Introduce fixup/relocation lists for dma buffers.

50 files changed, 2587 insertions, 2761 deletions

diff --git a/src/mesa/drivers/dri/i915/Makefile b/src/mesa/drivers/dri/i915/Makefile
index c46d9c2b630..d1dd2d6a513 100644
--- a/src/mesa/drivers/dri/i915/Makefile
+++ b/src/mesa/drivers/dri/i915/Makefile
@@ -7,6 +7,7 @@ LIBNAME = i915_dri.so
 DRIVER_SOURCES = \
 	bufmgr_fake.c \
 	intel_regions.c \
+	intel_batchbuffer.c \
 	intel_mipmap_tree.c \
 	i915_tex_layout.c \
 	intel_tex_image.c \
@@ -15,6 +16,11 @@ DRIVER_SOURCES = \
 	intel_tex_validate.c \
 	intel_tex_format.c \
 	intel_tex.c \
+	intel_pixel.c \
+	intel_pixel_copy.c \
+	intel_pixel_read.c \
+	intel_pixel_draw.c \
+	intel_buffers.c \
 	intel_blit.c \
 	i915_tex.c \
 	i915_texstate.c \
@@ -26,10 +32,8 @@ DRIVER_SOURCES = \
 	i915_state.c \
 	i915_texprog.c \
 	i915_vtbl.c \
-	intel_batchbuffer.c \
 	intel_context.c \
 	intel_ioctl.c \
-	intel_pixel.c \
 	intel_screen.c \
 	intel_span.c \
 	intel_state.c \
diff --git a/src/mesa/drivers/dri/i915/bufmgr.h b/src/mesa/drivers/dri/i915/bufmgr.h
index 64affa09697..bf740274255 100644
--- a/src/mesa/drivers/dri/i915/bufmgr.h
+++ b/src/mesa/drivers/dri/i915/bufmgr.h
@@ -10,9 +10,23 @@
 /* The buffer manager context.  Opaque.
  */
 struct bufmgr;
-struct bm_buffer_list;
 
-struct bufmgr *bm_fake_intel_Attach( struct intel_context *intel );
+#define BM_LIST_MAX 32
+
+/* List of buffers to validate.  Probably better managed by the client:
+ */
+struct bm_buffer_list {
+   struct {
+      unsigned buffer;
+      unsigned *offset_return;
+      unsigned *memtype_return;
+   } elem[BM_LIST_MAX];
+
+   unsigned nr;
+};
+
+
+struct bufmgr *bm_fake_intel_Attach( struct intel_context *intel ); 
 
 /* struct bufmgr *bmCreate( ... ); */
 /* struct bufmgr *bmAttach( ... ); */
@@ -58,7 +72,8 @@ int bmInitPool( struct bufmgr *,
 #define BM_NO_EVICT    0x40
 #define BM_NO_MOVE     0x80	/* not yet used */
 #define BM_NO_ALLOC    0x100	/* legacy "fixed" buffers only */
-
+#define BM_CLIENT      0x200	/* for map - pointer will be accessed
+				 * without dri lock */
 
 #define BM_MEM_MASK (BM_MEM_LOCAL|BM_MEM_AGP|BM_MEM_VRAM)
 
diff --git a/src/mesa/drivers/dri/i915/bufmgr_fake.c b/src/mesa/drivers/dri/i915/bufmgr_fake.c
index 5baebc02cb6..9e6fb36fd47 100644
--- a/src/mesa/drivers/dri/i915/bufmgr_fake.c
+++ b/src/mesa/drivers/dri/i915/bufmgr_fake.c
@@ -21,9 +21,6 @@ struct _mesa_HashTable;
 
 static int delayed_free( struct bufmgr *bm );
 
-/* Maximum number of buffers to pass to bmValidateBufferList:
- */
-#define BM_LIST_MAX 32
 #define BM_POOL_MAX 8
 
 
@@ -72,17 +69,6 @@ struct bufmgr {
 };
 
 
-/* List of buffers to validate: 
- */
-struct bm_buffer_list {
-   struct {
-      unsigned buffer;
-      unsigned *offset_return;
-      unsigned *memtype_return;
-   } elem[BM_LIST_MAX];
-
-   unsigned nr;
-};
 
 
 
@@ -145,20 +131,22 @@ static struct block *alloc_block( struct bufmgr *bm,
 {
    GLuint i;
 
-   for (i = 0; i < bm->nr_pools; i++) {
-      struct block *block;
-
-      if (bm->pool[i].flags & BM_NO_ALLOC)
-	 continue;
-
-      if ((bm->pool[i].flags & flags & BM_MEM_MASK) == 0)
-	 continue;
-      
-      block = alloc_from_pool(bm, i, size, align);
-      if (block)
-	 return block;
+   if (!(flags & BM_CLIENT)) {
+      for (i = 0; i < bm->nr_pools; i++) {
+	 struct block *block;
+	 
+	 if (bm->pool[i].flags & BM_NO_ALLOC)
+	    continue;
+	 
+	 if ((bm->pool[i].flags & flags & BM_MEM_MASK) == 0)
+	    continue;
+	 
+	 block = alloc_from_pool(bm, i, size, align);
+	 if (block)
+	    return block;
+      }
    }
-   
+
    if (flags & BM_MEM_LOCAL)
       return alloc_local(size);
 
@@ -166,14 +154,15 @@ static struct block *alloc_block( struct bufmgr *bm,
 }
 
 static int bmAllocMem( struct bufmgr *bm,
-		       struct buffer *buf )	
+		       struct buffer *buf,
+		       GLuint flags )	
 {
    delayed_free(bm);
 
    buf->block = alloc_block(bm, 
 			    buf->size,
 			    buf->alignment, 
-			    buf->flags);
+			    buf->flags | flags);
 
    if (buf->block)
       buf->block->buf = buf;
@@ -191,30 +180,43 @@ static int bmAllocMem( struct bufmgr *bm,
  */
 static void free_block( struct bufmgr *bm, struct block *block )
 {
+   DBG("free block %p\n", block);
+
    if (!block) 
       return;
 
-   remove_from_list(block);
+   DBG("free block (mem: %d, sz %d) from buf %d\n",
+       block->mem_type,
+       block->buf->size,
+       block->buf->id);
 
    switch (block->mem_type) {
    case BM_MEM_AGP:
    case BM_MEM_VRAM:
+      remove_from_list(block);
+
+      DBG("    - offset %x\n", block->mem->ofs);
+
       if (bmTestFence(bm, block->fence)) {
+	 DBG("    - free immediately\n");
          mmFreeMem(block->mem);
          free(block);
       }
       else {
+	 DBG("    - place on delayed_free list\n");
 	 block->buf = NULL;
          insert_at_tail(&block->pool->freed, block);
       }
       break;
 
    case BM_MEM_LOCAL:
+      DBG("    - free local memory\n");
       ALIGN_FREE(block->virtual);
       free(block);
       break;
 
    default:
+      DBG("    - unknown memory type\n");
       free(block);
       break;
    }
@@ -259,8 +261,8 @@ static int move_buffers( struct bufmgr *bm,
     */ 
    for (i = 0; i < nr; i++) {    
       if (!buffers[i]->block) {
-/* 	 if (flags & BM_NO_ALLOC) */
-/* 	    goto cleanup; */
+ 	 if (flags & BM_NO_ALLOC)
+ 	    goto cleanup;
 
 	 newMem[i] = alloc_block(bm, 
 				 buffers[i]->size,
@@ -276,8 +278,8 @@ static int move_buffers( struct bufmgr *bm,
 	    goto cleanup;
 
 	 /* Known issue: this assert will get hit on texture swapping.
-	  * There's not much to do about that at this stage - it's
-	  * tbd.
+	  * There's not much to do about that at this stage - it's a
+	  * todo item.
 	  */
  	 assert(!buffers[i]->mapped);
 
@@ -304,7 +306,7 @@ static int move_buffers( struct bufmgr *bm,
 	     * mechanisms in final version.  Memcpy (or sse_memcpy) is
 	     * probably pretty good for local->agp uploads.
 	     */
-	    _mesa_printf("* %d\n", buffers[i]->size);
+	    DBG("memcpy %d bytes\n", buffers[i]->size);
 	    memcpy(newMem[i]->virtual,
 		   buffers[i]->block->virtual, 
 		   buffers[i]->size);
@@ -535,25 +537,9 @@ unsigned bmBufferStatic(struct bufmgr *bm,
 }
 
 
-#if 0
-/* How wise/useful is this?
- */
-void bmBufferSetParams( struct bufmgr *bm,
-			unsigned buffer,
-			unsigned flags,
-			unsigned alignment )
-{
-   struct buffer *buf = (struct buffer *)_mesa_HashLookup( bm->hash, buffer );
-   assert(!buf->block);
-   buf->flags = flags;
-   buf->alignment = alignment;
-}
-#endif
-
-
 
-/* If buffer size changes, create new buffer in local memory.
- * Otherwise update in place.
+/* If buffer size changes, free and reallocate.  Otherwise update in
+ * place.
  */
 void bmBufferData(struct bufmgr *bm, 
 		  unsigned buffer, 
@@ -579,7 +565,7 @@ void bmBufferData(struct bufmgr *bm,
    buf->size = size;
 
    if (data != NULL) {      
-      bmAllocMem(bm, buf);
+      bmAllocMem(bm, buf, buf->flags | flags);
       memcpy(buf->block->virtual, data, size);
    }
 }
@@ -597,7 +583,7 @@ void bmBufferSubData(struct bufmgr *bm,
    DBG("bmBufferSubdata %d offset 0x%x sz 0x%x\n", buffer, offset, size);
 
    if (buf->block == 0)
-      bmAllocMem(bm, buf);
+      bmAllocMem(bm, buf, buf->flags);
 
    if (buf->block->mem_type != BM_MEM_LOCAL)
       bmFinishFence(bm, buf->block->fence);
@@ -611,7 +597,7 @@ void bmBufferSubData(struct bufmgr *bm,
  */
 void *bmMapBuffer( struct bufmgr *bm,
 		   unsigned buffer, 
-		   unsigned access )
+		   unsigned flags )
 {
    struct buffer *buf = (struct buffer *)_mesa_HashLookup( bm->hash, buffer );
 
@@ -620,14 +606,14 @@ void *bmMapBuffer( struct bufmgr *bm,
    if (buf->mapped)
       return NULL;
 
-   buf->mapped = 1;
-
    if (buf->block == 0)
-      bmAllocMem(bm, buf);
+      bmAllocMem(bm, buf, flags);
 
    if (buf->block == 0)
       return NULL;
 
+   buf->mapped = 1;
+
    /* Finish any outstanding operations to/from this memory:
     */
    if (buf->block->mem_type != BM_MEM_LOCAL) 
@@ -731,6 +717,7 @@ int bmValidateBufferList( struct bufmgr *bm,
       if (!delayed_free(bm) &&
 	  !evict_lru(bm, flags))
 	 return 0;
+      _mesa_printf("couldn't allocate sufficient texture memory\n");
       exit(1);
    }
 
@@ -739,6 +726,8 @@ int bmValidateBufferList( struct bufmgr *bm,
       DBG("%d: buf %d ofs 0x%x\n",
 		   i, bufs[i]->id, bufs[i]->block->mem->ofs);
 
+      assert(!bufs[i]->mapped);
+
       if (list->elem[i].offset_return)
 	 list->elem[i].offset_return[0] = bufs[i]->block->mem->ofs;
       
@@ -791,7 +780,6 @@ unsigned bmFenceBufferList( struct bufmgr *bm, struct bm_buffer_list *list )
  */
 unsigned bmSetFence( struct bufmgr *bm )
 {
-   assert(bm->intel->batch.space == bm->intel->batch.size);
    assert(bm->intel->locked);
 
    return intelEmitIrqLocked( bm->intel );
diff --git a/src/mesa/drivers/dri/i915/i830_context.c b/src/mesa/drivers/dri/i915/i830_context.c
index 3979f531a8b..56d03864ad4 100644
--- a/src/mesa/drivers/dri/i915/i830_context.c
+++ b/src/mesa/drivers/dri/i915/i830_context.c
@@ -59,7 +59,7 @@ GLboolean i830CreateContext( const __GLcontextModes *mesaVis,
 {
    struct dd_function_table functions;
    i830ContextPtr i830 = (i830ContextPtr) CALLOC_STRUCT(i830_context);
-   intelContextPtr intel = &i830->intel;
+   struct intel_context *intel = &i830->intel;
    GLcontext *ctx = &intel->ctx;
    if (!i830) return GL_FALSE;
 
diff --git a/src/mesa/drivers/dri/i915/i830_context.h b/src/mesa/drivers/dri/i915/i830_context.h
index 62e20d0a02e..6ab6ccb68ae 100644
--- a/src/mesa/drivers/dri/i915/i830_context.h
+++ b/src/mesa/drivers/dri/i915/i830_context.h
@@ -48,10 +48,8 @@
  */
 #define I830_DESTREG_CBUFADDR0 0
 #define I830_DESTREG_CBUFADDR1 1
-#define I830_DESTREG_CBUFADDR2 2
 #define I830_DESTREG_DBUFADDR0 3
 #define I830_DESTREG_DBUFADDR1 4
-#define I830_DESTREG_DBUFADDR2 5
 #define I830_DESTREG_DV0 6
 #define I830_DESTREG_DV1 7
 #define I830_DESTREG_SENABLE 8
@@ -160,7 +158,7 @@ i830CreateContext( const __GLcontextModes *mesaVis,
 /* i830_tex.c, i830_texstate.c
  */
 extern void 
-i830UpdateTextureState( intelContextPtr intel );
+i830UpdateTextureState( struct intel_context *intel );
 
 extern void 
 i830InitTextureFuncs( struct dd_function_table *functions );
@@ -206,7 +204,7 @@ i830TryTextureDrawPixels( GLcontext *ctx,
 			  const GLvoid *pixels );
 
 extern void 
-i830ClearWithTris( intelContextPtr intel, GLbitfield mask,
+i830ClearWithTris( struct intel_context *intel, GLbitfield mask,
 		   GLboolean all, GLint cx, GLint cy, GLint cw, GLint ch);
 
 
diff --git a/src/mesa/drivers/dri/i915/i830_metaops.c b/src/mesa/drivers/dri/i915/i830_metaops.c
index 8984b6c6ecd..635a73041a4 100644
--- a/src/mesa/drivers/dri/i915/i830_metaops.c
+++ b/src/mesa/drivers/dri/i915/i830_metaops.c
@@ -58,7 +58,7 @@ do {						\
  * current GL state and used for other purposes than simply rendering
  * incoming triangles.
  */
-static void set_initial_state( i830ContextPtr i830 )
+static void set_initial_state( struct intel_context *intel )
 {
    memcpy(&i830->meta, &i830->initial, sizeof(i830->meta) );
    i830->meta.active = ACTIVE;
@@ -66,7 +66,7 @@ static void set_initial_state( i830ContextPtr i830 )
 }
 
 
-static void set_no_depth_stencil_write( i830ContextPtr i830 )
+static void set_no_depth_stencil_write( struct intel_context *intel )
 {
    /* ctx->Driver.Enable( ctx, GL_STENCIL_TEST, GL_FALSE )
     */
@@ -88,7 +88,7 @@ static void set_no_depth_stencil_write( i830ContextPtr i830 )
 
 /* Set stencil unit to replace always with the reference value.
  */
-static void set_stencil_replace( i830ContextPtr i830,
+static void set_stencil_replace( struct intel_context *intel,
 				 GLuint s_mask,
 				 GLuint s_clear)
 {
@@ -140,7 +140,7 @@ static void set_stencil_replace( i830ContextPtr i830,
 }
 
 
-static void set_color_mask( i830ContextPtr i830, GLboolean state )
+static void set_color_mask( struct intel_context *intel, GLboolean state )
 {
    const GLuint mask = ((1 << WRITEMASK_RED_SHIFT) |
 			(1 << WRITEMASK_GREEN_SHIFT) |
@@ -161,7 +161,7 @@ static void set_color_mask( i830ContextPtr i830, GLboolean state )
 /* Installs a one-stage passthrough texture blend pipeline.  Is there
  * more that can be done to turn off texturing?
  */
-static void set_no_texture( i830ContextPtr i830 )
+static void set_no_texture( struct intel_context *intel )
 {
    static const struct gl_tex_env_combine_state comb = {
       GL_NONE, GL_NONE,
@@ -181,7 +181,7 @@ static void set_no_texture( i830ContextPtr i830 )
 /* Set up a single element blend stage for 'replace' texturing with no
  * funny ops.
  */
-static void enable_texture_blend_replace( i830ContextPtr i830,
+static void enable_texture_blend_replace( struct intel_context *intel,
 					  GLenum format )
 {
    static const struct gl_tex_env_combine_state comb = {
@@ -207,7 +207,7 @@ static void enable_texture_blend_replace( i830ContextPtr i830,
 /* Set up an arbitary piece of memory as a rectangular texture
  * (including the front or back buffer).
  */
-static void set_tex_rect_source( i830ContextPtr i830,
+static void set_tex_rect_source( struct intel_context *intel,
 				 GLuint offset,
 				 GLuint width, 
 				 GLuint height,
@@ -249,17 +249,17 @@ static void set_tex_rect_source( i830ContextPtr i830,
 
 /* Select between front and back draw buffers.
  */
-static void set_draw_offset( i830ContextPtr i830,
+static void set_draw_offset( struct intel_context *intel,
 			     GLuint offset )
 {
-   i830->meta.Buffer[I830_DESTREG_CBUFADDR2] = offset;
+/*    i830->meta.Buffer[I830_DESTREG_CBUFADDR2] = offset; */
    i830->meta.emitted &= ~I830_UPLOAD_BUFFERS;
 }
 
 /* Setup an arbitary draw format, useful for targeting
  * texture or agp memory.
  */
-static void set_draw_format( i830ContextPtr i830,
+static void set_draw_format( struct intel_context *intel,
 			     GLuint format,
 			     GLuint depth_format)
 {
@@ -271,7 +271,7 @@ static void set_draw_format( i830ContextPtr i830,
 }
 
 
-static void set_vertex_format( i830ContextPtr i830 )
+static void set_vertex_format( struct intel_context *intel )
 {
    i830->meta.Ctx[I830_CTXREG_VF] =  (_3DSTATE_VFT0_CMD |
 				      VFT0_TEX_COUNT(1) |
@@ -287,448 +287,6 @@ static void set_vertex_format( i830ContextPtr i830 )
 }
 
 
-static void draw_quad(i830ContextPtr i830, 
-		      GLfloat x0, GLfloat x1,
-		      GLfloat y0, GLfloat y1, 
-		      GLubyte red, GLubyte green,
-		      GLubyte blue, GLubyte alpha,
-		      GLfloat s0, GLfloat s1,
-		      GLfloat t0, GLfloat t1 )
-{
-#if 0
-   GLuint vertex_size = 8;
-   GLuint *vb = intelEmitInlinePrimitiveLocked( &i830->intel, 
-						PRIM3D_TRIFAN, 
-						4*vertex_size,
-						vertex_size );
-   intelVertex tmp;
-   int i;
-
-   
-/*    fprintf(stderr, "%s: %f,%f-%f,%f 0x%x%x%x%x %f,%f-%f,%f\n", */
-/* 	   __FUNCTION__, */
-/* 	   x0,y0,x1,y1,red,green,blue,alpha,s0,t0,s1,t1); */
-
-
-   /* initial vertex, left bottom */
-   tmp.v.x = x0;
-   tmp.v.y = y0;
-   tmp.v.z = 1.0;
-   tmp.v.w = 1.0;
-   tmp.v.color.red = red;
-   tmp.v.color.green = green;
-   tmp.v.color.blue = blue;
-   tmp.v.color.alpha = alpha;
-   tmp.v.specular.red = 0;
-   tmp.v.specular.green = 0;
-   tmp.v.specular.blue = 0;
-   tmp.v.specular.alpha = 0;
-   tmp.v.u0 = s0;
-   tmp.v.v0 = t0;
-   for (i = 0 ; i < 8 ; i++)
-      vb[i] = tmp.ui[i];
-
-   /* right bottom */
-   vb += 8;
-   tmp.v.x = x1;
-   tmp.v.u0 = s1;
-   for (i = 0 ; i < 8 ; i++)
-      vb[i] = tmp.ui[i];
-
-   /* right top */
-   vb += 8;
-   tmp.v.y = y1;
-   tmp.v.v0 = t1;
-   for (i = 0 ; i < 8 ; i++)
-      vb[i] = tmp.ui[i];
-
-   /* left top */
-   vb += 8;
-   tmp.v.x = x0;
-   tmp.v.u0 = s0;
-   for (i = 0 ; i < 8 ; i++)
-      vb[i] = tmp.ui[i];
-
-/*    fprintf(stderr, "%s: DV1: %x\n",  */
-/* 	   __FUNCTION__, i830->meta.Buffer[I830_DESTREG_DV1]); */
-#endif
-}
-
-void 
-i830ClearWithTris(intelContextPtr intel, GLbitfield mask,
-		  GLboolean all,
-		  GLint cx, GLint cy, GLint cw, GLint ch)
-{
-   i830ContextPtr i830 = I830_CONTEXT( intel );
-   __DRIdrawablePrivate *dPriv = intel->driDrawable;
-   intelScreenPrivate *screen = intel->intelScreen;
-   int x0, y0, x1, y1;
-
-
-   SET_STATE( i830, meta );
-   set_initial_state( i830 );
-   set_no_texture( i830 );
-   set_vertex_format( i830 ); 
-
-   LOCK_HARDWARE(intel);
-
-   if(!all) {
-      x0 = cx;
-      y0 = cy;
-      x1 = x0 + cw;
-      y1 = y0 + ch;
-   } else {
-      x0 = 0;
-      y0 = 0;
-      x1 = x0 + dPriv->w;
-      y1 = y0 + dPriv->h;
-   }
-
-   /* Don't do any clipping to screen - these are window coordinates.
-    * The active cliprects will be applied as for any other geometry.
-    */
-
-   if(mask & BUFFER_BIT_FRONT_LEFT) {
-      set_no_depth_stencil_write( i830 );
-      set_color_mask( i830, GL_TRUE );
-      set_draw_offset( i830, screen->front.offset );
-      draw_quad(i830, x0, x1, y0, y1,
-		intel->clear_red, intel->clear_green,
-		intel->clear_blue, intel->clear_alpha,
-		0, 0, 0, 0);
-   }
-
-   if(mask & BUFFER_BIT_BACK_LEFT) {
-      set_no_depth_stencil_write( i830 );
-      set_color_mask( i830, GL_TRUE );
-      set_draw_offset( i830, screen->back.offset );
-
-      draw_quad(i830, x0, x1, y0, y1,
-		intel->clear_red, intel->clear_green,
-		intel->clear_blue, intel->clear_alpha,
-		0, 0, 0, 0);
-   }
-
-   if(mask & BUFFER_BIT_STENCIL) {
-      set_stencil_replace( i830, 
-			   intel->ctx.Stencil.WriteMask[0], 
-			   intel->ctx.Stencil.Clear);
-
-      set_color_mask( i830, GL_FALSE );
-      set_draw_offset( i830, screen->front.offset );
-      draw_quad( i830, x0, x1, y0, y1, 0, 0, 0, 0, 0, 0, 0, 0 );
-   }
 
-   UNLOCK_HARDWARE(intel);
-
-   SET_STATE( i830, state );
-}
-
-
-
-
-GLboolean
-i830TryTextureReadPixels( GLcontext *ctx,
-			  GLint x, GLint y, GLsizei width, GLsizei height,
-			  GLenum format, GLenum type,
-			  const struct gl_pixelstore_attrib *pack,
-			  GLvoid *pixels )
-{
-   i830ContextPtr i830 = I830_CONTEXT(ctx);
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
-   intelScreenPrivate *screen = i830->intel.intelScreen;
-   GLint pitch = pack->RowLength ? pack->RowLength : width;
-   __DRIdrawablePrivate *dPriv = i830->intel.driDrawable;
-   int textureFormat;
-   GLenum glTextureFormat;
-   int src_offset = i830->meta.Buffer[I830_DESTREG_CBUFADDR2];
-   int destOffset = 0;
-   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) {
-      fprintf(stderr, "%s: check_color failed\n", __FUNCTION__);
-      return GL_FALSE;
-   }
-
-   switch (screen->fbFormat) {
-   case DV_PF_565:
-      textureFormat = MAPSURF_16BIT | MT_16BIT_RGB565;
-      glTextureFormat = GL_RGB;
-      break;
-   case DV_PF_555:
-      textureFormat = MAPSURF_16BIT | MT_16BIT_ARGB1555;
-      glTextureFormat = GL_RGBA;
-      break;
-   case DV_PF_8888:
-      textureFormat = MAPSURF_32BIT | MT_32BIT_ARGB8888;
-      glTextureFormat = GL_RGBA;
-      break;
-   default:
-      fprintf(stderr, "%s: textureFormat failed %x\n", __FUNCTION__,
-	      screen->fbFormat);
-      return GL_FALSE;
-   }
-
-
-   switch (type) {
-   case GL_UNSIGNED_SHORT_5_6_5: 
-      if (format != GL_RGB) return GL_FALSE;
-      destFormat = COLR_BUF_RGB565; 
-      depthFormat = DEPTH_FRMT_16_FIXED;
-      destPitch = pitch * 2;
-      break;
-   case GL_UNSIGNED_INT_8_8_8_8_REV: 
-      if (format != GL_BGRA) return GL_FALSE;
-      destFormat = COLR_BUF_ARGB8888; 
-      depthFormat = DEPTH_FRMT_24_FIXED_8_OTHER;
-      destPitch = pitch * 4;
-      break;
-   default:
-      fprintf(stderr, "%s: destFormat failed %s\n", __FUNCTION__,
-	      _mesa_lookup_enum_by_nr(type));
-      return GL_FALSE;
-   }
-
-   destFormat |= (0x02<<24);
-
-/*    fprintf(stderr, "type: %s destFormat: %x\n", */
-/* 	   _mesa_lookup_enum_by_nr(type), */
-/* 	   destFormat); */
-
-   intelFlush( ctx );
-
-   SET_STATE( i830, meta );
-   set_initial_state( i830 );
-   set_no_depth_stencil_write( i830 );
-
-   LOCK_HARDWARE( intel );
-   {
-      intelWaitForIdle( intel ); /* required by GL */
-
-      if (!driClipRectToFramebuffer(ctx->ReadBuffer, &x, &y, &width, &height)) {
-	 UNLOCK_HARDWARE( intel );
-	 SET_STATE(i830, state);
-	 fprintf(stderr, "%s: cliprect failed\n", __FUNCTION__);
-	 return GL_TRUE;
-      }
-
-#if 0
-      /* FIXME -- Just emit the correct state
-       */
-      if (i830SetParam(i830->driFd, I830_SETPARAM_CBUFFER_PITCH, 
-		      destPitch) != 0) {
-	 UNLOCK_HARDWARE( intel );
-	 SET_STATE(i830, state);
-	 fprintf(stderr, "%s: setparam failed\n", __FUNCTION__);
-	 return GL_FALSE;
-      }
-#endif
-
-
-      y = dPriv->h - y - height;
-      x += dPriv->x;
-      y += dPriv->y;
-
-
-      /* Set the frontbuffer up as a large rectangular texture.
-       */
-      set_tex_rect_source( i830, 
-			   src_offset, 
-			   screen->width, 
-			   screen->height, 
-			   screen->front.pitch, 
-			   textureFormat ); 
-   
-   
-      enable_texture_blend_replace( i830, glTextureFormat ); 
-
-
-      /* Set the 3d engine to draw into the agp memory
-       */
-
-      set_draw_offset( i830, destOffset ); 
-      set_draw_format( i830, destFormat, depthFormat );  
-
-
-      /* Draw a single quad, no cliprects:
-       */
-      i830->intel.numClipRects = 1;
-      i830->intel.pClipRects = &tmp;
-      i830->intel.pClipRects[0].x1 = 0;
-      i830->intel.pClipRects[0].y1 = 0;
-      i830->intel.pClipRects[0].x2 = width;
-      i830->intel.pClipRects[0].y2 = height;
-
-      draw_quad( i830, 
-		 0, width, 0, height, 
-		 0, 255, 0, 0, 
-		 x, x+width, y, y+height );
-
-      intelWindowMoved( intel );
-   }
-   UNLOCK_HARDWARE( intel );
-   intelFinish( ctx ); /* required by GL */
 
-   SET_STATE( i830, state );
-   return GL_TRUE;
-}
-
-
-GLboolean
-i830TryTextureDrawPixels( GLcontext *ctx,
-			  GLint x, GLint y, GLsizei width, GLsizei height,
-			  GLenum format, GLenum type,
-			  const struct gl_pixelstore_attrib *unpack,
-			  const GLvoid *pixels )
-{
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
-   i830ContextPtr i830 = I830_CONTEXT(ctx);
-   GLint pitch = unpack->RowLength ? unpack->RowLength : width;
-   __DRIdrawablePrivate *dPriv = intel->driDrawable;
-   int textureFormat;
-   GLenum glTextureFormat;
-   int dst_offset = i830->meta.Buffer[I830_DESTREG_CBUFADDR2];
-   int src_offset = 0;
-
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   /* Todo -- upload images that aren't in agp space, then texture
-    * from them.  
-    */
-
-   if ( !intelIsAgpMemory( intel, pixels, pitch*height ) ) {
-      fprintf(stderr, "%s: intelIsAgpMemory failed\n", __FUNCTION__);
-      return GL_FALSE;
-   }
-
-   /* Todo -- don't want to clobber all the drawing state like we do
-    * for readpixels -- most of this state can be handled just fine.
-    */
-   if (	ctx->_ImageTransferState ||
-	unpack->SwapBytes ||
-	unpack->LsbFirst ||
-	ctx->Color.AlphaEnabled || 
-	ctx->Depth.Test ||
-	ctx->Fog.Enabled ||
-	ctx->Scissor.Enabled ||
-	ctx->Stencil.Enabled ||
-	!ctx->Color.ColorMask[0] ||
-	!ctx->Color.ColorMask[1] ||
-	!ctx->Color.ColorMask[2] ||
-	!ctx->Color.ColorMask[3] ||
-	ctx->Color.ColorLogicOpEnabled ||
-	ctx->Texture._EnabledUnits) {
-      fprintf(stderr, "%s: other tests failed\n", __FUNCTION__);
-      return GL_FALSE;
-   }
-
-   /* Todo -- remove these restrictions:
-    */
-   if (ctx->Pixel.ZoomX != 1.0F ||
-       ctx->Pixel.ZoomY != -1.0F)
-      return GL_FALSE;
-
-
-
-   switch (type) {
-   case GL_UNSIGNED_SHORT_1_5_5_5_REV:
-      if (format != GL_BGRA) return GL_FALSE;
-      textureFormat = MAPSURF_16BIT | MT_16BIT_ARGB1555;
-      glTextureFormat = GL_RGBA;
-      break;
-   case GL_UNSIGNED_SHORT_5_6_5: 
-      if (format != GL_RGB) return GL_FALSE;
-      textureFormat = MAPSURF_16BIT | MT_16BIT_RGB565;
-      glTextureFormat = GL_RGB;
-      break;
-   case GL_UNSIGNED_SHORT_8_8_MESA: 
-      if (format != GL_YCBCR_MESA) return GL_FALSE;
-      textureFormat = (MAPSURF_422 | MT_422_YCRCB_SWAPY 
-/*  		       | TM0S1_COLORSPACE_CONVERSION */
-	 );
-      glTextureFormat = GL_YCBCR_MESA;
-      break;
-   case GL_UNSIGNED_SHORT_8_8_REV_MESA: 
-      if (format != GL_YCBCR_MESA) return GL_FALSE;
-      textureFormat = (MAPSURF_422 | MT_422_YCRCB_NORMAL 
-/* 		       | TM0S1_COLORSPACE_CONVERSION */
-	 );
-      glTextureFormat = GL_YCBCR_MESA;
-      break;
-   case GL_UNSIGNED_INT_8_8_8_8_REV: 
-      if (format != GL_BGRA) return GL_FALSE;
-      textureFormat = MAPSURF_32BIT | MT_32BIT_ARGB8888;
-      glTextureFormat = GL_RGBA;
-      break;
-   default:
-      fprintf(stderr, "%s: destFormat failed\n", __FUNCTION__);
-      return GL_FALSE;
-   }
-
-   intelFlush( ctx );
-
-   SET_STATE( i830, meta );
-
-   LOCK_HARDWARE( intel );
-   {
-      intelWaitForIdle( intel ); /* required by GL */
-
-      y -= height;			/* cope with pixel zoom */
-
-      if (!driClipRectToFramebuffer(ctx->ReadBuffer, &x, &y, &width, &height)) {
-	 UNLOCK_HARDWARE( intel );
-	 SET_STATE(i830, state);
-	 fprintf(stderr, "%s: cliprect failed\n", __FUNCTION__);
-	 return GL_TRUE;
-      }
-
-
-      y = dPriv->h - y - height;
-
-      set_initial_state( i830 );
-
-      /* Set the pixel image up as a rectangular texture.
-       */
-      set_tex_rect_source( i830, 
-			   src_offset, 
-			   width, 
-			   height, 
-			   pitch, /* XXXX!!!! -- /2 sometimes */
-			   textureFormat ); 
-   
-   
-      enable_texture_blend_replace( i830, glTextureFormat ); 
-
-   
-      /* Draw to the current draw buffer:
-       */
-      set_draw_offset( i830, dst_offset );
-
-      /* Draw a quad, use regular cliprects
-       */
-/*       fprintf(stderr, "x: %d y: %d width %d height %d\n", x, y, width, height); */
-
-      draw_quad( i830, 
-		 x, x+width, y, y+height,
-		 0, 255, 0, 0, 
-		 0, width, 0, height );
-
-      intelWindowMoved( intel );
-   }
-   UNLOCK_HARDWARE( intel );
-   intelFinish( ctx ); /* required by GL */
-   
-   SET_STATE(i830, state);
-
-   return GL_TRUE;
-}
 
diff --git a/src/mesa/drivers/dri/i915/i830_reg.h b/src/mesa/drivers/dri/i915/i830_reg.h
index c45b88ca520..ff08683d7e3 100644
--- a/src/mesa/drivers/dri/i915/i830_reg.h
+++ b/src/mesa/drivers/dri/i915/i830_reg.h
@@ -407,10 +407,10 @@
 #define LOGICOP_SET			0xf
 #define MODE4_ENABLE_STENCIL_TEST_MASK	((1<<17)|(0xff00))
 #define ENABLE_STENCIL_TEST_MASK	(1<<17)
-#define STENCIL_TEST_MASK(x)		((x)<<8)
+#define STENCIL_TEST_MASK(x)		(((x)&0xff)<<8)
 #define MODE4_ENABLE_STENCIL_WRITE_MASK	((1<<16)|(0x00ff))
 #define ENABLE_STENCIL_WRITE_MASK	(1<<16)
-#define STENCIL_WRITE_MASK(x)		(x)
+#define STENCIL_WRITE_MASK(x)		((x)&0xff)
 
 /* _3DSTATE_MODES_5, p196 */
 #define _3DSTATE_MODES_5_CMD		(CMD_3D|(0x0c<<24))
diff --git a/src/mesa/drivers/dri/i915/i830_state.c b/src/mesa/drivers/dri/i915/i830_state.c
index 605f400c06d..f810bd1bd7a 100644
--- a/src/mesa/drivers/dri/i915/i830_state.c
+++ b/src/mesa/drivers/dri/i915/i830_state.c
@@ -1012,13 +1012,12 @@ static void i830_init_packets( i830ContextPtr i830 )
       (BUF_3D_ID_DEPTH |
        BUF_3D_PITCH(screen->depth.pitch * screen->cpp) |
        BUF_3D_USE_FENCE);
-   i830->state.Buffer[I830_DESTREG_DBUFADDR2] = screen->depth.offset;
+/*    i830->state.Buffer[I830_DESTREG_DBUFADDR2] = screen->depth.offset; */
 
 
    i830->state.Buffer[I830_DESTREG_DV0] = _3DSTATE_DST_BUF_VARS_CMD;
 
    switch (screen->fbFormat) {
-   case DV_PF_555:
    case DV_PF_565:
       i830->state.Buffer[I830_DESTREG_DV1] = (DSTORG_HORT_BIAS(0x8) | /* .5 */
 					      DSTORG_VERT_BIAS(0x8) | /* .5 */
diff --git a/src/mesa/drivers/dri/i915/i830_texblend.c b/src/mesa/drivers/dri/i915/i830_texblend.c
index 49e0347643c..19560868963 100644
--- a/src/mesa/drivers/dri/i915/i830_texblend.c
+++ b/src/mesa/drivers/dri/i915/i830_texblend.c
@@ -101,7 +101,7 @@ static GLuint emit_factor( GLuint blendUnit, GLuint *state, GLuint count,
 }
 
 
-static __inline__ GLuint GetTexelOp(GLint unit)
+static inline GLuint GetTexelOp(GLint unit)
 {
    switch(unit) {
    case 0: return TEXBLENDARG_TEXEL0;
diff --git a/src/mesa/drivers/dri/i915/i830_texstate.c b/src/mesa/drivers/dri/i915/i830_texstate.c
index 85da0ee281f..d248eba2964 100644
--- a/src/mesa/drivers/dri/i915/i830_texstate.c
+++ b/src/mesa/drivers/dri/i915/i830_texstate.c
@@ -428,7 +428,7 @@ static GLboolean i830UpdateTexUnit( GLcontext *ctx, GLuint unit )
    struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
 
    if (texUnit->_ReallyEnabled &&
-       INTEL_CONTEXT(ctx)->intelScreen->tex.size < 2048 * 1024)
+       intel_context(ctx)->intelScreen->tex.size < 2048 * 1024)
       return GL_FALSE;
 
    switch(texUnit->_ReallyEnabled) {
@@ -450,7 +450,7 @@ static GLboolean i830UpdateTexUnit( GLcontext *ctx, GLuint unit )
 }
 
 
-void i830UpdateTextureState( intelContextPtr intel )
+void i830UpdateTextureState( struct intel_context *intel )
 {
    i830ContextPtr i830 = I830_CONTEXT(intel);
    GLcontext *ctx = &intel->ctx;
diff --git a/src/mesa/drivers/dri/i915/i830_vtbl.c b/src/mesa/drivers/dri/i915/i830_vtbl.c
index 6e73f83728c..47b53033cbb 100644
--- a/src/mesa/drivers/dri/i915/i830_vtbl.c
+++ b/src/mesa/drivers/dri/i915/i830_vtbl.c
@@ -34,7 +34,7 @@
 #include "tnl/t_context.h"
 #include "tnl/t_vertex.h"
 
-static GLboolean i830_check_vertex_size( intelContextPtr intel,
+static GLboolean i830_check_vertex_size( struct intel_context *intel,
 					 GLuint expected );
 
 #define SZ_TO_HW(sz)  ((sz-2)&0x3)
@@ -59,7 +59,7 @@ do {									\
 #define VRTX_TEX_SET_FMT(n, x)          ((x)<<((n)*2))
 #define TEXBIND_SET(n, x) 		((x)<<((n)*4))
 
-static void i830_render_start( intelContextPtr intel )
+static void i830_render_start( struct intel_context *intel )
 {
    GLcontext *ctx = &intel->ctx;
    i830ContextPtr i830 = I830_CONTEXT(intel);
@@ -186,7 +186,7 @@ static void i830_render_start( intelContextPtr intel )
    }
 }
 
-static void i830_reduced_primitive_state( intelContextPtr intel,
+static void i830_reduced_primitive_state( struct intel_context *intel,
 					  GLenum rprim )
 {
     i830ContextPtr i830 = I830_CONTEXT(intel);
@@ -217,7 +217,7 @@ static void i830_reduced_primitive_state( intelContextPtr intel,
 /* Pull apart the vertex format registers and figure out how large a
  * vertex is supposed to be. 
  */
-static GLboolean i830_check_vertex_size( intelContextPtr intel,
+static GLboolean i830_check_vertex_size( struct intel_context *intel,
 					 GLuint expected )
 {
    i830ContextPtr i830 = I830_CONTEXT(intel);
@@ -257,11 +257,11 @@ static GLboolean i830_check_vertex_size( intelContextPtr intel,
    return sz == expected;
 }
 
-static void i830_emit_invarient_state( intelContextPtr intel )
+static void i830_emit_invarient_state( struct intel_context *intel )
 {
    BATCH_LOCALS;
 
-   BEGIN_BATCH( 200 );
+   BEGIN_BATCH(200, 0);
 
    OUT_BATCH(_3DSTATE_MAP_CUBE | MAP_UNIT(0));
    OUT_BATCH(_3DSTATE_MAP_CUBE | MAP_UNIT(1));
@@ -355,7 +355,7 @@ static void i830_emit_invarient_state( intelContextPtr intel )
 #define emit( intel, state, size )			\
 do {							\
    int k;						\
-   BEGIN_BATCH( size / sizeof(GLuint));			\
+   BEGIN_BATCH(size / sizeof(GLuint), 0);			\
    for (k = 0 ; k < size / sizeof(GLuint) ; k++)	\
       OUT_BATCH(state[k]);				\
    ADVANCE_BATCH();					\
@@ -364,7 +364,7 @@ do {							\
 
 /* Push the state into the sarea and/or texture memory.
  */
-static void i830_emit_state( intelContextPtr intel )
+static void i830_emit_state( struct intel_context *intel )
 {
    i830ContextPtr i830 = I830_CONTEXT(intel);
    struct i830_hw_state *state = i830->current;
@@ -405,32 +405,32 @@ static void i830_emit_state( intelContextPtr intel )
    state->emitted |= dirty;
 }
 
-static void i830_destroy_context( intelContextPtr intel )
+static void i830_destroy_context( struct intel_context *intel )
 {
    _tnl_free_vertices(&intel->ctx);
 }
 
-static void i830_set_draw_offset( intelContextPtr intel, int offset )
+static void i830_set_draw_offset( struct intel_context *intel, int offset )
 {
    i830ContextPtr i830 = I830_CONTEXT(intel);
    I830_STATECHANGE( i830, I830_UPLOAD_BUFFERS );
-   i830->state.Buffer[I830_DESTREG_CBUFADDR2] = offset;
+/*    i830->state.Buffer[I830_DESTREG_CBUFADDR2] = offset; */
 }
 
 /* This isn't really handled at the moment.
  */
-static void i830_lost_hardware( intelContextPtr intel )
+static void i830_lost_hardware( struct intel_context *intel )
 {
    I830_CONTEXT(intel)->state.emitted = 0;
 }
 
 
 
-static void i830_emit_flush( intelContextPtr intel )
+static void i830_emit_flush( struct intel_context *intel )
 {
    BATCH_LOCALS;
 
-   BEGIN_BATCH(2);
+   BEGIN_BATCH(2, 0);
    OUT_BATCH( MI_FLUSH | FLUSH_MAP_CACHE ); 
    OUT_BATCH( 0 );
    ADVANCE_BATCH();
diff --git a/src/mesa/drivers/dri/i915/i915_context.c b/src/mesa/drivers/dri/i915/i915_context.c
index f1b11bf95b1..0fe88794df6 100644
--- a/src/mesa/drivers/dri/i915/i915_context.c
+++ b/src/mesa/drivers/dri/i915/i915_context.c
@@ -69,7 +69,7 @@ static void i915InvalidateState( GLcontext *ctx, GLuint new_state )
    _ac_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    _tnl_invalidate_vertex_state( ctx, new_state );
-   INTEL_CONTEXT(ctx)->NewGLState |= new_state;
+   intel_context(ctx)->NewGLState |= new_state;
 
    /* Todo: gather state values under which tracked parameters become
     * invalidated, add callbacks for things like
@@ -103,9 +103,8 @@ GLboolean i915CreateContext( const __GLcontextModes *mesaVis,
 			    void *sharedContextPrivate)
 {
    struct dd_function_table functions;
-   i915ContextPtr i915 = (i915ContextPtr) CALLOC_STRUCT(i915_context);
-   intelContextPtr intel = &i915->intel;
-   intelScreenPrivate *intelScreen;
+   struct i915_context *i915 = (struct i915_context *) CALLOC_STRUCT(i915_context);
+   struct intel_context *intel = &i915->intel;
    GLcontext *ctx = &intel->ctx;
 
    if (!i915) return GL_FALSE;
@@ -113,6 +112,7 @@ GLboolean i915CreateContext( const __GLcontextModes *mesaVis,
    _mesa_printf( "\ntexmem branch (i915, drop2)\n\n");
    
    i915InitVtbl( i915 );
+   i915InitMetaFuncs( i915 );
 
    i915InitDriverFunctions( &functions );
 
@@ -126,49 +126,6 @@ GLboolean i915CreateContext( const __GLcontextModes *mesaVis,
    ctx->Const.MaxTextureImageUnits = I915_TEX_UNITS;
    ctx->Const.MaxTextureCoordUnits = I915_TEX_UNITS;
 
-   intel->bm = bm_fake_intel_Attach( intel );
-
-   bmInitPool(intel->bm,
-              intel->intelScreen->tex.offset, /* low offset */
-              intel->intelScreen->tex.map, /* low virtual */
-              intel->intelScreen->tex.size,
-	      BM_MEM_AGP);
-
-   intelScreen = intel->intelScreen;
-
-   /* These are still static, but create regions for them.  
-    */
-   intel->front_region = 
-      intel_region_create_static(intel,
-				 BM_MEM_AGP,
-				 intelScreen->front.offset,
-				 intelScreen->front.map,
-				 intelScreen->cpp,
-				 intelScreen->front.pitch,
-				 intelScreen->height);
-
-
-   intel->back_region = 
-      intel_region_create_static(intel,
-				 BM_MEM_AGP,
-				 intelScreen->back.offset,
-				 intelScreen->back.map,
-				 intelScreen->cpp,
-				 intelScreen->back.pitch,
-				 intelScreen->height);
-
-   /* Still assuming front.cpp == depth.cpp
-    */
-   intel->depth_region = 
-      intel_region_create_static(intel,
-				 BM_MEM_AGP,
-				 intelScreen->depth.offset,
-				 intelScreen->depth.map,
-				 intelScreen->cpp,
-				 intelScreen->depth.pitch,
-				 intelScreen->height);
-
-   intelInitBatchBuffer(intel);
 
    /* Advertise the full hardware capabilities.  The new memory
     * manager should cope much better with overload situations:
diff --git a/src/mesa/drivers/dri/i915/i915_context.h b/src/mesa/drivers/dri/i915/i915_context.h
index 0db119cd413..a50b6275fbf 100644
--- a/src/mesa/drivers/dri/i915/i915_context.h
+++ b/src/mesa/drivers/dri/i915/i915_context.h
@@ -53,10 +53,8 @@
  */
 #define I915_DESTREG_CBUFADDR0 0
 #define I915_DESTREG_CBUFADDR1 1
-#define I915_DESTREG_CBUFADDR2 2
 #define I915_DESTREG_DBUFADDR0 3
 #define I915_DESTREG_DBUFADDR1 4
-#define I915_DESTREG_DBUFADDR2 5
 #define I915_DESTREG_DV0 6
 #define I915_DESTREG_DV1 7
 #define I915_DESTREG_SENABLE 8
@@ -87,7 +85,6 @@
 #define I915_STPREG_ST1        1
 #define I915_STP_SETUP_SIZE    2
 
-#define I915_TEXREG_MS2        0
 #define I915_TEXREG_MS3        1
 #define I915_TEXREG_MS4        2
 #define I915_TEXREG_SS2        3
@@ -198,6 +195,15 @@ struct i915_hw_state {
    GLuint ConstantSize;
    GLuint Program[I915_PROGRAM_SIZE];
    GLuint ProgramSize;
+   
+   /* Region pointers for relocation: 
+    */
+   struct intel_region *draw_region;
+   struct intel_region *depth_region;
+   struct intel_region *tex_region[I915_TEX_UNITS];
+   GLuint tex_offset[I915_TEX_UNITS];
+   
+
    GLuint active;		/* I915_UPLOAD_* */
    GLuint emitted;		/* I915_UPLOAD_* */
 };
@@ -222,20 +228,14 @@ struct i915_context
 };
 
 
-typedef struct i915_context *i915ContextPtr;
-
-
 #define I915_STATECHANGE(i915, flag)					\
 do {									\
-   if (0) fprintf(stderr, "I915_STATECHANGE %x in %s\n", flag, __FUNCTION__);	\
    INTEL_FIREVERTICES( &(i915)->intel );					\
    (i915)->state.emitted &= ~(flag);					\
 } while (0)
 
 #define I915_ACTIVESTATE(i915, flag, mode)			\
 do {								\
-   if (0) fprintf(stderr, "I915_ACTIVESTATE %x %d in %s\n",	\
-		  flag, mode, __FUNCTION__);			\
    INTEL_FIREVERTICES( &(i915)->intel );				\
    if (mode)							\
       (i915)->state.active |= (flag);				\
@@ -247,7 +247,7 @@ do {								\
 /*======================================================================
  * i915_vtbl.c
  */
-extern void i915InitVtbl( i915ContextPtr i915 );
+extern void i915InitVtbl( struct i915_context *i915 );
 
 
 
@@ -284,7 +284,7 @@ extern GLboolean i915CreateContext( const __GLcontextModes *mesaVis,
 /*======================================================================
  * i915_texprog.c
  */
-extern void i915ValidateTextureProgram( i915ContextPtr i915 );
+extern void i915ValidateTextureProgram( struct i915_context *i915 );
 
 
 /*======================================================================
@@ -298,42 +298,26 @@ extern void i915_print_ureg( const char *msg, GLuint ureg );
  * i915_state.c
  */
 extern void i915InitStateFunctions( struct dd_function_table *functions );
-extern void i915InitState( i915ContextPtr i915 );
+extern void i915InitState( struct i915_context *i915 );
 extern void i915_update_fog( GLcontext *ctx );
 
 
 /*======================================================================
  * i915_tex.c
  */
-extern void i915UpdateTextureState( intelContextPtr intel );
+extern void i915UpdateTextureState( struct intel_context *intel );
 extern void i915InitTextureFuncs( struct dd_function_table *functions );
 
 /*======================================================================
  * i915_metaops.c
  */
-extern GLboolean
-i915TryTextureReadPixels( GLcontext *ctx,
-			  GLint x, GLint y, GLsizei width, GLsizei height,
-			  GLenum format, GLenum type,
-			  const struct gl_pixelstore_attrib *pack,
-			  GLvoid *pixels );
-
-extern GLboolean
-i915TryTextureDrawPixels( GLcontext *ctx,
-			  GLint x, GLint y, GLsizei width, GLsizei height,
-			  GLenum format, GLenum type,
-			  const struct gl_pixelstore_attrib *unpack,
-			  const GLvoid *pixels );
-
-extern void 
-i915ClearWithTris( intelContextPtr intel, GLbitfield mask,
-		   GLboolean all, GLint cx, GLint cy, GLint cw, GLint ch);
+void i915InitMetaFuncs( struct i915_context *i915 );
 
 
 /*======================================================================
  * i915_fragprog.c
  */
-extern void i915ValidateFragmentProgram( i915ContextPtr i915 );
+extern void i915ValidateFragmentProgram( struct i915_context *i915 );
 extern void i915InitFragProgFuncs( struct dd_function_table *functions );
 
 /*======================================================================
diff --git a/src/mesa/drivers/dri/i915/i915_fragprog.c b/src/mesa/drivers/dri/i915/i915_fragprog.c
index 5cd6ea4de51..c35f72ba75c 100644
--- a/src/mesa/drivers/dri/i915/i915_fragprog.c
+++ b/src/mesa/drivers/dri/i915/i915_fragprog.c
@@ -806,7 +806,7 @@ static void check_wpos( struct i915_fragment_program *p )
 
 static void translate_program( struct i915_fragment_program *p )
 {
-   i915ContextPtr i915 = I915_CONTEXT(p->ctx);
+   struct i915_context *i915 = I915_CONTEXT(p->ctx);
    
    i915_init_program( i915, p );
    check_wpos( p ); 
@@ -840,7 +840,7 @@ static void i915BindProgram( GLcontext *ctx,
 			    struct program *prog )
 {
    if (target == GL_FRAGMENT_PROGRAM_ARB) {
-      i915ContextPtr i915 = I915_CONTEXT(ctx);
+      struct i915_context *i915 = I915_CONTEXT(ctx);
       struct i915_fragment_program *p = (struct i915_fragment_program *)prog;
 
       if (i915->current_program == p) 
@@ -896,7 +896,7 @@ static void i915DeleteProgram( GLcontext *ctx,
 			      struct program *prog )
 {
    if (prog->Target == GL_FRAGMENT_PROGRAM_ARB) {
-      i915ContextPtr i915 = I915_CONTEXT(ctx);
+      struct i915_context *i915 = I915_CONTEXT(ctx);
       struct i915_fragment_program *p = (struct i915_fragment_program *)prog;
       
       if (i915->current_program == p) 
@@ -940,10 +940,10 @@ static void i915ProgramStringNotify( GLcontext *ctx,
 }
 
 
-void i915ValidateFragmentProgram( i915ContextPtr i915 )
+void i915ValidateFragmentProgram( struct i915_context *i915 )
 {
    GLcontext *ctx = &i915->intel.ctx;
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
+   struct intel_context *intel = intel_context(ctx);
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    struct vertex_buffer *VB = &tnl->vb;
 
diff --git a/src/mesa/drivers/dri/i915/i915_metaops.c b/src/mesa/drivers/dri/i915/i915_metaops.c
index b24dfe8a02a..59bd64f4425 100644
--- a/src/mesa/drivers/dri/i915/i915_metaops.c
+++ b/src/mesa/drivers/dri/i915/i915_metaops.c
@@ -34,39 +34,32 @@
 #include "intel_screen.h"
 #include "intel_batchbuffer.h"
 #include "intel_ioctl.h"
+#include "intel_regions.h"
 
 #include "i915_context.h"
 #include "i915_reg.h"
 
 /* A large amount of state doesn't need to be uploaded.
  */
-#define ACTIVE (I915_UPLOAD_PROGRAM | 		\
+#define ACTIVE (I915_UPLOAD_INVARIENT | 	\
+                I915_UPLOAD_PROGRAM | 		\
 		I915_UPLOAD_STIPPLE |		\
 		I915_UPLOAD_CTX |		\
 		I915_UPLOAD_BUFFERS |		\
 		I915_UPLOAD_TEX(0))		
 
-#define SET_STATE( i915, STATE )			\
+#define SET_STATE( i915, STATE )		\
 do {						\
    i915->current->emitted &= ~ACTIVE;		\
-   i915->current = &i915->STATE;			\
+   i915->current = &i915->STATE;		\
    i915->current->emitted &= ~ACTIVE;		\
 } while (0)
 
-/* Operations where the 3D engine is decoupled temporarily from the
- * current GL state and used for other purposes than simply rendering
- * incoming triangles.
- */
-static void set_initial_state( i915ContextPtr i915 )
-{
-   memcpy(&i915->meta, &i915->initial, sizeof(i915->meta) );
-   i915->meta.active = ACTIVE;
-   i915->meta.emitted = 0;
-}
 
-
-static void set_no_depth_stencil_write( i915ContextPtr i915 )
+static void meta_no_depth_stencil_write( struct intel_context *intel )
 {
+   struct i915_context *i915 = i915_context(&intel->ctx);
+
    /* ctx->Driver.Enable( ctx, GL_STENCIL_TEST, GL_FALSE )
     */
    i915->meta.Ctx[I915_CTXREG_LIS5] &= ~(S5_STENCIL_TEST_ENABLE | 
@@ -82,10 +75,11 @@ static void set_no_depth_stencil_write( i915ContextPtr i915 )
 
 /* Set stencil unit to replace always with the reference value.
  */
-static void set_stencil_replace( i915ContextPtr i915,
+static void meta_stencil_replace( struct intel_context *intel,
 				 GLuint s_mask,
 				 GLuint s_clear)
 {
+   struct i915_context *i915 = i915_context(&intel->ctx);
    GLuint op = STENCILOP_REPLACE;
    GLuint func = COMPAREFUNC_ALWAYS;
 
@@ -100,7 +94,6 @@ static void set_stencil_replace( i915ContextPtr i915,
    i915->meta.Ctx[I915_CTXREG_LIS6] &= ~(S6_DEPTH_TEST_ENABLE |
 				       S6_DEPTH_WRITE_ENABLE);
 
-
    /* ctx->Driver.StencilMask( ctx, s_mask )
     */
    i915->meta.Ctx[I915_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_WRITE_MASK;
@@ -108,7 +101,6 @@ static void set_stencil_replace( i915ContextPtr i915,
    i915->meta.Ctx[I915_CTXREG_STATE4] |= (ENABLE_STENCIL_WRITE_MASK |
 					STENCIL_WRITE_MASK(s_mask));
 
-
    /* ctx->Driver.StencilOp( ctx, GL_REPLACE, GL_REPLACE, GL_REPLACE )
     */
    i915->meta.Ctx[I915_CTXREG_LIS5] &= ~(S5_STENCIL_FAIL_MASK |
@@ -137,8 +129,9 @@ static void set_stencil_replace( i915ContextPtr i915,
 }
 
 
-static void set_color_mask( i915ContextPtr i915, GLboolean state )
+static void meta_color_mask( struct intel_context *intel, GLboolean state )
 {
+   struct i915_context *i915 = i915_context(&intel->ctx);
    const GLuint mask = (S5_WRITEDISABLE_RED |
 			S5_WRITEDISABLE_GREEN |
 			S5_WRITEDISABLE_BLUE |
@@ -210,8 +203,10 @@ static void set_color_mask( i915ContextPtr i915, GLboolean state )
 
 
 
-static void set_no_texture( i915ContextPtr i915 )
+static void meta_no_texture( struct intel_context *intel )
 {
+   struct i915_context *i915 = i915_context(&intel->ctx);
+
    static const GLuint prog[] = {
       _3DSTATE_PIXEL_SHADER_PROGRAM,
 
@@ -240,9 +235,10 @@ static void set_no_texture( i915ContextPtr i915 )
    i915->meta.emitted &= ~I915_UPLOAD_PROGRAM;
 }
 
-#if 0
-static void enable_texture_blend_replace( i915ContextPtr i915 )
+static void meta_texture_blend_replace( struct intel_context *intel )
 {
+   struct i915_context *i915 = i915_context(&intel->ctx);
+
    static const GLuint prog[] = {
       _3DSTATE_PIXEL_SHADER_PROGRAM,
 
@@ -285,78 +281,86 @@ static void enable_texture_blend_replace( i915ContextPtr i915 )
 /* Set up an arbitary piece of memory as a rectangular texture
  * (including the front or back buffer).
  */
-static void set_tex_rect_source( i915ContextPtr i915,
-				 GLuint offset,
-				 GLuint width, 
-				 GLuint height,
-				 GLuint pitch,
+static void meta_tex_rect_source( struct intel_context *intel,
+				 struct intel_region *region,
 				 GLuint textureFormat )
 {
+   struct i915_context *i915 = i915_context(&intel->ctx);
    GLuint unit = 0;
    GLint numLevels = 1;
    GLuint *state = i915->meta.Tex[0];
 
-   pitch *= i915->intel.intelScreen->cpp;
+   GLuint pitch = region->pitch * region->cpp;
 
 /*    fprintf(stderr, "%s: offset: %x w: %d h: %d pitch %d format %x\n", */
 /* 	   __FUNCTION__, offset, width, height, pitch, textureFormat ); */
 
-   state[I915_TEXREG_MS2] = offset;
-   state[I915_TEXREG_MS3] = (((height - 1) << MS3_HEIGHT_SHIFT) |
-			    ((width - 1) << MS3_WIDTH_SHIFT) |
-			    textureFormat |
-			    MS3_USE_FENCE_REGS);
+   intel_region_release(intel, &i915->meta.tex_region[0]);
+   intel_region_reference(&i915->meta.tex_region[0], region);
+   i915->meta.tex_offset[0] = 0;
+
+   state[I915_TEXREG_MS3] = (((region->height - 1) << MS3_HEIGHT_SHIFT) |
+			     ((region->pitch - 1) << MS3_WIDTH_SHIFT) |
+			     textureFormat |
+			     MS3_USE_FENCE_REGS);
 
    state[I915_TEXREG_MS4] = ((((pitch / 4) - 1) << MS4_PITCH_SHIFT) | 
-			    MS4_CUBE_FACE_ENA_MASK |
-			    ((((numLevels-1) * 4)) << MS4_MAX_LOD_SHIFT));
+			     MS4_CUBE_FACE_ENA_MASK |
+			     ((((numLevels-1) * 4)) << MS4_MAX_LOD_SHIFT));
 
    state[I915_TEXREG_SS2] = ((FILTER_NEAREST << SS2_MIN_FILTER_SHIFT) |
-			    (MIPFILTER_NONE << SS2_MIP_FILTER_SHIFT) |
-			    (FILTER_NEAREST << SS2_MAG_FILTER_SHIFT));
+			     (MIPFILTER_NONE << SS2_MIP_FILTER_SHIFT) |
+			     (FILTER_NEAREST << SS2_MAG_FILTER_SHIFT));
+
    state[I915_TEXREG_SS3] = ((TEXCOORDMODE_WRAP << SS3_TCX_ADDR_MODE_SHIFT) |
-			    (TEXCOORDMODE_WRAP << SS3_TCY_ADDR_MODE_SHIFT) |
-			    (TEXCOORDMODE_WRAP << SS3_TCZ_ADDR_MODE_SHIFT) |
-			    (unit<<SS3_TEXTUREMAP_INDEX_SHIFT));
+			     (TEXCOORDMODE_WRAP << SS3_TCY_ADDR_MODE_SHIFT) |
+			     (TEXCOORDMODE_WRAP << SS3_TCZ_ADDR_MODE_SHIFT) |
+			     (unit<<SS3_TEXTUREMAP_INDEX_SHIFT));
 
    state[I915_TEXREG_SS4] = 0;
 
    i915->meta.emitted &= ~I915_UPLOAD_TEX(0);
 }
-#endif
 
 /* Select between front and back draw buffers.
  */
-static void set_draw_offset( i915ContextPtr i915,
-			     GLuint offset )
+static void meta_draw_region( struct intel_context *intel,
+			     struct intel_region *draw_region,
+			     struct intel_region *depth_region )
 {
-   i915->meta.Buffer[I915_DESTREG_CBUFADDR2] = offset;
+   struct i915_context *i915 = i915_context(&intel->ctx);
+
+   intel_region_release(intel, &i915->meta.draw_region);
+   intel_region_release(intel, &i915->meta.depth_region);
+   intel_region_reference(&i915->meta.draw_region, draw_region);
+   intel_region_reference(&i915->meta.depth_region, depth_region);
+
    i915->meta.emitted &= ~I915_UPLOAD_BUFFERS;
 }
 
-#if 0
 /* Setup an arbitary draw format, useful for targeting texture or agp
  * memory.
  */
-static void set_draw_format( i915ContextPtr i915,
+static void set_draw_format( struct intel_context *intel,
 			     GLuint format,
 			     GLuint depth_format)
 {
+   struct i915_context *i915 = i915_context(&intel->ctx);
+
    i915->meta.Buffer[I915_DESTREG_DV1] = (DSTORG_HORT_BIAS(0x8) | /* .5 */
-					DSTORG_VERT_BIAS(0x8) | /* .5 */
-					format |
-					LOD_PRECLAMP_OGL |
-					TEX_DEFAULT_COLOR_OGL |
-					depth_format);
+					  DSTORG_VERT_BIAS(0x8) | /* .5 */
+					  format |
+					  LOD_PRECLAMP_OGL |
+					  TEX_DEFAULT_COLOR_OGL |
+					  depth_format);
 
    i915->meta.emitted &= ~I915_UPLOAD_BUFFERS;
-/*    fprintf(stderr, "%s: DV1: %x\n",  */
-/* 	   __FUNCTION__, i915->meta.Buffer[I915_DESTREG_DV1]); */
 }
-#endif
 
-static void set_vertex_format( i915ContextPtr i915 )
+static void set_vertex_format( struct intel_context *intel )
 {
+   struct i915_context *i915 = i915_context(&intel->ctx);
+
    i915->meta.Ctx[I915_CTXREG_LIS2] = 
       (S2_TEXCOORD_FMT(0, TEXCOORDFMT_2D) |
        S2_TEXCOORD_FMT(1, TEXCOORDFMT_NOT_PRESENT) | 
@@ -371,148 +375,50 @@ static void set_vertex_format( i915ContextPtr i915 )
 
    i915->meta.Ctx[I915_CTXREG_LIS4] |= 
       (S4_VFMT_COLOR |
-       S4_VFMT_SPEC_FOG |
-       S4_VFMT_XYZW);
+       S4_VFMT_XYZ);
 
    i915->meta.emitted &= ~I915_UPLOAD_CTX;
-
 }
 
 
-static void draw_quad(i915ContextPtr i915, 
-		      GLfloat x0, GLfloat x1,
-		      GLfloat y0, GLfloat y1, 
-		      GLubyte red, GLubyte green,
-		      GLubyte blue, GLubyte alpha,
-		      GLfloat s0, GLfloat s1,
-		      GLfloat t0, GLfloat t1 )
-{
-#if 0
-   GLuint vertex_size = 8;
-   GLuint *vb = intelEmitInlinePrimitiveLocked( &i915->intel, 
-						PRIM3D_TRIFAN, 
-						4 * vertex_size,
-						vertex_size );
-   intelVertex tmp;
-   int i;
-
-   if (0)
-      fprintf(stderr, "%s: %f,%f-%f,%f 0x%x%x%x%x %f,%f-%f,%f\n",
-	      __FUNCTION__,
-	      x0,y0,x1,y1,red,green,blue,alpha,s0,t0,s1,t1);
-
-
-   /* initial vertex, left bottom */
-   tmp.v.x = x0;
-   tmp.v.y = y0;
-   tmp.v.z = 1.0;
-   tmp.v.w = 1.0; 
-   tmp.v.color.red = red;
-   tmp.v.color.green = green;
-   tmp.v.color.blue = blue;
-   tmp.v.color.alpha = alpha;
-   tmp.v.specular.red = 0;
-   tmp.v.specular.green = 0;
-   tmp.v.specular.blue = 0;
-   tmp.v.specular.alpha = 0;
-   tmp.v.u0 = s0;
-   tmp.v.v0 = t0;
-
-   for (i = 0 ; i < vertex_size ; i++)
-      vb[i] = tmp.ui[i];
-
-   /* right bottom */
-   vb += vertex_size;
-   tmp.v.x = x1;
-   tmp.v.u0 = s1;
-   for (i = 0 ; i < vertex_size ; i++)
-      vb[i] = tmp.ui[i];
-
-   /* right top */
-   vb += vertex_size;
-   tmp.v.y = y1;
-   tmp.v.v0 = t1;
-   for (i = 0 ; i < vertex_size ; i++)
-      vb[i] = tmp.ui[i];
-
-   /* left top */
-   vb += vertex_size;
-   tmp.v.x = x0;
-   tmp.v.u0 = s0;
-   for (i = 0 ; i < vertex_size ; i++)
-      vb[i] = tmp.ui[i];
-#endif
-}
 
-void 
-i915ClearWithTris(intelContextPtr intel, GLbitfield mask,
-		  GLboolean all,
-		  GLint cx, GLint cy, GLint cw, GLint ch)
+/* Operations where the 3D engine is decoupled temporarily from the
+ * current GL state and used for other purposes than simply rendering
+ * incoming triangles.
+ */
+static void install_meta_state( struct intel_context *intel )
 {
-   i915ContextPtr i915 = i915_context( &intel->ctx );
-   __DRIdrawablePrivate *dPriv = intel->driDrawable;
-   intelScreenPrivate *screen = intel->intelScreen;
-   int x0, y0, x1, y1;
-
-   SET_STATE( i915, meta ); 
-   set_initial_state( i915 ); 
-   set_no_texture( i915 ); 
-   set_vertex_format( i915 ); 
-
-   LOCK_HARDWARE(intel);
-
-   if(!all) {
-      x0 = cx;
-      y0 = cy;
-      x1 = x0 + cw;
-      y1 = y0 + ch;
-   } else {
-      x0 = 0;
-      y0 = 0;
-      x1 = x0 + dPriv->w;
-      y1 = y0 + dPriv->h;
-   }
-
-   /* Don't do any clipping to screen - these are window coordinates.
-    * The active cliprects will be applied as for any other geometry.
-    */
-
-   if (mask & BUFFER_BIT_FRONT_LEFT) { 
-      set_no_depth_stencil_write( i915 );
-      set_color_mask( i915, GL_TRUE );
-      set_draw_offset( i915, screen->front.offset );
-
-      draw_quad(i915, x0, x1, y0, y1,
-		intel->clear_red, intel->clear_green, 
- 		intel->clear_blue, intel->clear_alpha, 
-		0, 0, 0, 0);
-   }
-
-   if (mask & BUFFER_BIT_BACK_LEFT) {
-      set_no_depth_stencil_write( i915 );
-      set_color_mask( i915, GL_TRUE );
-      set_draw_offset( i915, screen->back.offset );
+   struct i915_context *i915 = i915_context(&intel->ctx);
+   memcpy(&i915->meta, &i915->initial, sizeof(i915->meta) );
+   i915->meta.active = ACTIVE;
+   i915->meta.emitted = 0;
 
-      draw_quad(i915, x0, x1, y0, y1,
-		intel->clear_red, intel->clear_green,
-		intel->clear_blue, intel->clear_alpha,
-		0, 0, 0, 0);
-   }
+   SET_STATE(i915, meta);
+   set_vertex_format(intel);
+   meta_no_texture(intel);
+}
 
-   if (mask & BUFFER_BIT_STENCIL) {
-      set_stencil_replace( i915, 
-			   intel->ctx.Stencil.WriteMask[0], 
-			   intel->ctx.Stencil.Clear);
-      
-      set_color_mask( i915, GL_FALSE );
-      set_draw_offset( i915, screen->front.offset ); /* could be either? */
+static void leave_meta_state( struct intel_context *intel )
+{
+   struct i915_context *i915 = i915_context(&intel->ctx);
+   intel_region_release(intel, &i915->meta.draw_region);
+   intel_region_release(intel, &i915->meta.depth_region);
+   intel_region_release(intel, &i915->meta.tex_region[0]);
+   SET_STATE(i915, state);
+}
 
-      draw_quad( i915, x0, x1, y0, y1, 0, 0, 0, 0, 0, 0, 0, 0 );
-   }
 
-   UNLOCK_HARDWARE(intel);
 
-   SET_STATE( i915, state );
+void i915InitMetaFuncs( struct i915_context *i915 )
+{
+   i915->intel.vtbl.install_meta_state = install_meta_state;
+   i915->intel.vtbl.leave_meta_state = leave_meta_state;
+   i915->intel.vtbl.meta_no_depth_stencil_write = meta_no_depth_stencil_write;
+   i915->intel.vtbl.meta_stencil_replace = meta_stencil_replace;
+   i915->intel.vtbl.meta_color_mask = meta_color_mask;
+   i915->intel.vtbl.meta_no_texture = meta_no_texture;
+   i915->intel.vtbl.meta_texture_blend_replace = meta_texture_blend_replace;
+   i915->intel.vtbl.meta_tex_rect_source = meta_tex_rect_source;
+   i915->intel.vtbl.meta_draw_region = meta_draw_region;
+   i915->intel.vtbl.meta_draw_format = set_draw_format;
 }
-
-
diff --git a/src/mesa/drivers/dri/i915/i915_program.c b/src/mesa/drivers/dri/i915/i915_program.c
index 45276fb6908..84f25a431a6 100644
--- a/src/mesa/drivers/dri/i915/i915_program.c
+++ b/src/mesa/drivers/dri/i915/i915_program.c
@@ -358,7 +358,7 @@ void i915_program_error( struct i915_fragment_program *p, const char *msg )
    p->error = 1;
 }
 
-void i915_init_program( i915ContextPtr i915, struct i915_fragment_program *p )
+void i915_init_program( struct i915_context *i915, struct i915_fragment_program *p )
 {
    GLcontext *ctx = &i915->intel.ctx;
    TNLcontext *tnl = TNL_CONTEXT( ctx );
@@ -431,7 +431,7 @@ void i915_fini_program( struct i915_fragment_program *p )
    p->declarations[0] |= program_size + decl_size - 2;
 }
 
-void i915_upload_program( i915ContextPtr i915, struct i915_fragment_program *p )
+void i915_upload_program( struct i915_context *i915, struct i915_fragment_program *p )
 {
    GLuint program_size = p->csr - p->program;
    GLuint decl_size = p->decl - p->declarations;
diff --git a/src/mesa/drivers/dri/i915/i915_program.h b/src/mesa/drivers/dri/i915/i915_program.h
index 8891a177855..8b9849e3401 100644
--- a/src/mesa/drivers/dri/i915/i915_program.h
+++ b/src/mesa/drivers/dri/i915/i915_program.h
@@ -84,7 +84,7 @@
 
 /* One neat thing about the UREG representation:  
  */
-static __inline int swizzle( int reg, int x, int y, int z, int w )
+static inline int swizzle( int reg, int x, int y, int z, int w )
 {
    return ((reg & ~UREG_XYZW_CHANNEL_MASK) |
 	   CHANNEL_SRC( GET_CHANNEL_SRC( reg, x ), 0 ) |
@@ -95,7 +95,7 @@ static __inline int swizzle( int reg, int x, int y, int z, int w )
 
 /* Another neat thing about the UREG representation:  
  */
-static __inline int negate( int reg, int x, int y, int z, int w )
+static inline int negate( int reg, int x, int y, int z, int w )
 {
    return reg ^ (((x&1)<<UREG_CHANNEL_X_NEGATE_SHIFT)|
 		 ((y&1)<<UREG_CHANNEL_Y_NEGATE_SHIFT)|
@@ -149,10 +149,10 @@ extern GLuint i915_emit_param4fv( struct i915_fragment_program *p,
 extern void i915_program_error( struct i915_fragment_program *p,
                                 const char *msg );
 
-extern void i915_init_program( i915ContextPtr i915,
+extern void i915_init_program( struct i915_context *i915,
 			      struct i915_fragment_program *p );
 
-extern void i915_upload_program( i915ContextPtr i915, 
+extern void i915_upload_program( struct i915_context *i915, 
 				struct i915_fragment_program *p );
 
 extern void i915_fini_program( struct i915_fragment_program *p );
diff --git a/src/mesa/drivers/dri/i915/i915_reg.h b/src/mesa/drivers/dri/i915/i915_reg.h
index 3241b36b660..3bc8a23c1b9 100644
--- a/src/mesa/drivers/dri/i915/i915_reg.h
+++ b/src/mesa/drivers/dri/i915/i915_reg.h
@@ -435,10 +435,10 @@
 #define LOGICOP_MASK			(0xf<<18)
 #define MODE4_ENABLE_STENCIL_TEST_MASK	((1<<17)|(0xff00))
 #define ENABLE_STENCIL_TEST_MASK	(1<<17)
-#define STENCIL_TEST_MASK(x)		((x)<<8)
+#define STENCIL_TEST_MASK(x)		(((x)&0xff)<<8)
 #define MODE4_ENABLE_STENCIL_WRITE_MASK	((1<<16)|(0x00ff))
 #define ENABLE_STENCIL_WRITE_MASK	(1<<16)
-#define STENCIL_WRITE_MASK(x)		(x)
+#define STENCIL_WRITE_MASK(x)		((x)&0xff)
 
 /* _3DSTATE_MODES_5, p220 */
 #define _3DSTATE_MODES_5_CMD		(CMD_3D|(0x0c<<24))
diff --git a/src/mesa/drivers/dri/i915/i915_state.c b/src/mesa/drivers/dri/i915/i915_state.c
index 8c67bea2ec9..2586219b82b 100644
--- a/src/mesa/drivers/dri/i915/i915_state.c
+++ b/src/mesa/drivers/dri/i915/i915_state.c
@@ -48,7 +48,7 @@ static void
 i915StencilFuncSeparate(GLcontext *ctx, GLenum face, GLenum func, GLint ref,
                         GLuint mask)
 {
-   i915ContextPtr i915 = I915_CONTEXT(ctx);
+   struct i915_context *i915 = I915_CONTEXT(ctx);
    int test = intel_translate_compare_func( func );
 
    mask = mask & 0xff;
@@ -73,7 +73,7 @@ i915StencilFuncSeparate(GLcontext *ctx, GLenum face, GLenum func, GLint ref,
 static void
 i915StencilMaskSeparate(GLcontext *ctx, GLenum face, GLuint mask)
 {
-   i915ContextPtr i915 = I915_CONTEXT(ctx);
+   struct i915_context *i915 = I915_CONTEXT(ctx);
 
    if (INTEL_DEBUG&DEBUG_DRI)
       fprintf(stderr, "%s : mask 0x%x\n", __FUNCTION__, mask);
@@ -91,7 +91,7 @@ static void
 i915StencilOpSeparate(GLcontext *ctx, GLenum face, GLenum fail, GLenum zfail,
                       GLenum zpass)
 {
-   i915ContextPtr i915 = I915_CONTEXT(ctx);
+   struct i915_context *i915 = I915_CONTEXT(ctx);
    int fop = intel_translate_stencil_op(fail); 
    int dfop = intel_translate_stencil_op(zfail); 
    int dpop = intel_translate_stencil_op(zpass);
@@ -116,7 +116,7 @@ i915StencilOpSeparate(GLcontext *ctx, GLenum face, GLenum fail, GLenum zfail,
 
 static void i915AlphaFunc(GLcontext *ctx, GLenum func, GLfloat ref)
 {
-   i915ContextPtr i915 = I915_CONTEXT(ctx);
+   struct i915_context *i915 = I915_CONTEXT(ctx);
    int test = intel_translate_compare_func( func );
    GLubyte refByte;
 
@@ -137,7 +137,7 @@ static void i915AlphaFunc(GLcontext *ctx, GLenum func, GLfloat ref)
  */
 static void i915EvalLogicOpBlendState(GLcontext *ctx)
 {
-   i915ContextPtr i915 = I915_CONTEXT(ctx);
+   struct i915_context *i915 = I915_CONTEXT(ctx);
 
    I915_STATECHANGE(i915, I915_UPLOAD_CTX);
 
@@ -157,7 +157,7 @@ static void i915EvalLogicOpBlendState(GLcontext *ctx)
 
 static void i915BlendColor(GLcontext *ctx, const GLfloat color[4])
 {
-   i915ContextPtr i915 = I915_CONTEXT(ctx);
+   struct i915_context *i915 = I915_CONTEXT(ctx);
    GLubyte r, g, b, a;
 
    if (INTEL_DEBUG&DEBUG_DRI)
@@ -194,7 +194,7 @@ static GLuint translate_blend_equation( GLenum mode )
 
 static void i915UpdateBlendState( GLcontext *ctx )
 {
-   i915ContextPtr i915 = I915_CONTEXT(ctx);
+   struct i915_context *i915 = I915_CONTEXT(ctx);
    GLuint iab = (i915->state.Ctx[I915_CTXREG_IAB] & 
 		 ~(IAB_SRC_FACTOR_MASK |
 		   IAB_DST_FACTOR_MASK |
@@ -261,7 +261,7 @@ static void i915BlendEquationSeparate(GLcontext *ctx, GLenum eqRGB,
 
 static void i915DepthFunc(GLcontext *ctx, GLenum func)
 {
-   i915ContextPtr i915 = I915_CONTEXT(ctx);
+   struct i915_context *i915 = I915_CONTEXT(ctx);
    int test = intel_translate_compare_func( func );
 
    if (INTEL_DEBUG&DEBUG_DRI)
@@ -274,7 +274,7 @@ static void i915DepthFunc(GLcontext *ctx, GLenum func)
 
 static void i915DepthMask(GLcontext *ctx, GLboolean flag)
 {
-   i915ContextPtr i915 = I915_CONTEXT(ctx);
+   struct i915_context *i915 = I915_CONTEXT(ctx);
 
    if (INTEL_DEBUG&DEBUG_DRI)
       fprintf(stderr, "%s flag (%d)\n", __FUNCTION__, flag);
@@ -295,7 +295,7 @@ static void i915DepthMask(GLcontext *ctx, GLboolean flag)
  */
 static void i915PolygonStipple( GLcontext *ctx, const GLubyte *mask )
 {
-   i915ContextPtr i915 = I915_CONTEXT(ctx);
+   struct i915_context *i915 = I915_CONTEXT(ctx);
    const GLubyte *m = mask;
    GLubyte p[4];
    int i,j,k;
@@ -348,7 +348,7 @@ static void i915PolygonStipple( GLcontext *ctx, const GLubyte *mask )
 static void i915Scissor(GLcontext *ctx, GLint x, GLint y, 
 			GLsizei w, GLsizei h)
 {
-   i915ContextPtr i915 = I915_CONTEXT(ctx);
+   struct i915_context *i915 = I915_CONTEXT(ctx);
    intelScreenPrivate *screen = i915->intel.intelScreen;
    int x1, y1, x2, y2;
 
@@ -382,7 +382,7 @@ static void i915Scissor(GLcontext *ctx, GLint x, GLint y,
 
 static void i915LogicOp(GLcontext *ctx, GLenum opcode)
 {
-   i915ContextPtr i915 = I915_CONTEXT(ctx);
+   struct i915_context *i915 = I915_CONTEXT(ctx);
    int tmp = intel_translate_logic_op(opcode);
 
    if (INTEL_DEBUG&DEBUG_DRI)
@@ -397,7 +397,7 @@ static void i915LogicOp(GLcontext *ctx, GLenum opcode)
 
 static void i915CullFaceFrontFace(GLcontext *ctx, GLenum unused)
 {
-   i915ContextPtr i915 = I915_CONTEXT(ctx);
+   struct i915_context *i915 = I915_CONTEXT(ctx);
    GLuint mode;
 
    if (INTEL_DEBUG&DEBUG_DRI)
@@ -425,7 +425,7 @@ static void i915CullFaceFrontFace(GLcontext *ctx, GLenum unused)
 
 static void i915LineWidth( GLcontext *ctx, GLfloat widthf )
 {
-   i915ContextPtr i915 = I915_CONTEXT( ctx );
+   struct i915_context *i915 = I915_CONTEXT( ctx );
    int lis4 = i915->state.Ctx[I915_CTXREG_LIS4] & ~S4_LINE_WIDTH_MASK;
    int width;
 
@@ -444,7 +444,7 @@ static void i915LineWidth( GLcontext *ctx, GLfloat widthf )
 
 static void i915PointSize(GLcontext *ctx, GLfloat size)
 {
-   i915ContextPtr i915 = I915_CONTEXT(ctx);
+   struct i915_context *i915 = I915_CONTEXT(ctx);
    int lis4 = i915->state.Ctx[I915_CTXREG_LIS4] & ~S4_POINT_WIDTH_MASK;
    GLint point_size = (int)size;
 
@@ -469,7 +469,7 @@ static void i915ColorMask(GLcontext *ctx,
 			 GLboolean r, GLboolean g,
 			 GLboolean b, GLboolean a)
 {
-   i915ContextPtr i915 = I915_CONTEXT( ctx );
+   struct i915_context *i915 = I915_CONTEXT( ctx );
    GLuint tmp = i915->state.Ctx[I915_CTXREG_LIS5] & ~S5_WRITEDISABLE_MASK;
 
    if (INTEL_DEBUG&DEBUG_DRI)
@@ -490,7 +490,7 @@ static void update_specular( GLcontext *ctx )
 {
    /* A hack to trigger the rebuild of the fragment program.
     */
-   INTEL_CONTEXT(ctx)->NewGLState |= _NEW_TEXTURE;
+   intel_context(ctx)->NewGLState |= _NEW_TEXTURE;
    I915_CONTEXT(ctx)->tex_program.translated = 0; 
 }
 
@@ -507,7 +507,7 @@ static void i915LightModelfv(GLcontext *ctx, GLenum pname,
 
 static void i915ShadeModel(GLcontext *ctx, GLenum mode)
 {
-   i915ContextPtr i915 = I915_CONTEXT(ctx);
+   struct i915_context *i915 = I915_CONTEXT(ctx);
    I915_STATECHANGE(i915, I915_UPLOAD_CTX);
 
    if (mode == GL_SMOOTH) {
@@ -526,7 +526,7 @@ static void i915ShadeModel(GLcontext *ctx, GLenum mode)
  */
 void i915_update_fog( GLcontext *ctx )
 {
-   i915ContextPtr i915 = I915_CONTEXT(ctx);
+   struct i915_context *i915 = I915_CONTEXT(ctx);
    GLenum mode;
    GLboolean enabled;
    GLboolean try_pixel_fog;
@@ -619,7 +619,7 @@ void i915_update_fog( GLcontext *ctx )
 
 static void i915Fogfv(GLcontext *ctx, GLenum pname, const GLfloat *param)
 {
-   i915ContextPtr i915 = I915_CONTEXT(ctx);
+   struct i915_context *i915 = I915_CONTEXT(ctx);
 
    switch (pname) {
    case GL_FOG_COORDINATE_SOURCE_EXT: 
@@ -671,7 +671,7 @@ static void i915Hint(GLcontext *ctx, GLenum target, GLenum state)
 
 static void i915Enable(GLcontext *ctx, GLenum cap, GLboolean state)
 {
-   i915ContextPtr i915 = I915_CONTEXT(ctx);
+   struct i915_context *i915 = I915_CONTEXT(ctx);
 
    switch(cap) {
    case GL_TEXTURE_2D:
@@ -785,7 +785,7 @@ static void i915Enable(GLcontext *ctx, GLenum cap, GLboolean state)
 }
 
 
-static void i915_init_packets( i915ContextPtr i915 )
+static void i915_init_packets( struct i915_context *i915 )
 {
    intelScreenPrivate *screen = i915->intel.intelScreen;
 
@@ -823,7 +823,6 @@ static void i915_init_packets( i915ContextPtr i915 )
 					   ENABLE_STENCIL_WRITE_MASK |
 					   STENCIL_WRITE_MASK(0xff));
 
-
       i915->state.Ctx[I915_CTXREG_IAB] = (_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD |
 					IAB_MODIFY_ENABLE |
 					IAB_MODIFY_FUNC |
@@ -862,19 +861,15 @@ static void i915_init_packets( i915ContextPtr i915 )
 	  BUF_3D_PITCH(screen->front.pitch * screen->cpp) |
 	  BUF_3D_USE_FENCE);
 
-
       i915->state.Buffer[I915_DESTREG_DBUFADDR0] = _3DSTATE_BUF_INFO_CMD;
       i915->state.Buffer[I915_DESTREG_DBUFADDR1] = 
 	 (BUF_3D_ID_DEPTH |
 	  BUF_3D_PITCH(screen->depth.pitch * screen->cpp) |
 	  BUF_3D_USE_FENCE);
-      i915->state.Buffer[I915_DESTREG_DBUFADDR2] = screen->depth.offset;
-
 
       i915->state.Buffer[I915_DESTREG_DV0] = _3DSTATE_DST_BUF_VARS_CMD;
 
       switch (screen->fbFormat) {
-      case DV_PF_555:
       case DV_PF_565:
 	 i915->state.Buffer[I915_DESTREG_DV1] = (DSTORG_HORT_BIAS(0x8) | /* .5 */
 					       DSTORG_VERT_BIAS(0x8) | /* .5 */
@@ -938,7 +933,7 @@ void i915InitStateFunctions( struct dd_function_table *functions )
 }
 
 
-void i915InitState( i915ContextPtr i915 )
+void i915InitState( struct i915_context *i915 )
 {
    GLcontext *ctx = &i915->intel.ctx;
 
diff --git a/src/mesa/drivers/dri/i915/i915_tex.c b/src/mesa/drivers/dri/i915/i915_tex.c
index 92bc94b8eef..7f47cf90445 100644
--- a/src/mesa/drivers/dri/i915/i915_tex.c
+++ b/src/mesa/drivers/dri/i915/i915_tex.c
@@ -48,7 +48,7 @@
 static void i915TexEnv( GLcontext *ctx, GLenum target, 
 			GLenum pname, const GLfloat *param )
 {
-   i915ContextPtr i915 = I915_CONTEXT( ctx );
+   struct i915_context *i915 = I915_CONTEXT( ctx );
 
    switch (pname) {
    case GL_TEXTURE_ENV_COLOR: 	/* Should be a tracked param */
diff --git a/src/mesa/drivers/dri/i915/i915_texprog.c b/src/mesa/drivers/dri/i915/i915_texprog.c
index 74ece96f8b7..e1ec7a6e8a1 100644
--- a/src/mesa/drivers/dri/i915/i915_texprog.c
+++ b/src/mesa/drivers/dri/i915/i915_texprog.c
@@ -536,7 +536,7 @@ static void emit_program_fini( struct i915_fragment_program *p )
 }
 
 
-static void i915EmitTextureProgram( i915ContextPtr i915 )
+static void i915EmitTextureProgram( struct i915_context *i915 )
 {
    GLcontext *ctx = &i915->intel.ctx;
    struct i915_fragment_program *p = &i915->tex_program;
@@ -570,9 +570,9 @@ static void i915EmitTextureProgram( i915ContextPtr i915 )
 }
 
 
-void i915ValidateTextureProgram( i915ContextPtr i915 )
+void i915ValidateTextureProgram( struct i915_context *i915 )
 {
-   intelContextPtr intel = &i915->intel;
+   struct intel_context *intel = &i915->intel;
    GLcontext *ctx = &intel->ctx;
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    struct vertex_buffer *VB = &tnl->vb;
diff --git a/src/mesa/drivers/dri/i915/i915_texstate.c b/src/mesa/drivers/dri/i915/i915_texstate.c
index f1db5f646cd..f498bcdea82 100644
--- a/src/mesa/drivers/dri/i915/i915_texstate.c
+++ b/src/mesa/drivers/dri/i915/i915_texstate.c
@@ -110,19 +110,31 @@ static GLuint translate_wrap_mode( GLenum wrap )
  * efficient, but this has gotten complex enough that we need
  * something which is understandable and reliable.
  */
-static GLboolean i915_update_tex_unit( GLcontext *ctx,
+static GLboolean i915_update_tex_unit( struct intel_context *intel,
 				       GLuint unit, 
 				       GLuint ss3 )
 {
+   GLcontext *ctx = &intel->ctx;
+   struct i915_context *i915 = i915_context(ctx);
    struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current;
    struct intel_texture_object *intelObj = intel_texture_object(tObj);
-   struct i915_context *i915 = i915_context(ctx);
-   GLuint state[I915_TEX_SETUP_SIZE];
    struct gl_texture_image *firstImage = tObj->Image[0][intelObj->firstLevel];
+   GLuint *state = i915->state.Tex[unit];
 
    memset(state, 0, sizeof(state));
 
-   state[I915_TEXREG_MS2] = 0;	/* will fixup later */
+   intel_region_release(intel, &i915->state.tex_region[unit]);
+
+   if (!intel_finalize_mipmap_tree(intel, unit))
+      return GL_FALSE;   
+
+   intel_region_reference(&i915->state.tex_region[unit],
+			  intelObj->mt->region);
+
+   i915->state.tex_offset[unit] = intel_miptree_image_offset(intelObj->mt, 0,
+							     intelObj->firstLevel); 
+
+
    state[I915_TEXREG_MS3] = 
       (((firstImage->Height - 1) << MS3_HEIGHT_SHIFT) |
        ((firstImage->Width - 1) << MS3_WIDTH_SHIFT) |
@@ -253,11 +265,11 @@ static GLboolean i915_update_tex_unit( GLcontext *ctx,
    
 
    I915_ACTIVESTATE(i915, I915_UPLOAD_TEX(unit), GL_TRUE);
+   /* memcmp was already disabled, but definitely won't work as the
+    * region might now change and that wouldn't be detected:
+    */
+			  I915_STATECHANGE( i915, I915_UPLOAD_TEX(unit) );
 
-   if (1 || memcmp(state, i915->state.Tex[unit], sizeof(state)) != 0) {
-      I915_STATECHANGE( i915, I915_UPLOAD_TEX(unit) );
-      memcpy(i915->state.Tex[unit], state, sizeof(state));
-   }
 
 #if 0
    DBG(TEXTURE, "state[I915_TEXREG_SS2] = 0x%x\n", state[I915_TEXREG_SS2]);
@@ -274,25 +286,24 @@ static GLboolean i915_update_tex_unit( GLcontext *ctx,
 
 
 
-void i915UpdateTextureState( intelContextPtr intel )
+void i915UpdateTextureState( struct intel_context *intel )
 {
-   GLcontext *ctx = &intel->ctx;
    GLboolean ok = GL_TRUE;
    GLuint i;
 
    for (i = 0 ; i < I915_TEX_UNITS && ok ; i++) {
-      switch (ctx->Texture.Unit[i]._ReallyEnabled) {
+      switch (intel->ctx.Texture.Unit[i]._ReallyEnabled) {
       case TEXTURE_1D_BIT:
       case TEXTURE_2D_BIT:
       case TEXTURE_CUBE_BIT:
-	 ok = i915_update_tex_unit( ctx, i, SS3_NORMALIZED_COORDS );
+	 ok = i915_update_tex_unit( intel, i, SS3_NORMALIZED_COORDS );
 	 break;
       case TEXTURE_RECT_BIT:
       case TEXTURE_3D_BIT:
-	 ok = i915_update_tex_unit( ctx, i, 0 );
+	 ok = i915_update_tex_unit( intel, i, 0 );
 	 break;
       case 0: {
-	 struct i915_context *i915 = i915_context(ctx);
+	 struct i915_context *i915 = i915_context(&intel->ctx);
 	 if (i915->state.active & I915_UPLOAD_TEX(i)) 
 	    I915_ACTIVESTATE(i915, I915_UPLOAD_TEX(i), GL_FALSE);
 	 break;
diff --git a/src/mesa/drivers/dri/i915/i915_vtbl.c b/src/mesa/drivers/dri/i915/i915_vtbl.c
index 4f4792c0eed..eb6af9056c3 100644
--- a/src/mesa/drivers/dri/i915/i915_vtbl.c
+++ b/src/mesa/drivers/dri/i915/i915_vtbl.c
@@ -38,14 +38,15 @@
 
 #include "intel_batchbuffer.h"
 #include "intel_tex.h"
+#include "intel_regions.h"
 
 #include "i915_reg.h"
 #include "i915_context.h"
 
-static void i915_render_start( intelContextPtr intel )
+static void i915_render_start( struct intel_context *intel )
 {
    GLcontext *ctx = &intel->ctx;
-   i915ContextPtr i915 = i915_context(&intel->ctx);
+   struct i915_context *i915 = i915_context(&intel->ctx);
 
    if (ctx->FragmentProgram._Active) 
       i915ValidateFragmentProgram( i915 );
@@ -54,42 +55,42 @@ static void i915_render_start( intelContextPtr intel )
 }
 
 
-static void i915_reduced_primitive_state( intelContextPtr intel,
+static void i915_reduced_primitive_state( struct intel_context *intel,
 					  GLenum rprim )
 {
-   i915ContextPtr i915 = i915_context(&intel->ctx);
-    GLuint st1 = i915->state.Stipple[I915_STPREG_ST1];
-
-    st1 &= ~ST1_ENABLE;
-
-    switch (rprim) {
-    case GL_TRIANGLES:
-       if (intel->ctx.Polygon.StippleFlag &&
-	   intel->hw_stipple)
-	  st1 |= ST1_ENABLE;
-       break;
-    case GL_LINES:
-    case GL_POINTS:
-    default:
-       break;
-    }
-
-    i915->intel.reduced_primitive = rprim;
-
-    if (st1 != i915->state.Stipple[I915_STPREG_ST1]) {
-       I915_STATECHANGE(i915, I915_UPLOAD_STIPPLE);
-       i915->state.Stipple[I915_STPREG_ST1] = st1;
-    }
+   struct i915_context *i915 = i915_context(&intel->ctx);
+   GLuint st1 = i915->state.Stipple[I915_STPREG_ST1];
+
+   st1 &= ~ST1_ENABLE;
+
+   switch (rprim) {
+   case GL_TRIANGLES:
+      if (intel->ctx.Polygon.StippleFlag &&
+	  intel->hw_stipple)
+	 st1 |= ST1_ENABLE;
+      break;
+   case GL_LINES:
+   case GL_POINTS:
+   default:
+      break;
+   }
+
+   i915->intel.reduced_primitive = rprim;
+
+   if (st1 != i915->state.Stipple[I915_STPREG_ST1]) {
+      I915_STATECHANGE(i915, I915_UPLOAD_STIPPLE);
+      i915->state.Stipple[I915_STPREG_ST1] = st1;
+   }
 }
 
 
 /* Pull apart the vertex format registers and figure out how large a
  * vertex is supposed to be. 
  */
-static GLboolean i915_check_vertex_size( intelContextPtr intel,
+static GLboolean i915_check_vertex_size( struct intel_context *intel,
 					 GLuint expected )
 {
-   i915ContextPtr i915 = i915_context(&intel->ctx);
+   struct i915_context *i915 = i915_context(&intel->ctx);
    int lis2 = i915->current->Ctx[I915_CTXREG_LIS2];
    int lis4 = i915->current->Ctx[I915_CTXREG_LIS4];
    int i, sz = 0;
@@ -133,11 +134,11 @@ static GLboolean i915_check_vertex_size( intelContextPtr intel,
 }
 
 
-static void i915_emit_invarient_state( intelContextPtr intel )
+static void i915_emit_invarient_state( struct intel_context *intel )
 {
    BATCH_LOCALS;
 
-   BEGIN_BATCH( 200 );
+   BEGIN_BATCH( 200, 0 );
 
    OUT_BATCH(_3DSTATE_AA_CMD |
 	     AA_LINE_ECAAR_WIDTH_ENABLE |
@@ -205,21 +206,15 @@ static void i915_emit_invarient_state( intelContextPtr intel )
 }
 
 
-#define emit( intel, state, size )			\
-do {							\
-   int k;						\
-   BEGIN_BATCH( (size) / sizeof(GLuint));		\
-   for (k = 0 ; k < (size) / sizeof(GLuint) ; k++)	\
-      OUT_BATCH((state)[k]);				\
-   ADVANCE_BATCH();					\
-} while (0);
+#define emit(intel, state, size )		     \
+   intel_batchbuffer_data(intel->batch, state, size, 0 )
 
 
 /* Push the state into the sarea and/or texture memory.
  */
-static void i915_emit_state( intelContextPtr intel )
+static void i915_emit_state( struct intel_context *intel )
 {
-   i915ContextPtr i915 = i915_context(&intel->ctx);
+   struct i915_context *i915 = i915_context(&intel->ctx);
    struct i915_hw_state *state = i915->current;
    int i;
    GLuint dirty;
@@ -234,32 +229,48 @@ static void i915_emit_state( intelContextPtr intel )
    dirty = state->active & ~state->emitted;
 
 
-   if (VERBOSE) 
+   if (INTEL_DEBUG & DEBUG_STATE) 
       fprintf(stderr, "%s dirty: %x\n", __FUNCTION__, dirty);
 
    if (dirty & I915_UPLOAD_INVARIENT) {
-      if (VERBOSE) fprintf(stderr, "I915_UPLOAD_INVARIENT:\n"); 
+      if (INTEL_DEBUG & DEBUG_STATE) fprintf(stderr, "I915_UPLOAD_INVARIENT:\n"); 
       i915_emit_invarient_state( intel );
    }
 
    if (dirty & I915_UPLOAD_CTX) {
-      if (VERBOSE) fprintf(stderr, "I915_UPLOAD_CTX:\n"); 
-      emit( i915, state->Ctx, sizeof(state->Ctx) );
+      if (INTEL_DEBUG & DEBUG_STATE) fprintf(stderr, "I915_UPLOAD_CTX:\n"); 
+      emit(intel, state->Ctx, sizeof(state->Ctx) );
    }
 
    if (dirty & I915_UPLOAD_BUFFERS) {
-      if (VERBOSE) fprintf(stderr, "I915_UPLOAD_BUFFERS:\n"); 
-      emit( i915, state->Buffer, sizeof(state->Buffer) );
+      if (INTEL_DEBUG & DEBUG_STATE) fprintf(stderr, "I915_UPLOAD_BUFFERS:\n"); 
+
+      BEGIN_BATCH(I915_DEST_SETUP_SIZE, 0);
+      OUT_BATCH(state->Buffer[I915_DESTREG_CBUFADDR0]);
+      OUT_BATCH(state->Buffer[I915_DESTREG_CBUFADDR1]);
+      OUT_RELOC(state->draw_region->buffer, BM_MEM_AGP|BM_WRITE, 0);
+
+      OUT_BATCH(state->Buffer[I915_DESTREG_DBUFADDR0]);
+      OUT_BATCH(state->Buffer[I915_DESTREG_DBUFADDR1]);
+      OUT_RELOC(state->depth_region->buffer, BM_MEM_AGP|BM_WRITE, 0);
+
+      OUT_BATCH(state->Buffer[I915_DESTREG_DV0]);
+      OUT_BATCH(state->Buffer[I915_DESTREG_DV1]);
+      OUT_BATCH(state->Buffer[I915_DESTREG_SENABLE]);
+      OUT_BATCH(state->Buffer[I915_DESTREG_SR0]);
+      OUT_BATCH(state->Buffer[I915_DESTREG_SR1]);
+      OUT_BATCH(state->Buffer[I915_DESTREG_SR2]);
+      ADVANCE_BATCH();
    }
 
    if (dirty & I915_UPLOAD_STIPPLE) {
-      if (VERBOSE) fprintf(stderr, "I915_UPLOAD_STIPPLE:\n"); 
-      emit( i915, state->Stipple, sizeof(state->Stipple) );
+      if (INTEL_DEBUG & DEBUG_STATE) fprintf(stderr, "I915_UPLOAD_STIPPLE:\n"); 
+      emit(intel, state->Stipple, sizeof(state->Stipple) );
    }
 
    if (dirty & I915_UPLOAD_FOG) {
-      if (VERBOSE) fprintf(stderr, "I915_UPLOAD_FOG:\n"); 
-      emit( i915, state->Fog, sizeof(state->Fog) );
+      if (INTEL_DEBUG & DEBUG_STATE) fprintf(stderr, "I915_UPLOAD_FOG:\n"); 
+      emit(intel, state->Fog, sizeof(state->Fog) );
    }
 
    /* Combine all the dirty texture state into a single command to
@@ -272,20 +283,29 @@ static void i915_emit_state( intelContextPtr intel )
 	 if (dirty & I915_UPLOAD_TEX(i)) 
 	    nr++;
 
-      BEGIN_BATCH(2+nr*3);
+      BEGIN_BATCH(2+nr*3, 0);
       OUT_BATCH(_3DSTATE_MAP_STATE | (3*nr));
       OUT_BATCH((dirty & I915_UPLOAD_TEX_ALL) >> I915_UPLOAD_TEX_0_SHIFT);
       for (i = 0 ; i < I915_TEX_UNITS ; i++)
 	 if (dirty & I915_UPLOAD_TEX(i)) {
-	    /* Emit zero texture offset, will fixup before firing */
-	    intel_add_texoffset_fixup(intel, i, (GLuint *)batch_ptr); 
-	    batch_ptr += 4;
+
+	    if (state->tex_region[i]) {
+	       OUT_RELOC(state->tex_region[i]->buffer,
+			 BM_MEM_AGP|BM_READ,
+			 state->tex_offset[i]);
+	    }
+	    else {
+	       assert(i == 0);
+	       assert(state == &i915->meta);
+	       OUT_BATCH(0);
+	    }
+
 	    OUT_BATCH(state->Tex[i][I915_TEXREG_MS3]);
 	    OUT_BATCH(state->Tex[i][I915_TEXREG_MS4]);
 	 }
       ADVANCE_BATCH();
 
-      BEGIN_BATCH(2+nr*3);
+      BEGIN_BATCH(2+nr*3, 0);
       OUT_BATCH(_3DSTATE_SAMPLER_STATE | (3*nr));
       OUT_BATCH((dirty & I915_UPLOAD_TEX_ALL) >> I915_UPLOAD_TEX_0_SHIFT);
       for (i = 0 ; i < I915_TEX_UNITS ; i++)
@@ -298,63 +318,69 @@ static void i915_emit_state( intelContextPtr intel )
    }
 
    if (dirty & I915_UPLOAD_CONSTANTS) {
-      if (VERBOSE) fprintf(stderr, "I915_UPLOAD_CONSTANTS:\n"); 
-      emit( i915, state->Constant, state->ConstantSize * sizeof(GLuint) );
+      if (INTEL_DEBUG & DEBUG_STATE) fprintf(stderr, "I915_UPLOAD_CONSTANTS:\n"); 
+      emit(intel, state->Constant, state->ConstantSize * sizeof(GLuint) );
    }
 
    if (dirty & I915_UPLOAD_PROGRAM) {
-      if (VERBOSE) fprintf(stderr, "I915_UPLOAD_PROGRAM:\n"); 
+      if (INTEL_DEBUG & DEBUG_STATE) fprintf(stderr, "I915_UPLOAD_PROGRAM:\n"); 
 
       assert((state->Program[0] & 0x1ff)+2 == state->ProgramSize);
       
-      emit( i915, state->Program, state->ProgramSize * sizeof(GLuint) );
-      if (VERBOSE)
+      emit(intel, state->Program, state->ProgramSize * sizeof(GLuint) );
+      if (INTEL_DEBUG & DEBUG_STATE)
 	 i915_disassemble_program( state->Program, state->ProgramSize );
    }
 
    state->emitted |= dirty;
 }
 
-static void i915_destroy_context( intelContextPtr intel )
+static void i915_destroy_context( struct intel_context *intel )
 {
    _tnl_free_vertices(&intel->ctx);
 }
 
-static void i915_set_draw_offset( intelContextPtr intel, int offset )
+static void i915_set_draw_region( struct intel_context *intel, 
+				  struct intel_region *draw_region,
+				  struct intel_region *depth_region)
 {
-   i915ContextPtr i915 = i915_context(&intel->ctx);
+   struct i915_context *i915 = i915_context(&intel->ctx);
+
+   intel_region_release(intel, &i915->state.draw_region);
+   intel_region_release(intel, &i915->state.depth_region);
+   intel_region_reference(&i915->state.draw_region, draw_region);
+   intel_region_reference(&i915->state.depth_region, depth_region);
+
    I915_STATECHANGE( i915, I915_UPLOAD_BUFFERS );
-   i915->state.Buffer[I915_DESTREG_CBUFADDR2] = offset;
 }
 
-static void i915_lost_hardware( intelContextPtr intel )
+static void i915_lost_hardware( struct intel_context *intel )
 {
-   i915ContextPtr i915 = i915_context(&intel->ctx);
+   struct i915_context *i915 = i915_context(&intel->ctx);
    i915->state.emitted = 0;
 }
 
-static void i915_emit_flush( intelContextPtr intel )
+static void i915_emit_flush( struct intel_context *intel )
 {
    BATCH_LOCALS;
 
-   BEGIN_BATCH(2);
+   BEGIN_BATCH(2, 0);
    OUT_BATCH( MI_FLUSH | FLUSH_MAP_CACHE | FLUSH_RENDER_CACHE ); 
    OUT_BATCH( 0 );
    ADVANCE_BATCH();
 }
 
 
-void i915InitVtbl( i915ContextPtr i915 )
+void i915InitVtbl( struct i915_context *i915 )
 {
    i915->intel.vtbl.check_vertex_size = i915_check_vertex_size;
-   i915->intel.vtbl.clear_with_tris = i915ClearWithTris;
    i915->intel.vtbl.destroy = i915_destroy_context;
    i915->intel.vtbl.emit_invarient_state = i915_emit_invarient_state;
    i915->intel.vtbl.emit_state = i915_emit_state;
    i915->intel.vtbl.lost_hardware = i915_lost_hardware;
    i915->intel.vtbl.reduced_primitive_state = i915_reduced_primitive_state;
    i915->intel.vtbl.render_start = i915_render_start;
-   i915->intel.vtbl.set_draw_offset = i915_set_draw_offset;
+   i915->intel.vtbl.set_draw_region = i915_set_draw_region;
    i915->intel.vtbl.update_texture_state = i915UpdateTextureState;
    i915->intel.vtbl.emit_flush = i915_emit_flush;
 }
diff --git a/src/mesa/drivers/dri/i915/intel_batchbuffer.c b/src/mesa/drivers/dri/i915/intel_batchbuffer.c
index 91fbaed4da5..fd04b054abe 100644
--- a/src/mesa/drivers/dri/i915/intel_batchbuffer.c
+++ b/src/mesa/drivers/dri/i915/intel_batchbuffer.c
@@ -1,6 +1,6 @@
 /**************************************************************************
  * 
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
  * All Rights Reserved.
  * 
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -25,79 +25,263 @@
  * 
  **************************************************************************/
 
+#include "intel_batchbuffer.h"
+#include "intel_ioctl.h"
+#include "bufmgr.h"
 
-#include <stdio.h>
-#include <errno.h>
+/* Relocations in kernel space:
+ *    - pass dma buffer seperately
+ *    - memory manager knows how to patch
+ *    - pass list of dependent buffers
+ *    - pass relocation list
+ *
+ * Either:
+ *    - get back an offset for buffer to fire
+ *    - memory manager knows how to fire buffer
+ *
+ * Really want the buffer to be AGP and pinned.
+ *
+ */
 
-#include "mtypes.h"
-#include "context.h"
-#include "enums.h"
+/* Cliprect fence: The highest fence protecting a dma buffer
+ * containing explicit cliprect information.  Like the old drawable
+ * lock but irq-driven.  X server must wait for this fence to expire
+ * before changing cliprects [and then doing sw rendering?].  For
+ * other dma buffers, the scheduler will grab current cliprect info
+ * and mix into buffer.  X server must hold the lock while changing
+ * cliprects???  Make per-drawable.  Need cliprects in shared memory
+ * -- beats storing them with every cmd buffer in the queue.
+ *
+ * ==> X server must wait for this fence to expire before touching the
+ * framebuffer with new cliprects.
+ *
+ * ==> Cliprect-dependent buffers associated with a
+ * cliprect-timestamp.  All of the buffers associated with a timestamp
+ * must go to hardware before any buffer with a newer timestamp.
+ *
+ * ==> Dma should be queued per-drawable for correct X/GL
+ * synchronization.  Or can fences be used for this?
+ *
+ * Applies to: Blit operations, metaops, X server operations -- X
+ * server automatically waits on its own dma to complete before
+ * modifying cliprects ???
+ */				
 
-#include "intel_reg.h"
-#include "intel_batchbuffer.h"
-#include "intel_context.h"
+static void intel_dump_batchbuffer( unsigned offset,
+				    int *ptr,
+				    int count )
+{
+   int i;
+   fprintf(stderr, "\n\n\nSTART BATCH (%d dwords):\n", count/4);
+   for (i = 0; i < count/4; i += 4) 
+      fprintf(stderr, "0x%x:\t0x%08x 0x%08x 0x%08x 0x%08x\n", 
+	      offset + i*4, ptr[i], ptr[i+1], ptr[i+2], ptr[i+3]);
+   fprintf(stderr, "END BATCH\n\n\n");
+}
 
-#include "bufmgr.h"
 
+static void intel_batchbuffer_reset( struct intel_batchbuffer *batch )
+{
+   bmBufferData(batch->bm,
+		batch->buffer,
+		BATCH_SZ,
+		NULL,
+		0);
+		
+   if (!batch->list) 
+      batch->list = bmNewBufferList();
+
+   batch->list->nr = 0;
+   batch->nr_relocs = 0;
+   batch->flags = 0;
 
+   bmAddBuffer( batch->list,
+		batch->buffer,
+		0,
+		NULL,
+		&batch->offset[batch->list->nr]);
 
-void intelDestroyBatchBuffer( struct intel_context *intel )
+   batch->map = bmMapBuffer(batch->bm, batch->buffer, 
+			    BM_MEM_AGP|BM_MEM_LOCAL|BM_CLIENT|BM_WRITE);
+   batch->ptr = batch->map;
+}
+
+/*======================================================================
+ * Public functions
+ */
+struct intel_batchbuffer *intel_batchbuffer_alloc( struct intel_context *intel )
 {
+   struct intel_batchbuffer *batch = calloc(sizeof(*batch), 1);
+
+   batch->intel = intel;
+   batch->bm = intel->bm;
+
+   bmGenBuffers(intel->bm, 1, &batch->buffer);
+   intel_batchbuffer_reset( batch );
+   return batch;
+}
+
+void intel_batchbuffer_free( struct intel_batchbuffer *batch )
+{
+   if (batch->map)
+      bmUnmapBuffer(batch->bm, batch->buffer);
+   
+   free(batch);
+}
+
+/* TODO: Push this whole function into bufmgr.
+ */
+static void do_flush_locked( struct intel_batchbuffer *batch,
+			     GLuint used,
+			     GLboolean ignore_cliprects)
+{
+   GLuint *ptr;
+   GLuint i;
+
+   bmValidateBufferList( batch->bm, 
+			 batch->list,
+			 BM_MEM_AGP );
+
+   /* Apply the relocations.  This nasty map indicates to me that the
+    * whole task should be done internally by the memory manager, and
+    * that dma buffers probably need to be pinned within agp space.
+    */
+   ptr = (GLuint *)bmMapBuffer(batch->bm, batch->buffer, 
+			       BM_NO_MOVE|BM_NO_UPLOAD|
+			       BM_NO_EVICT|BM_MEM_AGP|
+			       BM_WRITE);
+
+   for (i = 0; i < batch->nr_relocs; i++) {
+      struct buffer_reloc *r = &batch->reloc[i];
+
+      DBG("apply fixup at offset 0x%x, elem %d (buf %d, offset 0x%x), delta 0x%x\n",
+	  r->offset, r->elem, batch->list->elem[r->elem].buffer,
+	  batch->offset[r->elem], r->delta);
+
+      ptr[r->offset/4] = batch->offset[r->elem] + r->delta;
+   }
+
+   if (0)
+      intel_dump_batchbuffer( 0, ptr, used );
+
+
+
+   bmUnmapBuffer(batch->bm, batch->buffer);
+
+
+   /* Fire the batch buffer, which was uploaded above:
+    */
+   intel_batch_ioctl(batch->intel, 
+		     batch->offset[0],
+		     used,
+		     ignore_cliprects);
+   
+   batch->last_fence = bmFenceBufferList(batch->bm, batch->list);
 }
 
 
-void intelInstallBatchBuffer( struct intel_context *intel )
+
+GLuint intel_batchbuffer_flush( struct intel_batchbuffer *batch )
 {
-   assert(!intel->batch.ptr);
+   struct intel_context *intel = batch->intel;
+   GLuint used = batch->ptr - batch->map;
 
-   intel->alloc.current++;
-   intel->alloc.current %= INTEL_ALLOC_NR;
+   if (used == 0) 
+      return batch->last_fence;
 
-   DBG("%s: %d\n", __FUNCTION__, intel->alloc.current);
+   /* Add the MI_BATCH_BUFFER_END: 
+    */
+   if (intel_batchbuffer_space(batch) & 4) {
+      ((int *)batch->ptr)[0] = MI_BATCH_BUFFER_END;
+      used += 4;
+   }
+   else {
+      ((int *)batch->ptr)[0] = 0;
+      ((int *)batch->ptr)[1] = MI_BATCH_BUFFER_END;
+      used += 8;
+   }
 
-   intel->batch.size = INTEL_ALLOC_SIZE;
-   intel->batch.space = intel->batch.size;
-   intel->batch.start_offset = 0;
+   bmUnmapBuffer(batch->bm, batch->buffer);
+   batch->ptr = NULL;
+   batch->map = NULL;
 
-   intel->batch.ptr = bmMapBuffer( intel->bm, 
-				   intel->alloc.buffer[intel->alloc.current],
-				   BM_WRITE | BM_MEM_AGP );
+   /* TODO: Just pass the relocation list and dma buffer up to the
+    * kernel.
+    */
+   if (!intel->locked)
+   {
+      assert(!(batch->flags & INTEL_BATCH_NO_CLIPRECTS));
 
+      LOCK_HARDWARE(intel);
+      do_flush_locked(batch, used, GL_FALSE);
+      UNLOCK_HARDWARE(intel);
+   }
+   else {
+      assert(!(batch->flags & INTEL_BATCH_CLIPRECTS));
 
-   assert(!intel->buffer_list);
-   intel->buffer_list = bmNewBufferList();
-      
-   /* Add the batchbuffer 
+      do_flush_locked(batch, used, GL_TRUE);
+   }
+
+   /* Reset the buffer:
     */
-   bmAddBuffer(intel->buffer_list,
-	       intel->alloc.buffer[intel->alloc.current],
-	       BM_READ,
-	       NULL,
-	       &intel->batch.start_offset);
+   intel_batchbuffer_reset( batch );
+   return batch->last_fence;
+}
+
+void intel_batchbuffer_finish( struct intel_batchbuffer *batch )
+{   
+   bmFinishFence(batch->bm, 
+		 intel_batchbuffer_flush(batch));
 }
+   
 
-void intelInitBatchBuffer( struct intel_context *intel )
+/*  This is the only way buffers get added to the validate list.
+ */
+GLboolean intel_batchbuffer_emit_reloc( struct intel_batchbuffer *batch,
+					GLuint buffer,
+					GLuint flags,
+					GLuint delta )
 {
-   GLint i;
-
-   _mesa_printf("%s: %d\n", __FUNCTION__, intel->alloc.current);
-   bmGenBuffers(intel->bm,
-		INTEL_ALLOC_NR,
-		intel->alloc.buffer);
-
-   for (i = 0; i < INTEL_ALLOC_NR; i++)
-      bmBufferData(intel->bm, 
-		   intel->alloc.buffer[i],
-		   INTEL_ALLOC_SIZE,
-		   NULL,
-		   BM_MEM_AGP);
-		
+   GLuint i;
+
+   assert(batch->nr_relocs <= MAX_RELOCS);
+
+   for (i = 0; i < batch->list->nr; i++) 
+      if (buffer == batch->list->elem[i].buffer)
+	 break;
 
+   if (i == batch->list->nr) {
+      if (i == BM_LIST_MAX)
+	 return GL_FALSE;
+
+      bmAddBuffer(batch->list,
+		  buffer,
+		  flags,
+		  NULL,
+		  &batch->offset[i]);
+   }
+
+   {
+      struct buffer_reloc *r = &batch->reloc[batch->nr_relocs++];
+      r->offset = batch->ptr - batch->map;
+      r->delta = delta;
+      r->elem = i;
+   }
+
+   batch->ptr += 4;
+   return GL_TRUE;
 }
 
 
-void intelValidateBuffers( struct intel_context *intel )
+
+void intel_batchbuffer_data(struct intel_batchbuffer *batch,
+			    const void *data,
+			    GLuint bytes,
+			    GLuint flags)
 {
-   if (!bmValidateBufferList(intel->bm, intel->buffer_list, BM_MEM_AGP))
-      assert(0);
+   assert((bytes & 3) == 0);
+   intel_batchbuffer_require_space(batch, bytes, flags);
+   __memcpy(batch->ptr, data, bytes);
+   batch->ptr += bytes;
 }
+
diff --git a/src/mesa/drivers/dri/i915/intel_batchbuffer.h b/src/mesa/drivers/dri/i915/intel_batchbuffer.h
index 00ca9920cb1..e451a0e7284 100644
--- a/src/mesa/drivers/dri/i915/intel_batchbuffer.h
+++ b/src/mesa/drivers/dri/i915/intel_batchbuffer.h
@@ -1,74 +1,111 @@
-/**************************************************************************
- * 
- * 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_BATCHBUFFER_H
 #define INTEL_BATCHBUFFER_H
 
-#include "intel_context.h"
-#include "intel_ioctl.h"
+#include "mtypes.h"
+#include "bufmgr.h"
+
+struct intel_context;
+
+#define BATCH_SZ 4096
+#define MAX_RELOCS 100
+
+#define INTEL_BATCH_NO_CLIPRECTS 0x1
+#define INTEL_BATCH_CLIPRECTS    0x2
+
+struct buffer_reloc {
+   GLuint offset;
+   GLuint elem;			/* elem in buffer list, not buffer id */
+   GLuint delta;		/* not needed? */
+};
+
+struct intel_batchbuffer {
+   struct bufmgr *bm;
+   struct intel_context *intel;
+
+   GLuint buffer;
+   GLuint last_fence;
+   GLuint flags;
+
+   /* In progress:
+    */
+   GLuint offset[BM_LIST_MAX];
+   struct bm_buffer_list *list;   
+   GLubyte *map;
+   GLubyte *ptr; 
+
+   struct buffer_reloc reloc[MAX_RELOCS];
+   GLuint nr_relocs;
+};
+
+struct intel_batchbuffer *intel_batchbuffer_alloc( struct intel_context *intel );
+
+void intel_batchbuffer_free( struct intel_batchbuffer *batch );
+
+
+void intel_batchbuffer_finish( struct intel_batchbuffer *batch );
 
+GLuint intel_batchbuffer_flush( struct intel_batchbuffer *batch );
 
-#define BATCH_LOCALS	GLubyte *batch_ptr;
 
-#define VERBOSE 0
+/* Unlike bmBufferData, this currently requires the buffer be mapped.
+ * Consider it a convenience function wrapping multple
+ * intel_buffer_dword() calls.
+ */
+void intel_batchbuffer_data(struct intel_batchbuffer *batch,
+			    const void *data,
+			    GLuint bytes,
+			    GLuint flags);
 
+void intel_batchbuffer_release_space(struct intel_batchbuffer *batch,
+				   GLuint bytes);
 
-#define BEGIN_BATCH(n)							\
-do {									\
-   if (VERBOSE) fprintf(stderr, 					\
-			"BEGIN_BATCH(%d) in %s, %d dwords free\n",	\
-			(n), __FUNCTION__, intel->batch.space/4);	\
-   assert(intel->locked); \
-   if (intel->batch.space < (n)*4)					\
-      intelFlushBatch(intel, GL_TRUE);					\
-   batch_ptr = intel->batch.ptr;					\
-} while (0)
+GLboolean intel_batchbuffer_emit_reloc( struct intel_batchbuffer *batch,
+					GLuint buffer,
+					GLuint flags,
+					GLuint offset );
 
-#define OUT_BATCH(n)					\
-do {							\
-   *(GLuint *)batch_ptr = (n);				\
-   if (VERBOSE) fprintf(stderr, " -- %08x at %s/%d\n", (n), __FILE__, __LINE__);	\
-   batch_ptr += 4;					\
-} while (0)
+/* Inline functions - might actually be better off with these
+ * non-inlined.  Certainly better off switching all command packets to
+ * be passed as structs rather than dwords, but that's a little bit of
+ * work...
+ */
+static inline GLuint 
+intel_batchbuffer_space( struct intel_batchbuffer *batch )
+{
+   return (BATCH_SZ - 8) - (batch->ptr - batch->map);
+}
 
-#define ADVANCE_BATCH()						\
-do {								\
-   if (VERBOSE) fprintf(stderr, "ADVANCE_BATCH()\n");		\
-   intel->batch.space -= (batch_ptr - intel->batch.ptr);	\
-   intel->batch.ptr = batch_ptr;				\
-   assert(intel->batch.space >= 0);				\
-} while(0)
 
-extern void intelInitBatchBuffer( struct intel_context *intel );
-extern void intelDestroyBatchBuffer( struct intel_context *intel );
+static inline void 
+intel_batchbuffer_emit_dword(struct intel_batchbuffer *batch,
+			     GLuint dword)
+{
+   assert(batch->map);
+   assert(intel_batchbuffer_space(batch) >= 4);
+   *(GLuint *)(batch->ptr) = dword;
+   batch->ptr += 4;
+}
 
-void intelInstallBatchBuffer( struct intel_context *intel );
+static inline void 
+intel_batchbuffer_require_space(struct intel_batchbuffer *batch,
+				GLuint sz,
+				GLuint flags)
+{
+   assert(sz < BATCH_SZ - 8);
+   if (intel_batchbuffer_space(batch) < sz ||
+       (batch->flags != 0 && flags != 0 && batch->flags != flags))
+      intel_batchbuffer_flush(batch);
+   
+   batch->flags |= flags;
+}
 
+/* Here are the crusty old macros, to be removed:
+ */
+#define BATCH_LOCALS 
+#define BEGIN_BATCH(n, flags) intel_batchbuffer_require_space(intel->batch, n*4, flags)
+#define OUT_BATCH(d)  intel_batchbuffer_emit_dword(intel->batch, d)
+#define OUT_RELOC(buf,flags,delta) intel_batchbuffer_emit_reloc(intel->batch, buf, flags, delta)
+#define ADVANCE_BATCH() do { } while(0)
 
-void intelValidateBuffers( struct intel_context *intel );
 
 #endif
diff --git a/src/mesa/drivers/dri/i915/intel_blit.c b/src/mesa/drivers/dri/i915/intel_blit.c
index 47786a48f9b..147713dc349 100644
--- a/src/mesa/drivers/dri/i915/intel_blit.c
+++ b/src/mesa/drivers/dri/i915/intel_blit.c
@@ -37,6 +37,7 @@
 #include "intel_batchbuffer.h"
 #include "intel_context.h"
 #include "intel_blit.h"
+#include "intel_regions.h"
 
 #include "bufmgr.h"
 
@@ -48,7 +49,7 @@
  */
 void intelCopyBuffer( const __DRIdrawablePrivate *dPriv ) 
 {
-   intelContextPtr intel;
+   struct intel_context *intel;
 
    if (0)
       fprintf(stderr, "%s\n", __FUNCTION__);
@@ -57,13 +58,13 @@ void intelCopyBuffer( const __DRIdrawablePrivate *dPriv )
    assert(dPriv->driContextPriv);
    assert(dPriv->driContextPriv->driverPrivate);
 
-   intel = (intelContextPtr) dPriv->driContextPriv->driverPrivate;
-
+   intel = (struct intel_context *) dPriv->driContextPriv->driverPrivate;
    intelFlush( &intel->ctx );
-   LOCK_HARDWARE( intel );
-   intelInstallBatchBuffer(intel);   
-   intelValidateBuffers( intel );
 
+   /* The LOCK_HARDWARE is required for the cliprects.  Buffer offsets
+    * should work regardless.
+    */
+   LOCK_HARDWARE( intel );
    {
       intelScreenPrivate *intelScreen = intel->intelScreen;
       __DRIdrawablePrivate *dPriv = intel->driDrawable;
@@ -71,24 +72,18 @@ void intelCopyBuffer( const __DRIdrawablePrivate *dPriv )
       drm_clip_rect_t *pbox = dPriv->pClipRects;
       int pitch = intelScreen->front.pitch;
       int cpp = intelScreen->cpp;
+      int BR13, CMD;
       int i;
-      GLuint CMD, BR13;
-      BATCH_LOCALS;
 
-      switch(cpp) {
-      case 2: 
+
+      if (cpp == 2) {
 	 BR13 = (pitch * cpp) | (0xCC << 16) | (1<<24);
 	 CMD = XY_SRC_COPY_BLT_CMD;
-	 break;
-      case 4:
+      } 
+      else {
 	 BR13 = (pitch * cpp) | (0xCC << 16) | (1<<24) | (1<<25);
 	 CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA |
 		XY_SRC_COPY_BLT_WRITE_RGB);
-	 break;
-      default:
-	 BR13 = (pitch * cpp) | (0xCC << 16) | (1<<24);
-	 CMD = XY_SRC_COPY_BLT_CMD;
-	 break;
       }
 
       for (i = 0 ; i < nbox; i++, pbox++) 
@@ -99,44 +94,44 @@ void intelCopyBuffer( const __DRIdrawablePrivate *dPriv )
 	     pbox->y2 > intelScreen->height)
 	    continue;
  
-	 BEGIN_BATCH( 8);
+	 BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS);
 	 OUT_BATCH( CMD );
 	 OUT_BATCH( BR13 );
 	 OUT_BATCH( (pbox->y1 << 16) | pbox->x1 );
 	 OUT_BATCH( (pbox->y2 << 16) | pbox->x2 );
 
 	 if (intel->sarea->pf_current_page == 0) 
-	    OUT_BATCH( intelScreen->front.offset );
+	    OUT_RELOC( intel->front_region->buffer, BM_MEM_AGP|BM_WRITE, 0 );
 	 else
-	    OUT_BATCH( intelScreen->back.offset );			
+	    OUT_RELOC( intel->back_region->buffer, BM_MEM_AGP|BM_WRITE, 0 ); 
 
 	 OUT_BATCH( (pbox->y1 << 16) | pbox->x1 );
 	 OUT_BATCH( BR13 & 0xffff );
 
 	 if (intel->sarea->pf_current_page == 0) 
-	    OUT_BATCH( intelScreen->back.offset );			
+	    OUT_RELOC( intel->back_region->buffer, BM_MEM_AGP|BM_READ, 0 ); 
 	 else
-	    OUT_BATCH( intelScreen->front.offset );
+	    OUT_RELOC( intel->front_region->buffer, BM_MEM_AGP|BM_READ, 0 );
 
 	 ADVANCE_BATCH();
       }
    }
-   intelFlushBatchLocked( intel, GL_TRUE, GL_TRUE, GL_TRUE );
-   assert(intel->buffer_list == NULL);
 
+   intel_batchbuffer_flush( intel->batch );
    UNLOCK_HARDWARE( intel );
 }
 
 
 
 
-void intelEmitFillBlitLocked( intelContextPtr intel,
-			      GLuint cpp,
-			      GLshort dst_pitch,
-			      GLuint dst_offset,
-			      GLshort x, GLshort y, 
-			      GLshort w, GLshort h,
-			      GLuint color )
+void intelEmitFillBlit( struct intel_context *intel,
+			GLuint cpp,
+			GLshort dst_pitch,
+			GLuint dst_buffer,
+			GLuint dst_offset,
+			GLshort x, GLshort y, 
+			GLshort w, GLshort h,
+			GLuint color )
 {
    GLuint BR13, CMD;
    BATCH_LOCALS;
@@ -159,12 +154,12 @@ void intelEmitFillBlitLocked( intelContextPtr intel,
       return;
    }
 
-   BEGIN_BATCH( 6);
+   BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS);
    OUT_BATCH( CMD );
    OUT_BATCH( BR13 );
    OUT_BATCH( (y << 16) | x );
    OUT_BATCH( ((y+h) << 16) | (x+w) );
-   OUT_BATCH( dst_offset );
+   OUT_RELOC( dst_buffer, BM_MEM_AGP|BM_WRITE, dst_offset );
    OUT_BATCH( color );
    ADVANCE_BATCH();
 }
@@ -172,15 +167,17 @@ void intelEmitFillBlitLocked( intelContextPtr intel,
 
 /* Copy BitBlt
  */
-void intelEmitCopyBlitLocked( intelContextPtr intel,
-			      GLuint cpp,
-			      GLshort src_pitch,
-			      GLuint  src_offset,
-			      GLshort dst_pitch,
-			      GLuint  dst_offset,
-			      GLshort src_x, GLshort src_y,
-			      GLshort dst_x, GLshort dst_y,
-			      GLshort w, GLshort h )
+void intelEmitCopyBlit( struct intel_context *intel,
+			GLuint cpp,
+			GLshort src_pitch,
+			GLuint  src_buffer,
+			GLuint  src_offset,
+			GLshort dst_pitch,
+			GLuint  dst_buffer,
+			GLuint  dst_offset,
+			GLshort src_x, GLshort src_y,
+			GLshort dst_x, GLshort dst_y,
+			GLshort w, GLshort h )
 {
    GLuint CMD, BR13;
    int dst_y2 = dst_y + h;
@@ -188,11 +185,11 @@ void intelEmitCopyBlitLocked( intelContextPtr intel,
    BATCH_LOCALS;
 
 
-   if (0)
-      _mesa_printf("%s src:0x%x/%d %d,%d dst:0x%x/%d %d,%d sz:%dx%d\n",
+   if (1)
+      _mesa_printf("%s src:buf(%d)/%d %d,%d dst:buf(%d)/%d %d,%d sz:%dx%d\n",
 		   __FUNCTION__,
-		   src_offset, src_pitch, src_x, src_y,
-		   dst_offset, dst_pitch, dst_x, dst_y,
+		   src_buffer, src_pitch, src_x, src_y,
+		   dst_buffer, dst_pitch, dst_x, dst_y,
 		   w,h);
 
    src_pitch *= cpp;
@@ -223,30 +220,27 @@ void intelEmitCopyBlitLocked( intelContextPtr intel,
     * we adjust the offsets manually (below), it seems to work fine.
     */
    if (0) {
-      BEGIN_BATCH(8);
+      BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS);
       OUT_BATCH( CMD );
       OUT_BATCH( BR13 );
       OUT_BATCH( (dst_y << 16) | dst_x );
       OUT_BATCH( (dst_y2 << 16) | dst_x2 );
-      OUT_BATCH( dst_offset );	
+      OUT_RELOC( dst_buffer, BM_MEM_AGP|BM_WRITE, dst_offset );	
       OUT_BATCH( (src_y << 16) | src_x );
       OUT_BATCH( ((GLint)src_pitch&0xffff) );
-      OUT_BATCH( src_offset ); 
+      OUT_RELOC( src_buffer, BM_MEM_AGP|BM_READ, src_offset ); 
       ADVANCE_BATCH();
    }
    else {
-      src_offset += src_y * src_pitch;
-      dst_offset += dst_y * dst_pitch;
-
-      BEGIN_BATCH(8);
+      BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS);
       OUT_BATCH( CMD );
       OUT_BATCH( BR13 );
       OUT_BATCH( (0 << 16) | dst_x );
       OUT_BATCH( (h << 16) | dst_x2 );
-      OUT_BATCH( dst_offset );	
+      OUT_RELOC( dst_buffer, BM_MEM_AGP|BM_WRITE, dst_offset + dst_y * dst_pitch );	
       OUT_BATCH( (0 << 16) | src_x );
       OUT_BATCH( ((GLint)src_pitch&0xffff) );
-      OUT_BATCH( src_offset ); 
+      OUT_RELOC( src_buffer, BM_MEM_AGP|BM_READ, src_offset + src_y * src_pitch ); 
       ADVANCE_BATCH();
    }
 }
@@ -256,7 +250,7 @@ void intelEmitCopyBlitLocked( intelContextPtr intel,
 void intelClearWithBlit(GLcontext *ctx, GLbitfield flags, GLboolean all,
 		      GLint cx1, GLint cy1, GLint cw, GLint ch)
 {
-   intelContextPtr intel = INTEL_CONTEXT( ctx );
+   struct intel_context *intel = intel_context( ctx );
    intelScreenPrivate *intelScreen = intel->intelScreen;
    GLuint clear_depth, clear_color;
    GLint cx, cy;
@@ -300,8 +294,6 @@ void intelClearWithBlit(GLcontext *ctx, GLbitfield flags, GLboolean all,
 
    intelFlush( &intel->ctx );
    LOCK_HARDWARE( intel );
-   intelInstallBatchBuffer(intel);   
-   intelValidateBuffers( intel );
    {
       /* flip top to bottom */
       cy = intel->driDrawable->h-cy1-ch;
@@ -351,40 +343,40 @@ void intelClearWithBlit(GLcontext *ctx, GLbitfield flags, GLboolean all,
 	    continue;
 
 	 if ( flags & BUFFER_BIT_FRONT_LEFT ) {	    
-	    BEGIN_BATCH( 6);	    
+	    BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS);
 	    OUT_BATCH( CMD );
 	    OUT_BATCH( BR13 );
 	    OUT_BATCH( (b.y1 << 16) | b.x1 );
 	    OUT_BATCH( (b.y2 << 16) | b.x2 );
-	    OUT_BATCH( intelScreen->front.offset );
+	    OUT_RELOC( intel->front_region->buffer, BM_MEM_AGP|BM_WRITE, 0 );
 	    OUT_BATCH( clear_color );
 	    ADVANCE_BATCH();
 	 }
 
 	 if ( flags & BUFFER_BIT_BACK_LEFT ) {
-	    BEGIN_BATCH( 6); 
+	    BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS); 
 	    OUT_BATCH( CMD );
 	    OUT_BATCH( BR13 );
 	    OUT_BATCH( (b.y1 << 16) | b.x1 );
 	    OUT_BATCH( (b.y2 << 16) | b.x2 );
-	    OUT_BATCH( intelScreen->back.offset );
+	    OUT_RELOC( intel->back_region->buffer, BM_MEM_AGP|BM_WRITE, 0 );
 	    OUT_BATCH( clear_color );
 	    ADVANCE_BATCH();
 	 }
 
 	 if ( flags & (BUFFER_BIT_STENCIL | BUFFER_BIT_DEPTH) ) {
-	    BEGIN_BATCH( 6);
+	    BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS);
 	    OUT_BATCH( D_CMD );
 	    OUT_BATCH( BR13 );
 	    OUT_BATCH( (b.y1 << 16) | b.x1 );
 	    OUT_BATCH( (b.y2 << 16) | b.x2 );
-	    OUT_BATCH( intelScreen->depth.offset );
+	    OUT_RELOC( intel->depth_region->buffer, BM_MEM_AGP|BM_WRITE, 0 );
 	    OUT_BATCH( clear_depth );
 	    ADVANCE_BATCH();
 	 }      
       }
    }
-   intelFlushBatchLocked( intel, GL_TRUE, GL_FALSE, GL_TRUE );
+   intel_batchbuffer_flush( intel->batch );
    UNLOCK_HARDWARE( intel );
 }
 
diff --git a/src/mesa/drivers/dri/i915/intel_blit.h b/src/mesa/drivers/dri/i915/intel_blit.h
index dd75844730d..0bbdb0a7ed8 100644
--- a/src/mesa/drivers/dri/i915/intel_blit.h
+++ b/src/mesa/drivers/dri/i915/intel_blit.h
@@ -35,23 +35,26 @@ extern void intelCopyBuffer( const __DRIdrawablePrivate *dpriv );
 extern void intelClearWithBlit(GLcontext *ctx, GLbitfield mask, GLboolean all,
 			     GLint cx1, GLint cy1, GLint cw, GLint ch);
 
-extern void intelEmitCopyBlitLocked( intelContextPtr intel,
-				     GLuint cpp,
-				     GLshort src_pitch,
-				     GLuint  src_offset,
-				     GLshort dst_pitch,
-				     GLuint  dst_offset,
-				     GLshort srcx, GLshort srcy,
-				     GLshort dstx, GLshort dsty,
-				     GLshort w, GLshort h );
+extern void intelEmitCopyBlit( struct intel_context *intel,
+			       GLuint cpp,
+			       GLshort src_pitch,
+			       GLuint  src_buffer,
+			       GLuint  src_offset,
+			       GLshort dst_pitch,
+			       GLuint  dst_buffer,
+			       GLuint  dst_offset,
+			       GLshort srcx, GLshort srcy,
+			       GLshort dstx, GLshort dsty,
+			       GLshort w, GLshort h );
 
-extern void intelEmitFillBlitLocked( intelContextPtr intel,
-				     GLuint cpp,
-				     GLshort dst_pitch,
-				     GLuint dst_offset,
-				     GLshort x, GLshort y, 
-				     GLshort w, GLshort h,
-				     GLuint color );
+extern void intelEmitFillBlit( struct intel_context *intel,
+			       GLuint cpp,
+			       GLshort dst_pitch,
+			       GLuint dst_buffer,
+			       GLuint dst_offset,
+			       GLshort x, GLshort y, 
+			       GLshort w, GLshort h,
+			       GLuint color );
 
 
 #endif
diff --git a/src/mesa/drivers/dri/i915/intel_buffers.c b/src/mesa/drivers/dri/i915/intel_buffers.c
new file mode 100644
index 00000000000..fd937d0b0ba
--- /dev/null
+++ b/src/mesa/drivers/dri/i915/intel_buffers.c
@@ -0,0 +1,505 @@
+/**************************************************************************
+ * 
+ * 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_screen.h"
+#include "intel_context.h"
+#include "intel_blit.h"
+#include "intel_tris.h"
+#include "intel_regions.h"
+#include "intel_batchbuffer.h"
+#include "context.h"
+#include "framebuffer.h"
+#include "swrast/swrast.h"
+
+GLboolean intel_intersect_cliprects( drm_clip_rect_t *dst,
+				     const drm_clip_rect_t *a,
+				     const drm_clip_rect_t *b )
+{
+   GLint bx = b->x1;
+   GLint by = b->y1;
+   GLint bw = b->x2 - bx;
+   GLint bh = b->y2 - by;
+
+   if (bx < a->x1) bw -= a->x1 - bx, bx = a->x1;
+   if (by < a->y1) bh -= a->y1 - by, by = a->y1;
+   if (bx + bw > a->x2) bw = a->x2 - bx;
+   if (by + bh > a->y2) bh = a->y2 - by;
+   if (bw <= 0) return GL_FALSE;
+   if (bh <= 0) return GL_FALSE;
+
+   dst->x1 = bx;
+   dst->y1 = by;
+   dst->x2 = bx + bw;
+   dst->y2 = by + bh;
+
+   return GL_TRUE;
+}
+
+struct intel_region *intel_drawbuf_region( struct intel_context *intel )
+{
+   switch (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0]) {
+   case BUFFER_BIT_FRONT_LEFT:
+      return intel->front_region;
+   case BUFFER_BIT_BACK_LEFT:
+      return intel->back_region;
+   default:
+      /* Not necessary to fallback - could handle either NONE or
+       * FRONT_AND_BACK cases below.
+       */
+      return NULL;		
+   }
+}
+
+struct intel_region *intel_readbuf_region( struct intel_context *intel )
+{
+   GLcontext *ctx = &intel->ctx;
+
+   /* This will have to change to support EXT_fbo's, but is correct
+    * for now:
+    */
+   switch (ctx->ReadBuffer->_ColorReadBufferIndex) {
+   case BUFFER_FRONT_LEFT:
+      return intel->front_region;
+   case BUFFER_BACK_LEFT:
+      return intel->back_region;
+   default:
+      assert(0);
+      return NULL;
+   }
+}
+
+
+
+static void intelBufferSize(GLframebuffer *buffer,
+			    GLuint *width, 
+			    GLuint *height)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct intel_context *intel = intel_context(ctx);
+   /* Need to lock to make sure the driDrawable is uptodate.  This
+    * information is used to resize Mesa's software buffers, so it has
+    * to be correct.
+    */
+   LOCK_HARDWARE(intel);
+   if (intel->driDrawable) {
+      *width = intel->driDrawable->w;
+      *height = intel->driDrawable->h;
+   }
+   else {
+      *width = 0;
+      *height = 0;
+   }
+   UNLOCK_HARDWARE(intel);
+}
+
+
+static void intelSetFrontClipRects( struct intel_context *intel )
+{
+   __DRIdrawablePrivate *dPriv = intel->driDrawable;
+
+   if (!dPriv) return;
+
+   intel->numClipRects = dPriv->numClipRects;
+   intel->pClipRects = dPriv->pClipRects;
+   intel->drawX = dPriv->x;
+   intel->drawY = dPriv->y;
+}
+
+
+static void intelSetBackClipRects( struct intel_context *intel )
+{
+   __DRIdrawablePrivate *dPriv = intel->driDrawable;
+
+   if (!dPriv) return;
+
+   if (intel->sarea->pf_enabled == 0 && dPriv->numBackClipRects == 0) {
+      intel->numClipRects = dPriv->numClipRects;
+      intel->pClipRects = dPriv->pClipRects;
+      intel->drawX = dPriv->x;
+      intel->drawY = dPriv->y;
+   } else {
+      intel->numClipRects = dPriv->numBackClipRects;
+      intel->pClipRects = dPriv->pBackClipRects;
+      intel->drawX = dPriv->backX;
+      intel->drawY = dPriv->backY;
+      
+      if (dPriv->numBackClipRects == 1 &&
+	  dPriv->x == dPriv->backX &&
+	  dPriv->y == dPriv->backY) {
+      
+	 /* Repeat the calculation of the back cliprect dimensions here
+	  * as early versions of dri.a in the Xserver are incorrect.  Try
+	  * very hard not to restrict future versions of dri.a which
+	  * might eg. allocate truly private back buffers.
+	  */
+	 int x1, y1;
+	 int x2, y2;
+	 
+	 x1 = dPriv->x;
+	 y1 = dPriv->y;      
+	 x2 = dPriv->x + dPriv->w;
+	 y2 = dPriv->y + dPriv->h;
+	 
+	 if (x1 < 0) x1 = 0;
+	 if (y1 < 0) y1 = 0;
+	 if (x2 > intel->intelScreen->width) x2 = intel->intelScreen->width;
+	 if (y2 > intel->intelScreen->height) y2 = intel->intelScreen->height;
+
+	 if (x1 == dPriv->pBackClipRects[0].x1 &&
+	     y1 == dPriv->pBackClipRects[0].y1) {
+
+	    dPriv->pBackClipRects[0].x2 = x2;
+	    dPriv->pBackClipRects[0].y2 = y2;
+	 }
+      }
+   }
+}
+
+
+void intelWindowMoved( struct intel_context *intel )
+{
+   if (!intel->ctx.DrawBuffer) {
+      intelSetFrontClipRects( intel );
+   }
+   else {
+      switch (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0]) {
+      case BUFFER_BIT_FRONT_LEFT:
+	 intelSetFrontClipRects( intel );
+	 break;
+      case BUFFER_BIT_BACK_LEFT:
+	 intelSetBackClipRects( intel );
+	 break;
+      default:
+	 /* glDrawBuffer(GL_NONE or GL_FRONT_AND_BACK): software fallback */
+	 intelSetFrontClipRects( intel );
+      }
+   }
+
+   /* Set state we know depends on drawable parameters:
+    */
+   {
+      GLcontext *ctx = &intel->ctx;
+
+      ctx->Driver.Scissor( ctx, ctx->Scissor.X, ctx->Scissor.Y,
+			   ctx->Scissor.Width, ctx->Scissor.Height );
+      
+      ctx->Driver.DepthRange( ctx, 
+			      ctx->Viewport.Near,
+			      ctx->Viewport.Far );
+   }
+}
+
+static void emit_clip_rect_quads(struct intel_context *intel,
+				 const drm_clip_rect_t *clear)
+{
+   GLuint i;
+
+   for (i = 0; i < intel->numClipRects; i++) {
+      drm_clip_rect_t rect;
+
+      if (intel_intersect_cliprects(&rect, 
+				    clear,
+				    &intel->pClipRects[i]))
+      {
+	 intel_meta_draw_quad(intel, 
+			      rect.x1, rect.x2, 
+			      rect.y1, rect.y2, 
+			      0,
+			      intel->ClearColor, 
+			      0, 0, 0, 0,
+			      INTEL_BATCH_NO_CLIPRECTS);
+      }
+   }
+}
+
+
+/* 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,
+			       GLboolean all,
+			       GLint cx, GLint cy, 
+			       GLint cw, GLint ch)
+{
+   __DRIdrawablePrivate *dPriv = intel->driDrawable;
+   drm_clip_rect_t clear;
+
+   LOCK_HARDWARE(intel);
+   intel->vtbl.install_meta_state(intel);
+
+   if(!all) {
+      clear.x1 = cx;
+      clear.y1 = cy;
+      clear.x2 = cx + cw;
+      clear.y2 = cy + ch;
+   } else {
+      clear.x1 = 0;
+      clear.y1 = 0;
+      clear.x2 = dPriv->w;
+      clear.y2 = dPriv->h;
+   }
+
+   /* Back and stencil cliprects are the same.  Try and do both
+    * buffers at once:
+    */
+   if (mask & (BUFFER_BIT_BACK_LEFT|BUFFER_BIT_STENCIL)) { 
+      intel->vtbl.meta_draw_region(intel, 
+				   intel->back_region,
+				   intel->depth_region );
+
+      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_depth_stencil_write(intel);
+      
+      /* Do cliprects explicitly:
+       */
+      emit_clip_rect_quads(intel, &clear);
+   }
+
+   /* Front may have different cliprects: 
+    */
+   if (mask & BUFFER_BIT_FRONT_LEFT) {
+      intel->vtbl.meta_no_depth_stencil_write(intel);
+      intel->vtbl.meta_color_mask(intel, GL_TRUE );
+      intel->vtbl.meta_draw_region(intel, 
+				   intel->front_region,
+				   intel->depth_region);
+
+      emit_clip_rect_quads(intel, &clear);
+   }
+
+   intel->vtbl.leave_meta_state( intel );
+   intel_batchbuffer_flush( intel->batch );
+   UNLOCK_HARDWARE(intel);
+}
+
+
+
+
+
+static void intelClear(GLcontext *ctx, 
+		       GLbitfield mask, 
+		       GLboolean all,
+		       GLint cx, GLint cy, 
+		       GLint cw, GLint ch)
+{
+   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;
+
+   if (0)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+
+   if (mask & BUFFER_BIT_FRONT_LEFT) {
+      if (colorMask == ~0) {
+	 blit_mask |= BUFFER_BIT_FRONT_LEFT;
+      } 
+      else {
+	 tri_mask |= BUFFER_BIT_FRONT_LEFT;
+      }
+   }
+
+   if (mask & BUFFER_BIT_BACK_LEFT) {
+      if (colorMask == ~0) {
+	 blit_mask |= BUFFER_BIT_BACK_LEFT;
+      } 
+      else {
+	 tri_mask |= BUFFER_BIT_BACK_LEFT;
+      }
+   }
+
+   if (mask & BUFFER_BIT_DEPTH) {
+      blit_mask |= BUFFER_BIT_DEPTH;
+   }
+
+   if (mask & BUFFER_BIT_STENCIL) {
+      if (!intel->hw_stencil) {
+	 swrast_mask |= BUFFER_BIT_STENCIL;
+      }
+      else if (ctx->Stencil.WriteMask[0] != 0xff) {
+	 tri_mask |= BUFFER_BIT_STENCIL;
+      } 
+      else {
+	 blit_mask |= BUFFER_BIT_STENCIL;
+      }
+   }
+
+   swrast_mask |= (mask & BUFFER_BIT_ACCUM);
+
+   intelFlush( ctx );
+
+   if (blit_mask) 
+      intelClearWithBlit( ctx, blit_mask, all, cx, cy, cw, ch );
+
+   if (tri_mask) 
+      intelClearWithTris( intel, tri_mask, all, cx, cy, cw, ch);
+
+   if (swrast_mask)
+      _swrast_Clear( ctx, swrast_mask, all, cx, cy, cw, ch );
+}
+
+
+
+
+
+
+
+/* Flip the front & back buffers
+ */
+static void intelPageFlip( const __DRIdrawablePrivate *dPriv )
+{
+#if 0
+   struct intel_context *intel;
+   int tmp, ret;
+
+   if (INTEL_DEBUG & DEBUG_IOCTL)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   assert(dPriv);
+   assert(dPriv->driContextPriv);
+   assert(dPriv->driContextPriv->driverPrivate);
+
+   intel = (struct intel_context *) dPriv->driContextPriv->driverPrivate;
+
+   intelFlush( &intel->ctx );
+   LOCK_HARDWARE( intel );
+
+   if (dPriv->pClipRects) {
+      *(drm_clip_rect_t *)intel->sarea->boxes = dPriv->pClipRects[0];
+      intel->sarea->nbox = 1;
+   }
+
+   ret = drmCommandNone(intel->driFd, DRM_I830_FLIP); 
+   if (ret) {
+      fprintf(stderr, "%s: %d\n", __FUNCTION__, ret);
+      UNLOCK_HARDWARE( intel );
+      exit(1);
+   }
+
+   tmp = intel->sarea->last_enqueue;
+   intelRefillBatchLocked( intel );
+   UNLOCK_HARDWARE( intel );
+
+
+   intelSetDrawBuffer( &intel->ctx, intel->ctx.Color.DriverDrawBuffer );
+#endif
+}
+
+
+void intelSwapBuffers( __DRIdrawablePrivate *dPriv )
+{
+   if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
+      struct intel_context *intel;
+      GLcontext *ctx;
+      intel = (struct intel_context *) dPriv->driContextPriv->driverPrivate;
+      ctx = &intel->ctx;
+      if (ctx->Visual.doubleBufferMode) {
+	 _mesa_notifySwapBuffers( ctx );  /* flush pending rendering comands */
+	 if ( 0 /*intel->doPageFlip*/ ) { /* doPageFlip is never set !!! */
+	    intelPageFlip( dPriv );
+	 } else {
+	    intelCopyBuffer( dPriv );
+	 }
+      }
+   } 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__);
+   }
+}
+
+static void intelDrawBuffer(GLcontext *ctx, GLenum mode )
+{
+   struct intel_context *intel = intel_context(ctx);
+   int front = 0;
+ 
+   if (!ctx->DrawBuffer)
+      return;
+
+   switch ( ctx->DrawBuffer->_ColorDrawBufferMask[0] ) {
+   case BUFFER_BIT_FRONT_LEFT:
+      front = 1;
+      FALLBACK( intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE );
+      break;
+   case BUFFER_BIT_BACK_LEFT:
+      front = 0;
+      FALLBACK( intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE );
+      break;
+   default:
+      FALLBACK( intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE );
+      return;
+   }
+
+   if ( intel->sarea->pf_current_page == 1 ) 
+      front ^= 1;
+   
+   intelSetFrontClipRects( intel );
+
+
+   if (front) {
+      if (intel->draw_region != intel->front_region) {
+	 intel_region_release(intel, &intel->draw_region);
+	 intel_region_reference(&intel->draw_region, intel->front_region);
+      }
+   } else {
+      if (intel->draw_region != intel->back_region) {
+	 intel_region_release(intel, &intel->draw_region);
+	 intel_region_reference(&intel->draw_region, intel->back_region);
+      }
+   }
+
+   intel->vtbl.set_draw_region( intel, 
+				intel->draw_region,
+				intel->depth_region);
+}
+
+static void intelReadBuffer( GLcontext *ctx, GLenum mode )
+{
+   /* nothing, until we implement h/w glRead/CopyPixels or CopyTexImage */
+}
+
+
+
+void intelInitBufferFuncs( struct dd_function_table *functions )
+{
+   functions->Clear = intelClear;
+   functions->GetBufferSize = intelBufferSize;
+   functions->ResizeBuffers = _mesa_resize_framebuffer;
+   functions->DrawBuffer = intelDrawBuffer;
+   functions->ReadBuffer = intelReadBuffer;
+}
diff --git a/src/mesa/drivers/dri/i915/intel_context.c b/src/mesa/drivers/dri/i915/intel_context.c
index 53b28a8a6be..50510dac2b5 100644
--- a/src/mesa/drivers/dri/i915/intel_context.c
+++ b/src/mesa/drivers/dri/i915/intel_context.c
@@ -55,6 +55,8 @@
 #include "intel_ioctl.h"
 #include "intel_batchbuffer.h"
 #include "intel_blit.h"
+#include "intel_pixel.h"
+#include "intel_regions.h"
 
 #include "bufmgr.h"
 
@@ -80,9 +82,6 @@ int INTEL_DEBUG = (0);
 #define need_GL_NV_vertex_program
 #include "extension_helper.h"
 
-#ifndef VERBOSE
-int VERBOSE = 0;
-#endif
 
 #if DEBUG_LOCKING
 char *prevLockFile;
@@ -95,7 +94,7 @@ int prevLockLine;
 
 #define DRIVER_DATE                     "20060212"
 
-const GLubyte *intelGetString( GLcontext *ctx, GLenum name )
+static const GLubyte *intelGetString( GLcontext *ctx, GLenum name )
 {
    const char * chipset;
    static char buffer[128];
@@ -106,7 +105,7 @@ const GLubyte *intelGetString( GLcontext *ctx, GLenum name )
       break;
       
    case GL_RENDERER:
-      switch (INTEL_CONTEXT(ctx)->intelScreen->deviceID) {
+      switch (intel_context(ctx)->intelScreen->deviceID) {
       case PCI_CHIP_845_G:
 	 chipset = "Intel(R) 845G"; break;
       case PCI_CHIP_I830_M:
@@ -133,27 +132,6 @@ const GLubyte *intelGetString( GLcontext *ctx, GLenum name )
    }
 }
 
-static void intelBufferSize(GLframebuffer *buffer,
-			   GLuint *width, GLuint *height)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
-   /* Need to lock to make sure the driDrawable is uptodate.  This
-    * information is used to resize Mesa's software buffers, so it has
-    * to be correct.
-    */
-   LOCK_HARDWARE(intel);
-   if (intel->driDrawable) {
-      *width = intel->driDrawable->w;
-      *height = intel->driDrawable->h;
-   }
-   else {
-      *width = 0;
-      *height = 0;
-   }
-   UNLOCK_HARDWARE(intel);
-}
-
 
 /**
  * Extension strings exported by the intel driver.
@@ -252,7 +230,32 @@ static void intelInvalidateState( GLcontext *ctx, GLuint new_state )
    _ac_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    _tnl_invalidate_vertex_state( ctx, new_state );
-   INTEL_CONTEXT(ctx)->NewGLState |= new_state;
+   intel_context(ctx)->NewGLState |= new_state;
+}
+
+
+void intelFlush( GLcontext *ctx )
+{
+   struct intel_context *intel = intel_context( ctx );
+
+   if (intel->Fallback)
+      _swrast_flush( ctx );
+
+   INTEL_FIREVERTICES( intel );
+
+   if (intel->batch->map != intel->batch->ptr)
+      intel_batchbuffer_flush( intel->batch );
+
+   /* XXX: Need to do an MI_FLUSH here.  Actually, the bufmgr_fake.c
+    * code will have done one already.
+    */
+}
+
+void intelFinish( GLcontext *ctx ) 
+{
+   struct intel_context *intel = intel_context( ctx );
+   intelFlush( ctx );
+   bmFinishFence( intel->bm, intel->last_fence );
 }
 
 
@@ -261,10 +264,7 @@ void intelInitDriverFunctions( struct dd_function_table *functions )
    _mesa_init_driver_functions( functions );
 
    functions->Flush = intelFlush;
-   functions->Clear = intelClear;
    functions->Finish = intelFinish;
-   functions->GetBufferSize = intelBufferSize;
-   functions->ResizeBuffers = _mesa_resize_framebuffer;
    functions->GetString = intelGetString;
    functions->UpdateState = intelInvalidateState;
    functions->CopyColorTable = _swrast_CopyColorTable;
@@ -275,11 +275,12 @@ void intelInitDriverFunctions( struct dd_function_table *functions )
    intelInitTextureFuncs( functions );
    intelInitPixelFuncs( functions );
    intelInitStateFuncs( functions );
+   intelInitBufferFuncs( functions );
 }
 
 
 
-GLboolean intelInitContext( intelContextPtr intel,
+GLboolean intelInitContext( struct intel_context *intel,
 			    const __GLcontextModes *mesaVis,
 			    __DRIcontextPrivate *driContextPriv,
 			    void *sharedContextPrivate,
@@ -379,6 +380,51 @@ GLboolean intelInitContext( intelContextPtr intel,
 /* 		      GL_TRUE, */
 		      GL_FALSE);
 
+
+   /* Buffer manager: 
+    */
+   intel->bm = bm_fake_intel_Attach( intel );
+
+   bmInitPool(intel->bm,
+              intel->intelScreen->tex.offset, /* low offset */
+              intel->intelScreen->tex.map, /* low virtual */
+              intel->intelScreen->tex.size,
+	      BM_MEM_AGP);
+
+   /* These are still static, but create regions for them.  
+    */
+   intel->front_region = 
+      intel_region_create_static(intel,
+				 BM_MEM_AGP,
+				 intelScreen->front.offset,
+				 intelScreen->front.map,
+				 intelScreen->cpp,
+				 intelScreen->front.pitch,
+				 intelScreen->height);
+
+
+   intel->back_region = 
+      intel_region_create_static(intel,
+				 BM_MEM_AGP,
+				 intelScreen->back.offset,
+				 intelScreen->back.map,
+				 intelScreen->cpp,
+				 intelScreen->back.pitch,
+				 intelScreen->height);
+
+   /* Still assuming front.cpp == depth.cpp
+    */
+   intel->depth_region = 
+      intel_region_create_static(intel,
+				 BM_MEM_AGP,
+				 intelScreen->depth.offset,
+				 intelScreen->depth.map,
+				 intelScreen->cpp,
+				 intelScreen->depth.pitch,
+				 intelScreen->height);
+   
+   intel->batch = intel_batchbuffer_alloc( intel );
+
    if (intel->ctx.Mesa_DXTn) {
      _mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" );
      _mesa_enable_extension( ctx, "GL_S3_s3tc" );
@@ -393,24 +439,16 @@ GLboolean intelInitContext( intelContextPtr intel,
 /* 			  DRI_TEXMGR_DO_TEXTURE_RECT ); */
 
 
-   intel->prim.flush = NULL;
    intel->prim.primitive = ~0;
 
 
 #if DO_DEBUG
    INTEL_DEBUG  = driParseDebugString( getenv( "INTEL_DEBUG" ),
 				       debug_control );
-   INTEL_DEBUG |= driParseDebugString( getenv( "INTEL_DEBUG" ),
-				       debug_control );
 #endif
 
-#ifndef VERBOSE
-   if (getenv("INTEL_VERBOSE"))
-      VERBOSE=1;
-#endif
 
-   if (getenv("INTEL_NO_RAST") || 
-       getenv("INTEL_NO_RAST")) {
+   if (getenv("INTEL_NO_RAST")) {
       fprintf(stderr, "disabling 3D rasterization\n");
       FALLBACK(intel, INTEL_FALLBACK_USER, 1); 
    }
@@ -420,7 +458,7 @@ GLboolean intelInitContext( intelContextPtr intel,
 
 void intelDestroyContext(__DRIcontextPrivate *driContextPriv)
 {
-   intelContextPtr intel = (intelContextPtr) driContextPriv->driverPrivate;
+   struct intel_context *intel = (struct intel_context *) driContextPriv->driverPrivate;
 
    assert(intel); /* should never be null */
    if (intel) {
@@ -437,7 +475,7 @@ void intelDestroyContext(__DRIcontextPrivate *driContextPriv)
       _swrast_DestroyContext (&intel->ctx);
       intel->Fallback = 0;	/* don't call _swrast_Flush later */
 
-      intelDestroyBatchBuffer(intel);
+      intel_batchbuffer_free(intel->batch);
       
 
       if ( release_texture_heaps ) {
@@ -452,102 +490,6 @@ void intelDestroyContext(__DRIcontextPrivate *driContextPriv)
    }
 }
 
-void intelSetFrontClipRects( intelContextPtr intel )
-{
-   __DRIdrawablePrivate *dPriv = intel->driDrawable;
-
-   if (!dPriv) return;
-
-   intel->numClipRects = dPriv->numClipRects;
-   intel->pClipRects = dPriv->pClipRects;
-   intel->drawX = dPriv->x;
-   intel->drawY = dPriv->y;
-}
-
-
-void intelSetBackClipRects( intelContextPtr intel )
-{
-   __DRIdrawablePrivate *dPriv = intel->driDrawable;
-
-   if (!dPriv) return;
-
-   if (intel->sarea->pf_enabled == 0 && dPriv->numBackClipRects == 0) {
-      intel->numClipRects = dPriv->numClipRects;
-      intel->pClipRects = dPriv->pClipRects;
-      intel->drawX = dPriv->x;
-      intel->drawY = dPriv->y;
-   } else {
-      intel->numClipRects = dPriv->numBackClipRects;
-      intel->pClipRects = dPriv->pBackClipRects;
-      intel->drawX = dPriv->backX;
-      intel->drawY = dPriv->backY;
-      
-      if (dPriv->numBackClipRects == 1 &&
-	  dPriv->x == dPriv->backX &&
-	  dPriv->y == dPriv->backY) {
-      
-	 /* Repeat the calculation of the back cliprect dimensions here
-	  * as early versions of dri.a in the Xserver are incorrect.  Try
-	  * very hard not to restrict future versions of dri.a which
-	  * might eg. allocate truly private back buffers.
-	  */
-	 int x1, y1;
-	 int x2, y2;
-	 
-	 x1 = dPriv->x;
-	 y1 = dPriv->y;      
-	 x2 = dPriv->x + dPriv->w;
-	 y2 = dPriv->y + dPriv->h;
-	 
-	 if (x1 < 0) x1 = 0;
-	 if (y1 < 0) y1 = 0;
-	 if (x2 > intel->intelScreen->width) x2 = intel->intelScreen->width;
-	 if (y2 > intel->intelScreen->height) y2 = intel->intelScreen->height;
-
-	 if (x1 == dPriv->pBackClipRects[0].x1 &&
-	     y1 == dPriv->pBackClipRects[0].y1) {
-
-	    dPriv->pBackClipRects[0].x2 = x2;
-	    dPriv->pBackClipRects[0].y2 = y2;
-	 }
-      }
-   }
-}
-
-
-void intelWindowMoved( intelContextPtr intel )
-{
-   if (!intel->ctx.DrawBuffer) {
-      intelSetFrontClipRects( intel );
-   }
-   else {
-      switch (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0]) {
-      case BUFFER_BIT_FRONT_LEFT:
-	 intelSetFrontClipRects( intel );
-	 break;
-      case BUFFER_BIT_BACK_LEFT:
-	 intelSetBackClipRects( intel );
-	 break;
-      default:
-	 /* glDrawBuffer(GL_NONE or GL_FRONT_AND_BACK): software fallback */
-	 intelSetFrontClipRects( intel );
-      }
-   }
-
-   /* Set state we know depends on drawable parameters:
-    */
-   {
-      GLcontext *ctx = &intel->ctx;
-
-      ctx->Driver.Scissor( ctx, ctx->Scissor.X, ctx->Scissor.Y,
-			   ctx->Scissor.Width, ctx->Scissor.Height );
-      
-      ctx->Driver.DepthRange( ctx, 
-			      ctx->Viewport.Near,
-			      ctx->Viewport.Far );
-   }
-}
-
 GLboolean intelUnbindContext(__DRIcontextPrivate *driContextPriv)
 {
    return GL_TRUE;
@@ -559,7 +501,7 @@ GLboolean intelMakeCurrent(__DRIcontextPrivate *driContextPriv,
 {
 
    if (driContextPriv) {
-      intelContextPtr intel = (intelContextPtr) driContextPriv->driverPrivate;
+      struct intel_context *intel = (struct intel_context *) driContextPriv->driverPrivate;
 
       if ( intel->driDrawable != driDrawPriv ) {
 	 /* Shouldn't the readbuffer be stored also? */
@@ -579,7 +521,7 @@ GLboolean intelMakeCurrent(__DRIcontextPrivate *driContextPriv,
    return GL_TRUE;
 }
 
-void intelGetLock( intelContextPtr intel, GLuint flags )
+void intelGetLock( struct intel_context *intel, GLuint flags )
 {
    __DRIdrawablePrivate *dPriv = intel->driDrawable;
    __DRIscreenPrivate *sPriv = intel->driScreen;
@@ -611,117 +553,4 @@ void intelGetLock( intelContextPtr intel, GLuint flags )
    }
 }
 
-void intelSwapBuffers( __DRIdrawablePrivate *dPriv )
-{
-   if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
-      intelContextPtr intel;
-      GLcontext *ctx;
-      intel = (intelContextPtr) dPriv->driContextPriv->driverPrivate;
-      ctx = &intel->ctx;
-      if (ctx->Visual.doubleBufferMode) {
-	 _mesa_notifySwapBuffers( ctx );  /* flush pending rendering comands */
-	 if ( 0 /*intel->doPageFlip*/ ) { /* doPageFlip is never set !!! */
-	    intelPageFlip( dPriv );
-	 } else {
-	    intelCopyBuffer( dPriv );
-	 }
-      }
-   } 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 intelInitState( GLcontext *ctx )
-{
-   /* Mesa should do this for us:
-    */
-   ctx->Driver.AlphaFunc( ctx, 
-			  ctx->Color.AlphaFunc,
-			  ctx->Color.AlphaRef);
-
-   ctx->Driver.BlendColor( ctx,
-			   ctx->Color.BlendColor );
-
-   ctx->Driver.BlendEquationSeparate( ctx, 
-				      ctx->Color.BlendEquationRGB,
-				      ctx->Color.BlendEquationA);
-
-   ctx->Driver.BlendFuncSeparate( ctx,
-				  ctx->Color.BlendSrcRGB,
-				  ctx->Color.BlendDstRGB,
-				  ctx->Color.BlendSrcA,
-				  ctx->Color.BlendDstA);
-
-   ctx->Driver.ColorMask( ctx, 
-			  ctx->Color.ColorMask[RCOMP],
-			  ctx->Color.ColorMask[GCOMP],
-			  ctx->Color.ColorMask[BCOMP],
-			  ctx->Color.ColorMask[ACOMP]);
-
-   ctx->Driver.CullFace( ctx, ctx->Polygon.CullFaceMode );
-   ctx->Driver.DepthFunc( ctx, ctx->Depth.Func );
-   ctx->Driver.DepthMask( ctx, ctx->Depth.Mask );
-
-   ctx->Driver.Enable( ctx, GL_ALPHA_TEST, ctx->Color.AlphaEnabled );
-   ctx->Driver.Enable( ctx, GL_BLEND, ctx->Color.BlendEnabled );
-   ctx->Driver.Enable( ctx, GL_COLOR_LOGIC_OP, ctx->Color.ColorLogicOpEnabled );
-   ctx->Driver.Enable( ctx, GL_COLOR_SUM, ctx->Fog.ColorSumEnabled );
-   ctx->Driver.Enable( ctx, GL_CULL_FACE, ctx->Polygon.CullFlag );
-   ctx->Driver.Enable( ctx, GL_DEPTH_TEST, ctx->Depth.Test );
-   ctx->Driver.Enable( ctx, GL_DITHER, ctx->Color.DitherFlag );
-   ctx->Driver.Enable( ctx, GL_FOG, ctx->Fog.Enabled );
-   ctx->Driver.Enable( ctx, GL_LIGHTING, ctx->Light.Enabled );
-   ctx->Driver.Enable( ctx, GL_LINE_SMOOTH, ctx->Line.SmoothFlag );
-   ctx->Driver.Enable( ctx, GL_POLYGON_STIPPLE, ctx->Polygon.StippleFlag );
-   ctx->Driver.Enable( ctx, GL_SCISSOR_TEST, ctx->Scissor.Enabled );
-   ctx->Driver.Enable( ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled );
-   ctx->Driver.Enable( ctx, GL_TEXTURE_1D, GL_FALSE );
-   ctx->Driver.Enable( ctx, GL_TEXTURE_2D, GL_FALSE );
-   ctx->Driver.Enable( ctx, GL_TEXTURE_RECTANGLE_NV, GL_FALSE );
-   ctx->Driver.Enable( ctx, GL_TEXTURE_3D, GL_FALSE );
-   ctx->Driver.Enable( ctx, GL_TEXTURE_CUBE_MAP, GL_FALSE );
-
-   ctx->Driver.Fogfv( ctx, GL_FOG_COLOR, ctx->Fog.Color );
-   ctx->Driver.Fogfv( ctx, GL_FOG_MODE, 0 );
-   ctx->Driver.Fogfv( ctx, GL_FOG_DENSITY, &ctx->Fog.Density );
-   ctx->Driver.Fogfv( ctx, GL_FOG_START, &ctx->Fog.Start );
-   ctx->Driver.Fogfv( ctx, GL_FOG_END, &ctx->Fog.End );
-
-   ctx->Driver.FrontFace( ctx, ctx->Polygon.FrontFace );
-
-   {
-      GLfloat f = (GLfloat)ctx->Light.Model.ColorControl;
-      ctx->Driver.LightModelfv( ctx, GL_LIGHT_MODEL_COLOR_CONTROL, &f );
-   }
 
-   ctx->Driver.LineWidth( ctx, ctx->Line.Width );
-   ctx->Driver.LogicOpcode( ctx, ctx->Color.LogicOp );
-   ctx->Driver.PointSize( ctx, ctx->Point.Size );
-   ctx->Driver.PolygonStipple( ctx, (const GLubyte *)ctx->PolygonStipple );
-   ctx->Driver.Scissor( ctx, ctx->Scissor.X, ctx->Scissor.Y,
-			ctx->Scissor.Width, ctx->Scissor.Height );
-   ctx->Driver.ShadeModel( ctx, ctx->Light.ShadeModel );
-   ctx->Driver.StencilFuncSeparate( ctx, GL_FRONT,
-                                    ctx->Stencil.Function[0],
-                                    ctx->Stencil.Ref[0],
-                                    ctx->Stencil.ValueMask[0] );
-   ctx->Driver.StencilFuncSeparate( ctx, GL_BACK,
-                                    ctx->Stencil.Function[1],
-                                    ctx->Stencil.Ref[1],
-                                    ctx->Stencil.ValueMask[1] );
-   ctx->Driver.StencilMaskSeparate( ctx, GL_FRONT, ctx->Stencil.WriteMask[0] );
-   ctx->Driver.StencilMaskSeparate( ctx, GL_BACK, ctx->Stencil.WriteMask[1] );
-   ctx->Driver.StencilOpSeparate( ctx, GL_FRONT,
-                                  ctx->Stencil.FailFunc[0],
-                                  ctx->Stencil.ZFailFunc[0],
-                                  ctx->Stencil.ZPassFunc[0]);
-   ctx->Driver.StencilOpSeparate( ctx, GL_BACK,
-                                  ctx->Stencil.FailFunc[1],
-                                  ctx->Stencil.ZFailFunc[1],
-                                  ctx->Stencil.ZPassFunc[1]);
-
-
-   ctx->Driver.DrawBuffer( ctx, ctx->Color.DrawBuffer[0] );
-}
diff --git a/src/mesa/drivers/dri/i915/intel_context.h b/src/mesa/drivers/dri/i915/intel_context.h
index 1234a8241dc..8e6ec161802 100644
--- a/src/mesa/drivers/dri/i915/intel_context.h
+++ b/src/mesa/drivers/dri/i915/intel_context.h
@@ -48,28 +48,22 @@
 #define DV_PF_8888 (3<<8)
 
 struct intel_region;
+struct intel_context;
 
-typedef struct intel_context *intelContextPtr;
-
-typedef void (*intel_tri_func)(intelContextPtr, intelVertex *, intelVertex *,
+typedef void (*intel_tri_func)(struct intel_context *, intelVertex *, intelVertex *,
 							  intelVertex *);
-typedef void (*intel_line_func)(intelContextPtr, intelVertex *, intelVertex *);
-typedef void (*intel_point_func)(intelContextPtr, intelVertex *);
+typedef void (*intel_line_func)(struct intel_context *, intelVertex *, intelVertex *);
+typedef void (*intel_point_func)(struct intel_context *, intelVertex *);
 
 #define INTEL_FALLBACK_DRAW_BUFFER	 0x1
 #define INTEL_FALLBACK_READ_BUFFER	 0x2
 #define INTEL_FALLBACK_USER		 0x4
-#define INTEL_FALLBACK_NO_BATCHBUFFER	 0x8
-#define INTEL_FALLBACK_NO_TEXMEM	 0x10
-#define INTEL_FALLBACK_RENDERMODE	 0x20
+#define INTEL_FALLBACK_RENDERMODE	 0x8
 
-extern void intelFallback( intelContextPtr intel, GLuint bit, GLboolean mode );
+extern void intelFallback( struct intel_context *intel, GLuint bit, GLboolean mode );
 #define FALLBACK( intel, bit, mode ) intelFallback( intel, bit, mode )
 
 
-#define INTEL_TEX_MAXLEVELS 10
-
-
 
 struct intel_texture_object
 {
@@ -93,7 +87,8 @@ struct intel_texture_object
 
 
 
-struct intel_texture_image {
+struct intel_texture_image 
+{
    struct gl_texture_image base;
 
    /* These aren't stored in gl_texture_image 
@@ -109,12 +104,6 @@ struct intel_texture_image {
 };
 
 
-struct intel_reloc {
-   GLuint *value;
-   GLuint delta;
-   GLuint *dest;
-};
-
 #define INTEL_MAX_FIXUP 64
 
 struct intel_context
@@ -122,24 +111,59 @@ struct intel_context
    GLcontext ctx;		/* the parent class */
 
    struct {
-      void (*destroy)( intelContextPtr intel ); 
-      void (*emit_state)( intelContextPtr intel );
-      void (*emit_invarient_state)( intelContextPtr intel );
-      void (*lost_hardware)( intelContextPtr intel );
-      void (*update_texture_state)( intelContextPtr intel );
+      void (*destroy)( struct intel_context *intel ); 
+      void (*emit_state)( struct intel_context *intel );
+      void (*emit_invarient_state)( struct intel_context *intel );
+      void (*lost_hardware)( struct intel_context *intel );
+      void (*update_texture_state)( struct intel_context *intel );
 
-      void (*render_start)( intelContextPtr intel );
-      void (*set_draw_offset)( intelContextPtr intel, int offset );
-      void (*emit_flush)( intelContextPtr intel );
+      void (*render_start)( struct intel_context *intel );
+      void (*set_draw_region)( struct intel_context *intel, 
+			       struct intel_region *draw_region,
+			       struct intel_region *depth_region );
+      void (*emit_flush)( struct intel_context *intel );
 
-      void (*reduced_primitive_state)( intelContextPtr intel, GLenum rprim );
+      void (*reduced_primitive_state)( struct intel_context *intel, GLenum rprim );
 
-      GLboolean (*check_vertex_size)( intelContextPtr intel, GLuint expected );
+      GLboolean (*check_vertex_size)( struct intel_context *intel, GLuint expected );
 
-      void (*clear_with_tris)( intelContextPtr intel, GLbitfield mask,
+      void (*clear_with_tris)( struct intel_context *intel, GLbitfield mask,
 			       GLboolean all, 
 			       GLint cx, GLint cy, GLint cw, GLint ch);
 
+
+      /* Metaops: 
+       */
+      void (*install_meta_state)( struct intel_context *intel );
+      void (*leave_meta_state)( struct intel_context *intel );
+
+      void (*meta_draw_region)( struct intel_context *intel,
+				struct intel_region *draw_region,
+				struct intel_region *depth_region );
+
+      void (*meta_color_mask)( struct intel_context *intel,
+			       GLboolean );
+      
+      void (*meta_stencil_replace)( struct intel_context *intel,
+				    GLuint mask,
+				    GLuint clear );
+
+      void (*meta_no_depth_stencil_write)( struct intel_context *intel );
+
+      void (*meta_no_texture)( struct intel_context *intel );
+      void (*meta_texture_blend_replace)( struct intel_context *intel );
+
+      void (*meta_tex_rect_source)( struct intel_context *intel,
+				    struct intel_region *region,
+				    GLuint textureFormat );
+
+      void (*meta_draw_format)( struct intel_context *intel,
+				GLuint format,
+				GLuint depth_format );
+      
+      
+
+
    } vtbl;
 
    GLint refcount;   
@@ -148,42 +172,24 @@ struct intel_context
    
    GLuint last_fence;
 
-   struct {
-      GLuint start_offset;
-      GLint size;
-      GLint space;
-      GLubyte *ptr;
-   } batch;
-      
-#define INTEL_ALLOC_NR 64
-#define INTEL_ALLOC_SIZE 4096
-
-   struct {
-      GLuint buffer[INTEL_ALLOC_NR];
-      GLuint current;
-   } alloc;
+   struct intel_batchbuffer *batch;
 
    struct {
       GLuint id;
       GLuint primitive;
       GLubyte *start_ptr;      
-      void (*flush)( GLcontext * );
+      void (*flush)( struct intel_context * );
    } prim;
 
    GLboolean locked;
 
-   GLubyte clear_red;
-   GLubyte clear_green;
-   GLubyte clear_blue;
-   GLubyte clear_alpha;
    GLuint ClearColor;
    GLuint ClearDepth;
 
+   /* Offsets of fields within the current vertex:
+    */
    GLuint coloroffset;
    GLuint specoffset;
-
-   /* Support for duplicating XYZW as WPOS parameter (crutch for I915).
-    */
    GLuint wpos_offset;
    GLuint wpos_size;
 
@@ -201,10 +207,6 @@ struct intel_context
    /* AGP memory buffer manager:
     */
    struct bufmgr *bm;
-   struct bm_buffer_list *buffer_list;
-
-   struct intel_reloc fixup[INTEL_MAX_FIXUP];
-   GLuint nr_fixups;
 
 
    /* State for intelvb.c and inteltris.c.
@@ -231,7 +233,6 @@ struct intel_context
 
    /* These refer to the current draw (front vs. back) buffer:
     */
-   GLuint drawOffset;		/* agp offset of drawbuffer */
    int drawX;			/* origin of drawable in draw buffer */
    int drawY;
    GLuint numClipRects;		/* cliprects for that buffer */
@@ -347,8 +348,7 @@ do {									\
 
 #define INTEL_FIREVERTICES(intel)		\
 do {						\
-   if ((intel)->prim.flush)			\
-      (intel)->prim.flush(&(intel)->ctx);		\
+   assert(!(intel)->prim.flush);		\
 } while (0)
 
 /* ================================================================
@@ -382,7 +382,7 @@ do {						\
  * than COPY_DWORDS would:
  */
 #if defined(i386) || defined(__i386__)
-static __inline__ void * __memcpy(void * to, const void * from, size_t n)
+static inline void * __memcpy(void * to, const void * from, size_t n)
 {
    int d0, d1, d2;
    __asm__ __volatile__(
@@ -443,21 +443,19 @@ extern int INTEL_DEBUG;
  * intel_context.c:
  */
 
-extern void intelInitDriverFunctions( struct dd_function_table *functions );
-
-extern GLboolean intelInitContext( intelContextPtr intel, 
+extern GLboolean intelInitContext( struct intel_context *intel, 
 				   const __GLcontextModes *mesaVis,
 				   __DRIcontextPrivate *driContextPriv,
 				   void *sharedContextPrivate,
 				   struct dd_function_table *functions );
 
-extern void intelGetLock(intelContextPtr intel, GLuint flags);
-extern void intelSetBackClipRects(intelContextPtr intel);
-extern void intelSetFrontClipRects(intelContextPtr intel);
-extern void intelWindowMoved( intelContextPtr intel );
+extern void intelGetLock(struct intel_context *intel, GLuint flags);
 
 extern void intelInitState( GLcontext *ctx );
-extern const GLubyte *intelGetString( GLcontext *ctx, GLenum name );
+extern void intelFinish( GLcontext *ctx );
+extern void intelFlush( GLcontext *ctx );
+
+extern void intelInitDriverFunctions( struct dd_function_table *functions );
 
 
 /* ================================================================
@@ -517,6 +515,8 @@ extern void intelInitStateFuncs( struct dd_function_table *functions );
 #define BLENDFACT_INV_CONST_ALPHA	0x0f
 #define BLENDFACT_MASK          	0x0f
 
+#define MI_BATCH_BUFFER_END 	(0xA<<23)
+
 
 extern int intel_translate_compare_func( GLenum func );
 extern int intel_translate_stencil_op( GLenum op );
@@ -525,28 +525,19 @@ extern int intel_translate_logic_op( GLenum opcode );
 
 
 /* ================================================================
- * intel_ioctl.c:
+ * intel_buffers.c:
  */
-extern void intel_dump_batchbuffer( long offset,
-				    int *ptr,
-				    int count );
-
-
-/* ================================================================
- * intel_pixel.c:
- */	
-extern void intelInitPixelFuncs( struct dd_function_table *functions );
-
-GLboolean intel_check_color_per_fragment_ops( const GLcontext *ctx );
-
-GLboolean intel_clip_to_framebuffer( GLcontext *ctx,
-				     const GLframebuffer *buffer,
-				     GLint *x, GLint *y,
-				     GLsizei *width, GLsizei *height );
+void intelInitBufferFuncs( struct dd_function_table *functions );
 
 struct intel_region *intel_readbuf_region( struct intel_context *intel );
 struct intel_region *intel_drawbuf_region( struct intel_context *intel );
 
+extern void intelWindowMoved( struct intel_context *intel );
+
+extern GLboolean intel_intersect_cliprects( drm_clip_rect_t *dest,
+					    const drm_clip_rect_t *a,
+					    const drm_clip_rect_t *b );
+
 
 
 /*======================================================================
@@ -568,8 +559,6 @@ static inline struct intel_texture_image *intel_texture_image( struct gl_texture
    return (struct intel_texture_image *)img;
 }
 
-#define INTEL_CONTEXT(ctx)	intel_context(ctx)
-
 
 #endif
 
diff --git a/src/mesa/drivers/dri/i915/intel_ioctl.c b/src/mesa/drivers/dri/i915/intel_ioctl.c
index 4080782b5a4..4569c57d9a4 100644
--- a/src/mesa/drivers/dri/i915/intel_ioctl.c
+++ b/src/mesa/drivers/dri/i915/intel_ioctl.c
@@ -44,7 +44,7 @@
 #include "bufmgr.h"
 
 
-int intelEmitIrqLocked( intelContextPtr intel )
+int intelEmitIrqLocked( struct intel_context *intel )
 {
    drmI830IrqEmit ie;
    int ret, seq = 0;
@@ -54,14 +54,12 @@ int intelEmitIrqLocked( intelContextPtr intel )
 
    ie.irq_seq = &seq;
 
-#if 1
    ret = drmCommandWriteRead( intel->driFd, DRM_I830_IRQ_EMIT, 
 			      &ie, sizeof(ie) );
    if ( ret ) {
       fprintf( stderr, "%s: drmI830IrqEmit: %d\n", __FUNCTION__, ret );
       exit(1);
    }   
-#endif
 
    if (0)
       fprintf(stderr, "%s -->  %d\n", __FUNCTION__, seq );
@@ -69,7 +67,7 @@ int intelEmitIrqLocked( intelContextPtr intel )
    return seq;
 }
 
-void intelWaitIrq( intelContextPtr intel, int seq )
+void intelWaitIrq( struct intel_context *intel, int seq )
 {
    drmI830IrqWait iw;
    int ret;
@@ -79,7 +77,6 @@ void intelWaitIrq( intelContextPtr intel, int seq )
 
    iw.irq_seq = seq;
 	 
-#if 1
    do {
       ret = drmCommandWrite( intel->driFd, DRM_I830_IRQ_WAIT, &iw, sizeof(iw) );
    } while (ret == -EAGAIN || ret == -EINTR);
@@ -88,294 +85,61 @@ void intelWaitIrq( intelContextPtr intel, int seq )
       fprintf( stderr, "%s: drmI830IrqWait: %d\n", __FUNCTION__, ret );
       exit(1);
    }
-#endif
-}
-
-
-
-void intel_dump_batchbuffer( long offset,
-			     int *ptr,
-			     int count )
-{
-   int i;
-   fprintf(stderr, "\n\n\nSTART BATCH (%d dwords):\n", count);
-   for (i = 0; i < count/4; i += 4) 
-      fprintf(stderr, "\t0x%08x 0x%08x 0x%08x 0x%08x\n", 
-/* 	      (unsigned int)offset + i*4,  */
-	      ptr[i], ptr[i+1], ptr[i+2], ptr[i+3]);
-   fprintf(stderr, "END BATCH\n\n\n");
 }
 
 
-#define MI_BATCH_BUFFER_END 	(0xA<<23)
-
-
-void intelFlushBatchLocked( intelContextPtr intel, 
-			    GLboolean ignore_cliprects,
-			    GLboolean refill,
-			    GLboolean allow_unlock)
+void intel_batch_ioctl( struct intel_context *intel, 
+			GLuint start_offset,
+			GLuint used,
+			GLboolean ignore_cliprects)
 {
    drmI830BatchBuffer batch;
 
    assert(intel->locked);
-   assert(intel->buffer_list);
-   assert(intel->batch.ptr);
-
+   assert(used);
 
    if (0)
-      fprintf(stderr, "%s used %d of %d offset %x..%x refill %d\n",
+      fprintf(stderr, "%s used %d offset %x..%x ignore_cliprects %d\n",
 	      __FUNCTION__, 
-	      (intel->batch.size - intel->batch.space), 
-	      intel->batch.size,
-	      intel->batch.start_offset,
-	      intel->batch.start_offset + 
-	      (intel->batch.size - intel->batch.space), 
-	      refill);
+	      used, 
+	      start_offset,
+	      start_offset + used,
+	      ignore_cliprects);
 
    /* Throw away non-effective packets.  Won't work once we have
     * hardware contexts which would preserve statechanges beyond a
     * single buffer.
     */
    if (intel->numClipRects == 0 && !ignore_cliprects) {
-      /* Note that any state thought to have been emitted actually
-       * hasn't:
-       */
-      intel->batch.ptr -= (intel->batch.size - intel->batch.space);
-      intel->batch.space = intel->batch.size;
       intel->vtbl.lost_hardware( intel ); 
+      return;
    }
 
-   if (intel->batch.space != intel->batch.size) {
-      batch.start = intel->batch.start_offset;
-      batch.used = intel->batch.size - intel->batch.space;
-      batch.cliprects = intel->pClipRects;
-      batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects;
-      batch.DR1 = 0;
-      batch.DR4 = ((((GLuint)intel->drawX) & 0xffff) | 
-		   (((GLuint)intel->drawY) << 16));
+   batch.start = start_offset;
+   batch.used = used;
+   batch.cliprects = intel->pClipRects;
+   batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects;
+   batch.DR1 = 0;
+   batch.DR4 = ((((GLuint)intel->drawX) & 0xffff) | 
+		(((GLuint)intel->drawY) << 16));
       
-      if ((batch.used & 0x4) == 0) {
-	 ((int *)intel->batch.ptr)[0] = 0;
-	 ((int *)intel->batch.ptr)[1] = MI_BATCH_BUFFER_END;
-	 batch.used += 0x8;
-	 intel->batch.ptr += 0x8;
-      }
-      else {
-	 ((int *)intel->batch.ptr)[0] = MI_BATCH_BUFFER_END;
-	 batch.used += 0x4;
-	 intel->batch.ptr += 0x4;
-      }      
-
-      if (0)
- 	 intel_dump_batchbuffer( batch.start,
-				 (int *)(intel->batch.ptr - batch.used),
-				 batch.used );
-
-      if (0)
-	 fprintf(stderr, "%s: 0x%x..0x%x DR4: %x cliprects: %d\n",
-		 __FUNCTION__, 
-		 batch.start, 
-		 batch.start + batch.used * 4,
-		 batch.DR4, batch.num_cliprects);
-
+   if (0)
+      fprintf(stderr, "%s: 0x%x..0x%x DR4: %x cliprects: %d\n",
+	      __FUNCTION__, 
+	      batch.start, 
+	      batch.start + batch.used * 4,
+	      batch.DR4, batch.num_cliprects);
 #if 1
-      if (drmCommandWrite (intel->driFd, DRM_I830_BATCHBUFFER, &batch, 
-			   sizeof(batch))) {
-	 fprintf(stderr, "DRM_I830_BATCHBUFFER: %d\n",  -errno);
-	 UNLOCK_HARDWARE(intel);
-	 exit(1);
-      }
-#endif
-      
-      /* FIXME: use hardware contexts to avoid 'losing' hardware after
-       * each buffer flush.
-       */
-      intel->vtbl.lost_hardware( intel );
-   }
-
-   bmUnmapBuffer( intel->bm,
-		  intel->alloc.buffer[intel->alloc.current] );
-   intel->batch.ptr = NULL;
-   intel->batch.size = 0;
-   intel->batch.space = 0;
-
-   intel->last_fence = bmFenceBufferList(intel->bm, intel->buffer_list);
-   bmFreeBufferList(intel->buffer_list);
-
-
-   intel->buffer_list = NULL;
-}
-
-void intelFlushBatch( intelContextPtr intel, GLboolean refill )
-{
-   if (intel->locked) {
-      intelFlushBatchLocked( intel, GL_FALSE, refill, GL_FALSE );
-   } 
-   else {
-      assert(intel->batch.size == intel->batch.space);
-   }
-}
-
-
-
-
-
-static void wait_for_idle_locked( struct intel_context *intel )
-{
-   intelInstallBatchBuffer(intel);
-
-   bmAddBuffer(intel->buffer_list, intel->draw_region->buffer,
-	       BM_WRITE, NULL, NULL);
-   
-   intelValidateBuffers(intel);
-
-   intel->vtbl.emit_flush( intel );
-   intelFlushBatch( intel, GL_TRUE );
-
-   bmFinishFence( intel->bm, intel->last_fence );
-}
-
-void intelWaitForIdle( intelContextPtr intel )
-{   
-   if (intel->locked) {
-      wait_for_idle_locked( intel );
-   }
-   else {
-      LOCK_HARDWARE(intel);
-      wait_for_idle_locked( intel );
+   if (drmCommandWrite (intel->driFd, DRM_I830_BATCHBUFFER, &batch, 
+			sizeof(batch))) {
+      fprintf(stderr, "DRM_I830_BATCHBUFFER: %d\n",  -errno);
       UNLOCK_HARDWARE(intel);
-   }
-}
-
-
-
-void intelFlush( GLcontext *ctx )
-{
-   intelContextPtr intel = INTEL_CONTEXT( ctx );
-
-   if (intel->Fallback)
-      _swrast_flush( ctx );
-
-   INTEL_FIREVERTICES( intel );
-
-   if (intel->batch.size != intel->batch.space)
-      intelFlushBatch( intel, GL_FALSE );
-}
-
-void intelFinish( GLcontext *ctx  ) 
-{
-   intelContextPtr intel = INTEL_CONTEXT( ctx );
-   intelFlush( ctx );
-   intelWaitForIdle( intel );
-}
-
-
-void intelClear(GLcontext *ctx, GLbitfield mask, GLboolean all,
-		GLint cx, GLint cy, GLint cw, GLint ch)
-{
-   intelContextPtr intel = INTEL_CONTEXT( ctx );
-   const GLuint colorMask = *((GLuint *) &ctx->Color.ColorMask);
-   GLbitfield tri_mask = 0;
-   GLbitfield blit_mask = 0;
-   GLbitfield swrast_mask = 0;
-
-   if (0)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   /* Take care of cliprects, which are handled differently for
-    * clears, etc.
-    */
-   intelFlush( &intel->ctx );
-
-   if (mask & BUFFER_BIT_FRONT_LEFT) {
-      if (colorMask == ~0) {
-	 blit_mask |= BUFFER_BIT_FRONT_LEFT;
-      } 
-      else {
-	 tri_mask |= BUFFER_BIT_FRONT_LEFT;
-      }
-   }
-
-   if (mask & BUFFER_BIT_BACK_LEFT) {
-      if (colorMask == ~0) {
-	 blit_mask |= BUFFER_BIT_BACK_LEFT;
-      } 
-      else {
-	 tri_mask |= BUFFER_BIT_BACK_LEFT;
-      }
-   }
-
-   if (mask & BUFFER_BIT_DEPTH) {
-      blit_mask |= BUFFER_BIT_DEPTH;
-   }
-
-   if (mask & BUFFER_BIT_STENCIL) {
-      if (!intel->hw_stencil) {
-	 swrast_mask |= BUFFER_BIT_STENCIL;
-      }
-      else if (ctx->Stencil.WriteMask[0] != 0xff) {
-	 tri_mask |= BUFFER_BIT_STENCIL;
-      } 
-      else {
-	 blit_mask |= BUFFER_BIT_STENCIL;
-      }
-   }
-
-   swrast_mask |= (mask & BUFFER_BIT_ACCUM);
-
-   if (blit_mask) 
-      intelClearWithBlit( ctx, blit_mask, all, cx, cy, cw, ch );
-
-   if (tri_mask) 
-      intel->vtbl.clear_with_tris( intel, tri_mask, all, cx, cy, cw, ch);
-
-   if (swrast_mask)
-      _swrast_Clear( ctx, swrast_mask, all, cx, cy, cw, ch );
-}
-
-
-
-
-
-
-
-/* Flip the front & back buffes
- */
-void intelPageFlip( const __DRIdrawablePrivate *dPriv )
-{
-#if 0
-   intelContextPtr intel;
-   int tmp, ret;
-
-   if (INTEL_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   assert(dPriv);
-   assert(dPriv->driContextPriv);
-   assert(dPriv->driContextPriv->driverPrivate);
-
-   intel = (intelContextPtr) dPriv->driContextPriv->driverPrivate;
-
-   intelFlush( &intel->ctx );
-   LOCK_HARDWARE( intel );
-
-   if (dPriv->pClipRects) {
-      *(drm_clip_rect_t *)intel->sarea->boxes = dPriv->pClipRects[0];
-      intel->sarea->nbox = 1;
-   }
-
-   ret = drmCommandNone(intel->driFd, DRM_I830_FLIP); 
-   if (ret) {
-      fprintf(stderr, "%s: %d\n", __FUNCTION__, ret);
-      UNLOCK_HARDWARE( intel );
       exit(1);
    }
-
-   tmp = intel->sarea->last_enqueue;
-   intelRefillBatchLocked( intel );
-   UNLOCK_HARDWARE( intel );
-
-
-   intelSetDrawBuffer( &intel->ctx, intel->ctx.Color.DriverDrawBuffer );
 #endif
+      
+   /* FIXME: use hardware contexts to avoid 'losing' hardware after
+    * each buffer flush.
+    */
+   intel->vtbl.lost_hardware( intel );
 }
diff --git a/src/mesa/drivers/dri/i915/intel_ioctl.h b/src/mesa/drivers/dri/i915/intel_ioctl.h
index 89d2be06323..955681774b5 100644
--- a/src/mesa/drivers/dri/i915/intel_ioctl.h
+++ b/src/mesa/drivers/dri/i915/intel_ioctl.h
@@ -30,25 +30,13 @@
 
 #include "intel_context.h"
 
-extern void intelWaitAgeLocked( intelContextPtr intel, int age, GLboolean unlock );
+void intelWaitIrq( struct intel_context *intel, int seq );
+int intelEmitIrqLocked( struct intel_context *intel );
 
-extern void intelClear(GLcontext *ctx, GLbitfield mask, GLboolean all,
-		       GLint cx, GLint cy, GLint cw, GLint ch);
-
-extern void intelPageFlip( const __DRIdrawablePrivate *dpriv );
-extern void intelWaitForIdle( intelContextPtr intel );
-extern void intelFlushBatch( intelContextPtr intel, GLboolean refill );
-extern void intelFlushBatchLocked( intelContextPtr intel,
-				   GLboolean ignore_cliprects,
-				   GLboolean refill,
-				   GLboolean allow_unlock);
-extern void intelRefillBatchLocked( intelContextPtr intel, GLboolean allow_unlock );
-extern void intelFinish( GLcontext *ctx );
-extern void intelFlush( GLcontext *ctx );
-
-
-void intelWaitIrq( intelContextPtr intel, int seq );
-int intelEmitIrqLocked( intelContextPtr intel );
+void intel_batch_ioctl( struct intel_context *intel, 
+			GLuint start_offset,
+			GLuint used,
+			GLboolean ignore_cliprects);
 
 
 #endif
diff --git a/src/mesa/drivers/dri/i915/intel_pixel.c b/src/mesa/drivers/dri/i915/intel_pixel.c
index b2484975ccf..e38c6be576b 100644
--- a/src/mesa/drivers/dri/i915/intel_pixel.c
+++ b/src/mesa/drivers/dri/i915/intel_pixel.c
@@ -1,78 +1,40 @@
-/**************************************************************************
- * 
- * 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 "glheader.h"
-#include "enums.h"
-#include "mtypes.h"
-#include "macros.h"
 #include "swrast/swrast.h"
-
-#include "intel_screen.h"
 #include "intel_context.h"
-#include "intel_ioctl.h"
-#include "intel_batchbuffer.h"
-#include "intel_blit.h"
-#include "intel_regions.h"
-#include "bufmgr.h"
+#include "intel_pixel.h"
 
-static GLboolean
-check_color( GLcontext *ctx, GLenum type, GLenum format,
-	     const struct gl_pixelstore_attrib *packing,
-	     const void *pixels, GLint sz, GLint pitch )
-{
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
-   GLuint cpp = intel->intelScreen->cpp;
 
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
 
-   if (	(pitch & 63) ||
-	ctx->_ImageTransferState ||
-	packing->SwapBytes ||
-	packing->LsbFirst) {
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-	 fprintf(stderr, "%s: failed 1\n", __FUNCTION__);
-      return GL_FALSE;
-   }
+#if 0
+struct intel_client_region *intel_pack_region( struct intel_context *intel,
+					       const struct gl_pixelstore_attrib *pack,
+					       GLenum format,
+					       GLenum type,
+					       GLvoid *pixels )
+{
+}
 
-   if ( type == GL_UNSIGNED_INT_8_8_8_8_REV && 
-	cpp == 4 && 
-	format == GL_BGRA ) {
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-	 fprintf(stderr, "%s: passed 2\n", __FUNCTION__);
-      return GL_TRUE;
-   }
 
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      fprintf(stderr, "%s: failed\n", __FUNCTION__);
+struct intel_client_region *intel_unpack_region( struct intel_context *intel,
+						 const struct gl_pixelstore_attrib *pack,
+						 GLenum format,
+						 GLenum type,
+						 GLvoid *pixels )
+{
+   GLint pitch = unpack->RowLength ? unpack->RowLength : width;
+   /* XXX: Need to adjust pixels pointer for unpack->skip pixels/rows
+    * offsets.
+    */
 
-   return GL_FALSE;
 }
 
+void release_client_region( struct intel_context *intel,
+			    struct intel_region *region )
+{
+}
+#endif
+
+
+
 GLboolean intel_check_color_per_fragment_ops( const GLcontext *ctx )
 {
    return !(ctx->Color.AlphaEnabled || 
@@ -88,6 +50,33 @@ GLboolean intel_check_color_per_fragment_ops( const GLcontext *ctx )
 	    ctx->Texture._EnabledUnits);
 }
 
+#if 0
+/* The intel_region struct doesn't really do enough to capture the
+ * format of the pixels in the region.  For now this code assumes that
+ * the region is a display surface and hence is either ARGB8888 or
+ * RGB565.
+ */
+GLboolean intel_check_blit_format( struct intel_region *region,
+				   struct intel_client_region *client_region )
+{
+   if (region->cpp == 4 
+       client_region->cpp == 4 && 
+       client_region->type == GL_UNSIGNED_INT_8_8_8_8_REV &&        
+       client_region->format == GL_BGRA ) {
+      return GL_TRUE;
+   }
+   
+   if (region->cpp == 2 && 
+       client_region->cpp == 2 &&
+       client_region->type == GL_UNSIGNED_INT_5_6_5_REV && 
+       client_region->format == GL_BGR ) {
+      return GL_TRUE;
+   }
+   
+   fprintf(stderr, "%s: request doesn't match pixel format\n", __FUNCTION__);
+   return GL_FALSE;
+}
+#endif
 
 
 GLboolean intel_clip_to_framebuffer( GLcontext *ctx,
@@ -124,496 +113,6 @@ GLboolean intel_clip_to_framebuffer( GLcontext *ctx,
    return GL_TRUE;
 }
 
-static GLboolean
-intelTryReadPixels( GLcontext *ctx,
-		  GLint x, GLint y, GLsizei width, GLsizei height,
-		  GLenum format, GLenum type,
-		  const struct gl_pixelstore_attrib *pack,
-		  GLvoid *pixels )
-{
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
-   GLint size = 0;
-   GLint pitch = pack->RowLength ? pack->RowLength : width;
-
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   /* Only accelerate reading to agp buffers.
-    */
-   if ( 1 ) {
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-	 fprintf(stderr, "%s: dest not agp\n", __FUNCTION__);
-      return GL_FALSE;
-   }
-
-   /* Need GL_PACK_INVERT_MESA to cope with upsidedown results from
-    * blitter:
-    */
-   if (!pack->Invert) {
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-	 fprintf(stderr, "%s: MESA_PACK_INVERT not set\n", __FUNCTION__);
-      return GL_FALSE;
-   }
-
-   if (!check_color(ctx, type, format, pack, pixels, size, pitch))
-      return GL_FALSE;
-
-   switch ( intel->intelScreen->cpp ) {
-   case 4:
-      break;
-   default:
-      return GL_FALSE;
-   }
-
-
-   /* Although the blits go on the command buffer, need to do this and
-    * fire with lock held to guarentee cliprects and drawOffset are
-    * correct.
-    *
-    * This is an unusual situation however, as the code which flushes
-    * a full command buffer expects to be called unlocked.  As a
-    * workaround, immediately flush the buffer on aquiring the lock.
-    */
-   intelFlush( &intel->ctx );
-   LOCK_HARDWARE( intel );
-   {
-      __DRIdrawablePrivate *dPriv = intel->driDrawable;
-      int nbox = dPriv->numClipRects;
-      int src_offset = intel->drawOffset;
-      int src_pitch = intel->intelScreen->front.pitch;
-      int dst_offset = 0;
-      drm_clip_rect_t *box = dPriv->pClipRects;
-      int i;
-
-      if (!intel_clip_to_framebuffer(ctx, ctx->ReadBuffer, &x, &y, &width, &height)) {
-	 UNLOCK_HARDWARE( intel );
-	 if (INTEL_DEBUG & DEBUG_PIXEL)
-	    fprintf(stderr, "%s totally clipped -- nothing to do\n",
-		    __FUNCTION__);
-	 return GL_TRUE;
-      }
-
-
-      y = dPriv->h - y - height;
-      x += dPriv->x;
-      y += dPriv->y;
-
-
-      if (INTEL_DEBUG & DEBUG_PIXEL)
-	 fprintf(stderr, "readpixel blit src_pitch %d dst_pitch %d\n",
-		 src_pitch, pitch);
-
-      for (i = 0 ; i < nbox ; i++)
-      {
-	 GLint bx = box[i].x1;
-	 GLint by = box[i].y1;
-	 GLint bw = box[i].x2 - bx;
-	 GLint bh = box[i].y2 - by;
-	 
-	 if (bx < x) bw -= x - bx, bx = x;
-	 if (by < y) bh -= y - by, by = y;
-	 if (bx + bw > x + width) bw = x + width - bx;
-	 if (by + bh > y + height) bh = y + height - by;
-	 if (bw <= 0) continue;
-	 if (bh <= 0) continue;
-
-	 intelEmitCopyBlitLocked( intel,
-			    intel->intelScreen->cpp,
-			    src_pitch, src_offset,
-			    pitch, dst_offset,
-			    bx, by,
-			    bx - x, by - y,
-			    bw, bh );
-      }
-   }
-   UNLOCK_HARDWARE( intel );
-   intelFinish( &intel->ctx );
-
-   return GL_TRUE;
-}
-
-static 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__);
-
-   if (!intelTryReadPixels( ctx, x, y, width, height, format, type, pack, 
-			   pixels))
-      _swrast_ReadPixels( ctx, x, y, width, height, format, type, pack, 
-			  pixels);
-}
-
-
-
-
-static void do_draw_pix( GLcontext *ctx,
-			 GLint x, GLint y, GLsizei width, GLsizei height,
-			 GLint pitch,
-			 const void *pixels,
-			 GLuint dest )
-{
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
-   __DRIdrawablePrivate *dPriv = intel->driDrawable;
-   drm_clip_rect_t *box = dPriv->pClipRects;
-   int nbox = dPriv->numClipRects;
-   int i;
-   int src_offset = 0;
-   int src_pitch = pitch;
-
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   intelFlush( &intel->ctx );
-   LOCK_HARDWARE( intel );
-   if (ctx->DrawBuffer)
-   {
-      y -= height;			/* cope with pixel zoom */
-   
-      if (!intel_clip_to_framebuffer(ctx, ctx->DrawBuffer,
-			  &x, &y, &width, &height)) {
-	 UNLOCK_HARDWARE( intel );
-	 return;
-      }
-
-      y = dPriv->h - y - height; 	/* convert from gl to hardware coords */
-      x += dPriv->x;
-      y += dPriv->y;
-
-
-      for (i = 0 ; i < nbox ; i++ )
-      {
-	 GLint bx = box[i].x1;
-	 GLint by = box[i].y1;
-	 GLint bw = box[i].x2 - bx;
-	 GLint bh = box[i].y2 - by;
-
-	 if (bx < x) bw -= x - bx, bx = x;
-	 if (by < y) bh -= y - by, by = y;
-	 if (bx + bw > x + width) bw = x + width - bx;
-	 if (by + bh > y + height) bh = y + height - by;
-	 if (bw <= 0) continue;
-	 if (bh <= 0) continue;
-
-	 intelEmitCopyBlitLocked( intel,
-			    intel->intelScreen->cpp,
-			    src_pitch, src_offset,
-			    intel->intelScreen->front.pitch,
-			      intel->drawOffset,
-			    bx - x, by - y,
-			    bx, by,
-			    bw, bh );
-      }
-   }
-   UNLOCK_HARDWARE( intel );
-   intelFinish( &intel->ctx );
-}
-
-
-
-
-static GLboolean
-intelTryDrawPixels( GLcontext *ctx,
-		  GLint x, GLint y, GLsizei width, GLsizei height,
-		  GLenum format, GLenum type,
-		  const struct gl_pixelstore_attrib *unpack,
-		  const GLvoid *pixels )
-{
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
-   GLint pitch = unpack->RowLength ? unpack->RowLength : width;
-   GLuint dest;
-   GLuint cpp = intel->intelScreen->cpp;
-   GLint size = width * pitch * cpp;
-
-   /* XXX: Need to adjust pixels pointer for unpack->skip pixels/rows
-    * offsets.
-    */
-
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   switch (format) {
-   case GL_RGB:
-   case GL_RGBA:
-   case GL_BGRA:
-      dest = intel->drawOffset;
-
-      /* Planemask doesn't have full support in blits.
-       */
-      if (!ctx->Color.ColorMask[RCOMP] ||
-	  !ctx->Color.ColorMask[GCOMP] ||
-	  !ctx->Color.ColorMask[BCOMP] ||
-	  !ctx->Color.ColorMask[ACOMP]) {
-	 if (INTEL_DEBUG & DEBUG_PIXEL)
-	    fprintf(stderr, "%s: planemask\n", __FUNCTION__);
-	 return GL_FALSE;	
-      }
-
-      /* Can't do conversions on agp reads/draws. 
-       */
-      if ( 1 ) {
-	 if (INTEL_DEBUG & DEBUG_PIXEL)
-	    fprintf(stderr, "%s: not agp memory\n", __FUNCTION__);
-	 return GL_FALSE;
-      }
-
-      if (!check_color(ctx, type, format, unpack, pixels, size, pitch)) {
-	 return GL_FALSE;
-      }
-      if (!intel_check_color_per_fragment_ops(ctx)) {
-	 return GL_FALSE;
-      }
-      if (!ctx->Current.RasterPosValid)
-	 return GL_FALSE;
-
-      if (ctx->Pixel.ZoomX != 1.0F ||
-	  ctx->Pixel.ZoomY != -1.0F)
-	 return GL_FALSE;
-      break;
-
-   default:
-      return GL_FALSE;
-   }
-
-   if ( 0 )
-   {
-      do_draw_pix( ctx, x, y, width, height, pitch, pixels,
-		   dest );
-      return GL_TRUE;
-   }
-   else if (0)
-   {
-      /* Pixels is in regular memory -- get dma buffers and perform
-       * upload through them.  No point doing this for regular uploads
-       * but once we remove some of the restrictions above (colormask,
-       * pixelformat conversion, zoom?, etc), this could be a win.
-       */
-   }
-   else
-      return GL_FALSE;
-}
-
-static void
-intelDrawPixels( GLcontext *ctx,
-		 GLint x, GLint y, GLsizei width, GLsizei height,
-		 GLenum format, GLenum type,
-		 const struct gl_pixelstore_attrib *unpack,
-		 const GLvoid *pixels )
-{
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (!intelTryDrawPixels( ctx, x, y, width, height, format, type,
-			  unpack, pixels ))
-      _swrast_DrawPixels( ctx, x, y, width, height, format, type,
-			  unpack, pixels );
-}
-
-
-struct intel_region *intel_drawbuf_region( struct intel_context *intel )
-{
-   switch (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0]) {
-   case BUFFER_BIT_FRONT_LEFT:
-      return intel->front_region;
-   case BUFFER_BIT_BACK_LEFT:
-      return intel->back_region;
-   default:
-      /* Not necessary to fallback - could handle either NONE or
-       * FRONT_AND_BACK cases below.
-       */
-      return NULL;		
-   }
-}
-
-struct intel_region *intel_readbuf_region( struct intel_context *intel )
-{
-   GLcontext *ctx = &intel->ctx;
-
-   /* This will have to change to support EXT_fbo's, but is correct
-    * for now:
-    */
-   switch (ctx->ReadBuffer->_ColorReadBufferIndex) {
-   case BUFFER_FRONT_LEFT:
-      return intel->front_region;
-   case BUFFER_BACK_LEFT:
-      return intel->back_region;
-   default:
-      assert(0);
-      return NULL;
-   }
-}
-
-
-
-
-
-
-/**
- * Implement glCopyPixels for the front color buffer (or back buffer Pixmap)
- * for the color buffer.  Don't support zooming, pixel transfer, etc.
- * We do support copying from one window to another, ala glXMakeCurrentRead.
- */
-static GLboolean intelTryCopyPixels( 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 = NULL;
-
-   /* Copypixels can be more than a straight copy.  Ensure all the
-    * extra operations are disabled:
-    */
-   if (!intel_check_color_per_fragment_ops(ctx) ||
-       ctx->_ImageTransferState ||
-       ctx->Pixel.ZoomX != 1.0F || 
-       ctx->Pixel.ZoomY != 1.0F)
-      return GL_FALSE;
-
-   switch (type) {
-   case GL_COLOR:
-      src = intel_readbuf_region( intel );
-      /* No readbuffer, copypixels is a noop: 
-       */
-      if (!src)
-	 return GL_TRUE;
-      break;
-   case GL_DEPTH:
-      /* Don't think this is really possible execpt at 16bpp, when we have no stencil.
-       */
-      if (intel->intelScreen->cpp == 2)
-	 src = intel->depth_region;
-      break;
-   case GL_STENCIL:
-      /* Don't think this is really possible. 
-       */
-      break;
-   case GL_DEPTH_STENCIL_EXT:
-      /* Does it matter whether it is stencil/depth or depth/stencil?
-       */
-      src = intel->depth_region;
-      break;
-   default:
-      break;
-   }
-
-   if (!src || !dst) 
-      return GL_FALSE;
-
-
-
-   intelFlush( &intel->ctx );
-   LOCK_HARDWARE( intel );
-   intelInstallBatchBuffer( intel );
-   {
-      __DRIdrawablePrivate *dPriv = intel->driDrawable;
-      drm_clip_rect_t *box = dPriv->pClipRects;
-      GLint nbox = dPriv->numClipRects;
-      GLint delta_x = srcx - dstx;
-      GLint delta_y = srcy - dsty;
-      GLuint dst_offset = 0;
-      GLuint src_offset = 0;
-      GLuint i;
-
-#if 0
-      dsty -= height;			/* cope with pixel zoom */
-      srcy -= height;			/* cope with pixel zoom */
-#endif
-      if (!ctx->DrawBuffer)
-	 goto out;
-
-      if (!intel_clip_to_framebuffer(ctx, ctx->DrawBuffer, &dstx, &dsty, &width, &height)) 
-	 goto out;
-
-      /* Update source for clipped dest.  Need to also clip the source rect.
-       */
-      srcx = dstx + delta_x;
-      srcy = dsty + delta_y;
-
-      if (!intel_clip_to_framebuffer(ctx, ctx->DrawBuffer, &srcx, &srcy, &width, &height)) 
-	 goto out;
-
-      /* Update dest for clipped source:
-       */
-      dstx = srcx - delta_x;
-      dsty = srcy - delta_y;
-
-
-      srcy = dPriv->h - srcy - height; 	/* convert from gl to hardware coords */
-      dsty = dPriv->h - dsty - height; 	/* convert from gl to hardware coords */
-      srcx += dPriv->x;
-      dstx += dPriv->x;
-      srcy += dPriv->y;
-      dsty += dPriv->y;
-
-
-      bmAddBuffer(intel->buffer_list, dst->buffer, BM_NO_EVICT|BM_NO_UPLOAD|BM_WRITE, 
-		  NULL, &dst_offset);
-      bmAddBuffer(intel->buffer_list, src->buffer, BM_NO_EVICT|BM_NO_UPLOAD|BM_READ, 
-		  NULL, &src_offset);
-
-      if (!bmValidateBufferList(intel->bm, intel->buffer_list, BM_MEM_AGP)) 
-	 goto out;
-      
-      /* Could do slightly more clipping: Eg, take the intersection of
-       * the existing set of cliprects and those cliprects translated
-       * by delta_x, delta_y:
-       * 
-       * This code will not overwrite other windows, but will
-       * introduce garbage when copying from obscured window regions.
-       */
-      for (i = 0 ; i < nbox ; i++ )
-      {
-	 GLint bx = box[i].x1;
-	 GLint by = box[i].y1;
-	 GLint bw = box[i].x2 - bx;
-	 GLint bh = box[i].y2 - by;
-
-	 if (bx < dstx) bw -= dstx - bx, bx = dstx;
-	 if (by < dsty) bh -= dsty - by, by = dsty;
-	 if (bx + bw > dstx + width) bw = dstx + width - bx;
-	 if (by + bh > dsty + height) bh = dsty + height - by;
-	 if (bw <= 0) continue;
-	 if (bh <= 0) continue;
-
-	 assert(dst_offset == intel->drawOffset);
-
-	 intelEmitCopyBlitLocked( intel,
-				  dst->cpp,
-				  src->pitch, src_offset,
-				  dst->pitch, dst_offset,
-				  bx + delta_x, by - delta_y, /* srcx, srcy */
-				  bx, by, /* dstx, dsty */
-				  bw, bh );
-      }
-   }
- out:
-   intelFlushBatchLocked( intel, GL_TRUE, GL_FALSE, GL_FALSE);
-   UNLOCK_HARDWARE( intel );
-   return GL_TRUE;
-}
-
-
-static void
-intelCopyPixels( GLcontext *ctx,
-		 GLint srcx, GLint srcy, GLsizei width, GLsizei height,
-		 GLint destx, GLint desty, GLenum type )
-{
-   if (INTEL_DEBUG & DEBUG_PIXEL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (!intelTryCopyPixels( ctx, srcx, srcy, width, height, destx, desty, type)) {
-/*       if (INTEL_DEBUG & DEBUG_FALLBACKS) */
-	 _mesa_printf("fallback to _swrast_CopyPixels\n");
-      _swrast_CopyPixels( ctx, srcx, srcy, width, height, destx, desty, type);
-   }
-}
 
 
 
@@ -635,3 +134,4 @@ void intelInitPixelFuncs( struct dd_function_table *functions )
       functions->DrawPixels = _swrast_DrawPixels;
    }
 }
+
diff --git a/src/mesa/drivers/dri/i915/intel_pixel.h b/src/mesa/drivers/dri/i915/intel_pixel.h
new file mode 100644
index 00000000000..de101d7f297
--- /dev/null
+++ b/src/mesa/drivers/dri/i915/intel_pixel.h
@@ -0,0 +1,63 @@
+/**************************************************************************
+ * 
+ * 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_PIXEL_H
+#define INTEL_PIXEL_H
+
+#include "mtypes.h"
+
+void intelInitPixelFuncs( struct dd_function_table *functions );
+
+GLboolean intel_check_color_per_fragment_ops( const GLcontext *ctx );
+
+GLboolean intel_clip_to_framebuffer( GLcontext *ctx,
+				     const GLframebuffer *buffer,
+				     GLint *x, GLint *y,
+				     GLsizei *width, GLsizei *height );
+
+void intelReadPixels( GLcontext *ctx,
+		      GLint x, GLint y, 
+		      GLsizei width, GLsizei height,
+		      GLenum format, GLenum type,
+		      const struct gl_pixelstore_attrib *pack,
+		      GLvoid *pixels );
+
+void intelDrawPixels( GLcontext *ctx,
+		      GLint x, GLint y, 
+		      GLsizei width, GLsizei height,
+		      GLenum format, 
+		      GLenum type,
+		      const struct gl_pixelstore_attrib *unpack,
+		      const GLvoid *pixels );
+
+void intelCopyPixels( GLcontext *ctx,
+		      GLint srcx, GLint srcy, 
+		      GLsizei width, GLsizei height,
+		      GLint destx, GLint desty, 
+		      GLenum type );
+
+#endif
diff --git a/src/mesa/drivers/dri/i915/intel_pixel_copy.c b/src/mesa/drivers/dri/i915/intel_pixel_copy.c
new file mode 100644
index 00000000000..4f4f7fbaa6b
--- /dev/null
+++ b/src/mesa/drivers/dri/i915/intel_pixel_copy.c
@@ -0,0 +1,196 @@
+/**************************************************************************
+ * 
+ * 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 "glheader.h"
+#include "enums.h"
+#include "mtypes.h"
+#include "macros.h"
+#include "swrast/swrast.h"
+
+#include "intel_screen.h"
+#include "intel_context.h"
+#include "intel_ioctl.h"
+#include "intel_batchbuffer.h"
+#include "intel_blit.h"
+#include "intel_regions.h"
+#include "intel_pixel.h"
+#include "bufmgr.h"
+
+
+
+
+static GLboolean do_texture_copypixels( GLcontext *ctx,
+					GLint srcx, GLint srcy, 
+					GLsizei width, GLsizei height,
+					GLint dstx, GLint dsty, 
+					GLenum type )
+{
+   return GL_FALSE;
+}
+
+
+/**
+ * CopyPixels with the blitter.  Don't support zooming, pixel transfer, etc.
+ */
+static GLboolean do_blit_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 = NULL;
+
+   /* Copypixels can be more than a straight copy.  Ensure all the
+    * extra operations are disabled:
+    */
+   if (!intel_check_color_per_fragment_ops(ctx) ||
+       ctx->Pixel.ZoomX != 1.0F || 
+       ctx->Pixel.ZoomY != 1.0F)
+      return GL_FALSE;
+
+   switch (type) {
+   case GL_COLOR:
+      src = intel_readbuf_region( intel );
+      break;
+   case GL_DEPTH:
+      /* Don't think this is really possible execpt at 16bpp, when we have no stencil.
+       */
+      if (intel->intelScreen->cpp == 2)
+	 src = intel->depth_region;
+      break;
+   case GL_STENCIL:
+      /* Don't think this is really possible. 
+       */
+      break;
+   case GL_DEPTH_STENCIL_EXT:
+      /* Does it matter whether it is stencil/depth or depth/stencil?
+       */
+      src = intel->depth_region;
+      break;
+   default:
+      break;
+   }
+
+   if (!src || !dst) 
+      return GL_FALSE;
+
+
+
+   intelFlush( &intel->ctx );
+   LOCK_HARDWARE( intel );
+   {
+      __DRIdrawablePrivate *dPriv = intel->driDrawable;
+      drm_clip_rect_t *box = dPriv->pClipRects;
+      GLint nbox = dPriv->numClipRects;
+      GLint delta_x = srcx - dstx;
+      GLint delta_y = srcy - dsty;
+      GLuint i;
+
+      if (!intel_clip_to_framebuffer(ctx, ctx->DrawBuffer, &dstx, &dsty, &width, &height)) 
+	 goto out;
+
+      /* Update source for clipped dest.  Need to also clip the source rect.
+       */
+      srcx = dstx + delta_x;
+      srcy = dsty + delta_y;
+
+      if (!intel_clip_to_framebuffer(ctx, ctx->DrawBuffer, &srcx, &srcy, &width, &height)) 
+	 goto out;
+
+      /* Update dest for clipped source:
+       */
+      dstx = srcx - delta_x;
+      dsty = srcy - delta_y;
+
+
+      srcy = dPriv->h - srcy - height; 	/* convert from gl to hardware coords */
+      dsty = dPriv->h - dsty - height; 	/* convert from gl to hardware coords */
+      srcx += dPriv->x;
+      dstx += dPriv->x;
+      srcy += dPriv->y;
+      dsty += dPriv->y;
+
+
+      /* Could do slightly more clipping: Eg, take the intersection of
+       * the existing set of cliprects and those cliprects translated
+       * by delta_x, delta_y:
+       * 
+       * This code will not overwrite other windows, but will
+       * introduce garbage when copying from obscured window regions.
+       */
+      for (i = 0 ; i < nbox ; i++ )
+      {
+	 GLint bx = box[i].x1;
+	 GLint by = box[i].y1;
+	 GLint bw = box[i].x2 - bx;
+	 GLint bh = box[i].y2 - by;
+
+	 if (bx < dstx) bw -= dstx - bx, bx = dstx;
+	 if (by < dsty) bh -= dsty - by, by = dsty;
+	 if (bx + bw > dstx + width) bw = dstx + width - bx;
+	 if (by + bh > dsty + height) bh = dsty + height - by;
+	 if (bw <= 0) continue;
+	 if (bh <= 0) continue;
+
+	 intelEmitCopyBlit( intel,
+			    dst->cpp,
+			    src->pitch, src->buffer, 0,
+			    dst->pitch, dst->buffer, 0,
+			    bx + delta_x, by - delta_y, /* srcx, srcy */
+			    bx, by, /* dstx, dsty */
+			    bw, bh );
+      }
+   }
+ out:
+   intel_batchbuffer_flush( intel->batch );
+   UNLOCK_HARDWARE( intel );
+   return GL_TRUE;
+}
+
+
+void intelCopyPixels( GLcontext *ctx,
+		      GLint srcx, GLint srcy, 
+		      GLsizei width, GLsizei height,
+		      GLint destx, GLint desty, 
+		      GLenum type )
+{
+   if (INTEL_DEBUG & DEBUG_PIXEL)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   if (do_blit_copypixels( ctx, srcx, srcy, width, height, destx, desty, type))
+      return;
+
+   if (do_texture_copypixels( ctx, srcx, srcy, width, height, destx, desty, type))
+      return;
+
+   _mesa_printf("fallback to _swrast_CopyPixels\n");
+   _swrast_CopyPixels( ctx, srcx, srcy, width, height, destx, desty, type);
+}
+
+
diff --git a/src/mesa/drivers/dri/i915/intel_pixel_draw.c b/src/mesa/drivers/dri/i915/intel_pixel_draw.c
new file mode 100644
index 00000000000..081e6ba69da
--- /dev/null
+++ b/src/mesa/drivers/dri/i915/intel_pixel_draw.c
@@ -0,0 +1,242 @@
+/**************************************************************************
+ * 
+ * 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 "glheader.h"
+#include "enums.h"
+#include "mtypes.h"
+#include "macros.h"
+#include "swrast/swrast.h"
+
+#include "intel_screen.h"
+#include "intel_context.h"
+#include "intel_ioctl.h"
+#include "intel_batchbuffer.h"
+#include "intel_blit.h"
+#include "intel_regions.h"
+#include "intel_pixel.h"
+#include "bufmgr.h"
+
+
+
+static GLboolean do_texture_draw_pixels( struct intel_context *intel,
+					 GLint x, GLint y, 
+					 GLsizei width, GLsizei height,
+					 GLenum format, GLenum type,
+					 const struct gl_pixelstore_attrib *unpack,
+					 const GLvoid *pixels )
+{
+#if 0
+   GLint pitch = unpack->RowLength ? unpack->RowLength : width;
+   __DRIdrawablePrivate *dPriv = intel->driDrawable;
+   int textureFormat;
+   GLenum glTextureFormat;
+
+   if (INTEL_DEBUG & DEBUG_PIXEL)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+
+   if (	ctx->_ImageTransferState ||
+	unpack->SwapBytes ||
+	unpack->LsbFirst ||
+	ctx->Texture._EnabledUnits ||
+	ctx->FragmentProgram._Enabled) {
+      fprintf(stderr, "%s: cannot use texture path\n", __FUNCTION__);
+      return GL_FALSE;
+   }
+   
+   glGenTextures();
+   glTextureImage2D();
+   glBindTexture();
+   glEnable(GL_TEXTURE_RECTANGLE_NV);
+   
+   glDisable(GL_POLYGON_STIPPLE);
+   glDisable(GL_CULL);
+
+   _mesa_install_vp_passthrough(ctx);
+   _mesa_push_current(ctx);
+   
+   if (intel->Fallback)
+      goto Fail;
+
+   glBegin(GL_QUADS);
+   glVertex3f();
+   glTexCoord2f();
+   glVertex3f();
+   glTexCoord2f();
+   glVertex3f();
+   glTexCoord2f();
+   glVertex3f();
+   glTexCoord2f();
+   glEnd();
+   glFinish(); 
+   
+   ASSIGN_4V(ctx->Current.Atrrib[VERT_ATTRIB_TEX0], tex0);
+
+ fail:
+   glDisable(GL_TEXTURE_RECTANGLE_NV);
+   glDeleteTextures();
+   glBindTexture(old);
+
+
+#endif
+   return GL_FALSE;
+}
+
+
+
+
+
+/* Pros:  
+ *   - no waiting for idle before updating framebuffer.
+ *   
+ * Cons:
+ *   - if upload is by memcpy, this may actually be slower than fallback path.
+ *   - uploads the whole image even if destination is clipped
+ *   
+ * Need to benchmark.
+ */
+static GLboolean do_blit_draw_pixels( struct intel_context *intel,
+				      GLint x, GLint y, 
+				      GLsizei width, GLsizei height,
+				      GLenum format, GLenum type,
+				      const struct gl_pixelstore_attrib *unpack,
+				      const GLvoid *pixels )
+{
+#if 0
+   struct intel_region *dest = intel_drawbuf_region(intel);
+   struct intel_region *src = NULL;
+
+   if (INTEL_DEBUG & DEBUG_PIXEL)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   if (!src || !dest)
+      return GL_FALSE;
+
+   if (!intel_check_blit_format(dest, src)) 
+      return GL_FALSE;
+
+   if (!intel_check_blit_fragment_ops(ctx)) 
+      return GL_FALSE;
+
+   if (ctx->Pixel.ZoomX != 1.0F)
+      return GL_FALSE;
+
+   if (ctx->Pixel.ZoomY == -1.0F)
+      y -= height;			
+   else if (ctx->Pixel.ZoomY == 1.0F) {
+      src_pitch = -src_pitch;
+      src_y += height;
+   }
+   else
+      return GL_FALSE;
+
+
+   if (unpack->BufferObj->Name) {
+      src = intel_bufferobj_unpack_region(intel, unpack, 
+					  width, height, 
+					  format, type, 
+					  unpack->BufferObj);
+      src_offset = (unsigned long)pixels;
+   }
+   else {
+      src = intel_client_unpack_region(intel, unpack, 
+				       width, height, 
+				       format, type, pixels);
+      src_offset = 0;
+   }
+
+
+   intelFlush( &intel->ctx );
+   LOCK_HARDWARE( intel );
+   {
+      __DRIdrawablePrivate *dPriv = intel->driDrawable;
+      int nbox = dPriv->numClipRects;
+      drm_clip_rect_t *box = dPriv->pClipRects;
+      drm_clip_rect_t rect;
+      int i;
+      
+      y = dPriv->h - y - height; 	/* convert from gl to hardware coords */
+      x += dPriv->x;
+      y += dPriv->y;
+
+
+      for (i = 0 ; i < nbox ; i++ )
+      {
+	 if (!intel_intersect_cliprects(rect, db_rect, &box[i]))
+	    continue;
+
+	 intelEmitCopyBlit( intel,
+			    intel->intelScreen->cpp,
+			    src->pitch, src->buffer, src_offset,
+			    dst->pitch, dst->buffer, 0,
+			    rect->x1 - x, 
+			    rect->y1 - y,
+			    rect->x1, 
+			    rect->y1,
+			    rect->x2 - rect->x1,
+			    rect->y2 - rect->y1 );
+      }
+   }
+
+   intel_release_unpack_region( intel, src );
+   fence = intel_batchbuffer_flush( intel->batch );
+   UNLOCK_HARDWARE( intel );
+
+   bmWaitFence(intel->bm, fence);
+   intel_region_release(intel, &src);
+#endif
+   
+   return GL_FALSE;
+}
+
+
+
+void intelDrawPixels( 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);
+
+   if (INTEL_DEBUG & DEBUG_PIXEL)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   if (do_texture_draw_pixels( intel, x, y, width, height, format, type,
+			       unpack, pixels ))
+      return;
+
+   if (do_blit_draw_pixels( intel, x, y, width, height, format, type,
+			    unpack, pixels ))
+      return;
+
+   _swrast_DrawPixels( ctx, x, y, width, height, format, type,
+		       unpack, pixels );
+}
+
diff --git a/src/mesa/drivers/dri/i915/intel_pixel_read.c b/src/mesa/drivers/dri/i915/intel_pixel_read.c
new file mode 100644
index 00000000000..99bbcf69e03
--- /dev/null
+++ b/src/mesa/drivers/dri/i915/intel_pixel_read.c
@@ -0,0 +1,249 @@
+/**************************************************************************
+ * 
+ * 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 "glheader.h"
+#include "enums.h"
+#include "mtypes.h"
+#include "macros.h"
+#include "swrast/swrast.h"
+
+#include "intel_screen.h"
+#include "intel_context.h"
+#include "intel_ioctl.h"
+#include "intel_batchbuffer.h"
+#include "intel_blit.h"
+#include "intel_regions.h"
+#include "intel_pixel.h"
+#include "bufmgr.h"
+
+
+
+
+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) {
+      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))
+   {
+      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 );
+   intel->vtbl.install_meta_state(intel);
+   intel->vtbl.meta_no_depth_stencil_write(intel);
+
+   if (!driClipRectToFramebuffer(ctx->ReadBuffer, &x, &y, &width, &height)) {
+      UNLOCK_HARDWARE( intel );
+      SET_STATE(i830, state);
+      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 )
+{
+#if 0
+   struct intel_context *intel = intel_context(ctx);
+   GLint pitch = pack->RowLength ? pack->RowLength : width;
+   struct intel_region *src = intel_readbuf_region(intel);
+   struct intel_client_region *dst = intel_client_pack_region(intel,
+							      pack, 
+							      pixels);
+
+   if (ctx->_ImageTransferState ||
+       pack->SwapBytes ||
+       pack->LsbFirst) {
+      fprintf(stderr, "%s: failed 1\n", __FUNCTION__);
+      return GL_FALSE;
+   }
+
+   /* Need GL_PACK_INVERT_MESA to cope with upsidedown results from
+    * blitter:
+    */
+   if (!pack->Invert) {
+      fprintf(stderr, "%s: MESA_PACK_INVERT not set\n", __FUNCTION__);
+      return GL_FALSE;
+   }
+
+   if (!intel_check_blit_format(src, format, type))
+      return GL_FALSE;
+
+   /* 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 );
+   {
+      __DRIdrawablePrivate *dPriv = intel->driDrawable;
+      int nbox = dPriv->numClipRects;
+      drm_clip_rect_t *box = dPriv->pClipRects;
+      int i;
+
+      y = dPriv->h - y - height;
+      x += dPriv->x;
+      y += dPriv->y;
+
+
+      if (INTEL_DEBUG & DEBUG_PIXEL)
+	 fprintf(stderr, "readpixel blit src_pitch %d dst_pitch %d\n",
+		 src_pitch, pitch);
+
+      for (i = 0 ; i < nbox ; i++)
+      {
+	 GLint bx = box[i].x1;
+	 GLint by = box[i].y1;
+	 GLint bw = box[i].x2 - bx;
+	 GLint bh = box[i].y2 - by;
+	 
+	 if (bx < x) bw -= x - bx, bx = x;
+	 if (by < y) bh -= y - by, by = y;
+	 if (bx + bw > x + width) bw = x + width - bx;
+	 if (by + bh > y + height) bh = y + height - by;
+	 if (bw <= 0) continue;
+	 if (bh <= 0) continue;
+
+	 intelEmitCopyBlit( intel,
+			    src->cpp,
+			    src->pitch, src->buffer, 0,
+			    dst->pitch, dst->buffer, 0,
+			    bx, by,
+			    bx - x, by - y,
+			    bw, bh );
+      }
+
+      intel_batchbuffer_flush(intel->batch);
+      intel_client_region_release(intel, dst);
+   }
+   UNLOCK_HARDWARE( intel );
+   return GL_TRUE;
+#endif
+
+   return GL_FALSE;
+}
+
+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 );
+
+   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;
+
+   _swrast_ReadPixels( ctx, x, y, width, height, format, type, pack, pixels);
+}
+
diff --git a/src/mesa/drivers/dri/i915/intel_regions.c b/src/mesa/drivers/dri/i915/intel_regions.c
index f4907256edd..47957feedac 100644
--- a/src/mesa/drivers/dri/i915/intel_regions.c
+++ b/src/mesa/drivers/dri/i915/intel_regions.c
@@ -90,6 +90,7 @@ void intel_region_reference( struct intel_region **dst,
 			     struct intel_region *src)
 {
    src->refcount++;
+   assert(*dst == NULL);
    *dst = src;
 }
 
@@ -151,6 +152,7 @@ struct intel_region *intel_region_create_static( struct intel_context *intel,
 
 
 
+
 static void _mesa_copy_rect( GLubyte *dst,
 			     GLuint cpp,
 			     GLuint dst_pitch,
@@ -186,60 +188,6 @@ static void _mesa_copy_rect( GLubyte *dst,
 }
 
 
-
-/* Could make color a char * to handle deeper buffers.
- */
-static void _mesa_fill_rect( GLubyte *dst,
-			     GLuint cpp,
-			     GLuint dst_pitch,
-			     GLuint dst_x, 
-			     GLuint dst_y,
-			     GLuint width,
-			     GLuint height,
-			     GLuint color )
-{
-   GLuint i,j;
-
-   switch (cpp) {
-   case 1:
-      dst += dst_x;
-      dst += dst_y * dst_pitch;
-      for (i = 0; i < height; i++) { 
-	 memset(dst, color, width);
-	 dst += dst_pitch;
-      }
-      break;
-   case 2: {
-      GLushort color_short = color & 0xffff;
-      GLushort *dst_short = (GLushort *)dst;
-      dst_short += dst_x;
-      dst_short += dst_y * dst_pitch;
-
-      for (i = 0; i < height; i++) { 
-	 for (j = 0; j < width; j++) 
-	    dst_short[j] = color_short;
-      }
-      break;
-   }
-   case 4: {
-      GLuint *dst_int = (GLuint *)dst;
-      dst_int += dst_x;
-      dst_int += dst_y * dst_pitch;
-
-      for (i = 0; i < height; i++) { 
-	 for (j = 0; j < width; j++) 
-	    dst_int[j] = color;
-      }
-      break;
-   }
-   default:
-      assert(0);
-      return;
-   }
-}
-
-
-
 /* Upload data to a rectangular sub-region.  Lots of choices how to do this:
  *
  * - memcpy by span to current destination
@@ -283,47 +231,17 @@ void intel_region_copy( struct intel_context *intel,
 			GLuint srcx, GLuint srcy,
 			GLuint width, GLuint height )
 {
-   unsigned dst_offset;
-   unsigned src_offset;
-   struct bm_buffer_list *list = bmNewBufferList();
-
    DBG("%s\n", __FUNCTION__);
 
    assert(src->cpp == dst->cpp);
 
-   LOCK_HARDWARE(intel);
-   bmAddBuffer(list, dst->buffer, BM_WRITE, NULL, &dst_offset);
-   bmAddBuffer(list, src->buffer, BM_READ, NULL, &src_offset);
-
-   /* Query if both buffers are already uploaded:
-    */
-   if (bmValidateBufferList(intel->bm, list, BM_NO_EVICT|BM_NO_UPLOAD|BM_MEM_AGP)) {
-      intelEmitCopyBlitLocked(intel,
-			      dst->cpp,
-			      src->pitch, src_offset,
-			      dst->pitch, dst_offset, 
-			      srcx, srcy,
-			      dstx, dsty,
-			      width, height);
-
-      bmFenceBufferList(intel->bm, list);
-   }
-   else {
-      _mesa_copy_rect(intel_region_map(intel, dst),
-		      dst->cpp,
-		      dst->pitch,
-		      dstx, dsty,
-		      width, height,
-		      intel_region_map(intel, src),
-		      srcx, srcy,
-		      src->pitch);      
-
-      intel_region_unmap(intel, dst);
-      intel_region_unmap(intel, src);      
-   }
-   
-   bmFreeBufferList(list);
-   UNLOCK_HARDWARE(intel);
+   intelEmitCopyBlit(intel,
+		     dst->cpp,
+		     src->pitch, src->buffer, 0,
+		     dst->pitch, dst->buffer, 0, 
+		     srcx, srcy,
+		     dstx, dsty,
+		     width, height);
 }
 
 /* Fill a rectangular sub-region.  Need better logic about when to
@@ -335,37 +253,15 @@ void intel_region_fill( struct intel_context *intel,
 			GLuint width, GLuint height,
 			GLuint color )
 {
-   unsigned dst_offset;
-   struct bm_buffer_list *list = bmNewBufferList();
-
    DBG("%s\n", __FUNCTION__);
-
-   LOCK_HARDWARE(intel);
-   bmAddBuffer(list, dst->buffer, BM_WRITE, NULL, &dst_offset);
-
-   if (bmValidateBufferList(intel->bm, list, BM_NO_EVICT|BM_NO_UPLOAD|BM_MEM_AGP)) {
-      intelEmitFillBlitLocked(intel,
-			      dst->cpp,
-			      dst->pitch,
-			      dst_offset, 
-			      dstx, dsty,
-			      width, height,
-			      color );
-
-      bmFenceBufferList(intel->bm, list);
-   }
-   else {
-      _mesa_fill_rect(intel_region_map(intel, dst),
-		      dst->cpp,
-		      dst->pitch,
-		      dstx, dsty,
-		      width, height,
-		      color);      
-
-      intel_region_unmap(intel, dst);
-   }
    
-   bmFreeBufferList(list);
-   UNLOCK_HARDWARE(intel);
+   intelEmitFillBlit(intel,
+		     dst->cpp,
+		     dst->pitch,
+		     dst->buffer,
+		     0, 
+		     dstx, dsty,
+		     width, height,
+		     color );
 }
 
diff --git a/src/mesa/drivers/dri/i915/intel_render.c b/src/mesa/drivers/dri/i915/intel_render.c
index 1d7811eb696..3dfad52cd1c 100644
--- a/src/mesa/drivers/dri/i915/intel_render.c
+++ b/src/mesa/drivers/dri/i915/intel_render.c
@@ -106,7 +106,7 @@ static const int scale_prim[GL_POLYGON+1] = {
 };
 
 
-static void intelDmaPrimitive( intelContextPtr intel, GLenum prim )
+static void intelDmaPrimitive( struct intel_context *intel, GLenum prim )
 {
    if (0) fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim));
    INTEL_FIREVERTICES(intel);
@@ -115,7 +115,7 @@ static void intelDmaPrimitive( intelContextPtr intel, GLenum prim )
 }
 
 
-#define LOCAL_VARS intelContextPtr intel = INTEL_CONTEXT(ctx)
+#define LOCAL_VARS struct intel_context *intel = intel_context(ctx)
 #define INIT( prim ) 				\
 do {						\
    intelDmaPrimitive( intel, prim );		\
@@ -142,7 +142,7 @@ do {						\
 
 /* Heuristic to choose between the two render paths:  
  */
-static GLboolean choose_render( intelContextPtr intel,
+static GLboolean choose_render( struct intel_context *intel,
 				struct vertex_buffer *VB )
 {
    int vertsz = intel->vertex_size;
@@ -194,7 +194,7 @@ static GLboolean choose_render( intelContextPtr intel,
 static GLboolean intel_run_render( GLcontext *ctx, 
 				 struct tnl_pipeline_stage *stage )
 {
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
+   struct intel_context *intel = intel_context(ctx);
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    struct vertex_buffer *VB = &tnl->vb;
    GLuint i;
diff --git a/src/mesa/drivers/dri/i915/intel_screen.c b/src/mesa/drivers/dri/i915/intel_screen.c
index c8acc75d68e..6d8e1e4b7de 100644
--- a/src/mesa/drivers/dri/i915/intel_screen.c
+++ b/src/mesa/drivers/dri/i915/intel_screen.c
@@ -104,9 +104,9 @@ static GLboolean intelInitDriver(__DRIscreenPrivate *sPriv)
    intelScreen->cpp = gDRIPriv->cpp;
 			 
    switch (gDRIPriv->bitsPerPixel) {
-   case 15: intelScreen->fbFormat = DV_PF_555; break;
    case 16: intelScreen->fbFormat = DV_PF_565; break;
    case 32: intelScreen->fbFormat = DV_PF_8888; break;
+   default: exit(1); break;
    }
 
    intelScreen->front.pitch = gDRIPriv->fbStride;
diff --git a/src/mesa/drivers/dri/i915/intel_span.c b/src/mesa/drivers/dri/i915/intel_span.c
index f42e1fbd588..b8155d81cd6 100644
--- a/src/mesa/drivers/dri/i915/intel_span.c
+++ b/src/mesa/drivers/dri/i915/intel_span.c
@@ -42,7 +42,7 @@
 #define DBG 0
 
 #define LOCAL_VARS						\
-   intelContextPtr intel = INTEL_CONTEXT(ctx);			\
+   struct intel_context *intel = intel_context(ctx);			\
    __DRIdrawablePrivate *dPriv = intel->driDrawable;		\
    driRenderbuffer *drb = (driRenderbuffer *) rb;		\
    GLuint pitch = drb->pitch * drb->cpp;			\
@@ -54,7 +54,7 @@
    (void) buf; (void) p
 
 #define LOCAL_DEPTH_VARS					\
-   intelContextPtr intel = INTEL_CONTEXT(ctx);			\
+   struct intel_context *intel = intel_context(ctx);			\
    __DRIdrawablePrivate *dPriv = intel->driDrawable;		\
    driRenderbuffer *drb = (driRenderbuffer *) rb;		\
    GLuint pitch = drb->pitch * drb->cpp;			\
@@ -95,27 +95,6 @@ do {								\
 #define TAG(x) intel##x##_565
 #include "spantmp.h"
 
-/* 15 bit, 555 rgb color spanline and pixel functions
- */
-#define WRITE_RGBA( _x, _y, r, g, b, a )			\
-   *(GLushort *)(buf + _x*2 + _y*pitch)  = (((r & 0xf8) << 7) |	\
-		                            ((g & 0xf8) << 3) |	\
-                         		    ((b & 0xf8) >> 3))
-
-#define WRITE_PIXEL( _x, _y, p )  \
-   *(GLushort *)(buf + _x*2 + _y*pitch)  = p
-
-#define READ_RGBA( rgba, _x, _y )				\
-do {								\
-   GLushort p = *(GLushort *)(buf + _x*2 + _y*pitch);		\
-   rgba[0] = (p >> 7) & 0xf8;					\
-   rgba[1] = (p >> 3) & 0xf8;					\
-   rgba[2] = (p << 3) & 0xf8;					\
-   rgba[3] = 255;						\
-} while(0)
-
-#define TAG(x) intel##x##_555
-#include "spantmp.h"
 
 /* 16 bit depthbuffer functions.
  */
@@ -132,7 +111,7 @@ do {								\
 
 #undef LOCAL_VARS
 #define LOCAL_VARS						\
-   intelContextPtr intel = INTEL_CONTEXT(ctx);			\
+   struct intel_context *intel = intel_context(ctx);			\
    __DRIdrawablePrivate *dPriv = intel->driDrawable;		\
    driRenderbuffer *drb = (driRenderbuffer *) rb;		\
    GLuint pitch = drb->pitch * drb->cpp;			\
@@ -206,12 +185,11 @@ do {								\
  */
 void intelSpanRenderStart( GLcontext *ctx )
 {
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
+   struct intel_context *intel = intel_context(ctx);
    GLuint i;
 
    intelFlush(&intel->ctx);
    LOCK_HARDWARE(intel);
-   intelWaitForIdle(intel);
 
    /* Just map the framebuffer and all textures.  Bufmgr code will
     * take care of waiting on the necessary fences:
@@ -230,7 +208,7 @@ void intelSpanRenderStart( GLcontext *ctx )
 
 void intelSpanRenderFinish( GLcontext *ctx )
 {
-   intelContextPtr intel = INTEL_CONTEXT( ctx );
+   struct intel_context *intel = intel_context( ctx );
    GLuint i;
 
    _swrast_flush( ctx );
@@ -266,10 +244,7 @@ void
 intelSetSpanFunctions(driRenderbuffer *drb, const GLvisual *vis)
 {
    if (drb->Base.InternalFormat == GL_RGBA) {
-      if (vis->redBits == 5 && vis->greenBits == 5 && vis->blueBits == 5) {
-         intelInitPointers_555(&drb->Base);
-      }
-      else if (vis->redBits == 5 && vis->greenBits == 6 && vis->blueBits == 5) {
+      if (vis->redBits == 5 && vis->greenBits == 6 && vis->blueBits == 5) {
          intelInitPointers_565(&drb->Base);
       }
       else {
diff --git a/src/mesa/drivers/dri/i915/intel_state.c b/src/mesa/drivers/dri/i915/intel_state.c
index de40e540fd3..6ced0cb1465 100644
--- a/src/mesa/drivers/dri/i915/intel_state.c
+++ b/src/mesa/drivers/dri/i915/intel_state.c
@@ -30,6 +30,7 @@
 #include "context.h"
 #include "macros.h"
 #include "enums.h"
+#include "colormac.h"
 #include "dd.h"
 
 #include "intel_screen.h"
@@ -165,79 +166,26 @@ int intel_translate_logic_op( GLenum opcode )
    }
 }
 
-static void intelDrawBuffer(GLcontext *ctx, GLenum mode )
-{
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
-   intelScreenPrivate *screen = intel->intelScreen;
-   int front = 0;
- 
-   if (!ctx->DrawBuffer)
-      return;
-
-   switch ( ctx->DrawBuffer->_ColorDrawBufferMask[0] ) {
-   case BUFFER_BIT_FRONT_LEFT:
-      front = 1;
-      FALLBACK( intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE );
-      break;
-   case BUFFER_BIT_BACK_LEFT:
-      front = 0;
-      FALLBACK( intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE );
-      break;
-   default:
-      FALLBACK( intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE );
-      return;
-   }
-
-   if ( intel->sarea->pf_current_page == 1 ) 
-      front ^= 1;
-   
-   intelSetFrontClipRects( intel );
-
-
-   if (front) {
-      intel->drawOffset = screen->front.offset;
-      if (intel->draw_region != intel->front_region) {
-	 intel_region_release(intel, &intel->draw_region);
-	 intel_region_reference(&intel->draw_region, intel->front_region);
-      }
-   } else {
-      intel->drawOffset = screen->back.offset;
-      if (intel->draw_region != intel->back_region) {
-	 intel_region_release(intel, &intel->draw_region);
-	 intel_region_reference(&intel->draw_region, intel->back_region);
-      }
-   }
-
-   intel->vtbl.set_draw_offset( intel, intel->drawOffset );
-}
-
-static void intelReadBuffer( GLcontext *ctx, GLenum mode )
-{
-   /* nothing, until we implement h/w glRead/CopyPixels or CopyTexImage */
-}
-
 
 static void intelClearColor(GLcontext *ctx, const GLfloat color[4])
 {
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
+   struct intel_context *intel = intel_context(ctx);
    intelScreenPrivate *screen = intel->intelScreen;
+   GLubyte clear_chan[4];
 
-   CLAMPED_FLOAT_TO_UBYTE(intel->clear_red, color[0]);
-   CLAMPED_FLOAT_TO_UBYTE(intel->clear_green, color[1]);
-   CLAMPED_FLOAT_TO_UBYTE(intel->clear_blue, color[2]);
-   CLAMPED_FLOAT_TO_UBYTE(intel->clear_alpha, color[3]);
+   UNCLAMPED_FLOAT_TO_RGBA_CHAN(clear_chan, color);
 
    intel->ClearColor = INTEL_PACKCOLOR(screen->fbFormat,
-				       intel->clear_red, 
-				       intel->clear_green, 
-				       intel->clear_blue, 
-				       intel->clear_alpha);
+				       clear_chan[0], 
+				       clear_chan[1], 
+				       clear_chan[2], 
+				       clear_chan[3]);
 }
 
 
 static void intelCalcViewport( GLcontext *ctx )
 {
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
+   struct intel_context *intel = intel_context(ctx);
    const GLfloat *v = ctx->Viewport._WindowMap.m;
    GLfloat *m = intel->ViewportMatrix.m;
    GLint h = 0;
@@ -273,18 +221,111 @@ static void intelDepthRange( GLcontext *ctx,
  */
 static void intelRenderMode( GLcontext *ctx, GLenum mode )
 {
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
+   struct intel_context *intel = intel_context(ctx);
    FALLBACK( intel, INTEL_FALLBACK_RENDERMODE, (mode != GL_RENDER) );
 }
 
 
 void intelInitStateFuncs( struct dd_function_table *functions )
 {
-   functions->DrawBuffer = intelDrawBuffer;
-   functions->ReadBuffer = intelReadBuffer;
    functions->RenderMode = intelRenderMode;
    functions->Viewport = intelViewport;
    functions->DepthRange = intelDepthRange;
    functions->ClearColor = intelClearColor;
 }
 
+
+
+
+void intelInitState( GLcontext *ctx )
+{
+   /* Mesa should do this for us:
+    */
+   ctx->Driver.AlphaFunc( ctx, 
+			  ctx->Color.AlphaFunc,
+			  ctx->Color.AlphaRef);
+
+   ctx->Driver.BlendColor( ctx,
+			   ctx->Color.BlendColor );
+
+   ctx->Driver.BlendEquationSeparate( ctx, 
+				      ctx->Color.BlendEquationRGB,
+				      ctx->Color.BlendEquationA);
+
+   ctx->Driver.BlendFuncSeparate( ctx,
+				  ctx->Color.BlendSrcRGB,
+				  ctx->Color.BlendDstRGB,
+				  ctx->Color.BlendSrcA,
+				  ctx->Color.BlendDstA);
+
+   ctx->Driver.ColorMask( ctx, 
+			  ctx->Color.ColorMask[RCOMP],
+			  ctx->Color.ColorMask[GCOMP],
+			  ctx->Color.ColorMask[BCOMP],
+			  ctx->Color.ColorMask[ACOMP]);
+
+   ctx->Driver.CullFace( ctx, ctx->Polygon.CullFaceMode );
+   ctx->Driver.DepthFunc( ctx, ctx->Depth.Func );
+   ctx->Driver.DepthMask( ctx, ctx->Depth.Mask );
+
+   ctx->Driver.Enable( ctx, GL_ALPHA_TEST, ctx->Color.AlphaEnabled );
+   ctx->Driver.Enable( ctx, GL_BLEND, ctx->Color.BlendEnabled );
+   ctx->Driver.Enable( ctx, GL_COLOR_LOGIC_OP, ctx->Color.ColorLogicOpEnabled );
+   ctx->Driver.Enable( ctx, GL_COLOR_SUM, ctx->Fog.ColorSumEnabled );
+   ctx->Driver.Enable( ctx, GL_CULL_FACE, ctx->Polygon.CullFlag );
+   ctx->Driver.Enable( ctx, GL_DEPTH_TEST, ctx->Depth.Test );
+   ctx->Driver.Enable( ctx, GL_DITHER, ctx->Color.DitherFlag );
+   ctx->Driver.Enable( ctx, GL_FOG, ctx->Fog.Enabled );
+   ctx->Driver.Enable( ctx, GL_LIGHTING, ctx->Light.Enabled );
+   ctx->Driver.Enable( ctx, GL_LINE_SMOOTH, ctx->Line.SmoothFlag );
+   ctx->Driver.Enable( ctx, GL_POLYGON_STIPPLE, ctx->Polygon.StippleFlag );
+   ctx->Driver.Enable( ctx, GL_SCISSOR_TEST, ctx->Scissor.Enabled );
+   ctx->Driver.Enable( ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled );
+   ctx->Driver.Enable( ctx, GL_TEXTURE_1D, GL_FALSE );
+   ctx->Driver.Enable( ctx, GL_TEXTURE_2D, GL_FALSE );
+   ctx->Driver.Enable( ctx, GL_TEXTURE_RECTANGLE_NV, GL_FALSE );
+   ctx->Driver.Enable( ctx, GL_TEXTURE_3D, GL_FALSE );
+   ctx->Driver.Enable( ctx, GL_TEXTURE_CUBE_MAP, GL_FALSE );
+
+   ctx->Driver.Fogfv( ctx, GL_FOG_COLOR, ctx->Fog.Color );
+   ctx->Driver.Fogfv( ctx, GL_FOG_MODE, 0 );
+   ctx->Driver.Fogfv( ctx, GL_FOG_DENSITY, &ctx->Fog.Density );
+   ctx->Driver.Fogfv( ctx, GL_FOG_START, &ctx->Fog.Start );
+   ctx->Driver.Fogfv( ctx, GL_FOG_END, &ctx->Fog.End );
+
+   ctx->Driver.FrontFace( ctx, ctx->Polygon.FrontFace );
+
+   {
+      GLfloat f = (GLfloat)ctx->Light.Model.ColorControl;
+      ctx->Driver.LightModelfv( ctx, GL_LIGHT_MODEL_COLOR_CONTROL, &f );
+   }
+
+   ctx->Driver.LineWidth( ctx, ctx->Line.Width );
+   ctx->Driver.LogicOpcode( ctx, ctx->Color.LogicOp );
+   ctx->Driver.PointSize( ctx, ctx->Point.Size );
+   ctx->Driver.PolygonStipple( ctx, (const GLubyte *)ctx->PolygonStipple );
+   ctx->Driver.Scissor( ctx, ctx->Scissor.X, ctx->Scissor.Y,
+			ctx->Scissor.Width, ctx->Scissor.Height );
+   ctx->Driver.ShadeModel( ctx, ctx->Light.ShadeModel );
+   ctx->Driver.StencilFuncSeparate( ctx, GL_FRONT,
+                                    ctx->Stencil.Function[0],
+                                    ctx->Stencil.Ref[0],
+                                    ctx->Stencil.ValueMask[0] );
+   ctx->Driver.StencilFuncSeparate( ctx, GL_BACK,
+                                    ctx->Stencil.Function[1],
+                                    ctx->Stencil.Ref[1],
+                                    ctx->Stencil.ValueMask[1] );
+   ctx->Driver.StencilMaskSeparate( ctx, GL_FRONT, ctx->Stencil.WriteMask[0] );
+   ctx->Driver.StencilMaskSeparate( ctx, GL_BACK, ctx->Stencil.WriteMask[1] );
+   ctx->Driver.StencilOpSeparate( ctx, GL_FRONT,
+                                  ctx->Stencil.FailFunc[0],
+                                  ctx->Stencil.ZFailFunc[0],
+                                  ctx->Stencil.ZPassFunc[0]);
+   ctx->Driver.StencilOpSeparate( ctx, GL_BACK,
+                                  ctx->Stencil.FailFunc[1],
+                                  ctx->Stencil.ZFailFunc[1],
+                                  ctx->Stencil.ZPassFunc[1]);
+
+
+   ctx->Driver.DrawBuffer( ctx, ctx->Color.DrawBuffer[0] );
+}
diff --git a/src/mesa/drivers/dri/i915/intel_tex.h b/src/mesa/drivers/dri/i915/intel_tex.h
index be64cf03934..7860130c3fb 100644
--- a/src/mesa/drivers/dri/i915/intel_tex.h
+++ b/src/mesa/drivers/dri/i915/intel_tex.h
@@ -97,20 +97,7 @@ void intelCopyTexSubImage2D( GLcontext *ctx, GLenum target, GLint level,
 			     GLint xoffset, GLint yoffset,
 			     GLint x, GLint y, GLsizei width, GLsizei height );
 
-
-GLuint intel_validate_mipmap_tree( struct intel_context *intel,
-				   struct intel_texture_object *intelObj );
-
-void intel_add_texoffset_fixup( struct intel_context *intel,
-				GLuint unit,
-				GLuint *ptr );
-
-void intel_apply_fixups( struct intel_context *intel );
-
-GLboolean intel_prevalidate_buffers( struct intel_context *intel );
-GLboolean intel_validate_buffers( struct intel_context *intel );
-void intel_fence_buffers( struct intel_context *intel );
-
+GLuint intel_finalize_mipmap_tree( struct intel_context *intel, GLuint unit );
 
 void intel_tex_map_images( struct intel_context *intel,
 			   struct intel_texture_object *intelObj );
diff --git a/src/mesa/drivers/dri/i915/intel_tex_copy.c b/src/mesa/drivers/dri/i915/intel_tex_copy.c
index 50ba017d8bc..cd94f3db6df 100644
--- a/src/mesa/drivers/dri/i915/intel_tex_copy.c
+++ b/src/mesa/drivers/dri/i915/intel_tex_copy.c
@@ -37,6 +37,7 @@
 #include "intel_regions.h"
 #include "intel_tex.h"
 #include "intel_blit.h"
+#include "intel_pixel.h"
 #include "bufmgr.h"
 
 /* Do the best we can using the blitter.  A future project is to use
@@ -108,15 +109,13 @@ static GLboolean do_copy_texsubimage( struct intel_context *intel,
       return GL_FALSE;
  
 
+   intelFlush(ctx);
    LOCK_HARDWARE(intel);
-   intelInstallBatchBuffer(intel);
    {
       __DRIdrawablePrivate *dPriv = intel->driDrawable;
       GLuint image_offset = intel_miptree_image_offset(intelImage->mt, 
 						       intelImage->face,
 						       intelImage->level);
-      GLuint dst_offset = 0;
-      GLuint src_offset = 0;
       GLint orig_x = x;
       GLint orig_y = y;
       GLuint window_y;
@@ -138,35 +137,28 @@ static GLboolean do_copy_texsubimage( struct intel_context *intel,
       y = window_y + y;
 
 
-      bmAddBuffer(intel->buffer_list, 
-		  intelImage->mt->region->buffer, 
-		  BM_WRITE, NULL, &dst_offset);
-
-      bmAddBuffer(intel->buffer_list, 
-		  src->buffer, 
-		  BM_READ, NULL, &src_offset);
-
-      if (!bmValidateBufferList(intel->bm, intel->buffer_list, BM_MEM_AGP)) {
-	 ret = GL_FALSE;
-	 goto out;
-      }
 
       /* A bit of fiddling to get the blitter to work with -ve
        * pitches.  But we get a nice inverted blit this way, so it's
        * worth it:
        */
-      intelEmitCopyBlitLocked( intel,
-			       intelImage->mt->cpp,
-			       -src->pitch, 
-			       src_offset + intel->intelScreen->height * src->pitch * src->cpp, 
-			       intelImage->mt->pitch, 
-			       dst_offset + image_offset,
-			       x, y + height, 
-			       dstx, dsty,
-			       width, height );
+      intelEmitCopyBlit( intel,
+			 intelImage->mt->cpp,
+
+			 -src->pitch, 
+			 src->buffer,
+			 src->height * src->pitch * src->cpp, 
+
+			 intelImage->mt->pitch, 
+			 intelImage->mt->region->buffer,
+			 image_offset,
+
+			 x, y + height, 
+			 dstx, dsty,
+			 width, height );
 
    out:
-      intelFlushBatchLocked( intel, GL_TRUE, GL_FALSE, GL_FALSE);
+      intel_batchbuffer_flush( intel->batch );
    }
    
 
diff --git a/src/mesa/drivers/dri/i915/intel_tex_format.c b/src/mesa/drivers/dri/i915/intel_tex_format.c
index a645165fb27..ff02239e588 100644
--- a/src/mesa/drivers/dri/i915/intel_tex_format.c
+++ b/src/mesa/drivers/dri/i915/intel_tex_format.c
@@ -15,7 +15,7 @@ const struct gl_texture_format *
 intelChooseTextureFormat( GLcontext *ctx, GLint internalFormat,
                           GLenum format, GLenum type )
 {
-   intelContextPtr intel = INTEL_CONTEXT( ctx );
+   struct intel_context *intel = intel_context( ctx );
    const GLboolean do32bpt = (intel->intelScreen->cpp == 4);
 
    switch ( internalFormat ) {
diff --git a/src/mesa/drivers/dri/i915/intel_tex_image.c b/src/mesa/drivers/dri/i915/intel_tex_image.c
index c6f6df7c1cf..d14cb3fc65e 100644
--- a/src/mesa/drivers/dri/i915/intel_tex_image.c
+++ b/src/mesa/drivers/dri/i915/intel_tex_image.c
@@ -153,7 +153,7 @@ static void intelTexImage(GLcontext *ctx,
 			  GLint internalFormat,
 			  GLint width, GLint height, GLint border,
 			  GLenum format, GLenum type, const void *pixels,
-			  const struct gl_pixelstore_attrib *packing,
+			  const struct gl_pixelstore_attrib *unpack,
 			  struct gl_texture_object *texObj,
 			  struct gl_texture_image *texImage)
 {
@@ -165,11 +165,14 @@ static void intelTexImage(GLcontext *ctx,
    GLint texelBytes, sizeInBytes;
    GLuint dstRowStride;
 
+
    DBG("%s target %s level %d %dx%d border %d\n", __FUNCTION__,
 		_mesa_lookup_enum_by_nr(target),
 		level,
 		width, height, border);
 
+   intelFlush(ctx);
+
    intelImage->face = target_to_face( target );
    intelImage->level = level;
 
@@ -271,13 +274,11 @@ static void intelTexImage(GLcontext *ctx,
     */
    pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, 1, 
 					format, type,
-					pixels, packing, "glTexImage");
+					pixels, unpack, "glTexImage");
    if (!pixels) 
       return;
    
 
-
-
    LOCK_HARDWARE(intel);
    
    if (intelImage->mt) {
@@ -311,11 +312,11 @@ static void intelTexImage(GLcontext *ctx,
 					0, 0, 0,  /* dstX/Y/Zoffset */
 					dstRowStride, 0 /* dstImageStride */,
 					width, height, 1,
-					format, type, pixels, packing)) {
+					format, type, pixels, unpack)) {
       _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
    }
 
-   _mesa_unmap_teximage_pbo(ctx, packing);
+   _mesa_unmap_teximage_pbo(ctx, unpack);
 
    if (intelImage->mt) {
       intel_miptree_image_unmap(intel, intelImage->mt);
@@ -342,14 +343,14 @@ void intelTexImage2D(GLcontext *ctx,
 		     GLint internalFormat,
 		     GLint width, GLint height, GLint border,
 		     GLenum format, GLenum type, const void *pixels,
-		     const struct gl_pixelstore_attrib *packing,
+		     const struct gl_pixelstore_attrib *unpack,
 		     struct gl_texture_object *texObj,
 		     struct gl_texture_image *texImage)
 {
    intelTexImage( ctx, 2, target, level, 
 		internalFormat, width, height, border,
 		format, type, pixels,
-		packing, texObj, texImage );
+		unpack, texObj, texImage );
 }
 
 void intelTexImage1D(GLcontext *ctx, 
@@ -357,14 +358,14 @@ void intelTexImage1D(GLcontext *ctx,
 		     GLint internalFormat,
 		     GLint width, GLint border,
 		     GLenum format, GLenum type, const void *pixels,
-		     const struct gl_pixelstore_attrib *packing,
+		     const struct gl_pixelstore_attrib *unpack,
 		     struct gl_texture_object *texObj,
 		     struct gl_texture_image *texImage)
 {
    intelTexImage( ctx, 1, target, level, 
 		  internalFormat, width, 1, border,
 		  format, type, pixels,
-		  packing, texObj, texImage );
+		  unpack, texObj, texImage );
 }
 
 
diff --git a/src/mesa/drivers/dri/i915/intel_tex_subimage.c b/src/mesa/drivers/dri/i915/intel_tex_subimage.c
index c065e0c630d..2f38b901fea 100644
--- a/src/mesa/drivers/dri/i915/intel_tex_subimage.c
+++ b/src/mesa/drivers/dri/i915/intel_tex_subimage.c
@@ -57,6 +57,8 @@ static void intelTexSubimage (GLcontext *ctx,
 		xoffset, yoffset,
 		width, height);
 
+   intelFlush(ctx);
+
    pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, 1, format, type,
                                         pixels, packing, "glTexSubImage2D");
    if (!pixels)
diff --git a/src/mesa/drivers/dri/i915/intel_tex_validate.c b/src/mesa/drivers/dri/i915/intel_tex_validate.c
index bd225a484d8..a3e61fbc474 100644
--- a/src/mesa/drivers/dri/i915/intel_tex_validate.c
+++ b/src/mesa/drivers/dri/i915/intel_tex_validate.c
@@ -96,7 +96,7 @@ static void copy_image_data_to_tree( struct intel_context *intel,
 
 /*  
  */
-static GLuint intel_finalize_mipmap_tree( struct intel_context *intel, GLuint unit )
+GLuint intel_finalize_mipmap_tree( struct intel_context *intel, GLuint unit )
 {
    struct gl_texture_object *tObj = intel->ctx.Texture.Unit[unit]._Current;
    struct intel_texture_object *intelObj = intel_texture_object(tObj);
@@ -120,7 +120,7 @@ static GLuint intel_finalize_mipmap_tree( struct intel_context *intel, GLuint un
       if (intelObj->mt) {
 	 intel_miptree_release(intel, &intelObj->mt);
       }
-      return 0;
+      return GL_FALSE;
    }
 
 
@@ -191,113 +191,6 @@ static GLuint intel_finalize_mipmap_tree( struct intel_context *intel, GLuint un
    return GL_TRUE;
 }
 
-void intel_add_texoffset_fixup( struct intel_context *intel,
-				GLuint unit,
-				GLuint *ptr )
-{
-   struct gl_texture_object *tObj = intel->ctx.Texture.Unit[unit]._Current;
-   struct intel_texture_object *intelObj = intel_texture_object(tObj);
-
-#if 0
-   struct intel_reloc *f = &intel->fixup[intel->nr_fixups++];
-   assert(intel->nr_fixups <= INTEL_MAX_FIXUP);
-   f->dest = ptr;
-   f->value = &intelObj->textureOffset;
-   f->delta = (intel->intelScreen->tex.offset + 
-	       intel_miptree_image_offset(intelObj->mt, 0, intelObj->firstLevel));
-#else
-   *ptr = (intelObj->textureOffset + 	   
-	   intel_miptree_image_offset(intelObj->mt, 0, intelObj->firstLevel));
-#endif
-}
-
-/* Fix up the command buffer:
- */
-void intel_apply_fixups( struct intel_context *intel )
-{
-   GLuint i;
-
-   for (i = 0; i < intel->nr_fixups; i++) {
-      struct intel_reloc *f = &intel->fixup[i];
-      *f->dest = *f->value + f->delta;
-   }
-
-   intel->nr_fixups = 0;
-}
-
-
-
-/* One upshot of the new manager is that it should be possible to tell
- * ahead of time whether a certain set of buffers will cause a
- * fallback.  
- *
- * Unless we do this we either have to a) hold the DRI lock
- * while emitting all vertices and fire after each vertex buffer, or
- * b) build a fallback path that operates on i915 command streams
- * rather than the state in the GLcontext.
- */
-GLboolean intel_prevalidate_buffers( struct intel_context *intel )
-{
-   return GL_TRUE;		/* never fallback */
-}
-
-
-GLboolean intel_validate_buffers( struct intel_context *intel )
-{
-   GLcontext *ctx = &intel->ctx;
-   GLboolean ok = GL_TRUE;
-   GLuint i;
-
-   DBG("%s\n", __FUNCTION__);
-
-   assert(intel->locked);
-
-   /* Add the color and depth buffers so that fences associated with
-    * these buffers:
-    */
-   bmAddBuffer(intel->buffer_list,
-	       intel->draw_region->buffer,
-	       BM_WRITE,
-	       NULL,
-	       NULL);
-   
-   bmAddBuffer(intel->buffer_list,
-	       intel->depth_region->buffer,
-	       BM_WRITE,
-	       NULL,
-	       NULL);
-
-   /* Add each enabled texture:
-    */
-   for (i = 0 ; i < ctx->Const.MaxTextureUnits && ok ; i++) {
-      if (ctx->Texture.Unit[i]._ReallyEnabled) {
-	 struct gl_texture_object *tObj = intel->ctx.Texture.Unit[i]._Current;
-	 struct intel_texture_object *intelObj = intel_texture_object(tObj);
-
-	 ok = intel_finalize_mipmap_tree( intel, i );
-	 if (ok) {
-	    bmAddBuffer(intel->buffer_list,
-			intelObj->mt->region->buffer,
-			BM_READ,
-			NULL,
-			&intelObj->textureOffset);
-	 }
-      }
-   }
-
-   ok = bmValidateBufferList(intel->bm, intel->buffer_list, BM_MEM_AGP);
-   assert(ok);
-   return ok;
-}
-
-void intel_fence_buffers( struct intel_context *intel )
-{
-   assert(intel->locked);
-   assert(intel->buffer_list);
-}
-
-
-
 
 
 void intel_tex_map_images( struct intel_context *intel,
diff --git a/src/mesa/drivers/dri/i915/intel_tris.c b/src/mesa/drivers/dri/i915/intel_tris.c
index 1b0d31f3fdf..77694ac04f3 100644
--- a/src/mesa/drivers/dri/i915/intel_tris.c
+++ b/src/mesa/drivers/dri/i915/intel_tris.c
@@ -63,10 +63,9 @@ static void intelRasterPrimitive( GLcontext *ctx, GLenum rprim, GLuint hwprim );
  * of the locked region - vertex buffers, second batch buffer for
  * primitives, relocation fixups for texture addresses.
  */
-static void intel_flush_inline_primitive( GLcontext *ctx )
+static void intel_flush_inline_primitive( struct intel_context *intel )
 {
-   intelContextPtr intel = INTEL_CONTEXT( ctx );
-   GLuint used = intel->batch.ptr - intel->prim.start_ptr;
+   GLuint used = intel->batch->ptr - intel->prim.start_ptr;
 
    assert(intel->prim.primitive != ~0);
 
@@ -80,9 +79,7 @@ static void intel_flush_inline_primitive( GLcontext *ctx )
    goto finished;
    
  do_discard:
-   intel->batch.ptr -= used;
-   intel->batch.space += used;
-   assert(intel->batch.space >= 0);
+   intel->batch->ptr -= used;
 
  finished:
    intel->prim.primitive = ~0;
@@ -93,17 +90,19 @@ static void intel_flush_inline_primitive( GLcontext *ctx )
 
 /* Emit a primitive referencing vertices in a vertex buffer.
  */
-void intelStartInlinePrimitive( intelContextPtr intel, GLuint prim )
+void intelStartInlinePrimitive( struct intel_context *intel, 
+				GLuint prim, 
+				GLuint batch_flags )
 {
    BATCH_LOCALS;
    
    /* Emit a slot which will be filled with the inline primitive
     * command later.
     */
-   BEGIN_BATCH(2);
+   BEGIN_BATCH(2, batch_flags);
    OUT_BATCH( 0 );
 
-   intel->prim.start_ptr = batch_ptr;
+   intel->prim.start_ptr = intel->batch->ptr;
    intel->prim.primitive = prim;
    intel->prim.flush = intel_flush_inline_primitive;
 
@@ -112,31 +111,29 @@ void intelStartInlinePrimitive( intelContextPtr intel, GLuint prim )
 }
 
 
-void intelWrapInlinePrimitive( intelContextPtr intel )
+void intelWrapInlinePrimitive( struct intel_context *intel )
 {
    GLuint prim = intel->prim.primitive;
+   GLuint batchflags = intel->batch->flags;
+
+   intel_flush_inline_primitive(intel);
+   intel_batchbuffer_flush(intel->batch);
 
-   intel_flush_inline_primitive( &intel->ctx );
-   intelFlushBatch(intel, GL_TRUE);
-   intelInstallBatchBuffer( intel );    
-   intel_validate_buffers( intel );
    intel->vtbl.emit_state( intel );
-   intelStartInlinePrimitive( intel, prim );
+   intelStartInlinePrimitive( intel, prim, batchflags ); /* ??? */
 }
 
-GLuint *intelExtendInlinePrimitive( intelContextPtr intel, 
+GLuint *intelExtendInlinePrimitive( struct intel_context *intel, 
 				    GLuint dwords )
 {
    GLuint sz = dwords * sizeof(GLuint);
    GLuint *ptr;
 
-   if (intel->batch.space < sz) {
+   if (intel_batchbuffer_space(intel->batch) < sz) 
       intelWrapInlinePrimitive( intel );
-   }
 
-   ptr = (GLuint *)intel->batch.ptr;
-   intel->batch.ptr += sz;
-   intel->batch.space -= sz;
+   ptr = (GLuint *)intel->batch->ptr;
+   intel->batch->ptr += sz;
 
    return ptr;
 }
@@ -160,22 +157,18 @@ do {								\
 #else
 #define COPY_DWORDS( j, vb, vertsize, v )	\
 do {						\
-   if (0) fprintf(stderr, "\n");	\
    for ( j = 0 ; j < vertsize ; j++ ) {		\
-      if (0) fprintf(stderr, "   -- v(%d): %x/%f\n",j,	\
-	      ((GLuint *)v)[j],			\
-	      ((GLfloat *)v)[j]);		\
       vb[j] = ((GLuint *)v)[j];			\
    }						\
    vb += vertsize;				\
 } while (0)
 #endif
 
-static void __inline__ intel_draw_quad( intelContextPtr intel,
-					intelVertexPtr v0,
-					intelVertexPtr v1,
-					intelVertexPtr v2,
-					intelVertexPtr v3 )
+static void intel_draw_quad( struct intel_context *intel,
+			     intelVertexPtr v0,
+			     intelVertexPtr v1,
+			     intelVertexPtr v2,
+			     intelVertexPtr v3 )
 {
    GLuint vertsize = intel->vertex_size;
    GLuint *vb = intelExtendInlinePrimitive( intel, 6 * vertsize );
@@ -189,10 +182,10 @@ static void __inline__ intel_draw_quad( intelContextPtr intel,
    COPY_DWORDS( j, vb, vertsize, v3 );
 }
 
-static void __inline__ intel_draw_triangle( intelContextPtr intel,
-					    intelVertexPtr v0,
-					    intelVertexPtr v1,
-					    intelVertexPtr v2 )
+static void intel_draw_triangle( struct intel_context *intel,
+				 intelVertexPtr v0,
+				 intelVertexPtr v1,
+				 intelVertexPtr v2 )
 {
    GLuint vertsize = intel->vertex_size;
    GLuint *vb = intelExtendInlinePrimitive( intel, 3 * vertsize );
@@ -204,9 +197,9 @@ static void __inline__ intel_draw_triangle( intelContextPtr intel,
 }
 
 
-static __inline__ void intel_draw_line( intelContextPtr intel,
-					intelVertexPtr v0,
-					intelVertexPtr v1 )
+static void intel_draw_line( struct intel_context *intel,
+			     intelVertexPtr v0,
+			     intelVertexPtr v1 )
 {
    GLuint vertsize = intel->vertex_size;
    GLuint *vb = intelExtendInlinePrimitive( intel, 2 * vertsize );
@@ -217,8 +210,8 @@ static __inline__ void intel_draw_line( intelContextPtr intel,
 }
 
 
-static __inline__ void intel_draw_point( intelContextPtr intel,
-					 intelVertexPtr v0 )
+static void intel_draw_point( struct intel_context *intel,
+			      intelVertexPtr v0 )
 {
    GLuint vertsize = intel->vertex_size;
    GLuint *vb = intelExtendInlinePrimitive( intel, vertsize );
@@ -237,7 +230,7 @@ static __inline__ void intel_draw_point( intelContextPtr intel,
  *                Fixup for ARB_point_parameters                       *
  ***********************************************************************/
 
-static void intel_atten_point( intelContextPtr intel, intelVertexPtr v0 )
+static void intel_atten_point( struct intel_context *intel, intelVertexPtr v0 )
 {
    GLcontext *ctx = &intel->ctx;
    GLfloat psz[4], col[4], restore_psz, restore_alpha;
@@ -286,7 +279,7 @@ static void intel_atten_point( intelContextPtr intel, intelVertexPtr v0 )
 
 
 
-static void intel_wpos_triangle( intelContextPtr intel,
+static void intel_wpos_triangle( struct intel_context *intel,
 				 intelVertexPtr v0,
 				 intelVertexPtr v1,
 				 intelVertexPtr v2 )
@@ -302,7 +295,7 @@ static void intel_wpos_triangle( intelContextPtr intel,
 }
 
 
-static void intel_wpos_line( intelContextPtr intel,
+static void intel_wpos_line( struct intel_context *intel,
 			     intelVertexPtr v0,
 			     intelVertexPtr v1 )
 {
@@ -316,7 +309,7 @@ static void intel_wpos_line( intelContextPtr intel,
 }
 
 
-static void intel_wpos_point( intelContextPtr intel,
+static void intel_wpos_point( struct intel_context *intel,
 			      intelVertexPtr v0 )
 {
    GLuint offset = intel->wpos_offset;
@@ -446,7 +439,7 @@ do {							\
 #define VERT_RESTORE_SPEC( idx ) if (specoffset) v[idx]->ui[specoffset] = spec[idx]
 
 #define LOCAL_VARS(n)							\
-   intelContextPtr intel = INTEL_CONTEXT(ctx);				\
+   struct intel_context *intel = intel_context(ctx);				\
    GLuint color[n], spec[n];						\
    GLuint coloroffset = intel->coloroffset;		\
    GLboolean specoffset = intel->specoffset;			\
@@ -578,7 +571,7 @@ static void init_rast_tab( void )
  * primitives.
  */
 static void
-intel_fallback_tri( intelContextPtr intel,
+intel_fallback_tri( struct intel_context *intel,
 		   intelVertex *v0,
 		   intelVertex *v1,
 		   intelVertex *v2 )
@@ -599,7 +592,7 @@ intel_fallback_tri( intelContextPtr intel,
 
 
 static void
-intel_fallback_line( intelContextPtr intel,
+intel_fallback_line( struct intel_context *intel,
 		    intelVertex *v0,
 		    intelVertex *v1 )
 {
@@ -633,7 +626,7 @@ intel_fallback_line( intelContextPtr intel,
 #define INIT(x) intelRenderPrimitive( ctx, x )
 #undef LOCAL_VARS
 #define LOCAL_VARS						\
-    intelContextPtr intel = INTEL_CONTEXT(ctx);			\
+    struct intel_context *intel = intel_context(ctx);			\
     GLubyte *vertptr = (GLubyte *)intel->verts;			\
     const GLuint vertsize = intel->vertex_size;       	\
     const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;	\
@@ -659,7 +652,7 @@ intel_fallback_line( intelContextPtr intel,
 static void intelRenderClippedPoly( GLcontext *ctx, const GLuint *elts,
 				   GLuint n )
 {
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
+   struct intel_context *intel = intel_context(ctx);
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
    GLuint prim = intel->render_primitive;
@@ -690,7 +683,7 @@ static void intelRenderClippedLine( GLcontext *ctx, GLuint ii, GLuint jj )
 static void intelFastRenderClippedPoly( GLcontext *ctx, const GLuint *elts,
 				       GLuint n )
 {
-   intelContextPtr intel = INTEL_CONTEXT( ctx );
+   struct intel_context *intel = intel_context( ctx );
    const GLuint vertsize = intel->vertex_size;
    GLuint *vb = intelExtendInlinePrimitive( intel, (n-2) * 3 * vertsize );
    GLubyte *vertptr = (GLubyte *)intel->verts;
@@ -717,7 +710,7 @@ static void intelFastRenderClippedPoly( GLcontext *ctx, const GLuint *elts,
 void intelChooseRenderState(GLcontext *ctx)
 {
    TNLcontext *tnl = TNL_CONTEXT(ctx);
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
+   struct intel_context *intel = intel_context(ctx);
    GLuint flags = ctx->_TriangleCaps;
    struct fragment_program *fprog = ctx->FragmentProgram._Current;
    GLboolean have_wpos = (fprog && (fprog->Base.InputsRead & FRAG_BIT_WPOS));
@@ -814,7 +807,7 @@ static const GLenum reduced_prim[GL_POLYGON+1] = {
 
 static void intelRunPipeline( GLcontext *ctx )
 {
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
+   struct intel_context *intel = intel_context(ctx);
 
    if (intel->NewGLState) {
       if (intel->NewGLState & _NEW_TEXTURE) {
@@ -836,11 +829,7 @@ static void intelRenderStart( GLcontext *ctx )
 {
    struct intel_context *intel = intel_context(ctx);
 
-   intel->vtbl.render_start( INTEL_CONTEXT(ctx) );
-
-   LOCK_HARDWARE(intel);
-   intelInstallBatchBuffer( intel );    
-   intel_validate_buffers( intel );
+   intel->vtbl.render_start( intel_context(ctx) );
    intel->vtbl.emit_state( intel );
 }
 
@@ -852,10 +841,7 @@ static void intelRenderFinish( GLcontext *ctx )
       _swrast_flush( ctx );
 
    if (intel->prim.flush)
-      intel->prim.flush(ctx);
-
-   intelFlushBatch(intel, GL_TRUE); 
-   UNLOCK_HARDWARE(intel);
+      intel->prim.flush(intel);
 }
 
 
@@ -866,7 +852,7 @@ static void intelRenderFinish( GLcontext *ctx )
   */
 static void intelRasterPrimitive( GLcontext *ctx, GLenum rprim, GLuint hwprim )
 {
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
+   struct intel_context *intel = intel_context(ctx);
 
    if (0)
       fprintf(stderr, "%s %s %x\n", __FUNCTION__, 
@@ -877,7 +863,7 @@ static void intelRasterPrimitive( GLcontext *ctx, GLenum rprim, GLuint hwprim )
    /* Start a new primitive.  Arrange to have it flushed later on.
     */
    if (hwprim != intel->prim.primitive) 
-      intelStartInlinePrimitive( intel, hwprim );
+      intelStartInlinePrimitive( intel, hwprim, INTEL_BATCH_CLIPRECTS );
 }
 
 
@@ -885,7 +871,7 @@ static void intelRasterPrimitive( GLcontext *ctx, GLenum rprim, GLuint hwprim )
   */
 static void intelRenderPrimitive( GLcontext *ctx, GLenum prim )
 {
-   intelContextPtr intel = INTEL_CONTEXT(ctx);
+   struct intel_context *intel = intel_context(ctx);
 
    if (0)
       fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim));
@@ -938,7 +924,7 @@ static char *getFallbackString(GLuint bit)
 
 
 
-void intelFallback( intelContextPtr intel, GLuint bit, GLboolean mode )
+void intelFallback( struct intel_context *intel, GLuint bit, GLboolean mode )
 {
    GLcontext *ctx = &intel->ctx;
    TNLcontext *tnl = TNL_CONTEXT(ctx);
@@ -982,6 +968,70 @@ void intelFallback( intelContextPtr intel, GLuint bit, GLboolean mode )
 
 
 
+/**********************************************************************/
+/*             Used only with the metaops callbacks.                  */
+/**********************************************************************/
+void intel_meta_draw_quad(struct intel_context *intel, 
+			  GLfloat x0, GLfloat x1,
+			  GLfloat y0, GLfloat y1, 
+			  GLfloat z,
+			  GLuint color,
+			  GLfloat s0, GLfloat s1,
+			  GLfloat t0, GLfloat t1,
+			  GLuint flags)
+{
+   union fi *vb;
+
+   if (0)
+      fprintf(stderr, "%s: %f,%f-%f,%f 0x%x %f,%f-%f,%f\n",
+	      __FUNCTION__,
+	      x0,y0,x1,y1,color,s0,t0,s1,t1);
+
+   intel->vtbl.emit_state( intel );
+
+   intelStartInlinePrimitive( intel, PRIM3D_TRIFAN, flags );  
+   vb = (union fi *)intelExtendInlinePrimitive( intel, 4 * 6 );
+
+
+   /* initial vertex, left bottom */
+   vb[0].f = x0;
+   vb[1].f = y0;
+   vb[2].f = z;
+   vb[3].i = color;
+   vb[4].f = s0;
+   vb[5].f = t0;
+   vb += 6;
+
+   /* right bottom */
+   vb[0].f = x1;
+   vb[1].f = y0;
+   vb[2].f = z;
+   vb[3].i = color;
+   vb[4].f = s1;
+   vb[5].f = t0;
+   vb += 6;
+
+   /* right top */
+   vb[0].f = x1;
+   vb[1].f = y1;
+   vb[2].f = z;
+   vb[3].i = color;
+   vb[4].f = s1;
+   vb[5].f = t1;
+   vb += 6;
+
+   /* left top */
+   vb[0].f = x0;
+   vb[1].f = y1;
+   vb[2].f = z;
+   vb[3].i = color;
+   vb[4].f = s0;
+   vb[5].f = t1;
+
+   if (intel->prim.flush)
+      intel->prim.flush(intel);
+}
+
 
 /**********************************************************************/
 /*                            Initialization.                         */
diff --git a/src/mesa/drivers/dri/i915/intel_tris.h b/src/mesa/drivers/dri/i915/intel_tris.h
index 1a6164f8063..5fe5b2feb99 100644
--- a/src/mesa/drivers/dri/i915/intel_tris.h
+++ b/src/mesa/drivers/dri/i915/intel_tris.h
@@ -42,16 +42,23 @@
 
 extern void intelInitTriFuncs( GLcontext *ctx );
 
-extern void intelPrintRenderState( const char *msg, GLuint state );
 extern void intelChooseRenderState( GLcontext *ctx );
 
-extern void intelStartInlinePrimitive( intelContextPtr intel, GLuint prim );
-extern void intelWrapInlinePrimitive( intelContextPtr intel );
-extern GLuint *intelEmitInlinePrimitiveLocked(intelContextPtr intel, 
-					      int primitive, int dwords,
-					      int vertex_size);
+extern void intelStartInlinePrimitive( struct intel_context *intel, GLuint prim, GLuint flags );
+extern void intelWrapInlinePrimitive( struct intel_context *intel );
 
-GLuint *intelExtendInlinePrimitive( intelContextPtr intel, 
+GLuint *intelExtendInlinePrimitive( struct intel_context *intel, 
 				    GLuint dwords );
 
+
+void intel_meta_draw_quad(struct intel_context *intel, 
+			  GLfloat x0, GLfloat x1,
+			  GLfloat y0, GLfloat y1, 
+			  GLfloat z,
+			  GLuint color,
+			  GLfloat s0, GLfloat s1,
+			  GLfloat t0, GLfloat t1,
+			  GLuint flags );
+
+
 #endif