointers_ARGB4444(rb);
 	 break;
       case I915_TILING_X:
-	 intel_XTile_InitPointers_xRGB8888(rb);
+	 intel_XTile_InitPointers_ARGB4444(rb);
 	 break;
       case I915_TILING_Y:
-	 intel_YTile_InitPointers_xRGB8888(rb);
+	 intel_YTile_InitPointers_ARGB4444(rb);
 	 break;
       }
-   }
-   else if (rb->_ActualFormat == GL_RGBA8) {
-      /* 8888 RGBA */
+      break;
+   case MESA_FORMAT_ARGB1555:
       switch (tiling) {
       case I915_TILING_NONE:
       default:
-	 intelInitPointers_ARGB8888(rb);
+	 intelInitPointers_ARGB1555(rb);
 	 break;
       case I915_TILING_X:
-	 intel_XTile_InitPointers_ARGB8888(rb);
+	 intel_XTile_InitPointers_ARGB1555(rb);
 	 break;
       case I915_TILING_Y:
-	 intel_YTile_InitPointers_ARGB8888(rb);
+	 intel_YTile_InitPointers_ARGB1555(rb);
 	 break;
       }
-   }
-   else if (rb->_ActualFormat == GL_DEPTH_COMPONENT16) {
+      break;
+   case MESA_FORMAT_ARGB8888:
+      if (rb->AlphaBits == 0) { /* XXX: Need xRGB8888 Mesa format */
+	 /* 8888 RGBx */
+	 switch (tiling) {
+	 case I915_TILING_NONE:
+	 default:
+	    intelInitPointers_xRGB8888(rb);
+	    break;
+	 case I915_TILING_X:
+	    intel_XTile_InitPointers_xRGB8888(rb);
+	    break;
+	 case I915_TILING_Y:
+	    intel_YTile_InitPointers_xRGB8888(rb);
+	    break;
+	 }
+      } else {
+	 /* 8888 RGBA */
+	 switch (tiling) {
+	 case I915_TILING_NONE:
+	 default:
+	    intelInitPointers_ARGB8888(rb);
+	    break;
+	 case I915_TILING_X:
+	    intel_XTile_InitPointers_ARGB8888(rb);
+	    break;
+	 case I915_TILING_Y:
+	    intel_YTile_InitPointers_ARGB8888(rb);
+	    break;
+	 }
+      }
+      break;
+   case MESA_FORMAT_Z16:
       switch (tiling) {
       case I915_TILING_NONE:
       default:
@@ -754,38 +668,60 @@ intel_set_span_functions(struct intel_context *intel,
 	 intel_YTile_InitDepthPointers_z16(rb);
 	 break;
       }
-   }
-   else if (rb->_ActualFormat == GL_DEPTH_COMPONENT24 ||        /* XXX FBO remove */
-            rb->_ActualFormat == GL_DEPTH24_STENCIL8_EXT) {
-      switch (tiling) {
-      case I915_TILING_NONE:
-      default:
-	 intelInitDepthPointers_z24_s8(rb);
-	 break;
-      case I915_TILING_X:
-	 intel_XTile_InitDepthPointers_z24_s8(rb);
-	 break;
-      case I915_TILING_Y:
-	 intel_YTile_InitDepthPointers_z24_s8(rb);
-	 break;
-      }
-   }
-   else if (rb->_ActualFormat == GL_STENCIL_INDEX8_EXT) {
-      switch (tiling) {
-      case I915_TILING_NONE:
-      default:
-	 intelInitStencilPointers_z24_s8(rb);
-	 break;
-      case I915_TILING_X:
-	 intel_XTile_InitStencilPointers_z24_s8(rb);
-	 break;
-      case I915_TILING_Y:
-	 intel_YTile_InitStencilPointers_z24_s8(rb);
-	 break;
+      break;
+   case MESA_FORMAT_S8_Z24:
+      /* There are a few different ways SW asks us to access the S8Z24 data:
+       * Z24 depth-only depth reads
+       * S8Z24 depth reads
+       * S8Z24 stencil reads.
+       */
+      if (rb->_ActualFormat == GL_DEPTH_COMPONENT24) {
+	 switch (tiling) {
+	 case I915_TILING_NONE:
+	 default:
+	    intelInitDepthPointers_z24(rb);
+	    break;
+	 case I915_TILING_X:
+	    intel_XTile_InitDepthPointers_z24(rb);
+	    break;
+	 case I915_TILING_Y:
+	    intel_YTile_InitDepthPointers_z24(rb);
+	    break;
+	 }
+      } else if (rb->_ActualFormat == GL_DEPTH24_STENCIL8_EXT) {
+	 switch (tiling) {
+	 case I915_TILING_NONE:
+	 default:
+	    intelInitDepthPointers_z24_s8(rb);
+	    break;
+	 case I915_TILING_X:
+	    intel_XTile_InitDepthPointers_z24_s8(rb);
+	    break;
+	 case I915_TILING_Y:
+	    intel_YTile_InitDepthPointers_z24_s8(rb);
+	    break;
+	 }
+      } else if (rb->_ActualFormat == GL_STENCIL_INDEX8_EXT) {
+	 switch (tiling) {
+	 case I915_TILING_NONE:
+	 default:
+	    intelInitStencilPointers_z24_s8(rb);
+	    break;
+	 case I915_TILING_X:
+	    intel_XTile_InitStencilPointers_z24_s8(rb);
+	    break;
+	 case I915_TILING_Y:
+	    intel_YTile_InitStencilPointers_z24_s8(rb);
+	    break;
+	 }
+      } else {
+	 _mesa_problem(NULL,
+		       "Unexpected ActualFormat in intelSetSpanFunctions");
       }
-   }
-   else {
+      break;
+   default:
       _mesa_problem(NULL,
-                    "Unexpected _ActualFormat in intelSetSpanFunctions");
+                    "Unexpected MesaFormat in intelSetSpanFunctions");
+      break;
    }
 }
diff --git a/src/mesa/drivers/dri/intel/intel_spantmp.h b/src/mesa/drivers/dri/intel/intel_spantmp.h
new file mode 100644
index 0000000000..ead0b1c168
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_spantmp.h
@@ -0,0 +1,61 @@
+/*
+ * Copyright © 2009 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Eric Anholt <eric@anholt.net>
+ *
+ */
+
+/**
+ * Wrapper around the spantmp.h macrofest to generate spans code for
+ * all the tiling styles.
+ */
+
+#define SPANTMP_PIXEL_FMT INTEL_PIXEL_FMT
+#define SPANTMP_PIXEL_TYPE INTEL_PIXEL_TYPE
+#define PUT_VALUE(_x, _y, v) INTEL_WRITE_VALUE(NO_TILE(_x, _y), v)
+#define GET_VALUE(_x, _y) INTEL_READ_VALUE(NO_TILE(_x, _y))
+#define TAG(x) INTEL_TAG(intel##x)
+#define TAG2(x, y) INTEL_TAG(intel##x)##y
+#include "spantmp2.h"
+
+#define SPANTMP_PIXEL_FMT INTEL_PIXEL_FMT
+#define SPANTMP_PIXEL_TYPE INTEL_PIXEL_TYPE
+#define PUT_VALUE(_x, _y, v) INTEL_WRITE_VALUE(X_TILE(_x, _y), v)
+#define GET_VALUE(_x, _y) INTEL_READ_VALUE(X_TILE(_x, _y))
+#define TAG(x) INTEL_TAG(intel_XTile_##x)
+#define TAG2(x, y) INTEL_TAG(intel_XTile_##x)##y
+#include "spantmp2.h"
+
+#define SPANTMP_PIXEL_FMT INTEL_PIXEL_FMT
+#define SPANTMP_PIXEL_TYPE INTEL_PIXEL_TYPE
+#define PUT_VALUE(_x, _y, v) INTEL_WRITE_VALUE(X_TILE(_x, _y), v)
+#define GET_VALUE(_x, _y) INTEL_READ_VALUE(X_TILE(_x, _y))
+#define TAG(x) INTEL_TAG(intel_YTile_##x)
+#define TAG2(x, y) INTEL_TAG(intel_YTile_##x)##y
+#include "spantmp2.h"
+
+#undef INTEL_PIXEL_FMT
+#undef INTEL_PIXEL_TYPE
+#undef INTEL_WRITE_VALUE
+#undef INTEL_READ_VALUE
+#undef INTEL_TAG
diff --git a/src/mesa/drivers/dri/intel/intel_state.c b/src/mesa/drivers/dri/intel/intel_state.c
new file mode 100644
index 0000000000..4ee742377d
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_state.c
@@ -0,0 +1,233 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/macros.h"
+#include "main/enums.h"
+#include "main/colormac.h"
+#include "main/dd.h"
+
+#include "intel_screen.h"
+#include "intel_context.h"
+#include "intel_regions.h"
+#include "swrast/swrast.h"
+
+int
+intel_translate_shadow_compare_func(GLenum func)
+{
+   switch (func) {
+   case GL_NEVER: 
+       return COMPAREFUNC_ALWAYS;
+   case GL_LESS: 
+       return COMPAREFUNC_LEQUAL;
+   case GL_LEQUAL: 
+       return COMPAREFUNC_LESS;
+   case GL_GREATER: 
+       return COMPAREFUNC_GEQUAL;
+   case GL_GEQUAL: 
+      return COMPAREFUNC_GREATER;
+   case GL_NOTEQUAL: 
+      return COMPAREFUNC_EQUAL;
+   case GL_EQUAL: 
+      return COMPAREFUNC_NOTEQUAL;
+   case GL_ALWAYS: 
+       return COMPAREFUNC_NEVER;
+   }
+
+   fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func);
+   return COMPAREFUNC_NEVER;
+}
+
+int
+intel_translate_compare_func(GLenum func)
+{
+   switch (func) {
+   case GL_NEVER:
+      return COMPAREFUNC_NEVER;
+   case GL_LESS:
+      return COMPAREFUNC_LESS;
+   case GL_LEQUAL:
+      return COMPAREFUNC_LEQUAL;
+   case GL_GREATER:
+      return COMPAREFUNC_GREATER;
+   case GL_GEQUAL:
+      return COMPAREFUNC_GEQUAL;
+   case GL_NOTEQUAL:
+      return COMPAREFUNC_NOTEQUAL;
+   case GL_EQUAL:
+      return COMPAREFUNC_EQUAL;
+   case GL_ALWAYS:
+      return COMPAREFUNC_ALWAYS;
+   }
+
+   fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func);
+   return COMPAREFUNC_ALWAYS;
+}
+
+int
+intel_translate_stencil_op(GLenum op)
+{
+   switch (op) {
+   case GL_KEEP:
+      return STENCILOP_KEEP;
+   case GL_ZERO:
+      return STENCILOP_ZERO;
+   case GL_REPLACE:
+      return STENCILOP_REPLACE;
+   case GL_INCR:
+      return STENCILOP_INCRSAT;
+   case GL_DECR:
+      return STENCILOP_DECRSAT;
+   case GL_INCR_WRAP:
+      return STENCILOP_INCR;
+   case GL_DECR_WRAP:
+      return STENCILOP_DECR;
+   case GL_INVERT:
+      return STENCILOP_INVERT;
+   default:
+      return STENCILOP_ZERO;
+   }
+}
+
+int
+intel_translate_blend_factor(GLenum factor)
+{
+   switch (factor) {
+   case GL_ZERO:
+      return BLENDFACT_ZERO;
+   case GL_SRC_ALPHA:
+      return BLENDFACT_SRC_ALPHA;
+   case GL_ONE:
+      return BLENDFACT_ONE;
+   case GL_SRC_COLOR:
+      return BLENDFACT_SRC_COLR;
+   case GL_ONE_MINUS_SRC_COLOR:
+      return BLENDFACT_INV_SRC_COLR;
+   case GL_DST_COLOR:
+      return BLENDFACT_DST_COLR;
+   case GL_ONE_MINUS_DST_COLOR:
+      return BLENDFACT_INV_DST_COLR;
+   case GL_ONE_MINUS_SRC_ALPHA:
+      return BLENDFACT_INV_SRC_ALPHA;
+   case GL_DST_ALPHA:
+      return BLENDFACT_DST_ALPHA;
+   case GL_ONE_MINUS_DST_ALPHA:
+      return BLENDFACT_INV_DST_ALPHA;
+   case GL_SRC_ALPHA_SATURATE:
+      return BLENDFACT_SRC_ALPHA_SATURATE;
+   case GL_CONSTANT_COLOR:
+      return BLENDFACT_CONST_COLOR;
+   case GL_ONE_MINUS_CONSTANT_COLOR:
+      return BLENDFACT_INV_CONST_COLOR;
+   case GL_CONSTANT_ALPHA:
+      return BLENDFACT_CONST_ALPHA;
+   case GL_ONE_MINUS_CONSTANT_ALPHA:
+      return BLENDFACT_INV_CONST_ALPHA;
+   }
+
+   fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, factor);
+   return BLENDFACT_ZERO;
+}
+
+int
+intel_translate_logic_op(GLenum opcode)
+{
+   switch (opcode) {
+   case GL_CLEAR:
+      return LOGICOP_CLEAR;
+   case GL_AND:
+      return LOGICOP_AND;
+   case GL_AND_REVERSE:
+      return LOGICOP_AND_RVRSE;
+   case GL_COPY:
+      return LOGICOP_COPY;
+   case GL_COPY_INVERTED:
+      return LOGICOP_COPY_INV;
+   case GL_AND_INVERTED:
+      return LOGICOP_AND_INV;
+   case GL_NOOP:
+      return LOGICOP_NOOP;
+   case GL_XOR:
+      return LOGICOP_XOR;
+   case GL_OR:
+      return LOGICOP_OR;
+   case GL_OR_INVERTED:
+      return LOGICOP_OR_INV;
+   case GL_NOR:
+      return LOGICOP_NOR;
+   case GL_EQUIV:
+      return LOGICOP_EQUIV;
+   case GL_INVERT:
+      return LOGICOP_INV;
+   case GL_OR_REVERSE:
+      return LOGICOP_OR_RVRSE;
+   case GL_NAND:
+      return LOGICOP_NAND;
+   case GL_SET:
+      return LOGICOP_SET;
+   default:
+      return LOGICOP_SET;
+   }
+}
+
+
+static void
+intelClearColor(GLcontext *ctx, const GLfloat color[4])
+{
+   struct intel_context *intel = intel_context(ctx);
+   GLubyte clear[4];
+
+   CLAMPED_FLOAT_TO_UBYTE(clear[0], color[0]);
+   CLAMPED_FLOAT_TO_UBYTE(clear[1], color[1]);
+   CLAMPED_FLOAT_TO_UBYTE(clear[2], color[2]);
+   CLAMPED_FLOAT_TO_UBYTE(clear[3], color[3]);
+
+   /* compute both 32 and 16-bit clear values */
+   intel->ClearColor8888 = INTEL_PACKCOLOR8888(clear[0], clear[1],
+                                               clear[2], clear[3]);
+   intel->ClearColor565 = INTEL_PACKCOLOR565(clear[0], clear[1], clear[2]);
+}
+
+
+/* Fallback to swrast for select and feedback.
+ */
+static void
+intelRenderMode(GLcontext *ctx, GLenum mode)
+{
+   struct intel_context *intel = intel_context(ctx);
+   FALLBACK(intel, INTEL_FALLBACK_RENDERMODE, (mode != GL_RENDER));
+}
+
+
+void
+intelInitStateFuncs(struct dd_function_table *functions)
+{
+   functions->RenderMode = intelRenderMode;
+   functions->ClearColor = intelClearColor;
+}
diff --git a/src/mesa/drivers/dri/intel/intel_swapbuffers.c b/src/mesa/drivers/dri/intel/intel_swapbuffers.c
new file mode 100644
index 0000000000..7d035b9f6e
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_swapbuffers.c
@@ -0,0 +1,248 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include "intel_blit.h"
+#include "intel_buffers.h"
+#include "intel_swapbuffers.h"
+#include "intel_fbo.h"
+#include "intel_batchbuffer.h"
+#include "drirenderbuffer.h"
+#include "vblank.h"
+#include "i915_drm.h"
+
+
+
+/*
+ * Correct a drawablePrivate's set of vblank flags WRT the current context.
+ * When considering multiple crtcs.
+ */
+GLuint
+intelFixupVblank(struct intel_context *intel, __DRIdrawablePrivate *dPriv)
+{
+   if (!intel->intelScreen->driScrnPriv->dri2.enabled &&
+       intel->intelScreen->driScrnPriv->ddx_version.minor >= 7) {
+      volatile drm_i915_sarea_t *sarea = intel->sarea;
+      drm_clip_rect_t drw_rect = { .x1 = dPriv->x, .x2 = dPriv->x + dPriv->w,
+				   .y1 = dPriv->y, .y2 = dPriv->y + dPriv->h };
+      drm_clip_rect_t planeA_rect = { .x1 = sarea->planeA_x, .y1 = sarea->planeA_y,
+				     .x2 = sarea->planeA_x + sarea->planeA_w,
+				     .y2 = sarea->planeA_y + sarea->planeA_h };
+      drm_clip_rect_t planeB_rect = { .x1 = sarea->planeB_x, .y1 = sarea->planeB_y,
+				     .x2 = sarea->planeB_x + sarea->planeB_w,
+				     .y2 = sarea->planeB_y + sarea->planeB_h };
+      GLint areaA = driIntersectArea( drw_rect, planeA_rect );
+      GLint areaB = driIntersectArea( drw_rect, planeB_rect );
+      GLuint flags = dPriv->vblFlags;
+
+      /* Update vblank info
+       */
+      if (areaB > areaA || (areaA == areaB && areaB > 0)) {
+	 flags = dPriv->vblFlags | VBLANK_FLAG_SECONDARY;
+      } else {
+	 flags = dPriv->vblFlags & ~VBLANK_FLAG_SECONDARY;
+      }
+
+      /* Do the stupid test: Is one of them actually disabled?
+       */
+      if (sarea->planeA_w == 0 || sarea->planeA_h == 0) {
+	 flags = dPriv->vblFlags | VBLANK_FLAG_SECONDARY;
+      } else if (sarea->planeB_w == 0 || sarea->planeB_h == 0) {
+	 flags = dPriv->vblFlags & ~VBLANK_FLAG_SECONDARY;
+      }
+
+      return flags;
+   } else {
+      return dPriv->vblFlags & ~VBLANK_FLAG_SECONDARY;
+   }
+}
+
+
+/**
+ * Called from driSwapBuffers()
+ */
+void
+intelSwapBuffers(__DRIdrawablePrivate * dPriv)
+{
+   __DRIscreenPrivate *psp = dPriv->driScreenPriv;
+
+   if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
+      GET_CURRENT_CONTEXT(ctx);
+      struct intel_context *intel;
+
+      if (ctx == NULL)
+	 return;
+
+      intel = intel_context(ctx);
+
+      if (ctx->Visual.doubleBufferMode) {
+	 GLboolean missed_target;
+	 struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
+	 int64_t ust;
+         
+	 _mesa_notifySwapBuffers(ctx);  /* flush pending rendering comands */
+
+	/*
+	 * The old swapping ioctl was incredibly racy, just wait for vblank
+	 * and do the swap ourselves.
+	 */
+	 driWaitForVBlank(dPriv, &missed_target);
+
+	 /*
+	  * Update each buffer's vbl_pending so we don't get too out of
+	  * sync
+	  */
+	 intel_get_renderbuffer(&intel_fb->Base,
+		   		BUFFER_BACK_LEFT)->vbl_pending = dPriv->vblSeq;
+         intel_get_renderbuffer(&intel_fb->Base,
+		   		BUFFER_FRONT_LEFT)->vbl_pending = dPriv->vblSeq;
+
+	 intelCopyBuffer(dPriv, NULL);
+
+	 intel_fb->swap_count++;
+	 (*psp->systemTime->getUST) (&ust);
+	 if (missed_target) {
+	    intel_fb->swap_missed_count++;
+	    intel_fb->swap_missed_ust = ust - intel_fb->swap_ust;
+	 }
+
+	 intel_fb->swap_ust = ust;
+      }
+      drmCommandNone(intel->driFd, DRM_I915_GEM_THROTTLE);
+   }
+   else {
+      /* XXX this shouldn't be an error but we can't handle it for now */
+      fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__);
+   }
+}
+
+
+/**
+ * Called from driCopySubBuffer()
+ */
+void
+intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h)
+{
+   if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
+      struct intel_context *intel =
+         (struct intel_context *) dPriv->driContextPriv->driverPrivate;
+      GLcontext *ctx = &intel->ctx;
+
+      if (ctx->Visual.doubleBufferMode) {
+         drm_clip_rect_t rect;
+         rect.x1 = x + dPriv->x;
+         rect.y1 = (dPriv->h - y - h) + dPriv->y;
+         rect.x2 = rect.x1 + w;
+         rect.y2 = rect.y1 + h;
+         _mesa_notifySwapBuffers(ctx);  /* flush pending rendering comands */
+         intelCopyBuffer(dPriv, &rect);
+      }
+   }
+   else {
+      /* XXX this shouldn't be an error but we can't handle it for now */
+      fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__);
+   }
+}
+
+
+/**
+ * This will be called whenever the currently bound window is moved/resized.
+ * XXX: actually, it seems to NOT be called when the window is only moved (BP).
+ */
+void
+intelWindowMoved(struct intel_context *intel)
+{
+   GLcontext *ctx = &intel->ctx;
+   __DRIdrawablePrivate *dPriv = intel->driDrawable;
+   struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
+
+   if (!intel->intelScreen->driScrnPriv->dri2.enabled &&
+       intel->intelScreen->driScrnPriv->ddx_version.minor >= 7) {
+      GLuint flags = intelFixupVblank(intel, dPriv);
+
+      /* Check to see if we changed pipes */
+      if (flags != dPriv->vblFlags && dPriv->vblFlags &&
+	  !(dPriv->vblFlags & VBLANK_FLAG_NO_IRQ)) {
+	 int64_t count;
+	 drmVBlank vbl;
+	 int i;
+
+	 /*
+	  * Deal with page flipping
+	  */
+	 vbl.request.type = DRM_VBLANK_ABSOLUTE;
+
+	 if ( dPriv->vblFlags & VBLANK_FLAG_SECONDARY ) {
+	    vbl.request.type |= DRM_VBLANK_SECONDARY;
+	 }
+
+	 for (i = 0; i < 2; i++) {
+	    if (!intel_fb->color_rb[i] ||
+		(intel_fb->vbl_waited - intel_fb->color_rb[i]->vbl_pending) <=
+		(1<<23))
+	       continue;
+
+	    vbl.request.sequence = intel_fb->color_rb[i]->vbl_pending;
+	    drmWaitVBlank(intel->driFd, &vbl);
+	 }
+
+	 /*
+	  * Update msc_base from old pipe
+	  */
+	 driDrawableGetMSC32(dPriv->driScreenPriv, dPriv, &count);
+	 dPriv->msc_base = count;
+	 /*
+	  * Then get new vblank_base and vblSeq values
+	  */
+	 dPriv->vblFlags = flags;
+	 driGetCurrentVBlank(dPriv);
+	 dPriv->vblank_base = dPriv->vblSeq;
+
+	 intel_fb->vbl_waited = dPriv->vblSeq;
+
+	 for (i = 0; i < 2; i++) {
+	    if (intel_fb->color_rb[i])
+	       intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_waited;
+	 }
+      }
+   } else {
+      dPriv->vblFlags &= ~VBLANK_FLAG_SECONDARY;
+   }
+
+   /* Update Mesa's notion of window size */
+   driUpdateFramebufferSize(ctx, dPriv);
+   intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */
+
+   /* Update hardware scissor */
+   if (ctx->Driver.Scissor != NULL) {
+      ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y,
+			  ctx->Scissor.Width, ctx->Scissor.Height);
+   }
+
+   /* Re-calculate viewport related state */
+   if (ctx->Driver.DepthRange != NULL)
+      ctx->Driver.DepthRange( ctx, ctx->Viewport.Near, ctx->Viewport.Far );
+}
diff --git a/src/mesa/drivers/dri/intel/intel_swapbuffers.h b/src/mesa/drivers/dri/intel/intel_swapbuffers.h
new file mode 100644
index 0000000000..75bb6242ff
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_swapbuffers.h
@@ -0,0 +1,52 @@
+
+/**************************************************************************
+ * 
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#ifndef INTEL_SWAPBUFFERS_H
+#define INTEL_SWAPBUFFERS_H
+
+#include "dri_util.h"
+#include "drm.h"
+
+struct intel_context;
+struct intel_framebuffer;
+
+
+extern void
+intelSwapBuffers(__DRIdrawablePrivate * dPriv);
+
+extern void
+intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h);
+
+extern GLuint
+intelFixupVblank(struct intel_context *intel, __DRIdrawablePrivate *dPriv);
+
+extern void
+intelWindowMoved(struct intel_context *intel);
+
+
+#endif /* INTEL_SWAPBUFFERS_H */
diff --git a/src/mesa/drivers/dri/intel/intel_syncobj.c b/src/mesa/drivers/dri/intel/intel_syncobj.c
new file mode 100644
index 0000000000..b6ea56d547
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_syncobj.c
@@ -0,0 +1,142 @@
+/*
+ * Copyright © 2008 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Eric Anholt <eric@anholt.net>
+ *
+ */
+
+/** @file intel_syncobj.c
+ *
+ * Support for ARB_sync
+ *
+ * ARB_sync is implemented by flushing the current batchbuffer and keeping a
+ * reference on it.  We can then check for completion or wait for compeltion
+ * using the normal buffer object mechanisms.  This does mean that if an
+ * application is using many sync objects, it will emit small batchbuffers
+ * which may end up being a significant overhead.  In other tests of removing
+ * gratuitous batchbuffer syncs in Mesa, it hasn't appeared to be a significant
+ * performance bottleneck, though.
+ */
+
+#include "main/simple_list.h"
+#include "main/imports.h"
+
+#include "intel_context.h"
+#include "intel_batchbuffer.h"
+#include "intel_reg.h"
+
+#if FEATURE_ARB_sync
+
+static struct gl_sync_object *
+intel_new_sync_object(GLcontext *ctx, GLuint id)
+{
+   struct intel_sync_object *sync;
+
+   sync = _mesa_calloc(sizeof(struct intel_sync_object));
+
+   return &sync->Base;
+}
+
+static void
+intel_delete_sync_object(GLcontext *ctx, struct gl_sync_object *s)
+{
+   struct intel_sync_object *sync = (struct intel_sync_object *)s;
+
+   drm_intel_bo_unreference(sync->bo);
+   _mesa_free(sync);
+}
+
+static void
+intel_fence_sync(GLcontext *ctx, struct gl_sync_object *s,
+	       GLenum condition, GLbitfield flags)
+{
+   struct intel_context *intel = intel_context(ctx);
+   struct intel_sync_object *sync = (struct intel_sync_object *)s;
+
+   assert(condition == GL_SYNC_GPU_COMMANDS_COMPLETE);
+   intel_batchbuffer_emit_mi_flush(intel->batch);
+
+   sync->bo = intel->batch->buf;
+   drm_intel_bo_reference(sync->bo);
+
+   intelFlush(ctx);
+}
+
+/* We ignore the user-supplied timeout.  This is weaselly -- we're allowed to
+ * round to an implementation-dependent accuracy, and right now our
+ * implementation "rounds" to the wait-forever value.
+ *
+ * The fix would be a new kernel function to do the GTT transition with a
+ * timeout.
+ */
+static void intel_client_wait_sync(GLcontext *ctx, struct gl_sync_object *s,
+				 GLbitfield flags, GLuint64 timeout)
+{
+   struct intel_sync_object *sync = (struct intel_sync_object *)s;
+
+   if (sync->bo) {
+      drm_intel_bo_wait_rendering(sync->bo);
+      s->StatusFlag = 1;
+      drm_intel_bo_unreference(sync->bo);
+      sync->bo = NULL;
+   }
+}
+
+/* We have nothing to do for WaitSync.  Our GL command stream is sequential,
+ * so given that the sync object has already flushed the batchbuffer,
+ * any batchbuffers coming after this waitsync will naturally not occur until
+ * the previous one is done.
+ */
+static void intel_server_wait_sync(GLcontext *ctx, struct gl_sync_object *s,
+				 GLbitfield flags, GLuint64 timeout)
+{
+}
+
+static void intel_check_sync(GLcontext *ctx, struct gl_sync_object *s)
+{
+   struct intel_sync_object *sync = (struct intel_sync_object *)s;
+
+   if (sync->bo && drm_intel_bo_busy(sync->bo)) {
+      drm_intel_bo_unreference(sync->bo);
+      sync->bo = NULL;
+      s->StatusFlag = 1;
+   }
+}
+
+void intel_init_syncobj_functions(struct dd_function_table *functions)
+{
+   functions->NewSyncObject = intel_new_sync_object;
+   functions->DeleteSyncObject = intel_delete_sync_object;
+   functions->FenceSync = intel_fence_sync;
+   functions->CheckSync = intel_check_sync;
+   functions->ClientWaitSync = intel_client_wait_sync;
+   functions->ServerWaitSync = intel_server_wait_sync;
+}
+
+#else /* FEATURE_ARB_sync */
+
+void intel_init_syncobj_functions(struct dd_function_table *functions)
+{
+}
+
+#endif
diff --git a/src/mesa/drivers/dri/intel/intel_tex.c b/src/mesa/drivers/dri/intel/intel_tex.c
index e64d8a1556..df63f29a42 100644
--- a/src/mesa/drivers/dri/intel/intel_tex.c
+++ b/src/mesa/drivers/dri/intel/intel_tex.c
@@ -93,7 +93,7 @@ intelFreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage)
 static void *
 do_memcpy(void *dest, const void *src, size_t n)
 {
-   if ((((unsigned) src) & 63) || (((unsigned) dest) & 63)) {
+   if ((((unsigned long) src) & 63) || (((unsigned long) dest) & 63)) {
       return __memcpy(dest, src, n);
    }
    else
@@ -158,81 +158,11 @@ timed_memcpy(void *dest, const void *src, size_t n)
 }
 #endif /* DO_DEBUG */
 
-/**
- * Generate new mipmap data from BASE+1 to BASE+p (the minimally-sized mipmap
- * level).
- *
- * The texture object's miptree must be mapped.
- *
- * It would be really nice if this was just called by Mesa whenever mipmaps
- * needed to be regenerated, rather than us having to remember to do so in
- * each texture image modification path.
- *
- * This function should also include an accelerated path.
- */
-void
-intel_generate_mipmap(GLcontext *ctx, GLenum target,
-                      struct gl_texture_object *texObj)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct intel_texture_object *intelObj = intel_texture_object(texObj);
-   GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
-   int face, i;
-
-   _mesa_generate_mipmap(ctx, target, texObj);
-
-   /* Update the level information in our private data in the new images, since
-    * it didn't get set as part of a normal TexImage path.
-    */
-   for (face = 0; face < nr_faces; face++) {
-      for (i = texObj->BaseLevel + 1; i < texObj->MaxLevel; i++) {
-         struct intel_texture_image *intelImage;
-
-	 intelImage = intel_texture_image(texObj->Image[face][i]);
-	 if (intelImage == NULL)
-	    break;
-
-	 intelImage->level = i;
-	 intelImage->face = face;
-	 /* Unreference the miptree to signal that the new Data is a bare
-	  * pointer from mesa.
-	  */
-	 intel_miptree_release(intel, &intelImage->mt);
-      }
-   }
-}
-
-static void intelGenerateMipmap(GLcontext *ctx, GLenum target, struct gl_texture_object *texObj)
-{
-   struct intel_context *intel = intel_context(ctx);
-   struct intel_texture_object *intelObj = intel_texture_object(texObj);
-
-   intel_tex_map_level_images(intel, intelObj, texObj->BaseLevel);
-   intel_generate_mipmap(ctx, target, texObj);
-   intel_tex_unmap_level_images(intel, intelObj, texObj->BaseLevel);
-}
-
 void
 intelInitTextureFuncs(struct dd_function_table *functions)
 {
    functions->ChooseTextureFormat = intelChooseTextureFormat;
-   functions->TexImage1D = intelTexImage1D;
-   functions->TexImage2D = intelTexImage2D;
-   functions->TexImage3D = intelTexImage3D;
-   functions->TexSubImage1D = intelTexSubImage1D;
-   functions->TexSubImage2D = intelTexSubImage2D;
-   functions->TexSubImage3D = intelTexSubImage3D;
-   functions->CopyTexImage1D = intelCopyTexImage1D;
-   functions->CopyTexImage2D = intelCopyTexImage2D;
-   functions->CopyTexSubImage1D = intelCopyTexSubImage1D;
-   functions->CopyTexSubImage2D = intelCopyTexSubImage2D;
-   functions->GetTexImage = intelGetTexImage;
-   functions->GenerateMipmap = intelGenerateMipmap;
-
-   /* compressed texture functions */
-   functions->CompressedTexImage2D = intelCompressedTexImage2D;
-   functions->CompressedTexSubImage2D = intelCompressedTexSubImage2D;
-   functions->GetCompressedTexImage = intelGetCompressedTexImage;
+   functions->GenerateMipmap = intel_generate_mipmap;
 
    functions->NewTextureObject = intelNewTextureObject;
    functions->NewTextureImage = intelNewTextureImage;
diff --git a/src/mesa/drivers/dri/intel/intel_tex.h b/src/mesa/drivers/dri/intel/intel_tex.h
index 742ccc043a..471aa2a240 100644
--- a/src/mesa/drivers/dri/intel/intel_tex.h
+++ b/src/mesa/drivers/dri/intel/intel_tex.h
@@ -35,120 +35,23 @@
 
 void intelInitTextureFuncs(struct dd_function_table *functions);
 
+void intelInitTextureImageFuncs(struct dd_function_table *functions);
+
+void intelInitTextureSubImageFuncs(struct dd_function_table *functions);
+
+void intelInitTextureCopyImageFuncs(struct dd_function_table *functions);
+
 const struct gl_texture_format *intelChooseTextureFormat(GLcontext * ctx,
                                                          GLint internalFormat,
                                                          GLenum format,
                                                          GLenum type);
 
-
-void intelTexImage3D(GLcontext * ctx,
-                     GLenum target, GLint level,
-                     GLint internalFormat,
-                     GLint width, GLint height, GLint depth,
-                     GLint border,
-                     GLenum format, GLenum type, const void *pixels,
-                     const struct gl_pixelstore_attrib *packing,
-                     struct gl_texture_object *texObj,
-                     struct gl_texture_image *texImage);
-
-void intelTexSubImage3D(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);
-
-void intelTexImage2D(GLcontext * ctx,
-                     GLenum target, GLint level,
-                     GLint internalFormat,
-                     GLint width, GLint height, GLint border,
-                     GLenum format, GLenum type, const void *pixels,
-                     const struct gl_pixelstore_attrib *packing,
-                     struct gl_texture_object *texObj,
-                     struct gl_texture_image *texImage);
-
-void intelTexSubImage2D(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);
-
-void intelTexImage1D(GLcontext * ctx,
-                     GLenum target, GLint level,
-                     GLint internalFormat,
-                     GLint width, GLint border,
-                     GLenum format, GLenum type, const void *pixels,
-                     const struct gl_pixelstore_attrib *packing,
-                     struct gl_texture_object *texObj,
-                     struct gl_texture_image *texImage);
-
-void intelTexSubImage1D(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);
-
-void intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level,
-                         GLenum internalFormat,
-                         GLint x, GLint y, GLsizei width, GLint border);
-
-void intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level,
-                         GLenum internalFormat,
-                         GLint x, GLint y, GLsizei width, GLsizei height,
-                         GLint border);
-
-void intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level,
-                            GLint xoffset, GLint x, GLint y, GLsizei width);
-
-void intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
-                            GLint xoffset, GLint yoffset,
-                            GLint x, GLint y, GLsizei width, GLsizei height);
-
-void intelGetTexImage(GLcontext * ctx, GLenum target, GLint level,
-                      GLenum format, GLenum type, GLvoid * pixels,
-                      struct gl_texture_object *texObj,
-                      struct gl_texture_image *texImage);
-
-void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level,
-				GLint internalFormat,
-				GLint width, GLint height, GLint border,
-				GLsizei imageSize, const GLvoid *data,
-				struct gl_texture_object *texObj,
-				struct gl_texture_image *texImage );
-
-void intelCompressedTexSubImage2D(GLcontext * ctx,
-				  GLenum target,
-				  GLint level,
-				  GLint xoffset, GLint yoffset,
-				  GLsizei width, GLsizei height,
-				  GLenum format, GLsizei imageSize,
-				  const GLvoid * pixels,
-				  struct gl_texture_object *texObj,
-				  struct gl_texture_image *texImage);
-
-void intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level,
-				GLvoid *pixels,
-				struct gl_texture_object *texObj,
-				struct gl_texture_image *texImage);
-
 void intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname,
 		       unsigned long long offset, GLint depth, GLuint pitch);
 void intelSetTexBuffer(__DRIcontext *pDRICtx,
 		       GLint target, __DRIdrawable *pDraw);
+void intelSetTexBuffer2(__DRIcontext *pDRICtx,
+			GLint target, GLint format, __DRIdrawable *pDraw);
 
 GLuint intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit);
 
diff --git a/src/mesa/drivers/dri/intel/intel_tex_copy.c b/src/mesa/drivers/dri/intel/intel_tex_copy.c
index 08437aa0e2..028b49c14d 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_copy.c
@@ -73,11 +73,8 @@ get_teximage_source(struct intel_context *intel, GLenum internalFormat)
       return NULL;
    case GL_RGBA:
    case GL_RGBA8:
-      return intel_readbuf_region(intel);
    case GL_RGB:
-      if (intel->ctx.Visual.rgbBits == 16)
-         return intel_readbuf_region(intel);
-      return NULL;
+      return intel_readbuf_region(intel);
    default:
       return NULL;
    }
@@ -99,14 +96,24 @@ do_copy_texsubimage(struct intel_context *intel,
 
    if (!intelImage->mt || !src) {
       if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "%s fail %p %p\n",
-		 __FUNCTION__, intelImage->mt, src);
+	 fprintf(stderr, "%s fail %p %p (0x%08x)\n",
+		 __FUNCTION__, intelImage->mt, src, internalFormat);
+      return GL_FALSE;
+   }
+
+   if (intelImage->mt->cpp != src->cpp) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "%s fail %d vs %d cpp\n",
+		 __FUNCTION__, intelImage->mt->cpp, src->cpp);
       return GL_FALSE;
    }
 
    intelFlush(ctx);
    LOCK_HARDWARE(intel);
    {
+      drm_intel_bo *dst_bo = intel_region_buffer(intel,
+						 intelImage->mt->region,
+						 INTEL_WRITE_PART);
       GLuint image_offset = intel_miptree_image_offset(intelImage->mt,
                                                        intelImage->face,
                                                        intelImage->level);
@@ -118,8 +125,12 @@ do_copy_texsubimage(struct intel_context *intel,
       dstx += x - orig_x;
       dsty += y - orig_y;
 
-      /* image_offset may be non-page-aligned, but that's illegal for tiling. */
-      assert(intelImage->mt->region->tiling == I915_TILING_NONE);
+      /* Can't blit to tiled buffers with non-tile-aligned offset. */
+      if (intelImage->mt->region->tiling != I915_TILING_NONE &&
+	  (image_offset & 4095) != 0) {
+	 UNLOCK_HARDWARE(intel);
+	 return GL_FALSE;
+      }
 
       if (ctx->ReadBuffer->Name == 0) {
 	 /* reading from a window, adjust x, y */
@@ -140,35 +151,35 @@ do_copy_texsubimage(struct intel_context *intel,
 	 src_pitch = src->pitch;
       }
 
-      intelEmitCopyBlit(intel,
-			intelImage->mt->cpp,
-			src_pitch,
-			src->buffer,
-			0,
-			src->tiling,
-			intelImage->mt->pitch,
-			intelImage->mt->region->buffer,
-			image_offset,
-			intelImage->mt->region->tiling,
-			x, y, dstx, dsty, width, height,
-			GL_COPY);
+      if (!intelEmitCopyBlit(intel,
+			     intelImage->mt->cpp,
+			     src_pitch,
+			     src->buffer,
+			     0,
+			     src->tiling,
+			     intelImage->mt->pitch,
+			     dst_bo,
+			     image_offset,
+			     intelImage->mt->region->tiling,
+			     x, y, dstx, dsty, width, height,
+			     GL_COPY)) {
+	 UNLOCK_HARDWARE(intel);
+	 return GL_FALSE;
+      }
    }
 
    UNLOCK_HARDWARE(intel);
 
    /* GL_SGIS_generate_mipmap */
    if (intelImage->level == texObj->BaseLevel && texObj->GenerateMipmap) {
-      ctx->Driver.GenerateMipmap(ctx, target, texObj);
+      intel_generate_mipmap(ctx, target, texObj);
    }
 
    return GL_TRUE;
 }
 
 
-
-
-
-void
+static void
 intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level,
                     GLenum internalFormat,
                     GLint x, GLint y, GLsizei width, GLint border)
@@ -214,7 +225,8 @@ intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level,
                            width, border);
 }
 
-void
+
+static void
 intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level,
                     GLenum internalFormat,
                     GLint x, GLint y, GLsizei width, GLsizei height,
@@ -231,6 +243,14 @@ intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level,
    if (border)
       goto fail;
 
+   /* Setup or redefine the texture object, mipmap tree and texture
+    * image.  Don't populate yet.
+    */
+   ctx->Driver.TexImage2D(ctx, target, level, internalFormat,
+                          width, height, border,
+                          GL_RGBA, CHAN_TYPE, NULL,
+                          &ctx->DefaultPacking, texObj, texImage);
+
    srcx = x;
    srcy = y;
    dstx = 0;
@@ -241,15 +261,6 @@ intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level,
 				   &width, &height))
       return;
 
-   /* Setup or redefine the texture object, mipmap tree and texture
-    * image.  Don't populate yet.  
-    */
-   ctx->Driver.TexImage2D(ctx, target, level, internalFormat,
-                          width, height, border,
-                          GL_RGBA, CHAN_TYPE, NULL,
-                          &ctx->DefaultPacking, texObj, texImage);
-
-
    if (!do_copy_texsubimage(intel_context(ctx), target,
                             intel_texture_image(texImage),
                             internalFormat, 0, 0, x, y, width, height))
@@ -263,7 +274,7 @@ intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level,
 }
 
 
-void
+static void
 intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level,
                        GLint xoffset, GLint x, GLint y, GLsizei width)
 {
@@ -288,8 +299,7 @@ intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level,
 }
 
 
-
-void
+static void
 intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
                        GLint xoffset, GLint yoffset,
                        GLint x, GLint y, GLsizei width, GLsizei height)
@@ -302,7 +312,6 @@ intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
       _mesa_select_tex_image(ctx, texObj, target, level);
    GLenum internalFormat = texImage->InternalFormat;
 
-
    /* Need to check texture is compatible with source format. 
     */
 
@@ -317,3 +326,13 @@ intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level,
                                  xoffset, yoffset, x, y, width, height);
    }
 }
+
+
+void
+intelInitTextureCopyImageFuncs(struct dd_function_table *functions)
+{
+   functions->CopyTexImage1D = intelCopyTexImage1D;
+   functions->CopyTexImage2D = intelCopyTexImage2D;
+   functions->CopyTexSubImage1D = intelCopyTexSubImage1D;
+   functions->CopyTexSubImage2D = intelCopyTexSubImage2D;
+}
diff --git a/src/mesa/drivers/dri/intel/intel_tex_format.c b/src/mesa/drivers/dri/intel/intel_tex_format.c
index 5e418ac446..3322a71130 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_format.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_format.c
@@ -1,13 +1,18 @@
 #include "intel_context.h"
 #include "intel_tex.h"
+#include "intel_chipset.h"
 #include "main/texformat.h"
 #include "main/enums.h"
 
-/* It works out that this function is fine for all the supported
+
+/**
+ * Choose hardware texture format given the user's glTexImage parameters.
+ *
+ * It works out that this function is fine for all the supported
  * hardware.  However, there is still a need to map the formats onto
  * hardware descriptors.
- */
-/* Note that the i915 can actually support many more formats than
+ *
+ * Note that the i915 can actually support many more formats than
  * these if we take the step of simply swizzling the colors
  * immediately after sampling...
  */
@@ -16,7 +21,12 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat,
                          GLenum format, GLenum type)
 {
    struct intel_context *intel = intel_context(ctx);
-   const GLboolean do32bpt = (intel->ctx.Visual.rgbBits == 32);
+   const GLboolean do32bpt = (intel->ctx.Visual.rgbBits >= 24);
+
+#if 0
+   printf("%s intFmt=0x%x format=0x%x type=0x%x\n",
+          __FUNCTION__, internalFormat, format, type);
+#endif
 
    switch (internalFormat) {
    case 4:
@@ -151,20 +161,36 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat,
    case GL_SRGB8_EXT:
    case GL_SRGB_ALPHA_EXT:
    case GL_SRGB8_ALPHA8_EXT:
-   case GL_SLUMINANCE_EXT:
-   case GL_SLUMINANCE8_EXT:
-   case GL_SLUMINANCE_ALPHA_EXT:
-   case GL_SLUMINANCE8_ALPHA8_EXT:
    case GL_COMPRESSED_SRGB_EXT:
    case GL_COMPRESSED_SRGB_ALPHA_EXT:
    case GL_COMPRESSED_SLUMINANCE_EXT:
    case GL_COMPRESSED_SLUMINANCE_ALPHA_EXT:
-       return &_mesa_texformat_srgba8;
+      return &_mesa_texformat_sargb8;
+   case GL_SLUMINANCE_EXT:
+   case GL_SLUMINANCE8_EXT:
+      if (IS_G4X(intel->intelScreen->deviceID))
+         return &_mesa_texformat_sl8;
+      else
+         return &_mesa_texformat_sargb8;
+   case GL_SLUMINANCE_ALPHA_EXT:
+   case GL_SLUMINANCE8_ALPHA8_EXT:
+      if (IS_G4X(intel->intelScreen->deviceID))
+         return &_mesa_texformat_sla8;
+      else
+         return &_mesa_texformat_sargb8;
    case GL_COMPRESSED_SRGB_S3TC_DXT1_EXT:
    case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT1_EXT:
    case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT3_EXT:
    case GL_COMPRESSED_SRGB_ALPHA_S3TC_DXT5_EXT:
       return &_mesa_texformat_srgb_dxt1;
+
+   /* i915 could also do this */
+   case GL_DUDV_ATI:
+   case GL_DU8DV8_ATI:
+      return &_mesa_texformat_dudv8;
+   case GL_RGBA_SNORM:
+   case GL_RGBA8_SNORM:
+      return &_mesa_texformat_signed_rgba8888_rev;
 #endif
 
    default:
diff --git a/src/mesa/drivers/dri/intel/intel_tex_image.c b/src/mesa/drivers/dri/intel/intel_tex_image.c
index 2ac7dceb0f..a206fe6805 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_image.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_image.c
@@ -1,17 +1,14 @@
 
-#include <stdlib.h>
-#include <stdio.h>
-
 #include "main/glheader.h"
 #include "main/macros.h"
 #include "main/mtypes.h"
 #include "main/enums.h"
-#include "main/colortab.h"
+#include "main/bufferobj.h"
 #include "main/convolve.h"
 #include "main/context.h"
-#include "main/simple_list.h"
 #include "main/texcompress.h"
 #include "main/texformat.h"
+#include "main/texgetimage.h"
 #include "main/texobj.h"
 #include "main/texstore.h"
 #include "main/teximage.h"
@@ -62,7 +59,8 @@ logbase2(int n)
 static void
 guess_and_alloc_mipmap_tree(struct intel_context *intel,
                             struct intel_texture_object *intelObj,
-                            struct intel_texture_image *intelImage)
+                            struct intel_texture_image *intelImage,
+			    GLboolean expect_accelerated_upload)
 {
    GLuint firstLevel;
    GLuint lastLevel;
@@ -129,6 +127,7 @@ guess_and_alloc_mipmap_tree(struct intel_context *intel,
       comp_byte = intel_compressed_num_bytes(intelImage->base.TexFormat->MesaFormat);
    intelObj->mt = intel_miptree_create(intel,
                                        intelObj->base.Target,
+                                       intelImage->base._BaseFormat,
                                        intelImage->base.InternalFormat,
                                        firstLevel,
                                        lastLevel,
@@ -136,7 +135,8 @@ guess_and_alloc_mipmap_tree(struct intel_context *intel,
                                        height,
                                        depth,
                                        intelImage->base.TexFormat->TexelBytes,
-                                       comp_byte);
+                                       comp_byte,
+				       expect_accelerated_upload);
 
    DBG("%s - success\n", __FUNCTION__);
 }
@@ -202,14 +202,15 @@ try_pbo_upload(struct intel_context *intel,
    GLuint src_offset, src_stride;
    GLuint dst_offset, dst_stride;
 
-   if (!pbo ||
+   if (!_mesa_is_bufferobj(unpack->BufferObj) ||
        intel->ctx._ImageTransferState ||
        unpack->SkipPixels || unpack->SkipRows) {
-      _mesa_printf("%s: failure 1\n", __FUNCTION__);
+      DBG("%s: failure 1\n", __FUNCTION__);
       return GL_FALSE;
    }
 
-   src_offset = (GLuint) pixels;
+   /* note: potential 64-bit ptr to 32-bit int cast */
+   src_offset = (GLuint) (unsigned long) pixels;
 
    if (unpack->RowLength > 0)
       src_stride = unpack->RowLength;
@@ -231,12 +232,15 @@ try_pbo_upload(struct intel_context *intel,
 					       INTEL_WRITE_FULL);
 
 
-      intelEmitCopyBlit(intel,
-                        intelImage->mt->cpp,
-                        src_stride, src_buffer, src_offset, GL_FALSE,
-                        dst_stride, dst_buffer, dst_offset, GL_FALSE,
-                        0, 0, 0, 0, width, height,
-			GL_COPY);
+      if (!intelEmitCopyBlit(intel,
+			     intelImage->mt->cpp,
+			     src_stride, src_buffer, src_offset, GL_FALSE,
+			     dst_stride, dst_buffer, dst_offset, GL_FALSE,
+			     0, 0, 0, 0, width, height,
+			     GL_COPY)) {
+	 UNLOCK_HARDWARE(intel);
+	 return GL_FALSE;
+      }
    }
    UNLOCK_HARDWARE(intel);
 
@@ -244,7 +248,6 @@ try_pbo_upload(struct intel_context *intel,
 }
 
 
-
 static GLboolean
 try_pbo_zcopy(struct intel_context *intel,
               struct intel_texture_image *intelImage,
@@ -257,14 +260,15 @@ try_pbo_zcopy(struct intel_context *intel,
    GLuint src_offset, src_stride;
    GLuint dst_offset, dst_stride;
 
-   if (!pbo ||
+   if (!_mesa_is_bufferobj(unpack->BufferObj) ||
        intel->ctx._ImageTransferState ||
        unpack->SkipPixels || unpack->SkipRows) {
-      _mesa_printf("%s: failure 1\n", __FUNCTION__);
+      DBG("%s: failure 1\n", __FUNCTION__);
       return GL_FALSE;
    }
 
-   src_offset = (GLuint) pixels;
+   /* note: potential 64-bit ptr to 32-bit int cast */
+   src_offset = (GLuint) (unsigned long) pixels;
 
    if (unpack->RowLength > 0)
       src_stride = unpack->RowLength;
@@ -278,7 +282,7 @@ try_pbo_zcopy(struct intel_context *intel,
    dst_stride = intelImage->mt->pitch;
 
    if (src_stride != dst_stride || dst_offset != 0 || src_offset != 0) {
-      _mesa_printf("%s: failure 2\n", __FUNCTION__);
+      DBG("%s: failure 2\n", __FUNCTION__);
       return GL_FALSE;
    }
 
@@ -288,10 +292,6 @@ try_pbo_zcopy(struct intel_context *intel,
 }
 
 
-
-
-
-
 static void
 intelTexImage(GLcontext * ctx,
               GLint dims,
@@ -302,7 +302,8 @@ intelTexImage(GLcontext * ctx,
               GLenum format, GLenum type, const void *pixels,
               const struct gl_pixelstore_attrib *unpack,
               struct gl_texture_object *texObj,
-              struct gl_texture_image *texImage, GLsizei imageSize, int compressed)
+              struct gl_texture_image *texImage, GLsizei imageSize,
+              GLboolean compressed)
 {
    struct intel_context *intel = intel_context(ctx);
    struct intel_texture_object *intelObj = intel_texture_object(texObj);
@@ -310,8 +311,7 @@ intelTexImage(GLcontext * ctx,
    GLint postConvWidth = width;
    GLint postConvHeight = height;
    GLint texelBytes, sizeInBytes;
-   GLuint dstRowStride, srcRowStride = texImage->RowStride;
-
+   GLuint dstRowStride = 0, srcRowStride = texImage->RowStride;
 
    DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__,
        _mesa_lookup_enum_by_nr(target), level, width, height, depth, border);
@@ -383,7 +383,7 @@ intelTexImage(GLcontext * ctx,
    }
 
    if (!intelObj->mt) {
-      guess_and_alloc_mipmap_tree(intel, intelObj, intelImage);
+      guess_and_alloc_mipmap_tree(intel, intelObj, intelImage, pixels == NULL);
       if (!intelObj->mt) {
 	 DBG("guess_and_alloc_mipmap_tree: failed\n");
       }
@@ -409,11 +409,13 @@ intelTexImage(GLcontext * ctx,
        * a miptree, so create one just for our level and store it in the image.
        * It'll get moved into the object miptree at validate time.
        */
-      intelImage->mt = intel_miptree_create(intel, target, internalFormat,
+      intelImage->mt = intel_miptree_create(intel, target,
+					    intelImage->base.TexFormat->BaseFormat,
+					    internalFormat,
 					    level, level,
 					    width, height, depth,
 					    intelImage->base.TexFormat->TexelBytes,
-					    comp_byte);
+					    comp_byte, pixels == NULL);
 
    }
 
@@ -421,7 +423,7 @@ intelTexImage(GLcontext * ctx,
     */
    if (dims <= 2 &&
        intelImage->mt &&
-       intel_buffer_object(unpack->BufferObj) &&
+       _mesa_is_bufferobj(unpack->BufferObj) &&
        check_pbo_format(internalFormat, format,
                         type, intelImage->base.TexFormat)) {
 
@@ -459,8 +461,6 @@ intelTexImage(GLcontext * ctx,
       DBG("pbo upload failed\n");
    }
 
-
-
    /* intelCopyTexImage calls this function with pixels == NULL, with
     * the expectation that the mipmap tree will be set up but nothing
     * more will be done.  This is where those calls return:
@@ -478,12 +478,13 @@ intelTexImage(GLcontext * ctx,
    LOCK_HARDWARE(intel);
 
    if (intelImage->mt) {
-      texImage->Data = intel_miptree_image_map(intel,
-                                               intelImage->mt,
-                                               intelImage->face,
-                                               intelImage->level,
-                                               &dstRowStride,
-                                               intelImage->base.ImageOffsets);
+      if (pixels != NULL)
+         texImage->Data = intel_miptree_image_map(intel,
+                                                  intelImage->mt,
+                                                  intelImage->face,
+                                                  intelImage->level,
+                                                  &dstRowStride,
+                                                  intelImage->base.ImageOffsets);
       texImage->RowStride = dstRowStride / intelImage->mt->cpp;
    }
    else {
@@ -503,8 +504,9 @@ intelTexImage(GLcontext * ctx,
    }
 
    DBG("Upload image %dx%dx%d row_len %d "
-       "pitch %d\n",
-       width, height, depth, width * texelBytes, dstRowStride);
+       "pitch %d pixels %d compressed %d\n",
+       width, height, depth, width * texelBytes, dstRowStride,
+       pixels ? 1 : 0, compressed);
 
    /* Copy data.  Would like to know when it's ok for us to eg. use
     * the blitter to copy.  Or, use the hardware to do the format
@@ -517,7 +519,7 @@ intelTexImage(GLcontext * ctx,
 	       _mesa_copy_rect(texImage->Data, dst->cpp, dst->pitch,
 			       0, 0,
 			       intelImage->mt->level[level].width,
-			       intelImage->mt->level[level].height/4,
+			       (intelImage->mt->level[level].height+3)/4,
 			       pixels,
 			       srcRowStride,
 			       0, 0);
@@ -535,22 +537,24 @@ intelTexImage(GLcontext * ctx,
        }
    }
 
-   /* GL_SGIS_generate_mipmap */
-   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
-      intel_generate_mipmap(ctx, target, texObj);
-   }
-
    _mesa_unmap_teximage_pbo(ctx, unpack);
 
    if (intelImage->mt) {
-      intel_miptree_image_unmap(intel, intelImage->mt);
+      if (pixels != NULL)
+         intel_miptree_image_unmap(intel, intelImage->mt);
       texImage->Data = NULL;
    }
 
    UNLOCK_HARDWARE(intel);
+
+   /* GL_SGIS_generate_mipmap */
+   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+      intel_generate_mipmap(ctx, target, texObj);
+   }
 }
 
-void
+
+static void
 intelTexImage3D(GLcontext * ctx,
                 GLenum target, GLint level,
                 GLint internalFormat,
@@ -563,11 +567,11 @@ intelTexImage3D(GLcontext * ctx,
 {
    intelTexImage(ctx, 3, target, level,
                  internalFormat, width, height, depth, border,
-                 format, type, pixels, unpack, texObj, texImage, 0, 0);
+                 format, type, pixels, unpack, texObj, texImage, 0, GL_FALSE);
 }
 
 
-void
+static void
 intelTexImage2D(GLcontext * ctx,
                 GLenum target, GLint level,
                 GLint internalFormat,
@@ -579,10 +583,11 @@ intelTexImage2D(GLcontext * ctx,
 {
    intelTexImage(ctx, 2, target, level,
                  internalFormat, width, height, 1, border,
-                 format, type, pixels, unpack, texObj, texImage, 0, 0);
+                 format, type, pixels, unpack, texObj, texImage, 0, GL_FALSE);
 }
 
-void
+
+static void
 intelTexImage1D(GLcontext * ctx,
                 GLenum target, GLint level,
                 GLint internalFormat,
@@ -594,21 +599,24 @@ intelTexImage1D(GLcontext * ctx,
 {
    intelTexImage(ctx, 1, target, level,
                  internalFormat, width, 1, 1, border,
-                 format, type, pixels, unpack, texObj, texImage, 0, 0);
+                 format, type, pixels, unpack, texObj, texImage, 0, GL_FALSE);
 }
 
-void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level,
-				GLint internalFormat,
-				GLint width, GLint height, GLint border,
-				GLsizei imageSize, const GLvoid *data,
-				struct gl_texture_object *texObj,
-				struct gl_texture_image *texImage )
+
+static void
+intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level,
+                           GLint internalFormat,
+                           GLint width, GLint height, GLint border,
+                           GLsizei imageSize, const GLvoid *data,
+                           struct gl_texture_object *texObj,
+                           struct gl_texture_image *texImage )
 {
    intelTexImage(ctx, 2, target, level,
 		 internalFormat, width, height, 1, border,
-		 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, 1);
+		 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, GL_TRUE);
 }
 
+
 /**
  * Need to map texture image into memory before copying image data,
  * then unmap it.
@@ -617,11 +625,17 @@ static void
 intel_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)
 {
    struct intel_context *intel = intel_context(ctx);
    struct intel_texture_image *intelImage = intel_texture_image(texImage);
 
+   /* If we're reading from a texture that has been rendered to, need to
+    * make sure rendering is complete.
+    * We could probably predicate this on texObj->_RenderToTexture
+    */
+   intelFlush(ctx);
+
    /* Map */
    if (intelImage->mt) {
       /* Image is stored in hardware format in a buffer managed by the
@@ -665,28 +679,29 @@ intel_get_tex_image(GLcontext * ctx, GLenum target, GLint level,
    }
 }
 
-void
+
+static void
 intelGetTexImage(GLcontext * ctx, GLenum target, GLint level,
                  GLenum format, GLenum type, GLvoid * pixels,
                  struct gl_texture_object *texObj,
                  struct gl_texture_image *texImage)
 {
    intel_get_tex_image(ctx, target, level, format, type, pixels,
-		       texObj, texImage, 0);
-
-
+		       texObj, texImage, GL_FALSE);
 }
 
-void
+
+static void
 intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level,
 			   GLvoid *pixels,
 			   struct gl_texture_object *texObj,
 			   struct gl_texture_image *texImage)
 {
    intel_get_tex_image(ctx, target, level, 0, 0, pixels,
-		       texObj, texImage, 1);
+		       texObj, texImage, GL_TRUE);
 }
 
+
 void
 intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname,
 		  unsigned long long offset, GLint depth, GLuint pitch)
@@ -710,7 +725,9 @@ intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname,
 }
 
 void
-intelSetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv)
+intelSetTexBuffer2(__DRIcontext *pDRICtx, GLint target,
+		   GLint glx_texture_format,
+		   __DRIdrawable *dPriv)
 {
    struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
    struct intel_context *intel = pDRICtx->driverPrivate;
@@ -741,7 +758,10 @@ intelSetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv)
 
    type = GL_BGRA;
    format = GL_UNSIGNED_BYTE;
-   internalFormat = (rb->region->cpp == 3 ? 3 : 4);
+   if (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT)
+      internalFormat = GL_RGB;
+   else
+      internalFormat = GL_RGBA;
 
    mt = intel_miptree_create_for_region(intel, target,
 					internalFormat,
@@ -751,16 +771,21 @@ intelSetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv)
 
    _mesa_lock_texture(&intel->ctx, texObj);
 
+   texImage = _mesa_get_tex_image(&intel->ctx, texObj, target, level);
+   intelImage = intel_texture_image(texImage);
+
+   if (intelImage->mt) {
+      intel_miptree_release(intel, &intelImage->mt);
+      assert(!texImage->Data);
+   }
    if (intelObj->mt)
       intel_miptree_release(intel, &intelObj->mt);
 
    intelObj->mt = mt;
-   texImage = _mesa_get_tex_image(&intel->ctx, texObj, target, level);
    _mesa_init_teximage_fields(&intel->ctx, target, texImage,
 			      rb->region->width, rb->region->height, 1,
 			      0, internalFormat);
 
-   intelImage = intel_texture_image(texImage);
    intelImage->face = target_to_face(target);
    intelImage->level = level;
    texImage->TexFormat = intelChooseTextureFormat(&intel->ctx, internalFormat,
@@ -776,3 +801,25 @@ intelSetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv)
 
    _mesa_unlock_texture(&intel->ctx, texObj);
 }
+
+void
+intelSetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv)
+{
+   /* The old interface didn't have the format argument, so copy our
+    * implementation's behavior at the time.
+    */
+   intelSetTexBuffer2(pDRICtx, target, GLX_TEXTURE_FORMAT_RGBA_EXT, dPriv);
+}
+
+
+void
+intelInitTextureImageFuncs(struct dd_function_table *functions)
+{
+   functions->TexImage1D = intelTexImage1D;
+   functions->TexImage2D = intelTexImage2D;
+   functions->TexImage3D = intelTexImage3D;
+   functions->GetTexImage = intelGetTexImage;
+
+   functions->CompressedTexImage2D = intelCompressedTexImage2D;
+   functions->GetCompressedTexImage = intelGetCompressedTexImage;
+}
diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.c b/src/mesa/drivers/dri/intel/intel_tex_layout.c
index e6f9a41779..7d69ea4484 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_layout.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_layout.c
@@ -35,26 +35,39 @@
 #include "intel_context.h"
 #include "main/macros.h"
 
-GLuint intel_compressed_alignment(GLenum internalFormat)
+void intel_get_texture_alignment_unit(GLenum internalFormat, GLuint *w, GLuint *h)
 {
-    GLuint alignment = 4;
-
     switch (internalFormat) {
     case GL_COMPRESSED_RGB_FXT1_3DFX:
     case GL_COMPRESSED_RGBA_FXT1_3DFX:
-        alignment = 8;
+        *w = 8;
+        *h = 4;
+        break;
+
+    case GL_RGB_S3TC:
+    case GL_RGB4_S3TC:
+    case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
+    case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT:
+    case GL_RGBA_S3TC:
+    case GL_RGBA4_S3TC:
+    case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT:
+    case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT:
+        *w = 4;
+        *h = 4;
         break;
 
     default:
+        *w = 4;
+        *h = 2;
         break;
     }
-
-    return alignment;
 }
 
-void i945_miptree_layout_2d( struct intel_context *intel, struct intel_mipmap_tree *mt )
+void i945_miptree_layout_2d( struct intel_context *intel,
+			     struct intel_mipmap_tree *mt,
+			     uint32_t tiling )
 {
-   GLint align_h = 2, align_w = 4;
+   GLuint align_h = 2, align_w = 4;
    GLuint level;
    GLuint x = 0;
    GLuint y = 0;
@@ -62,9 +75,9 @@ void i945_miptree_layout_2d( struct intel_context *intel, struct intel_mipmap_tr
    GLuint height = mt->height0;
 
    mt->pitch = mt->width0;
+   intel_get_texture_alignment_unit(mt->internal_format, &align_w, &align_h);
 
    if (mt->compressed) {
-       align_w = intel_compressed_alignment(mt->internal_format);
        mt->pitch = ALIGN(mt->width0, align_w);
    }
 
@@ -92,7 +105,7 @@ void i945_miptree_layout_2d( struct intel_context *intel, struct intel_mipmap_tr
    /* Pitch must be a whole number of dwords, even though we
     * express it in texels.
     */
-   mt->pitch = intel_miptree_pitch_align (intel, mt, mt->pitch);
+   mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, mt->pitch);
    mt->total_height = 0;
 
    for ( level = mt->first_level ; level <= mt->last_level ; level++ ) {
diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.h b/src/mesa/drivers/dri/intel/intel_tex_layout.h
index dbc90e6f9b..c9de9b5678 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_layout.h
+++ b/src/mesa/drivers/dri/intel/intel_tex_layout.h
@@ -38,5 +38,7 @@ static GLuint minify( GLuint d )
    return MAX2(1, d>>1);
 }
 
-extern void i945_miptree_layout_2d( struct intel_context *intel, struct intel_mipmap_tree *mt );
-extern GLuint intel_compressed_alignment(GLenum);
+extern void i945_miptree_layout_2d(struct intel_context *intel,
+				   struct intel_mipmap_tree *mt,
+				   uint32_t tiling);
+extern void intel_get_texture_alignment_unit(GLenum, GLuint *, GLuint *);
diff --git a/src/mesa/drivers/dri/intel/intel_tex_subimage.c b/src/mesa/drivers/dri/intel/intel_tex_subimage.c
index f86de56897..89037073f8 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_subimage.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_subimage.c
@@ -44,10 +44,12 @@ intelTexSubimage(GLcontext * ctx,
                  GLenum target, GLint level,
                  GLint xoffset, GLint yoffset, GLint zoffset,
                  GLint width, GLint height, GLint depth,
+                 GLsizei imageSize,
                  GLenum format, GLenum type, const void *pixels,
                  const struct gl_pixelstore_attrib *packing,
                  struct gl_texture_object *texObj,
-                 struct gl_texture_image *texImage)
+                 struct gl_texture_image *texImage,
+                 GLboolean compressed)
 {
    struct intel_context *intel = intel_context(ctx);
    struct intel_texture_image *intelImage = intel_texture_image(texImage);
@@ -59,9 +61,14 @@ intelTexSubimage(GLcontext * ctx,
 
    intelFlush(ctx);
 
-   pixels =
-      _mesa_validate_pbo_teximage(ctx, dims, width, height, depth, format,
-                                  type, pixels, packing, "glTexSubImage2D");
+   if (compressed)
+      pixels = _mesa_validate_pbo_compressed_teximage(ctx, imageSize,
+                                                      pixels, packing,
+                                                      "glCompressedTexImage");
+   else
+      pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, depth,
+                                           format, type, pixels, packing,
+                                           "glTexSubImage");
    if (!pixels)
       return;
 
@@ -90,20 +97,28 @@ intelTexSubimage(GLcontext * ctx,
 
    assert(dstRowStride);
 
-   if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat,
-                                        texImage->TexFormat,
-                                        texImage->Data,
-                                        xoffset, yoffset, zoffset,
-                                        dstRowStride,
-                                        texImage->ImageOffsets,
-                                        width, height, depth,
-                                        format, type, pixels, packing)) {
-      _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage");
+   if (compressed) {
+      if (intelImage->mt) {
+         struct intel_region *dst = intelImage->mt->region;
+         
+         _mesa_copy_rect(texImage->Data, dst->cpp, dst->pitch,
+                         xoffset, yoffset / 4,
+                         (width + 3)  & ~3, (height + 3) / 4,
+                         pixels, (width + 3) & ~3, 0, 0);
+      } else
+        memcpy(texImage->Data, pixels, imageSize);
    }
-
-   /* GL_SGIS_generate_mipmap */
-   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
-      intel_generate_mipmap(ctx, target, texObj);
+   else {
+      if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat,
+                                           texImage->TexFormat,
+                                           texImage->Data,
+                                           xoffset, yoffset, zoffset,
+                                           dstRowStride,
+                                           texImage->ImageOffsets,
+                                           width, height, depth,
+                                           format, type, pixels, packing)) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage");
+      }
    }
 
    _mesa_unmap_teximage_pbo(ctx, packing);
@@ -114,13 +129,15 @@ intelTexSubimage(GLcontext * ctx,
    }
 
    UNLOCK_HARDWARE(intel);
-}
-
-
 
+   /* GL_SGIS_generate_mipmap */
+   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+      intel_generate_mipmap(ctx, target, texObj);
+   }
+}
 
 
-void
+static void
 intelTexSubImage3D(GLcontext * ctx,
                    GLenum target,
                    GLint level,
@@ -132,18 +149,15 @@ intelTexSubImage3D(GLcontext * ctx,
                    struct gl_texture_object *texObj,
                    struct gl_texture_image *texImage)
 {
-
    intelTexSubimage(ctx, 3,
                     target, level,
                     xoffset, yoffset, zoffset,
-                    width, height, depth,
-                    format, type, pixels, packing, texObj, texImage);
-
+                    width, height, depth, 0,
+                    format, type, pixels, packing, texObj, texImage, GL_FALSE);
 }
 
 
-
-void
+static void
 intelTexSubImage2D(GLcontext * ctx,
                    GLenum target,
                    GLint level,
@@ -155,17 +169,15 @@ intelTexSubImage2D(GLcontext * ctx,
                    struct gl_texture_object *texObj,
                    struct gl_texture_image *texImage)
 {
-
    intelTexSubimage(ctx, 2,
                     target, level,
                     xoffset, yoffset, 0,
-                    width, height, 1,
-                    format, type, pixels, packing, texObj, texImage);
-
+                    width, height, 1, 0,
+                    format, type, pixels, packing, texObj, texImage, GL_FALSE);
 }
 
 
-void
+static void
 intelTexSubImage1D(GLcontext * ctx,
                    GLenum target,
                    GLint level,
@@ -180,12 +192,11 @@ intelTexSubImage1D(GLcontext * ctx,
    intelTexSubimage(ctx, 1,
                     target, level,
                     xoffset, 0, 0,
-                    width, 1, 1,
-                    format, type, pixels, packing, texObj, texImage);
-
+                    width, 1, 1, 0,
+                    format, type, pixels, packing, texObj, texImage, GL_FALSE);
 }
 
-void
+static void
 intelCompressedTexSubImage2D(GLcontext * ctx,
 			     GLenum target,
 			     GLint level,
@@ -196,6 +207,20 @@ intelCompressedTexSubImage2D(GLcontext * ctx,
 			     struct gl_texture_object *texObj,
 			     struct gl_texture_image *texImage)
 {
-   fprintf(stderr, "stubbed CompressedTexSubImage2D: %dx%d@%dx%d\n",
-	   width, height, xoffset, yoffset);
+   intelTexSubimage(ctx, 2,
+                    target, level,
+                    xoffset, yoffset, 0,
+                    width, height, 1, imageSize,
+                    format, 0, pixels, &ctx->Unpack, texObj, texImage, GL_TRUE);
+}
+
+
+
+void
+intelInitTextureSubImageFuncs(struct dd_function_table *functions)
+{
+   functions->TexSubImage1D = intelTexSubImage1D;
+   functions->TexSubImage2D = intelTexSubImage2D;
+   functions->TexSubImage3D = intelTexSubImage3D;
+   functions->CompressedTexSubImage2D = intelCompressedTexSubImage2D;
 }
diff --git a/src/mesa/drivers/dri/intel/intel_tex_validate.c b/src/mesa/drivers/dri/intel/intel_tex_validate.c
index 820683d42e..a284d5475f 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_validate.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_validate.c
@@ -199,6 +199,7 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit)
    if (!intelObj->mt) {
       intelObj->mt = intel_miptree_create(intel,
                                           intelObj->base.Target,
+                                          firstImage->base._BaseFormat,
                                           firstImage->base.InternalFormat,
                                           intelObj->firstLevel,
                                           intelObj->lastLevel,
@@ -206,7 +207,8 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit)
                                           firstImage->base.Height,
                                           firstImage->base.Depth,
                                           cpp,
-                                          comp_byte);
+                                          comp_byte,
+					  GL_TRUE);
    }
 
    /* Pull in any images not in the object's tree:
@@ -240,7 +242,7 @@ intel_tex_map_level_images(struct intel_context *intel,
       struct intel_texture_image *intelImage =
 	 intel_texture_image(intelObj->base.Image[face][level]);
 
-      if (intelImage->mt) {
+      if (intelImage && intelImage->mt) {
 	 intelImage->base.Data =
 	    intel_miptree_image_map(intel,
 				    intelImage->mt,
@@ -267,7 +269,7 @@ intel_tex_unmap_level_images(struct intel_context *intel,
       struct intel_texture_image *intelImage =
 	 intel_texture_image(intelObj->base.Image[face][level]);
 
-      if (intelImage->mt) {
+      if (intelImage && intelImage->mt) {
 	 intel_miptree_image_unmap(intel, intelImage->mt);
 	 intelImage->base.Data = NULL;
       }