Merge remote branch 'origin/master' into gallium_draw_llvm

86 files changed, 1045 insertions, 736 deletions

diff --git a/src/mesa/drivers/common/meta.c b/src/mesa/drivers/common/meta.c
index b97b760f18..84a2a5fcb3 100644
--- a/src/mesa/drivers/common/meta.c
+++ b/src/mesa/drivers/common/meta.c
@@ -429,13 +429,15 @@ _mesa_meta_begin(GLcontext *ctx, GLbitfield state)
    if (state & META_SHADER) {
       if (ctx->Extensions.ARB_vertex_program) {
          save->VertexProgramEnabled = ctx->VertexProgram.Enabled;
-         save->VertexProgram = ctx->VertexProgram.Current;
+         _mesa_reference_vertprog(ctx, &save->VertexProgram,
+				  ctx->VertexProgram.Current);
          _mesa_set_enable(ctx, GL_VERTEX_PROGRAM_ARB, GL_FALSE);
       }
 
       if (ctx->Extensions.ARB_fragment_program) {
          save->FragmentProgramEnabled = ctx->FragmentProgram.Enabled;
-         save->FragmentProgram = ctx->FragmentProgram.Current;
+         _mesa_reference_fragprog(ctx, &save->FragmentProgram,
+				  ctx->FragmentProgram.Current);
          _mesa_set_enable(ctx, GL_FRAGMENT_PROGRAM_ARB, GL_FALSE);
       }
 
@@ -663,6 +665,7 @@ _mesa_meta_end(GLcontext *ctx)
                           save->VertexProgramEnabled);
          _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, 
                                   save->VertexProgram);
+	 _mesa_reference_vertprog(ctx, &save->VertexProgram, NULL);
       }
 
       if (ctx->Extensions.ARB_fragment_program) {
@@ -670,6 +673,7 @@ _mesa_meta_end(GLcontext *ctx)
                           save->FragmentProgramEnabled);
          _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current,
                                   save->FragmentProgram);
+	 _mesa_reference_fragprog(ctx, &save->FragmentProgram, NULL);
       }
 
       if (ctx->Extensions.ARB_shader_objects) {
@@ -720,6 +724,7 @@ _mesa_meta_end(GLcontext *ctx)
       for (tgt = 0; tgt < NUM_TEXTURE_TARGETS; tgt++) {
          _mesa_reference_texobj(&ctx->Texture.Unit[0].CurrentTex[tgt],
                                 save->CurrentTexture[tgt]);
+         _mesa_reference_texobj(&save->CurrentTexture[tgt], NULL);
       }
 
       /* Re-enable textures, texgen */
@@ -2062,8 +2067,10 @@ _mesa_meta_Bitmap(GLcontext *ctx,
    }
 
    bitmap1 = _mesa_map_pbo_source(ctx, &unpackSave, bitmap1);
-   if (!bitmap1)
+   if (!bitmap1) {
+      _mesa_meta_end(ctx);
       return;
+   }
 
    bitmap8 = (GLubyte *) calloc(1, width * height);
    if (bitmap8) {
diff --git a/src/mesa/drivers/dri/Makefile.template b/src/mesa/drivers/dri/Makefile.template
index a0c25d26cd..f19cc039b2 100644
--- a/src/mesa/drivers/dri/Makefile.template
+++ b/src/mesa/drivers/dri/Makefile.template
@@ -51,9 +51,12 @@ lib: symlinks subdirs depend
 	@$(MAKE) $(LIBNAME) $(TOP)/$(LIB_DIR)/$(LIBNAME)
 
 $(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(EXTRA_MODULES) Makefile \
-		$(TOP)/src/mesa/drivers/dri/Makefile.template
-	$(MKLIB) -o $@ -noprefix -linker '$(CC)' -ldflags '$(LDFLAGS)' \
+		$(TOP)/src/mesa/drivers/dri/Makefile.template $(TOP)/src/mesa/drivers/dri/common/dri_test.o
+	$(MKLIB) -o $@.tmp -noprefix -linker '$(CC)' -ldflags '$(LDFLAGS)' \
 		$(OBJECTS) $(MESA_MODULES) $(EXTRA_MODULES) $(DRI_LIB_DEPS)
+	$(CC) -o $@.test $(TOP)/src/mesa/drivers/dri/common/dri_test.o $@.tmp $(DRI_LIB_DEPS)
+	@rm -f $@.test
+	mv -f $@.tmp $@
 
 
 $(TOP)/$(LIB_DIR)/$(LIBNAME): $(LIBNAME)
diff --git a/src/mesa/drivers/dri/common/dri_test.c b/src/mesa/drivers/dri/common/dri_test.c
new file mode 100644
index 0000000000..793f0c37d7
--- /dev/null
+++ b/src/mesa/drivers/dri/common/dri_test.c
@@ -0,0 +1,89 @@
+#include "main/glheader.h"
+#include "main/compiler.h"
+#include "glapi/glapi.h"
+
+/* This is just supposed to make sure we get a reference to
+   the driver entry symbol that the compiler doesn't optimize away */
+
+extern char __driDriverExtensions[];
+
+/* provide glapi symbols */
+
+#if defined(GLX_USE_TLS)
+
+PUBLIC __thread struct _glapi_table * _glapi_tls_Dispatch
+    __attribute__((tls_model("initial-exec")));
+
+PUBLIC __thread void * _glapi_tls_Context
+    __attribute__((tls_model("initial-exec")));
+
+PUBLIC const struct _glapi_table *_glapi_Dispatch;
+PUBLIC const void *_glapi_Context;
+
+#else
+
+PUBLIC struct _glapi_table *_glapi_Dispatch;
+PUBLIC void *_glapi_Context;
+
+#endif
+
+PUBLIC void
+_glapi_check_multithread(void)
+{}
+
+PUBLIC void
+_glapi_set_context(void *context)
+{}
+
+PUBLIC void *
+_glapi_get_context(void)
+{
+	return 0;
+}
+
+PUBLIC void
+_glapi_set_dispatch(struct _glapi_table *dispatch)
+{}
+
+PUBLIC struct _glapi_table *
+_glapi_get_dispatch(void)
+{
+	return 0;
+}
+
+PUBLIC int
+_glapi_add_dispatch( const char * const * function_names,
+		     const char * parameter_signature )
+{
+	return 0;
+}
+
+PUBLIC GLint
+_glapi_get_proc_offset(const char *funcName)
+{
+	return 0;
+}
+
+PUBLIC _glapi_proc
+_glapi_get_proc_address(const char *funcName)
+{
+	return 0;
+}
+
+PUBLIC GLuint
+_glapi_get_dispatch_table_size(void)
+{
+	return 0;
+}
+
+PUBLIC unsigned long
+_glthread_GetID(void)
+{
+   return 0;
+}
+
+int main(int argc, char** argv)
+{
+   void* p = __driDriverExtensions;
+   return (int)(unsigned long)p;
+}
diff --git a/src/mesa/drivers/dri/common/dri_util.c b/src/mesa/drivers/dri/common/dri_util.c
index badbb5ff82..f1bbd38612 100644
--- a/src/mesa/drivers/dri/common/dri_util.c
+++ b/src/mesa/drivers/dri/common/dri_util.c
@@ -129,11 +129,6 @@ static int driUnbindContext(__DRIcontext *pcp)
      */
     pcp->driDrawablePriv = pcp->driReadablePriv = NULL;
 
-#if 0
-    /* Unbind the drawable */
-    pdp->driContextPriv = &psp->dummyContextPriv;
-#endif
-
     return GL_TRUE;
 }
 
@@ -433,7 +428,6 @@ driCreateNewDrawable(__DRIscreen *psp, const __DRIconfig *config,
     pdp->vblFlags = 0;
 
     pdp->driScreenPriv = psp;
-    pdp->driContextPriv = &psp->dummyContextPriv;
 
     if (!(*psp->DriverAPI.CreateBuffer)(psp, pdp, &config->modes,
 					renderType == GLX_PIXMAP_BIT)) {
@@ -567,17 +561,6 @@ driCreateNewContext(__DRIscreen *psp, const __DRIconfig *config,
     
     pcp->dri2.draw_stamp = 0;
     pcp->dri2.read_stamp = 0;
-    /* When the first context is created for a screen, initialize a "dummy"
-     * context.
-     */
-
-    if (!psp->dri2.enabled && !psp->dummyContextPriv.driScreenPriv) {
-        psp->dummyContextPriv.hHWContext = psp->pSAREA->dummy_context;
-        psp->dummyContextPriv.driScreenPriv = psp;
-        psp->dummyContextPriv.driDrawablePriv = NULL;
-        psp->dummyContextPriv.driverPrivate = NULL;
-	/* No other fields should be used! */
-    }
 
     pcp->hHWContext = hwContext;
 
@@ -734,13 +717,6 @@ driCreateNewScreen(int scrn,
     psp->myNum = scrn;
     psp->dri2.enabled = GL_FALSE;
 
-    /*
-    ** Do not init dummy context here; actual initialization will be
-    ** done when the first DRI context is created.  Init screen priv ptr
-    ** to NULL to let CreateContext routine that it needs to be inited.
-    */
-    psp->dummyContextPriv.driScreenPriv = NULL;
-
     psp->DriverAPI = driDriverAPI;
 
     *driver_modes = driDriverAPI.InitScreen(psp);
diff --git a/src/mesa/drivers/dri/common/dri_util.h b/src/mesa/drivers/dri/common/dri_util.h
index f63583cebc..038a81604f 100644
--- a/src/mesa/drivers/dri/common/dri_util.h
+++ b/src/mesa/drivers/dri/common/dri_util.h
@@ -399,11 +399,6 @@ struct __DRIcontextRec {
     void *driverPrivate;
 
     /**
-     * Pointer back to the \c __DRIcontext that contains this structure.
-     */
-    __DRIcontext *pctx;
-
-    /**
      * Pointer to drawable currently bound to this context for drawing.
      */
     __DRIdrawable *driDrawablePriv;
@@ -511,29 +506,12 @@ struct __DRIscreenRec {
     /*@}*/
 
     /**
-     * Dummy context to which drawables are bound when not bound to any
-     * other context. 
-     *
-     * A dummy hHWContext is created for this context, and is used by the GL
-     * core when a hardware lock is required but the drawable is not currently
-     * bound (e.g., potentially during a SwapBuffers request).  The dummy
-     * context is created when the first "real" context is created on this
-     * screen.
-     */
-    __DRIcontext dummyContextPriv;
-
-    /**
      * Device-dependent private information (not stored in the SAREA).
      * 
      * This pointer is never touched by the DRI layer.
      */
     void *private;
 
-    /**
-     * Pointer back to the \c __DRIscreen that contains this structure.
-     */
-    __DRIscreen *psc;
-
     /* Extensions provided by the loader. */
     const __DRIgetDrawableInfoExtension *getDrawableInfo;
     const __DRIsystemTimeExtension *systemTime;
diff --git a/src/mesa/drivers/dri/common/dri_sw.c b/src/mesa/drivers/dri/common/drisw_util.c
index b7f9036f47..8d08b93bfb 100644
--- a/src/mesa/drivers/dri/common/dri_sw.c
+++ b/src/mesa/drivers/dri/common/drisw_util.c
@@ -22,12 +22,12 @@
  */
 
 /**
- * \file dri_sw.c
+ * \file drisw_util.c
  *
  * DRISW utility functions, i.e. dri_util.c stripped from drm-specific bits.
  */
 
-#include "dri_sw.h"
+#include "drisw_util.h"
 #include "utils.h"
 
 
@@ -63,6 +63,7 @@ driCreateNewScreen(int scrn, const __DRIextension **extensions,
     setupLoaderExtensions(psp, extensions);
 
     psp->extensions = emptyExtensionList;
+    psp->fd = -1;
     psp->myNum = scrn;
 
     *driver_configs = driDriverAPI.InitScreen(psp);
@@ -146,6 +147,7 @@ static int driBindContext(__DRIcontext *pcp,
 	pcp->driDrawablePriv = pdp;
 	pcp->driReadablePriv = prp;
 	if (pdp) {
+	    pdp->driContextPriv = pcp;
 	    dri_get_drawable(pdp);
 	}
 	if ( prp && pdp != prp ) {
@@ -221,6 +223,7 @@ driCreateNewDrawable(__DRIscreen *psp,
     pdp->loaderPrivate = data;
 
     pdp->driScreenPriv = psp;
+    pdp->driContextPriv = NULL;
 
     dri_get_drawable(pdp);
 
@@ -229,6 +232,8 @@ driCreateNewDrawable(__DRIscreen *psp,
 	return NULL;
     }
 
+    pdp->lastStamp = 1; /* const */
+
     return pdp;
 }
 
diff --git a/src/mesa/drivers/dri/common/dri_sw.h b/src/mesa/drivers/dri/common/drisw_util.h
index e353e26b34..08d5a116e9 100644
--- a/src/mesa/drivers/dri/common/dri_sw.h
+++ b/src/mesa/drivers/dri/common/drisw_util.h
@@ -21,13 +21,25 @@
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 
+/**
+ * @file
+ * Binding of the DRI interface (dri_interface.h) for DRISW.
+ *
+ * The DRISW structs are 'base classes' of the corresponding DRI1 / DRI2 (DRM)
+ * structs. The bindings for SW and DRM can be unified by making the DRM structs
+ * 'sub-classes' of the SW structs, either proper or with field re-ordering.
+ *
+ * The code can also be unified but that requires cluttering the common code
+ * with ifdef's and guarding with (__DRIscreen::fd >= 0) for DRM.
+ */
 
-#ifndef _DRI_SW_H
-#define _DRI_SW_H
+#ifndef _DRISW_UTIL_H
+#define _DRISW_UTIL_H
 
 #include <GL/gl.h>
 #include <GL/internal/glcore.h>
 #include <GL/internal/dri_interface.h>
+typedef struct _drmLock drmLock;
 
 
 /**
@@ -38,6 +50,39 @@ extern const __DRIswrastExtension driSWRastExtension;
 
 
 /**
+ * Driver callback functions
+ */
+struct __DriverAPIRec {
+    const __DRIconfig **(*InitScreen) (__DRIscreen * priv);
+
+    void (*DestroyScreen)(__DRIscreen *driScrnPriv);
+
+    GLboolean (*CreateContext)(const __GLcontextModes *glVis,
+                               __DRIcontext *driContextPriv,
+                               void *sharedContextPrivate);
+
+    void (*DestroyContext)(__DRIcontext *driContextPriv);
+
+    GLboolean (*CreateBuffer)(__DRIscreen *driScrnPriv,
+                              __DRIdrawable *driDrawPriv,
+                              const __GLcontextModes *glVis,
+                              GLboolean pixmapBuffer);
+
+    void (*DestroyBuffer)(__DRIdrawable *driDrawPriv);
+
+    void (*SwapBuffers)(__DRIdrawable *driDrawPriv);
+
+    GLboolean (*MakeCurrent)(__DRIcontext *driContextPriv,
+                             __DRIdrawable *driDrawPriv,
+                             __DRIdrawable *driReadPriv);
+
+    GLboolean (*UnbindContext)(__DRIcontext *driContextPriv);
+};
+
+extern const struct __DriverAPIRec driDriverAPI;
+
+
+/**
  * Data types
  */
 struct __DRIscreenRec {
@@ -71,42 +116,17 @@ struct __DRIdrawableRec {
 
     void *loaderPrivate;
 
+    __DRIcontext *driContextPriv;
+
     __DRIscreen *driScreenPriv;
 
     int refcount;
-};
-
-
-/**
- * Driver callback functions
- */
-struct __DriverAPIRec {
-    const __DRIconfig **(*InitScreen) (__DRIscreen * priv);
-
-    void (*DestroyScreen)(__DRIscreen *driScrnPriv);
-
-    GLboolean (*CreateContext)(const __GLcontextModes *glVis,
-                               __DRIcontext *driContextPriv,
-                               void *sharedContextPrivate);
 
-    void (*DestroyContext)(__DRIcontext *driContextPriv);
+    /* gallium */
+    unsigned int lastStamp;
 
-    GLboolean (*CreateBuffer)(__DRIscreen *driScrnPriv,
-                              __DRIdrawable *driDrawPriv,
-                              const __GLcontextModes *glVis,
-                              GLboolean pixmapBuffer);
-
-    void (*DestroyBuffer)(__DRIdrawable *driDrawPriv);
-
-    void (*SwapBuffers)(__DRIdrawable *driDrawPriv);
-
-    GLboolean (*MakeCurrent)(__DRIcontext *driContextPriv,
-                             __DRIdrawable *driDrawPriv,
-                             __DRIdrawable *driReadPriv);
-
-    GLboolean (*UnbindContext)(__DRIcontext *driContextPriv);
+    int w;
+    int h;
 };
 
-extern const struct __DriverAPIRec driDriverAPI;
-
-#endif /* _DRI_SW_H */
+#endif /* _DRISW_UTIL_H */
diff --git a/src/mesa/drivers/dri/i915/i830_texstate.c b/src/mesa/drivers/dri/i915/i830_texstate.c
index e8f7e378ec..a28073919c 100644
--- a/src/mesa/drivers/dri/i915/i830_texstate.c
+++ b/src/mesa/drivers/dri/i915/i830_texstate.c
@@ -146,15 +146,15 @@ i830_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
 
    dri_bo_reference(intelObj->mt->region->buffer);
    i830->state.tex_buffer[unit] = intelObj->mt->region->buffer;
+   pitch = intelObj->mt->region->pitch * intelObj->mt->cpp;
+
    /* XXX: This calculation is probably broken for tiled images with
     * a non-page-aligned offset.
     */
-   i830->state.tex_offset[unit] = (dst_x + dst_y * intelObj->mt->pitch) *
-      intelObj->mt->cpp;
+   i830->state.tex_offset[unit] = dst_x * intelObj->mt->cpp + dst_y * pitch;
 
    format = translate_texture_format(firstImage->TexFormat,
 				     firstImage->InternalFormat);
-   pitch = intelObj->mt->pitch * intelObj->mt->cpp;
 
    state[I830_TEXREG_TM0LI] = (_3DSTATE_LOAD_STATE_IMMEDIATE_2 |
                                (LOAD_TEXTURE_MAP0 << unit) | 4);
diff --git a/src/mesa/drivers/dri/i915/i915_state.c b/src/mesa/drivers/dri/i915/i915_state.c
index 7275617a6f..91b228d52b 100644
--- a/src/mesa/drivers/dri/i915/i915_state.c
+++ b/src/mesa/drivers/dri/i915/i915_state.c
@@ -374,7 +374,7 @@ intelCalcViewport(GLcontext * ctx)
    else {
       /* window buffer, y=0=top */
       yScale = -1.0;
-      yBias = (intel->driDrawable) ? intel->driDrawable->h : 0.0F;
+      yBias = ctx->DrawBuffer->Height;
    }
 
    m[MAT_SX] = v[MAT_SX];
diff --git a/src/mesa/drivers/dri/i915/i915_tex_layout.c b/src/mesa/drivers/dri/i915/i915_tex_layout.c
index fe3908f580..702655283b 100644
--- a/src/mesa/drivers/dri/i915/i915_tex_layout.c
+++ b/src/mesa/drivers/dri/i915/i915_tex_layout.c
@@ -123,13 +123,12 @@ i915_miptree_layout_cube(struct intel_context *intel,
    assert(lvlWidth == lvlHeight); /* cubemap images are square */
 
    /* double pitch for cube layouts */
-   mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, dim * 2);
+   mt->total_width = dim * 2;
    mt->total_height = dim * 4;
 
    for (level = mt->first_level; level <= mt->last_level; level++) {
       intel_miptree_set_level_info(mt, level, 6,
 				   0, 0,
-				   /*OLD: mt->pitch, mt->total_height,*/
 				   lvlWidth, lvlHeight,
 				   1);
       lvlWidth /= 2;
@@ -167,7 +166,7 @@ i915_miptree_layout_3d(struct intel_context *intel,
    GLint level;
 
    /* Calculate the size of a single slice. */
-   mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, mt->width0);
+   mt->total_width = mt->width0;
 
    /* XXX: hardware expects/requires 9 levels at minimum. */
    for (level = mt->first_level; level <= MAX2(8, mt->last_level); level++) {
@@ -210,7 +209,7 @@ i915_miptree_layout_2d(struct intel_context *intel,
    GLuint img_height;
    GLint level;
 
-   mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, mt->width0);
+   mt->total_width = mt->width0;
    mt->total_height = 0;
 
    for (level = mt->first_level; level <= mt->last_level; level++) {
@@ -251,9 +250,8 @@ i915_miptree_layout(struct intel_context *intel, struct intel_mipmap_tree * mt,
       break;
    }
 
-   DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__,
-       mt->pitch,
-       mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp);
+   DBG("%s: %dx%dx%d\n", __FUNCTION__,
+       mt->total_width, mt->total_height, mt->cpp);
 
    return GL_TRUE;
 }
@@ -336,9 +334,9 @@ i945_miptree_layout_cube(struct intel_context *intel,
     * or the final row of 4x4, 2x2 and 1x1 faces below this.
     */
    if (dim > 32)
-      mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, dim * 2);
+      mt->total_width = dim * 2;
    else
-      mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, 14 * 8);
+      mt->total_width = 14 * 8;
 
    if (dim >= 4)
       mt->total_height = dim * 4 + 4;
@@ -423,11 +421,11 @@ i945_miptree_layout_3d(struct intel_context *intel,
    GLuint pack_y_pitch;
    GLuint level;
 
-   mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, mt->width0);
+   mt->total_width = mt->width0;
    mt->total_height = 0;
 
    pack_y_pitch = MAX2(mt->height0, 2);
-   pack_x_pitch = mt->pitch;
+   pack_x_pitch = mt->total_width;
    pack_x_nr = 1;
 
    for (level = mt->first_level; level <= mt->last_level; level++) {
@@ -454,7 +452,7 @@ i945_miptree_layout_3d(struct intel_context *intel,
       if (pack_x_pitch > 4) {
 	 pack_x_pitch >>= 1;
 	 pack_x_nr <<= 1;
-	 assert(pack_x_pitch * pack_x_nr <= mt->pitch);
+	 assert(pack_x_pitch * pack_x_nr <= mt->total_width);
       }
 
       if (pack_y_pitch > 2) {
@@ -491,9 +489,8 @@ i945_miptree_layout(struct intel_context *intel, struct intel_mipmap_tree * mt,
       break;
    }
 
-   DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__,
-       mt->pitch,
-       mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp);
+   DBG("%s: %dx%dx%d\n", __FUNCTION__,
+       mt->total_width, mt->total_height, mt->cpp);
 
    return GL_TRUE;
 }
diff --git a/src/mesa/drivers/dri/i915/i915_texstate.c b/src/mesa/drivers/dri/i915/i915_texstate.c
index a1ab8f8b6d..ff9ab88c5a 100644
--- a/src/mesa/drivers/dri/i915/i915_texstate.c
+++ b/src/mesa/drivers/dri/i915/i915_texstate.c
@@ -165,7 +165,7 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
    format = translate_texture_format(firstImage->TexFormat,
 				     firstImage->InternalFormat,
 				     tObj->DepthMode);
-   pitch = intelObj->mt->pitch * intelObj->mt->cpp;
+   pitch = intelObj->mt->region->pitch * intelObj->mt->cpp;
 
    state[I915_TEXREG_MS3] =
       (((firstImage->Height - 1) << MS3_HEIGHT_SHIFT) |
diff --git a/src/mesa/drivers/dri/i915/intel_tris.c b/src/mesa/drivers/dri/i915/intel_tris.c
index fb191fe346..9449a158dc 100644
--- a/src/mesa/drivers/dri/i915/intel_tris.c
+++ b/src/mesa/drivers/dri/i915/intel_tris.c
@@ -251,7 +251,7 @@ void intel_flush_prim(struct intel_context *intel)
       BEGIN_BATCH(5);
       OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 |
 		I1_LOAD_S(0) | I1_LOAD_S(1) | 1);
-      assert((offset & !S0_VB_OFFSET_MASK) == 0);
+      assert((offset & ~S0_VB_OFFSET_MASK) == 0);
       OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0, offset);
       OUT_BATCH((intel->vertex_size << S1_VERTEX_WIDTH_SHIFT) |
 		(intel->vertex_size << S1_VERTEX_PITCH_SHIFT));
@@ -270,7 +270,7 @@ void intel_flush_prim(struct intel_context *intel)
       OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 |
 		I1_LOAD_S(0) | I1_LOAD_S(2) | 1);
       /* S0 */
-      assert((offset & !S0_VB_OFFSET_MASK_830) == 0);
+      assert((offset & ~S0_VB_OFFSET_MASK_830) == 0);
       OUT_RELOC(vb_bo, I915_GEM_DOMAIN_VERTEX, 0,
 		offset | (intel->vertex_size << S0_VB_PITCH_SHIFT_830) |
 		S0_VB_ENABLE_830);
@@ -488,9 +488,9 @@ intel_wpos_triangle(struct intel_context *intel,
    __memcpy(v1_wpos, v1, size);
    __memcpy(v2_wpos, v2, size);
 
-   v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
-   v1_wpos[1] = -v1_wpos[1] + intel->driDrawable->h;
-   v2_wpos[1] = -v2_wpos[1] + intel->driDrawable->h;
+   v0_wpos[1] = -v0_wpos[1] + intel->ctx.DrawBuffer->Height;
+   v1_wpos[1] = -v1_wpos[1] + intel->ctx.DrawBuffer->Height;
+   v2_wpos[1] = -v2_wpos[1] + intel->ctx.DrawBuffer->Height;
 
 
    intel_draw_triangle(intel, v0, v1, v2);
@@ -509,8 +509,8 @@ intel_wpos_line(struct intel_context *intel,
    __memcpy(v0_wpos, v0, size);
    __memcpy(v1_wpos, v1, size);
 
-   v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
-   v1_wpos[1] = -v1_wpos[1] + intel->driDrawable->h;
+   v0_wpos[1] = -v0_wpos[1] + intel->ctx.DrawBuffer->Height;
+   v1_wpos[1] = -v1_wpos[1] + intel->ctx.DrawBuffer->Height;
 
    intel_draw_line(intel, v0, v1);
 }
@@ -524,7 +524,7 @@ intel_wpos_point(struct intel_context *intel, intelVertexPtr v0)
    GLfloat *v0_wpos = (GLfloat *)((char *)v0 + offset);
 
    __memcpy(v0_wpos, v0, size);
-   v0_wpos[1] = -v0_wpos[1] + intel->driDrawable->h;
+   v0_wpos[1] = -v0_wpos[1] + intel->ctx.DrawBuffer->Height;
 
    intel_draw_point(intel, v0);
 }
diff --git a/src/mesa/drivers/dri/i965/brw_defines.h b/src/mesa/drivers/dri/i965/brw_defines.h
index 984e56d00c..6fe574b22e 100644
--- a/src/mesa/drivers/dri/i965/brw_defines.h
+++ b/src/mesa/drivers/dri/i965/brw_defines.h
@@ -831,8 +831,8 @@
 #define CMD_URB					0x7805 /* GEN6+ */
 # define GEN6_URB_VS_SIZE_SHIFT				16
 # define GEN6_URB_VS_ENTRIES_SHIFT			0
-# define GEN6_URB_GS_SIZE_SHIFT				8
-# define GEN6_URB_GS_ENTRIES_SHIFT			0
+# define GEN6_URB_GS_ENTRIES_SHIFT			8
+# define GEN6_URB_GS_SIZE_SHIFT				0
 
 #define CMD_VIEWPORT_STATE_POINTERS			0x780d /* GEN6+ */
 # define GEN6_CC_VIEWPORT_MODIFY			(1 << 12)
diff --git a/src/mesa/drivers/dri/i965/brw_disasm.c b/src/mesa/drivers/dri/i965/brw_disasm.c
index ad61770212..db3fc50a63 100644
--- a/src/mesa/drivers/dri/i965/brw_disasm.c
+++ b/src/mesa/drivers/dri/i965/brw_disasm.c
@@ -57,6 +57,7 @@ struct {
     [BRW_OPCODE_DPH] = { .name = "dph", .nsrc = 2, .ndst = 1 },
     [BRW_OPCODE_DP3] = { .name = "dp3", .nsrc = 2, .ndst = 1 },
     [BRW_OPCODE_DP2] = { .name = "dp2", .nsrc = 2, .ndst = 1 },
+    [BRW_OPCODE_MATH] = { .name = "math", .nsrc = 2, .ndst = 1 },
 
     [BRW_OPCODE_AVG] = { .name = "avg", .nsrc = 2, .ndst = 1 },
     [BRW_OPCODE_ADD] = { .name = "add", .nsrc = 2, .ndst = 1 },
@@ -797,7 +798,11 @@ int brw_disasm (FILE *file, struct brw_instruction *inst)
     err |= control (file, "saturate", saturate, inst->header.saturate, NULL);
     err |= control (file, "debug control", debug_ctrl, inst->header.debug_control, NULL);
 
-    if (inst->header.opcode != BRW_OPCODE_SEND)
+    if (inst->header.opcode == BRW_OPCODE_MATH) {
+	string (file, " ");
+	err |= control (file, "function", math_function,
+			inst->header.destreg__conditionalmod, NULL);
+    } else if (inst->header.opcode != BRW_OPCODE_SEND)
 	err |= control (file, "conditional modifier", conditional_modifier,
 			inst->header.destreg__conditionalmod, NULL);
 
diff --git a/src/mesa/drivers/dri/i965/brw_eu_emit.c b/src/mesa/drivers/dri/i965/brw_eu_emit.c
index d2395dec28..c33c3def30 100644
--- a/src/mesa/drivers/dri/i965/brw_eu_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_eu_emit.c
@@ -1416,7 +1416,10 @@ void brw_urb_WRITE(struct brw_compile *p,
     * and the first message register index comes from src0.
     */
    if (intel->gen >= 6) {
+      brw_push_insn_state(p);
+      brw_set_mask_control( p, BRW_MASK_DISABLE );
       brw_MOV(p, brw_message_reg(msg_reg_nr), src0);
+      brw_pop_insn_state(p);
       src0 = brw_message_reg(msg_reg_nr);
    }
 
diff --git a/src/mesa/drivers/dri/i965/brw_tex_layout.c b/src/mesa/drivers/dri/i965/brw_tex_layout.c
index 09edfd81fd..b1f56a8e64 100644
--- a/src/mesa/drivers/dri/i965/brw_tex_layout.c
+++ b/src/mesa/drivers/dri/i965/brw_tex_layout.c
@@ -58,12 +58,12 @@ GLboolean brw_miptree_layout(struct intel_context *intel,
           GLuint qpitch = 0;
           GLuint y_pitch = 0;
 
-          mt->pitch = mt->width0;
+          mt->total_width = mt->width0;
           intel_get_texture_alignment_unit(mt->internal_format, &align_w, &align_h);
           y_pitch = ALIGN(height, align_h);
 
           if (mt->compressed) {
-              mt->pitch = ALIGN(mt->width0, align_w);
+              mt->total_width = ALIGN(mt->width0, align_w);
           }
 
           if (mt->first_level != mt->last_level) {
@@ -77,13 +77,11 @@ GLboolean brw_miptree_layout(struct intel_context *intel,
                       + minify(minify(mt->width0));
               }
 
-              if (mip1_width > mt->pitch) {
-                  mt->pitch = mip1_width;
+              if (mip1_width > mt->total_width) {
+                  mt->total_width = mip1_width;
               }
           }
 
-          mt->pitch = intel_miptree_pitch_align(intel, mt, tiling, mt->pitch);
-
           if (mt->compressed) {
               qpitch = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h) / 4;
               mt->total_height = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h) / 4 * 6;
@@ -137,10 +135,10 @@ GLboolean brw_miptree_layout(struct intel_context *intel,
       intel_get_texture_alignment_unit(mt->internal_format, &align_w, &align_h);
 
       if (mt->compressed) {
-          mt->pitch = ALIGN(width, align_w);
+          mt->total_width = ALIGN(width, align_w);
           pack_y_pitch = (height + 3) / 4;
       } else {
-	 mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, mt->width0);
+	 mt->total_width = mt->width0;
 	 pack_y_pitch = ALIGN(mt->height0, align_h);
       }
 
@@ -184,7 +182,7 @@ GLboolean brw_miptree_layout(struct intel_context *intel,
 	    if (pack_x_pitch > 4) {
 	       pack_x_pitch >>= 1;
 	       pack_x_nr <<= 1;
-	       assert(pack_x_pitch * pack_x_nr <= mt->pitch);
+	       assert(pack_x_pitch * pack_x_nr <= mt->total_width);
 	    }
 
 	    if (pack_y_pitch > 2) {
@@ -211,11 +209,8 @@ GLboolean brw_miptree_layout(struct intel_context *intel,
       i945_miptree_layout_2d(intel, mt, tiling);
       break;
    }
-   DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__,
-		mt->pitch,
-		mt->total_height,
-		mt->cpp,
-		mt->pitch * mt->total_height * mt->cpp );
+   DBG("%s: %dx%dx%d\n", __FUNCTION__,
+       mt->total_width, mt->total_height, mt->cpp);
 
    return GL_TRUE;
 }
diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c
index d16e916832..227261409c 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c
@@ -384,8 +384,9 @@ static void emit_sop( struct brw_vs_compile *c,
 {
    struct brw_compile *p = &c->func;
 
-   brw_CMP(p, brw_null_reg(), cond, arg1, arg0);
-   brw_SEL(p, dst, brw_null_reg(), brw_imm_f(1.0f));
+   brw_MOV(p, dst, brw_imm_f(0.0f));
+   brw_CMP(p, brw_null_reg(), cond, arg0, arg1);
+   brw_MOV(p, dst, brw_imm_f(1.0f));
    brw_set_predicate_control_flag_value(p, 0xff);
 }
 
@@ -1360,31 +1361,6 @@ static void emit_vertex_write( struct brw_vs_compile *c)
    }
 }
 
-
-/**
- * Called after code generation to resolve subroutine calls and the
- * END instruction.
- * \param end_inst  points to brw code for END instruction
- * \param last_inst  points to last instruction emitted before vertex write
- */
-static void 
-post_vs_emit( struct brw_vs_compile *c,
-              struct brw_instruction *end_inst,
-              struct brw_instruction *last_inst )
-{
-   GLint offset;
-
-   brw_resolve_cals(&c->func);
-
-   /* patch up the END code to jump past subroutines, etc */
-   offset = last_inst - end_inst;
-   if (offset > 1) {
-      brw_set_src1(end_inst, brw_imm_d(offset * 16));
-   } else {
-      end_inst->header.opcode = BRW_OPCODE_NOP;
-   }
-}
-
 static GLboolean
 accumulator_contains(struct brw_vs_compile *c, struct brw_reg val)
 {
@@ -1465,8 +1441,6 @@ void brw_vs_emit(struct brw_vs_compile *c )
    struct intel_context *intel = &brw->intel;
    const GLuint nr_insns = c->vp->program.Base.NumInstructions;
    GLuint insn, if_depth = 0, loop_depth = 0;
-   GLuint end_offset = 0;
-   struct brw_instruction *end_inst, *last_inst;
    struct brw_instruction *if_inst[MAX_IF_DEPTH], *loop_inst[MAX_LOOP_DEPTH] = { 0 };
    const struct brw_indirect stack_index = brw_indirect(0, 0);   
    GLuint index;
@@ -1750,12 +1724,8 @@ void brw_vs_emit(struct brw_vs_compile *c )
          brw_MOV(p, brw_ip_reg(), deref_1d(stack_index, 0));
 	 brw_set_access_mode(p, BRW_ALIGN_16);
 	 break;
-      case OPCODE_END:	
-         end_offset = p->nr_insn;
-         /* this instruction will get patched later to jump past subroutine
-          * code, etc.
-          */
-         brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16));
+      case OPCODE_END:
+	 emit_vertex_write(c);
          break;
       case OPCODE_PRINT:
          /* no-op */
@@ -1816,13 +1786,7 @@ void brw_vs_emit(struct brw_vs_compile *c )
       release_tmps(c);
    }
 
-   end_inst = &p->store[end_offset];
-   last_inst = &p->store[p->nr_insn];
-
-   /* The END instruction will be patched to jump to this code */
-   emit_vertex_write(c);
-
-   post_vs_emit(c, end_inst, last_inst);
+   brw_resolve_cals(p);
 
    brw_optimize(p);
 
diff --git a/src/mesa/drivers/dri/i965/brw_wm.h b/src/mesa/drivers/dri/i965/brw_wm.h
index 47b764d24d..5aade1c4e6 100644
--- a/src/mesa/drivers/dri/i965/brw_wm.h
+++ b/src/mesa/drivers/dri/i965/brw_wm.h
@@ -286,7 +286,7 @@ void brw_wm_pass0( struct brw_wm_compile *c );
 void brw_wm_pass1( struct brw_wm_compile *c );
 void brw_wm_pass2( struct brw_wm_compile *c );
 void brw_wm_emit( struct brw_wm_compile *c );
-
+GLboolean brw_wm_arg_can_be_immediate(enum prog_opcode, int arg);
 void brw_wm_print_value( struct brw_wm_compile *c,
 			 struct brw_wm_value *value );
 
diff --git a/src/mesa/drivers/dri/i965/brw_wm_emit.c b/src/mesa/drivers/dri/i965/brw_wm_emit.c
index 05e464d4b6..f79de5dbff 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_emit.c
@@ -61,6 +61,44 @@ static INLINE struct brw_reg sechalf( struct brw_reg reg )
    return reg;
 }
 
+/* Return the SrcReg index of the channels that can be immediate float operands
+ * instead of usage of PROGRAM_CONSTANT values through push/pull.
+ */
+GLboolean
+brw_wm_arg_can_be_immediate(enum prog_opcode opcode, int arg)
+{
+   int opcode_array[] = {
+      [OPCODE_ADD] = 2,
+      [OPCODE_CMP] = 3,
+      [OPCODE_DP3] = 2,
+      [OPCODE_DP4] = 2,
+      [OPCODE_DPH] = 2,
+      [OPCODE_MAX] = 2,
+      [OPCODE_MIN] = 2,
+      [OPCODE_MOV] = 1,
+      [OPCODE_MUL] = 2,
+      [OPCODE_SEQ] = 2,
+      [OPCODE_SGE] = 2,
+      [OPCODE_SGT] = 2,
+      [OPCODE_SLE] = 2,
+      [OPCODE_SLT] = 2,
+      [OPCODE_SNE] = 2,
+      [OPCODE_XPD] = 2,
+   };
+
+   /* These opcodes get broken down in a way that allow two
+    * args to be immediates.
+    */
+   if (opcode == OPCODE_MAD || opcode == OPCODE_LRP) {
+      if (arg == 1 || arg == 2)
+	 return GL_TRUE;
+   }
+
+   if (opcode > ARRAY_SIZE(opcode_array))
+      return GL_FALSE;
+
+   return arg == opcode_array[opcode] - 1;
+}
 
 /**
  * Computes the screen-space x,y position of the pixels.
@@ -545,8 +583,11 @@ void emit_sop(struct brw_compile *p,
    for (i = 0; i < 4; i++) {
       if (mask & (1<<i)) {	
 	 brw_push_insn_state(p);
-	 brw_CMP(p, brw_null_reg(), cond, arg1[i], arg0[i]);
-	 brw_SEL(p, dst[i], brw_null_reg(), brw_imm_f(1.0));
+	 brw_CMP(p, brw_null_reg(), cond, arg0[i], arg1[i]);
+	 brw_set_predicate_control(p, BRW_PREDICATE_NONE);
+	 brw_MOV(p, dst[i], brw_imm_f(0));
+	 brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
+	 brw_MOV(p, dst[i], brw_imm_f(1.0));
 	 brw_pop_insn_state(p);
       }
    }
@@ -617,14 +658,10 @@ void emit_cmp(struct brw_compile *p,
 
    for (i = 0; i < 4; i++) {
       if (mask & (1<<i)) {	
-	 brw_set_saturate(p, (mask & SATURATE) ? 1 : 0);
-	 brw_MOV(p, dst[i], arg2[i]);
-	 brw_set_saturate(p, 0);
-
 	 brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, arg0[i], brw_imm_f(0));
 
 	 brw_set_saturate(p, (mask & SATURATE) ? 1 : 0);
-	 brw_MOV(p, dst[i], arg1[i]);
+	 brw_SEL(p, dst[i], arg1[i], arg2[i]);
 	 brw_set_saturate(p, 0);
 	 brw_set_predicate_control_flag_value(p, 0xff);
       }
diff --git a/src/mesa/drivers/dri/i965/brw_wm_glsl.c b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
index 0b66cc6c9f..3b7e421b16 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_glsl.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
@@ -23,6 +23,9 @@ GLboolean brw_wm_is_glsl(const struct gl_fragment_program *fp)
 {
     int i;
 
+    if (INTEL_DEBUG & DEBUG_GLSL_FORCE)
+       return GL_TRUE;
+
     for (i = 0; i < fp->Base.NumInstructions; i++) {
 	const struct prog_instruction *inst = &fp->Base.Instructions[i];
 	switch (inst->Opcode) {
@@ -567,12 +570,35 @@ static struct brw_reg get_src_reg(struct brw_wm_compile *c,
     const GLuint nr = 1;
     const GLuint component = GET_SWZ(src->Swizzle, channel);
 
-    /* Extended swizzle terms */
-    if (component == SWIZZLE_ZERO) {
-       return brw_imm_f(0.0F);
-    }
-    else if (component == SWIZZLE_ONE) {
-       return brw_imm_f(1.0F);
+    /* Only one immediate value can be used per native opcode, and it
+     * has be in the src1 slot, so not all Mesa instructions will get
+     * to take advantage of immediate constants.
+     */
+    if (brw_wm_arg_can_be_immediate(inst->Opcode, srcRegIndex)) {
+       const struct gl_program_parameter_list *params;
+
+       params = c->fp->program.Base.Parameters;
+
+       /* Extended swizzle terms */
+       if (component == SWIZZLE_ZERO) {
+	  return brw_imm_f(0.0F);
+       } else if (component == SWIZZLE_ONE) {
+	  if (src->Negate)
+	     return brw_imm_f(-1.0F);
+	  else
+	     return brw_imm_f(1.0F);
+       }
+
+       if (src->File == PROGRAM_CONSTANT) {
+	  float f = params->ParameterValues[src->Index][component];
+
+	  if (src->Abs)
+	     f = fabs(f);
+	  if (src->Negate)
+	     f = -f;
+
+	  return brw_imm_f(f);
+       }
     }
 
     if (c->fp->use_const_buffer &&
diff --git a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
index ce0bf0b97d..6b9e566886 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
@@ -261,7 +261,7 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit )
 
    key.format = firstImage->TexFormat;
    key.internal_format = firstImage->InternalFormat;
-   key.pitch = intelObj->mt->pitch;
+   key.pitch = intelObj->mt->region->pitch;
    key.depth = firstImage->Depth;
    key.bo = intelObj->mt->region->buffer;
    key.offset = 0;
diff --git a/src/mesa/drivers/dri/i965/gen6_clip_state.c b/src/mesa/drivers/dri/i965/gen6_clip_state.c
index 06f8145e32..acc4b7f101 100644
--- a/src/mesa/drivers/dri/i965/gen6_clip_state.c
+++ b/src/mesa/drivers/dri/i965/gen6_clip_state.c
@@ -55,7 +55,7 @@ upload_clip_state(struct brw_context *brw)
    OUT_BATCH(GEN6_CLIP_STATISTICS_ENABLE);
    OUT_BATCH(GEN6_CLIP_ENABLE |
 	     GEN6_CLIP_API_OGL |
-	     GEN6_CLIP_MODE_REJECT_ALL | /* XXX: debug: get VS working */
+	     GEN6_CLIP_MODE_NORMAL |
 	     GEN6_CLIP_XY_TEST |
 	     depth_clamp |
 	     provoking);
diff --git a/src/mesa/drivers/dri/i965/gen6_gs_state.c b/src/mesa/drivers/dri/i965/gen6_gs_state.c
index 161e7b85c2..cefc93ba48 100644
--- a/src/mesa/drivers/dri/i965/gen6_gs_state.c
+++ b/src/mesa/drivers/dri/i965/gen6_gs_state.c
@@ -50,7 +50,8 @@ upload_gs_state(struct brw_context *brw)
       BEGIN_BATCH(7);
       OUT_BATCH(CMD_3D_GS_STATE << 16 | (7 - 2));
       OUT_RELOC(brw->gs.prog_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0);
-      OUT_BATCH((0 << GEN6_GS_SAMPLER_COUNT_SHIFT) |
+      OUT_BATCH(GEN6_GS_SPF_MODE |
+		(0 << GEN6_GS_SAMPLER_COUNT_SHIFT) |
 		(0 << GEN6_GS_BINDING_TABLE_ENTRY_COUNT_SHIFT));
       OUT_BATCH(0); /* scratch space base offset */
       OUT_BATCH((1 << GEN6_GS_DISPATCH_START_GRF_SHIFT) |
diff --git a/src/mesa/drivers/dri/i965/gen6_vs_state.c b/src/mesa/drivers/dri/i965/gen6_vs_state.c
index fe597dfb94..5916a13994 100644
--- a/src/mesa/drivers/dri/i965/gen6_vs_state.c
+++ b/src/mesa/drivers/dri/i965/gen6_vs_state.c
@@ -100,7 +100,8 @@ upload_vs_state(struct brw_context *brw)
 	     (brw->vs.prog_data->urb_read_length << GEN6_VS_URB_READ_LENGTH_SHIFT) |
 	     (0 << GEN6_VS_URB_ENTRY_READ_OFFSET_SHIFT));
    OUT_BATCH((0 << GEN6_VS_MAX_THREADS_SHIFT) |
-	     GEN6_VS_STATISTICS_ENABLE);
+	     GEN6_VS_STATISTICS_ENABLE |
+	     GEN6_VS_ENABLE);
    ADVANCE_BATCH();
 
    intel_batchbuffer_emit_mi_flush(intel->batch);
diff --git a/src/mesa/drivers/dri/intel/intel_blit.c b/src/mesa/drivers/dri/intel/intel_blit.c
index f2769aa3e8..167140d274 100644
--- a/src/mesa/drivers/dri/intel/intel_blit.c
+++ b/src/mesa/drivers/dri/intel/intel_blit.c
@@ -119,24 +119,8 @@ intelEmitCopyBlit(struct intel_context *intel,
            break;
    } while (pass < 2);
 
-   intel_prepare_render(intel);
-
-   if (pass >= 2) {
-      drm_intel_gem_bo_map_gtt(dst_buffer);
-      drm_intel_gem_bo_map_gtt(src_buffer);
-      _mesa_copy_rect((GLubyte *)dst_buffer->virtual + dst_offset,
-		      cpp,
-		      dst_pitch,
-		      dst_x, dst_y,
-		      w, h,
-		      (GLubyte *)src_buffer->virtual + src_offset,
-		      src_pitch,
-		      src_x, src_y);
-      drm_intel_gem_bo_unmap_gtt(src_buffer);
-      drm_intel_gem_bo_unmap_gtt(dst_buffer);
-
-      return GL_TRUE;
-   }
+   if (pass >= 2)
+      return GL_FALSE;
 
    intel_batchbuffer_require_space(intel->batch, 8 * 4);
    DBG("%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
@@ -482,6 +466,7 @@ intel_emit_linear_blit(struct intel_context *intel,
 		       unsigned int size)
 {
    GLuint pitch, height;
+   GLboolean ok;
 
    /* Blits are in a different ringbuffer so we don't use them. */
    assert(intel->gen < 6);
@@ -489,25 +474,27 @@ intel_emit_linear_blit(struct intel_context *intel,
    /* The pitch is a signed value. */
    pitch = MIN2(size, (1 << 15) - 1);
    height = size / pitch;
-   intelEmitCopyBlit(intel, 1,
-		     pitch, src_bo, src_offset, I915_TILING_NONE,
-		     pitch, dst_bo, dst_offset, I915_TILING_NONE,
-		     0, 0, /* src x/y */
-		     0, 0, /* dst x/y */
-		     pitch, height, /* w, h */
-		     GL_COPY);
+   ok = intelEmitCopyBlit(intel, 1,
+			  pitch, src_bo, src_offset, I915_TILING_NONE,
+			  pitch, dst_bo, dst_offset, I915_TILING_NONE,
+			  0, 0, /* src x/y */
+			  0, 0, /* dst x/y */
+			  pitch, height, /* w, h */
+			  GL_COPY);
+   assert(ok);
 
    src_offset += pitch * height;
    dst_offset += pitch * height;
    size -= pitch * height;
    assert (size < (1 << 15));
    if (size != 0) {
-      intelEmitCopyBlit(intel, 1,
-			size, src_bo, src_offset, I915_TILING_NONE,
-			size, dst_bo, dst_offset, I915_TILING_NONE,
-			0, 0, /* src x/y */
-			0, 0, /* dst x/y */
-			size, 1, /* w, h */
-			GL_COPY);
+      ok = intelEmitCopyBlit(intel, 1,
+			     size, src_bo, src_offset, I915_TILING_NONE,
+			     size, dst_bo, dst_offset, I915_TILING_NONE,
+			     0, 0, /* src x/y */
+			     0, 0, /* dst x/y */
+			     size, 1, /* w, h */
+			     GL_COPY);
+      assert(ok);
    }
 }
diff --git a/src/mesa/drivers/dri/intel/intel_buffers.c b/src/mesa/drivers/dri/intel/intel_buffers.c
index b10693050a..0480770ba1 100644
--- a/src/mesa/drivers/dri/intel/intel_buffers.c
+++ b/src/mesa/drivers/dri/intel/intel_buffers.c
@@ -226,7 +226,7 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
        * only changes with _NEW_STENCIL (which seems sensible).  So flag it
        * here since this is the _NEW_BUFFERS path.
        */
-      ctx->NewState |= (_NEW_DEPTH | _NEW_STENCIL);
+      intel->NewGLState |= (_NEW_DEPTH | _NEW_STENCIL);
    }
 
    intel->vtbl.set_draw_region(intel, colorRegions, depthRegion, 
@@ -236,7 +236,7 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
 #ifdef I915
    intelCalcViewport(ctx);
 #else
-   ctx->NewState |= _NEW_VIEWPORT;
+   intel->NewGLState |= _NEW_VIEWPORT;
 #endif
    /* Set state we know depends on drawable parameters:
     */
@@ -256,7 +256,7 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
    if (ctx->Driver.FrontFace)
       ctx->Driver.FrontFace(ctx, ctx->Polygon.FrontFace);
    else
-      ctx->NewState |= _NEW_POLYGON;
+      intel->NewGLState |= _NEW_POLYGON;
 }
 
 
diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c
index d6a1ba6952..9077a61132 100644
--- a/src/mesa/drivers/dri/intel/intel_context.c
+++ b/src/mesa/drivers/dri/intel/intel_context.c
@@ -63,7 +63,7 @@ int INTEL_DEBUG = (0);
 #endif
 
 
-#define DRIVER_DATE                     "20091221 DEVELOPMENT"
+#define DRIVER_DATE                     "20100330 DEVELOPMENT"
 #define DRIVER_DATE_GEM                 "GEM " DRIVER_DATE
 
 
@@ -389,7 +389,7 @@ intel_prepare_render(struct intel_context *intel)
    __DRIcontext *driContext = intel->driContext;
    __DRIdrawable *drawable;
 
-   drawable = intel->driDrawable;
+   drawable = driContext->driDrawablePriv;
    if (drawable->dri2.stamp != driContext->dri2.draw_stamp) {
       if (drawable->lastStamp != drawable->dri2.stamp)
 	 intel_update_renderbuffers(driContext, drawable);
@@ -397,7 +397,7 @@ intel_prepare_render(struct intel_context *intel)
       driContext->dri2.draw_stamp = drawable->dri2.stamp;
    }
 
-   drawable = intel->driReadDrawable;
+   drawable = driContext->driReadablePriv;
    if (drawable->dri2.stamp != driContext->dri2.read_stamp) {
       if (drawable->lastStamp != drawable->dri2.stamp)
 	 intel_update_renderbuffers(driContext, drawable);
@@ -444,6 +444,7 @@ static const struct dri_debug_control debug_control[] = {
    { "sing",  DEBUG_SINGLE_THREAD },
    { "thre",  DEBUG_SINGLE_THREAD },
    { "wm",    DEBUG_WM },
+   { "glsl_force", DEBUG_GLSL_FORCE },
    { "urb",   DEBUG_URB },
    { "vs",    DEBUG_VS },
    { NULL,    0 }
@@ -471,6 +472,7 @@ void
 intel_flush(GLcontext *ctx, GLboolean needs_mi_flush)
 {
    struct intel_context *intel = intel_context(ctx);
+    __DRIcontext *driContext = intel->driContext;
 
    if (intel->Fallback)
       _swrast_flush(ctx);
@@ -487,9 +489,10 @@ intel_flush(GLcontext *ctx, GLboolean needs_mi_flush)
       if (screen->dri2.loader &&
           (screen->dri2.loader->base.version >= 2)
 	  && (screen->dri2.loader->flushFrontBuffer != NULL) &&
-          intel->driDrawable && intel->driDrawable->loaderPrivate) {
-	 (*screen->dri2.loader->flushFrontBuffer)(intel->driDrawable,
-						  intel->driDrawable->loaderPrivate);
+          driContext->driDrawablePriv &&
+	  driContext->driDrawablePriv->loaderPrivate) {
+	 (*screen->dri2.loader->flushFrontBuffer)(driContext->driDrawablePriv,
+						  driContext->driDrawablePriv->loaderPrivate);
 
 	 /* Only clear the dirty bit if front-buffer rendering is no longer
 	  * enabled.  This is done so that the dirty bit can only be set in
@@ -606,7 +609,6 @@ intelInitContext(struct intel_context *intel,
 
    driContextPriv->driverPrivate = intel;
    intel->intelScreen = intelScreen;
-   intel->driScreen = sPriv;
    intel->driContext = driContextPriv;
    intel->driFd = sPriv->fd;
 
@@ -635,8 +637,7 @@ intelInitContext(struct intel_context *intel,
    }
 
    driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache,
-                       intel->driScreen->myNum,
-		       (intel->gen >= 4) ? "i965" : "i915");
+                       sPriv->myNum, (intel->gen >= 4) ? "i965" : "i915");
    if (intelScreen->deviceID == PCI_CHIP_I865_G)
       intel->maxBatchSize = 4096;
    else
@@ -844,14 +845,6 @@ intelDestroyContext(__DRIcontext * driContextPriv)
 GLboolean
 intelUnbindContext(__DRIcontext * driContextPriv)
 {
-   struct intel_context *intel =
-      (struct intel_context *) driContextPriv->driverPrivate;
-
-   /* Deassociate the context with the drawables.
-    */
-   intel->driDrawable = NULL;
-   intel->driReadDrawable = NULL;
-
    return GL_TRUE;
 }
 
@@ -880,12 +873,10 @@ intelMakeCurrent(__DRIcontext * driContextPriv,
       struct gl_framebuffer *fb = driDrawPriv->driverPrivate;
       struct gl_framebuffer *readFb = driReadPriv->driverPrivate;
 
-      _mesa_make_current(&intel->ctx, fb, readFb);
-      intel->driReadDrawable = driReadPriv;
-      intel->driDrawable = driDrawPriv;
       driContextPriv->dri2.draw_stamp = driDrawPriv->dri2.stamp - 1;
       driContextPriv->dri2.read_stamp = driReadPriv->dri2.stamp - 1;
       intel_prepare_render(intel);
+      _mesa_make_current(&intel->ctx, fb, readFb);
    }
    else {
       _mesa_make_current(NULL, NULL, NULL);
diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h
index 22736a9327..c4bb2bed8e 100644
--- a/src/mesa/drivers/dri/intel/intel_context.h
+++ b/src/mesa/drivers/dri/intel/intel_context.h
@@ -243,9 +243,6 @@ struct intel_context
    int driFd;
 
    __DRIcontext *driContext;
-   __DRIdrawable *driDrawable;
-   __DRIdrawable *driReadDrawable;
-   __DRIscreen *driScreen;
    struct intel_screen *intelScreen;
 
    /**
@@ -342,6 +339,7 @@ extern int INTEL_DEBUG;
 #define DEBUG_WM        0x800000
 #define DEBUG_URB       0x1000000
 #define DEBUG_VS        0x2000000
+#define DEBUG_GLSL_FORCE 0x4000000
 
 #define DBG(...) do {						\
 	if (INTEL_DEBUG & FILE_DEBUG_FLAG)			\
diff --git a/src/mesa/drivers/dri/intel/intel_fbo.c b/src/mesa/drivers/dri/intel/intel_fbo.c
index a429f8d003..8278d12bb9 100644
--- a/src/mesa/drivers/dri/intel/intel_fbo.c
+++ b/src/mesa/drivers/dri/intel/intel_fbo.c
@@ -104,7 +104,6 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
    struct intel_context *intel = intel_context(ctx);
    struct intel_renderbuffer *irb = intel_renderbuffer(rb);
    int cpp;
-   GLuint pitch;
 
    ASSERT(rb->Name != 0);
 
@@ -176,15 +175,11 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
 
    /* allocate new memory region/renderbuffer */
 
-   /* Choose a pitch to match hardware requirements:
-    */
-   pitch = ((cpp * width + 63) & ~63) / cpp;
-
    /* alloc hardware renderbuffer */
-   DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width, height, pitch);
+   DBG("Allocating %d x %d Intel RBO\n", width, height);
 
    irb->region = intel_region_alloc(intel, I915_TILING_NONE, cpp,
-				    width, height, pitch, GL_TRUE);
+				    width, height, GL_TRUE);
    if (!irb->region)
       return GL_FALSE;       /* out of memory? */
 
@@ -573,7 +568,7 @@ intel_render_texture(GLcontext * ctx,
 				  att->Zoffset,
 				  &dst_x, &dst_y);
 
-   intel_image->mt->region->draw_offset = (dst_y * intel_image->mt->pitch +
+   intel_image->mt->region->draw_offset = (dst_y * intel_image->mt->region->pitch +
 					   dst_x) * intel_image->mt->cpp;
    intel_image->mt->region->draw_x = dst_x;
    intel_image->mt->region->draw_y = dst_y;
diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
index 4f14946ec7..ef1966ea7e 100644
--- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
+++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
@@ -83,7 +83,6 @@ intel_miptree_create_internal(struct intel_context *intel,
    mt->cpp = compress_byte ? compress_byte : cpp;
    mt->compressed = compress_byte ? 1 : 0;
    mt->refcount = 1; 
-   mt->pitch = 0;
 
 #ifdef I915
    if (intel->is_945)
@@ -136,7 +135,7 @@ intel_miptree_create(struct intel_context *intel,
    /*
     * pitch == 0 || height == 0  indicates the null texture
     */
-   if (!mt || !mt->pitch || !mt->total_height) {
+   if (!mt || !mt->total_height) {
       free(mt);
       return NULL;
    }
@@ -144,9 +143,8 @@ intel_miptree_create(struct intel_context *intel,
    mt->region = intel_region_alloc(intel,
 				   tiling,
 				   mt->cpp,
-				   mt->pitch,
+				   mt->total_width,
 				   mt->total_height,
-				   mt->pitch,
 				   expect_accelerated_upload);
 
    if (!mt->region) {
@@ -177,81 +175,12 @@ intel_miptree_create_for_region(struct intel_context *intel,
 				      I915_TILING_NONE);
    if (!mt)
       return mt;
-#if 0
-   if (mt->pitch != region->pitch) {
-      fprintf(stderr,
-	      "region pitch (%d) doesn't match mipmap tree pitch (%d)\n",
-	      region->pitch, mt->pitch);
-      free(mt);
-      return NULL;
-   }
-#else
-   /* The mipmap tree pitch is aligned to 64 bytes to make sure render
-    * to texture works, but we don't need that for texturing from a
-    * pixmap.  Just override it here. */
-   mt->pitch = region->pitch;
-#endif
 
    intel_region_reference(&mt->region, region);
 
    return mt;
 }
 
-
-/**
- * intel_miptree_pitch_align:
- *
- * @intel: intel context pointer
- *
- * @mt: the miptree to compute pitch alignment for
- *
- * @pitch: the natural pitch value
- *
- * Given @pitch, compute a larger value which accounts for
- * any necessary alignment required by the device
- */
-int intel_miptree_pitch_align (struct intel_context *intel,
-			       struct intel_mipmap_tree *mt,
-			       uint32_t tiling,
-			       int pitch)
-{
-#ifdef I915
-   GLcontext *ctx = &intel->ctx;
-#endif
-
-   if (!mt->compressed) {
-      int pitch_align;
-
-      /* XXX: Align pitch to multiple of 64 bytes for now to allow
-       * render-to-texture to work in all cases. This should probably be
-       * replaced at some point by some scheme to only do this when really
-       * necessary.
-       */
-      pitch_align = 64;
-
-      if (tiling == I915_TILING_X)
-	 pitch_align = 512;
-      else if (tiling == I915_TILING_Y)
-	 pitch_align = 128;
-
-      pitch = ALIGN(pitch * mt->cpp, pitch_align);
-
-#ifdef I915
-      /* Do a little adjustment to linear allocations so that we avoid
-       * hitting the same channel of memory for 2 different pages when
-       * reading a 2x2 subspan or doing bilinear filtering.
-       */
-      if (tiling == I915_TILING_NONE && !(pitch & 511) &&
-	 (pitch + pitch_align) < (1 << ctx->Const.MaxTextureLevels))
-	 pitch += pitch_align;
-#endif
-
-      pitch /= mt->cpp;
-   }
-   return pitch;
-}
-
-
 void
 intel_miptree_reference(struct intel_mipmap_tree **dst,
                         struct intel_mipmap_tree *src)
@@ -351,13 +280,12 @@ intel_miptree_set_level_info(struct intel_mipmap_tree *mt,
    mt->level[level].width = w;
    mt->level[level].height = h;
    mt->level[level].depth = d;
-   mt->level[level].level_offset = (x + y * mt->pitch) * mt->cpp;
    mt->level[level].level_x = x;
    mt->level[level].level_y = y;
    mt->level[level].nr_images = nr_images;
 
-   DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__,
-       level, w, h, d, x, y, mt->level[level].level_offset);
+   DBG("%s level %d size: %d,%d,%d offset %d,%d\n", __FUNCTION__,
+       level, w, h, d, x, y);
 
    assert(nr_images);
    assert(!mt->level[level].x_offset);
@@ -424,7 +352,7 @@ intel_miptree_image_map(struct intel_context * intel,
    DBG("%s \n", __FUNCTION__);
 
    if (row_stride)
-      *row_stride = mt->pitch * mt->cpp;
+      *row_stride = mt->region->pitch * mt->cpp;
 
    if (mt->target == GL_TEXTURE_3D) {
       int i;
@@ -433,7 +361,7 @@ intel_miptree_image_map(struct intel_context * intel,
 
 	 intel_miptree_get_image_offset(mt, level, face, i,
 					&x, &y);
-	 image_offsets[i] = x + y * mt->pitch;
+	 image_offsets[i] = x + y * mt->region->pitch;
       }
 
       return intel_region_map(intel, mt->region);
@@ -444,7 +372,7 @@ intel_miptree_image_map(struct intel_context * intel,
       image_offsets[0] = 0;
 
       return intel_region_map(intel, mt->region) +
-	 (x + y * mt->pitch) * mt->cpp;
+	 (x + y * mt->region->pitch) * mt->cpp;
    }
 }
 
@@ -520,12 +448,15 @@ intel_miptree_image_copy(struct intel_context *intel,
        width = ALIGN(width, align_w);
    }
 
+   intel_prepare_render(intel);
+
    for (i = 0; i < depth; i++) {
       intel_miptree_get_image_offset(src, level, face, i, &src_x, &src_y);
       intel_miptree_get_image_offset(dst, level, face, i, &dst_x, &dst_y);
       success = intel_region_copy(intel,
 				  dst->region, 0, dst_x, dst_y,
-				  src->region, 0, src_x, src_y, width, height,
+				  src->region, 0, src_x, src_y,
+				  width, height, GL_FALSE,
 				  GL_COPY);
       if (!success) {
 	 GLubyte *src_ptr, *dst_ptr;
@@ -533,13 +464,13 @@ intel_miptree_image_copy(struct intel_context *intel,
 	 src_ptr = intel_region_map(intel, src->region);
 	 dst_ptr = intel_region_map(intel, dst->region);
 
-	 _mesa_copy_rect(dst_ptr + dst->cpp * (dst_x + dst_y * dst->pitch),
+	 _mesa_copy_rect(dst_ptr,
 			 dst->cpp,
-			 dst->pitch,
-			 0, 0, width, height,
-			 src_ptr + src->cpp * (src_x + src_y * src->pitch),
-			 src->pitch,
-			 0, 0);
+			 dst->region->pitch,
+			 dst_x, dst_y, width, height,
+			 src_ptr,
+			 src->region->pitch,
+			 src_x, src_y);
 	 intel_region_unmap(intel, src->region);
 	 intel_region_unmap(intel, dst->region);
       }
diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.h b/src/mesa/drivers/dri/intel/intel_mipmap_tree.h
index b19c548def..21db2f4d3b 100644
--- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.h
+++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.h
@@ -62,14 +62,6 @@
  */
 struct intel_mipmap_level
 {
-   /**
-    * Byte offset to the base of this level.
-    *
-    * This is used for mipmap levels of 1D/2D/3D textures.  However, CUBE
-    * layouts spread images around the whole tree, so the level offset is
-    * always zero in that case.
-    */
-   GLuint level_offset;
    /** Offset to this miptree level, used in computing x_offset. */
    GLuint level_x;
    /** Offset to this miptree level, used in computing y_offset. */
@@ -81,8 +73,8 @@ struct intel_mipmap_level
    /** Number of images at this level: 1 for 1D/2D, 6 for CUBE, depth for 3D */
    GLuint nr_images;
 
-   /**
-    * Byte offset from level_offset to the image for each cube face or depth
+   /** @{
+    * offsets from level_[xy] to the image for each cube face or depth
     * level.
     *
     * Pretty much have to accept that hardware formats
@@ -91,6 +83,7 @@ struct intel_mipmap_level
     * so have to store them as a lookup table.
     */
    GLuint *x_offset, *y_offset;
+   /** @} */
 };
 
 struct intel_mipmap_tree
@@ -109,8 +102,7 @@ struct intel_mipmap_tree
 
    /* Derived from the above:
     */
-   GLuint pitch;
-   GLuint depth_pitch;          /* per-image on i945? */
+   GLuint total_width;
    GLuint total_height;
 
    /* Includes image offset tables:
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_copy.c b/src/mesa/drivers/dri/intel/intel_pixel_copy.c
index f4f3fd6d88..56faf076c7 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_copy.c
@@ -108,14 +108,15 @@ do_blit_copypixels(GLcontext * ctx,
                    GLint dstx, GLint dsty, GLenum type)
 {
    struct intel_context *intel = intel_context(ctx);
-   struct intel_region *dst = intel_drawbuf_region(intel);
-   struct intel_region *src = copypix_src_region(intel, type);
+   struct intel_region *dst;
+   struct intel_region *src;
    struct gl_framebuffer *fb = ctx->DrawBuffer;
    struct gl_framebuffer *read_fb = ctx->ReadBuffer;
    GLint orig_dstx;
    GLint orig_dsty;
    GLint orig_srcx;
    GLint orig_srcy;
+   GLboolean flip = GL_FALSE;
 
    if (type == GL_DEPTH || type == GL_STENCIL) {
       if (INTEL_DEBUG & DEBUG_FALLBACKS)
@@ -133,15 +134,16 @@ do_blit_copypixels(GLcontext * ctx,
        ctx->Pixel.ZoomX != 1.0F || ctx->Pixel.ZoomY != 1.0F)
       return GL_FALSE;
 
+   intel_prepare_render(intel);
+
+   dst = intel_drawbuf_region(intel);
+   src = copypix_src_region(intel, type);
+
    if (!src || !dst)
       return GL_FALSE;
 
    intelFlush(&intel->ctx);
 
-   intel_prepare_render(intel);
-
-   /* XXX: We fail to handle different inversion between read and draw framebuffer. */
-
    /* Clip to destination buffer. */
    orig_dstx = dstx;
    orig_dsty = dsty;
@@ -164,23 +166,23 @@ do_blit_copypixels(GLcontext * ctx,
    dstx += srcx - orig_srcx;
    dsty += srcy - orig_srcy;
 
-   /* Convert from GL to hardware coordinates: */
+   /* Flip dest Y if it's a window system framebuffer. */
    if (fb->Name == 0) {
-      /* copypixels to a system framebuffer */
+      /* copypixels to a window system framebuffer */
       dsty = fb->Height - dsty - height;
-   } else {
-      /* copypixels to a user framebuffer object */
-      dsty = dsty;
+      flip = !flip;
    }
 
-   /* Flip source Y if it's a system framebuffer. */
-   if (read_fb->Name == 0)
-      srcy = fb->Height - srcy - height;
+   /* Flip source Y if it's a window system framebuffer. */
+   if (read_fb->Name == 0) {
+      srcy = read_fb->Height - srcy - height;
+      flip = !flip;
+   }
 
    if (!intel_region_copy(intel,
 			  dst, 0, dstx, dsty,
 			  src, 0, srcx, srcy,
-			  width, height,
+			  width, height, flip,
 			  ctx->Color.ColorLogicOpEnabled ?
 			  ctx->Color.LogicOp : GL_COPY)) {
       DBG("%s: blit failure\n", __FUNCTION__);
diff --git a/src/mesa/drivers/dri/intel/intel_reg.h b/src/mesa/drivers/dri/intel/intel_reg.h
index d19f1bae34..36d8180598 100644
--- a/src/mesa/drivers/dri/intel/intel_reg.h
+++ b/src/mesa/drivers/dri/intel/intel_reg.h
@@ -70,8 +70,10 @@
 
 /** @{
  * 915 definitions
+ *
+ * 915 documents say that bits 31:28 and 1 are "undefined, must be zero."
  */
-#define S0_VB_OFFSET_MASK		0xffffffc0
+#define S0_VB_OFFSET_MASK		0x0ffffffc
 #define S0_AUTO_CACHE_INV_DISABLE	(1<<0)
 /** @} */
 
diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c
index f042bcbc28..1172de90b1 100644
--- a/src/mesa/drivers/dri/intel/intel_regions.c
+++ b/src/mesa/drivers/dri/intel/intel_regions.c
@@ -164,7 +164,6 @@ intel_region_alloc_internal(struct intel_context *intel,
 
    /* Default to no tiling */
    region->tiling = I915_TILING_NONE;
-   region->bit_6_swizzle = I915_BIT_6_SWIZZLE_NONE;
 
    _DBG("%s <-- %p\n", __FUNCTION__, region);
    return region;
@@ -173,7 +172,7 @@ intel_region_alloc_internal(struct intel_context *intel,
 struct intel_region *
 intel_region_alloc(struct intel_context *intel,
 		   uint32_t tiling,
-                   GLuint cpp, GLuint width, GLuint height, GLuint pitch,
+                   GLuint cpp, GLuint width, GLuint height,
 		   GLboolean expect_accelerated_upload)
 {
    dri_bo *buffer;
@@ -187,19 +186,10 @@ intel_region_alloc(struct intel_context *intel,
    buffer = drm_intel_bo_alloc_tiled(intel->bufmgr, "region",
 				     width, height, cpp,
 				     &tiling, &aligned_pitch, flags);
-   /* We've already chosen a pitch as part of miptree layout.  It had
-    * better be the same.
-    */
-   assert(aligned_pitch == pitch * cpp);
 
    region = intel_region_alloc_internal(intel, cpp, width, height,
-					pitch, buffer);
-
-   if (tiling != I915_TILING_NONE) {
-      assert(((pitch * cpp) & 127) == 0);
-      drm_intel_bo_set_tiling(buffer, &tiling, pitch * cpp);
-      drm_intel_bo_get_tiling(buffer, &region->tiling, &region->bit_6_swizzle);
-   }
+					aligned_pitch / cpp, buffer);
+   region->tiling = tiling;
 
    return region;
 }
@@ -213,6 +203,7 @@ intel_region_alloc_for_handle(struct intel_context *intel,
    struct intel_region *region, *dummy;
    dri_bo *buffer;
    int ret;
+   uint32_t bit_6_swizzle;
 
    region = _mesa_HashLookup(intel->intelScreen->named_regions, handle);
    if (region != NULL) {
@@ -236,7 +227,7 @@ intel_region_alloc_for_handle(struct intel_context *intel,
       return region;
 
    ret = dri_bo_get_tiling(region->buffer, &region->tiling,
-			   &region->bit_6_swizzle);
+			   &bit_6_swizzle);
    if (ret != 0) {
       fprintf(stderr, "Couldn't get tiling of buffer %d (%s): %s\n",
 	      handle, name, strerror(-ret));
@@ -316,7 +307,7 @@ _mesa_copy_rect(GLubyte * dst,
    dst += dst_x * cpp;
    src += src_x * cpp;
    dst += dst_y * dst_pitch;
-   src += src_y * dst_pitch;
+   src += src_y * src_pitch;
    width *= cpp;
 
    if (width == dst_pitch && width == src_pitch)
@@ -380,8 +371,11 @@ intel_region_copy(struct intel_context *intel,
                   struct intel_region *src,
                   GLuint src_offset,
                   GLuint srcx, GLuint srcy, GLuint width, GLuint height,
+		  GLboolean flip,
 		  GLenum logicop)
 {
+   uint32_t src_pitch = src->pitch;
+
    _DBG("%s\n", __FUNCTION__);
 
    if (intel == NULL)
@@ -397,9 +391,12 @@ intel_region_copy(struct intel_context *intel,
 
    assert(src->cpp == dst->cpp);
 
+   if (flip)
+      src_pitch = -src_pitch;
+
    return intelEmitCopyBlit(intel,
 			    dst->cpp,
-			    src->pitch, src->buffer, src_offset, src->tiling,
+			    src_pitch, src->buffer, src_offset, src->tiling,
 			    dst->pitch, dst->buffer, dst_offset, dst->tiling,
 			    srcx, srcy, dstx, dsty, width, height,
 			    logicop);
diff --git a/src/mesa/drivers/dri/intel/intel_regions.h b/src/mesa/drivers/dri/intel/intel_regions.h
index 7ee6a988ea..2459c9a924 100644
--- a/src/mesa/drivers/dri/intel/intel_regions.h
+++ b/src/mesa/drivers/dri/intel/intel_regions.h
@@ -65,7 +65,6 @@ struct intel_region
    GLuint draw_x, draw_y; /**< Offset of drawing within the region */
 
    uint32_t tiling; /**< Which tiling mode the region is in */
-   uint32_t bit_6_swizzle; /**< GEM flag for address swizzling requirement */
    struct intel_buffer_object *pbo;     /* zero-copy uploads */
 
    uint32_t name; /**< Global name for the bo */
@@ -79,7 +78,7 @@ struct intel_region
 struct intel_region *intel_region_alloc(struct intel_context *intel,
                                         uint32_t tiling,
 					GLuint cpp, GLuint width,
-                                        GLuint height, GLuint pitch,
+                                        GLuint height,
 					GLboolean expect_accelerated_upload);
 
 struct intel_region *
@@ -122,6 +121,7 @@ intel_region_copy(struct intel_context *intel,
 		  struct intel_region *src,
 		  GLuint src_offset,
 		  GLuint srcx, GLuint srcy, GLuint width, GLuint height,
+		  GLboolean flip,
 		  GLenum logicop);
 
 /* Helpers for zerocopy uploads, particularly texture image uploads:
diff --git a/src/mesa/drivers/dri/intel/intel_screen.c b/src/mesa/drivers/dri/intel/intel_screen.c
index 6e4bb64365..5e3f40836d 100644
--- a/src/mesa/drivers/dri/intel/intel_screen.c
+++ b/src/mesa/drivers/dri/intel/intel_screen.c
@@ -312,18 +312,13 @@ intelCreateBuffer(__DRIscreen * driScrnPriv,
       }
 
       if (mesaVis->depthBits == 24) {
-	 if (mesaVis->stencilBits == 8) {
-	    /* combined depth/stencil buffer */
-	    struct intel_renderbuffer *depthStencilRb
-	       = intel_create_renderbuffer(MESA_FORMAT_S8_Z24);
-	    /* note: bind RB to two attachment points */
-	    _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &depthStencilRb->Base);
-	    _mesa_add_renderbuffer(fb, BUFFER_STENCIL, &depthStencilRb->Base);
-	 } else {
-	    struct intel_renderbuffer *depthRb
-	       = intel_create_renderbuffer(MESA_FORMAT_X8_Z24);
-	    _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &depthRb->Base);
-	 }
+	 assert(mesaVis->stencilBits == 8);
+	 /* combined depth/stencil buffer */
+	 struct intel_renderbuffer *depthStencilRb
+	    = intel_create_renderbuffer(MESA_FORMAT_S8_Z24);
+	 /* note: bind RB to two attachment points */
+	 _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &depthStencilRb->Base);
+	 _mesa_add_renderbuffer(fb, BUFFER_STENCIL, &depthStencilRb->Base);
       }
       else if (mesaVis->depthBits == 16) {
          /* just 16-bit depth buffer, no hw stencil */
diff --git a/src/mesa/drivers/dri/intel/intel_span.c b/src/mesa/drivers/dri/intel/intel_span.c
index fb5c01bc4d..c1e15d1b0f 100644
--- a/src/mesa/drivers/dri/intel/intel_span.c
+++ b/src/mesa/drivers/dri/intel/intel_span.c
@@ -48,11 +48,11 @@ intel_set_span_functions(struct intel_context *intel,
 
 #define LOCAL_VARS							\
    struct intel_renderbuffer *irb = intel_renderbuffer(rb);		\
-   const GLint yScale = ctx->DrawBuffer->Name ? 1 : -1;			\
-   const GLint yBias = ctx->DrawBuffer->Name ? 0 : irb->Base.Height - 1;\
+   const GLint yScale = rb->Name ? 1 : -1;				\
+   const GLint yBias = rb->Name ? 0 : rb->Height - 1;			\
    int minx = 0, miny = 0;						\
-   int maxx = ctx->DrawBuffer->Width;					\
-   int maxy = ctx->DrawBuffer->Height;					\
+   int maxx = rb->Width;						\
+   int maxy = rb->Height;						\
    int pitch = irb->region->pitch * irb->region->cpp;			\
    void *buf = irb->region->buffer->virtual;				\
    GLuint p;								\
@@ -108,11 +108,11 @@ intel_set_span_functions(struct intel_context *intel,
 
 #define LOCAL_DEPTH_VARS						\
    struct intel_renderbuffer *irb = intel_renderbuffer(rb);		\
-   const GLint yScale = ctx->DrawBuffer->Name ? 1 : -1;			\
-   const GLint yBias = ctx->DrawBuffer->Name ? 0 : irb->Base.Height - 1;\
+   const GLint yScale = rb->Name ? 1 : -1;				\
+   const GLint yBias = rb->Name ? 0 : rb->Height - 1;			\
    int minx = 0, miny = 0;						\
-   int maxx = ctx->DrawBuffer->Width;					\
-   int maxy = ctx->DrawBuffer->Height;					\
+   int maxx = rb->Width;						\
+   int maxy = rb->Height;						\
    int pitch = irb->region->pitch * irb->region->cpp;			\
    void *buf = irb->region->buffer->virtual;				\
    (void)buf; (void)pitch; /* unused for non-gttmap. */			\
@@ -134,7 +134,7 @@ intel_set_span_functions(struct intel_context *intel,
    (*(uint32_t *)(irb->region->buffer->virtual + NO_TILE(_x, _y)) = d)
 #define READ_DEPTH(d, _x, _y) \
    d = *(uint32_t *)(irb->region->buffer->virtual + NO_TILE(_x, _y))
-#define TAG(x) intel_##x##_z24_x8
+#define TAG(x) intel_##x##_z24_s8
 #include "depthtmp.h"
 
 void
@@ -361,7 +361,7 @@ intel_set_span_functions(struct intel_context *intel,
       break;
    case MESA_FORMAT_X8_Z24:
    case MESA_FORMAT_S8_Z24:
-      intel_InitDepthPointers_z24_x8(rb);
+      intel_InitDepthPointers_z24_s8(rb);
       break;
    default:
       _mesa_problem(NULL,
diff --git a/src/mesa/drivers/dri/intel/intel_tex_copy.c b/src/mesa/drivers/dri/intel/intel_tex_copy.c
index 13b8bcfa86..62e1e78f59 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_copy.c
@@ -130,18 +130,8 @@ do_copy_texsubimage(struct intel_context *intel,
       }
 
       if (ctx->ReadBuffer->Name == 0) {
-	 /* reading from a window, adjust x, y */
-	 const __DRIdrawable *dPriv = intel->driReadDrawable;
-	 y = dPriv->y + (dPriv->h - (y + height));
-	 x += dPriv->x;
-
-	 /* Invert the data coming from the source rectangle due to GL
-	  * and hardware disagreeing on where y=0 is.
-	  *
-	  * It appears that our offsets and pitches get mangled
-	  * appropriately by the hardware, and we don't need to adjust them
-	  * on our own.
-	  */
+	 /* Flip vertical orientation for system framebuffers */
+	 y = ctx->ReadBuffer->Height - (y + height);
 	 src_pitch = -src->pitch;
       } else {
 	 /* reading from a FBO, y is already oriented the way we like */
@@ -155,7 +145,7 @@ do_copy_texsubimage(struct intel_context *intel,
 			     src->buffer,
 			     0,
 			     src->tiling,
-			     intelImage->mt->pitch,
+			     intelImage->mt->region->pitch,
 			     dst_bo,
 			     0,
 			     intelImage->mt->region->tiling,
diff --git a/src/mesa/drivers/dri/intel/intel_tex_image.c b/src/mesa/drivers/dri/intel/intel_tex_image.c
index bac36eeb56..9db96acdc0 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_image.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_image.c
@@ -236,7 +236,7 @@ try_pbo_upload(struct intel_context *intel,
 				  intelImage->face, 0,
 				  &dst_x, &dst_y);
 
-   dst_stride = intelImage->mt->pitch;
+   dst_stride = intelImage->mt->region->pitch;
 
    if (drm_intel_bo_references(intel->batch->buf, dst_buffer))
       intelFlush(&intel->ctx);
@@ -290,7 +290,7 @@ try_pbo_zcopy(struct intel_context *intel,
 				  intelImage->face, 0,
 				  &dst_x, &dst_y);
 
-   dst_stride = intelImage->mt->pitch;
+   dst_stride = intelImage->mt->region->pitch;
 
    if (src_stride != dst_stride || dst_x != 0 || dst_y != 0 ||
        src_offset != 0) {
diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.c b/src/mesa/drivers/dri/intel/intel_tex_layout.c
index 7d69ea4484..d132e19e83 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_layout.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_layout.c
@@ -74,14 +74,14 @@ void i945_miptree_layout_2d( struct intel_context *intel,
    GLuint width = mt->width0;
    GLuint height = mt->height0;
 
-   mt->pitch = mt->width0;
+   mt->total_width = mt->width0;
    intel_get_texture_alignment_unit(mt->internal_format, &align_w, &align_h);
 
    if (mt->compressed) {
-       mt->pitch = ALIGN(mt->width0, align_w);
+       mt->total_width = ALIGN(mt->width0, align_w);
    }
 
-   /* May need to adjust pitch to accomodate the placement of
+   /* May need to adjust width to accomodate the placement of
     * the 2nd mipmap.  This occurs when the alignment
     * constraints of mipmap placement push the right edge of the
     * 2nd mipmap out past the width of its parent.
@@ -97,15 +97,11 @@ void i945_miptree_layout_2d( struct intel_context *intel,
                + minify(minify(mt->width0));
        }
 
-       if (mip1_width > mt->pitch) {
-           mt->pitch = mip1_width;
+       if (mip1_width > mt->total_width) {
+           mt->total_width = mip1_width;
        }
    }
 
-   /* Pitch must be a whole number of dwords, even though we
-    * express it in texels.
-    */
-   mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, mt->pitch);
    mt->total_height = 0;
 
    for ( level = mt->first_level ; level <= mt->last_level ; level++ ) {
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_context.c b/src/mesa/drivers/dri/nouveau/nouveau_context.c
index be57d48b8d..42bec659d7 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_context.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_context.c
@@ -54,6 +54,7 @@ static const struct dri_extension nouveau_extensions[] = {
 	{ "GL_ARB_texture_env_dot3",	NULL },
 	{ "GL_ARB_texture_mirrored_repeat", NULL },
 	{ "GL_EXT_fog_coord",		GL_EXT_fog_coord_functions },
+	{ "GL_EXT_framebuffer_blit",	NULL },
 	{ "GL_EXT_framebuffer_object",	GL_EXT_framebuffer_object_functions },
 	{ "GL_EXT_secondary_color",	GL_EXT_secondary_color_functions },
 	{ "GL_EXT_stencil_wrap",	NULL },
@@ -337,6 +338,8 @@ nouveau_validate_framebuffer(GLcontext *ctx)
 		update_framebuffer(dri_ctx, dri_read,
 				   &dri_ctx->dri2.read_stamp);
 
-	if (nouveau_next_dirty_state(ctx) >= 0)
+	if (nouveau_next_dirty_state(ctx) >= 0) {
+		nouveau_state_emit(ctx);
 		FIRE_RING(context_chan(ctx));
+	}
 }
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_driver.c b/src/mesa/drivers/dri/nouveau/nouveau_driver.c
index 1d12f43741..4ec864c181 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_driver.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_driver.c
@@ -135,4 +135,8 @@ nouveau_driver_functions_init(struct dd_function_table *functions)
 	functions->Flush = nouveau_flush;
 	functions->Finish = nouveau_finish;
 	functions->Clear = nouveau_clear;
+	functions->DrawPixels = _mesa_meta_DrawPixels;
+	functions->CopyPixels = _mesa_meta_CopyPixels;
+	functions->Bitmap = _mesa_meta_Bitmap;
+	functions->BlitFramebuffer = _mesa_meta_BlitFramebuffer;
 }
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_fbo.c b/src/mesa/drivers/dri/nouveau/nouveau_fbo.c
index 2ec3dc9242..8be7edb150 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_fbo.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_fbo.c
@@ -236,7 +236,7 @@ nouveau_render_texture(GLcontext *ctx, struct gl_framebuffer *fb,
 	/* Allocate a renderbuffer object for the texture if we
 	 * haven't already done so. */
 	if (!rb) {
-		rb = nouveau_renderbuffer_new(ctx, 0);
+		rb = nouveau_renderbuffer_new(ctx, ~0);
 		assert(rb);
 
 		rb->AllocStorage = NULL;
@@ -259,11 +259,7 @@ static void
 nouveau_finish_render_texture(GLcontext *ctx,
 			      struct gl_renderbuffer_attachment *att)
 {
-	struct nouveau_renderbuffer *nrb
-		= to_nouveau_renderbuffer(att->Renderbuffer);
-
 	texture_dirty(att->Texture);
-	nouveau_surface_ref(NULL, &nrb->surface);
 }
 
 void
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_gldefs.h b/src/mesa/drivers/dri/nouveau/nouveau_gldefs.h
index 00007a9a35..fbeed3baea 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_gldefs.h
+++ b/src/mesa/drivers/dri/nouveau/nouveau_gldefs.h
@@ -260,4 +260,23 @@ nvgl_filter_mode(unsigned filter)
 	}
 }
 
+static inline unsigned
+nvgl_texgen_mode(unsigned mode)
+{
+	switch (mode) {
+	case GL_EYE_LINEAR:
+		return 0x2400;
+	case GL_OBJECT_LINEAR:
+		return 0x2401;
+	case GL_SPHERE_MAP:
+		return 0x2402;
+	case GL_NORMAL_MAP:
+		return 0x8511;
+	case GL_REFLECTION_MAP:
+		return 0x8512;
+	default:
+		assert(0);
+	}
+}
+
 #endif
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_render.h b/src/mesa/drivers/dri/nouveau/nouveau_render.h
index bff0ccfd76..923b79b2cf 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_render.h
+++ b/src/mesa/drivers/dri/nouveau/nouveau_render.h
@@ -32,8 +32,8 @@
 struct nouveau_array_state;
 
 typedef void (*dispatch_t)(GLcontext *, unsigned int, int, unsigned int);
-typedef unsigned (*extract_u_t)(struct nouveau_array_state *a, int i, int j);
-typedef float (*extract_f_t)(struct nouveau_array_state *a, int i, int j);
+typedef unsigned (*extract_u_t)(struct nouveau_array_state *, int, int);
+typedef float (*extract_f_t)(struct nouveau_array_state *, int, int);
 
 struct nouveau_attr_info {
 	int vbo_index;
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_render_t.c b/src/mesa/drivers/dri/nouveau/nouveau_render_t.c
index c0505781cf..7ccd7e6416 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_render_t.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_render_t.c
@@ -254,7 +254,7 @@ get_scratch_vbo(GLcontext *ctx, unsigned size, struct nouveau_bo **bo,
  */
 static inline unsigned
 get_max_vertices(GLcontext *ctx, const struct _mesa_index_buffer *ib,
-		 unsigned n)
+		 int n)
 {
 	struct nouveau_render_state *render = to_render_state(ctx);
 
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_span.c b/src/mesa/drivers/dri/nouveau/nouveau_span.c
index f1a56dd03a..1bfdecc6a2 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_span.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_span.c
@@ -32,7 +32,6 @@
 #include "swrast/swrast.h"
 
 #define LOCAL_VARS							\
-	struct gl_framebuffer *fb = ctx->DrawBuffer;			\
 	struct nouveau_surface *s = &to_nouveau_renderbuffer(rb)->surface; \
 	GLuint p;							\
 	(void)p;
@@ -45,12 +44,12 @@
 #define HW_CLIPLOOP() {							\
 	int minx = 0;							\
 	int miny = 0;							\
-	int maxx = fb->Width;						\
-	int maxy = fb->Height;
+	int maxx = rb->Width;						\
+	int maxy = rb->Height;
 
 #define HW_ENDCLIPLOOP() }
 
-#define Y_FLIP(y) (fb->Name ? (y) : rb->Height - 1 - (y))
+#define Y_FLIP(y) (rb->Name ? (y) : rb->Height - 1 - (y))
 
 /* RGB565 span functions */
 #define SPANTMP_PIXEL_FMT GL_RGB
@@ -144,17 +143,28 @@ texture_unit_map_unmap(GLcontext *ctx, struct gl_texture_unit *u, GLboolean map)
 }
 
 static void
-span_map_unmap(GLcontext *ctx, GLboolean map)
+framebuffer_map_unmap(struct gl_framebuffer *fb, GLboolean map)
 {
 	int i;
 
-	for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++)
-		renderbuffer_map_unmap(ctx->DrawBuffer->_ColorDrawBuffers[i], map);
+	for (i = 0; i < fb->_NumColorDrawBuffers; i++)
+		renderbuffer_map_unmap(fb->_ColorDrawBuffers[i], map);
+
+	renderbuffer_map_unmap(fb->_ColorReadBuffer, map);
+
+	if (fb->_DepthBuffer)
+		renderbuffer_map_unmap(fb->_DepthBuffer->Wrapped, map);
+}
+
+static void
+span_map_unmap(GLcontext *ctx, GLboolean map)
+{
+	int i;
 
-	renderbuffer_map_unmap(ctx->DrawBuffer->_ColorReadBuffer, map);
+	framebuffer_map_unmap(ctx->DrawBuffer, map);
 
-	if (ctx->DrawBuffer->_DepthBuffer)
-		renderbuffer_map_unmap(ctx->DrawBuffer->_DepthBuffer->Wrapped, map);
+	if (ctx->ReadBuffer != ctx->DrawBuffer)
+		framebuffer_map_unmap(ctx->ReadBuffer, map);
 
 	for (i = 0; i < ctx->Const.MaxTextureUnits; i++)
 		texture_unit_map_unmap(ctx, &ctx->Texture.Unit[i], map);
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_state.c b/src/mesa/drivers/dri/nouveau/nouveau_state.c
index bc610451b4..a57df2d9dc 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_state.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_state.c
@@ -189,6 +189,7 @@ nouveau_enable(GLcontext *ctx, GLenum cap, GLboolean state)
 	case GL_LIGHTING:
 		context_dirty(ctx, FRAG);
 		context_dirty(ctx, MODELVIEW);
+		context_dirty(ctx, LIGHT_MODEL);
 		context_dirty(ctx, LIGHT_ENABLE);
 
 		for (i = 0; i < MAX_LIGHTS; i++) {
@@ -231,9 +232,17 @@ nouveau_enable(GLcontext *ctx, GLenum cap, GLboolean state)
 	case GL_TEXTURE_1D:
 	case GL_TEXTURE_2D:
 	case GL_TEXTURE_3D:
+	case GL_TEXTURE_RECTANGLE:
 		context_dirty_i(ctx, TEX_ENV, ctx->Texture.CurrentUnit);
 		context_dirty_i(ctx, TEX_OBJ, ctx->Texture.CurrentUnit);
 		break;
+	case GL_TEXTURE_GEN_S:
+	case GL_TEXTURE_GEN_T:
+	case GL_TEXTURE_GEN_R:
+	case GL_TEXTURE_GEN_Q:
+		context_dirty_i(ctx, TEX_GEN, ctx->Texture.CurrentUnit);
+		context_dirty(ctx, MODELVIEW);
+		break;
 	}
 }
 
@@ -368,7 +377,15 @@ static void
 nouveau_tex_gen(GLcontext *ctx, GLenum coord, GLenum pname,
 		const GLfloat *params)
 {
-	context_dirty_i(ctx, TEX_GEN, ctx->Texture.CurrentUnit);
+	switch (pname) {
+	case GL_TEXTURE_GEN_MODE:
+		context_dirty_i(ctx, TEX_GEN, ctx->Texture.CurrentUnit);
+		context_dirty(ctx, MODELVIEW);
+		break;
+	default:
+		context_dirty_i(ctx, TEX_GEN, ctx->Texture.CurrentUnit);
+		break;
+	}
 }
 
 static void
@@ -454,12 +471,19 @@ nouveau_state_emit(GLcontext *ctx)
 static void
 nouveau_update_state(GLcontext *ctx, GLbitfield new_state)
 {
+	int i;
+
 	if (new_state & (_NEW_PROJECTION | _NEW_MODELVIEW))
 		context_dirty(ctx, PROJECTION);
 
 	if (new_state & _NEW_MODELVIEW)
 		context_dirty(ctx, MODELVIEW);
 
+	if (new_state & _NEW_TEXTURE_MATRIX) {
+		for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++)
+			context_dirty_i(ctx, TEX_MAT, i);
+	}
+
 	if (new_state & _NEW_CURRENT_ATTRIB &&
 	    new_state & _NEW_LIGHT) {
 		context_dirty(ctx, MATERIAL_FRONT_AMBIENT);
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_state.h b/src/mesa/drivers/dri/nouveau/nouveau_state.h
index d01d962c9f..38ac9753c8 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_state.h
+++ b/src/mesa/drivers/dri/nouveau/nouveau_state.h
@@ -89,6 +89,10 @@ enum {
 	NOUVEAU_STATE_TEX_GEN1,
 	NOUVEAU_STATE_TEX_GEN2,
 	NOUVEAU_STATE_TEX_GEN3,
+	NOUVEAU_STATE_TEX_MAT0,
+	NOUVEAU_STATE_TEX_MAT1,
+	NOUVEAU_STATE_TEX_MAT2,
+	NOUVEAU_STATE_TEX_MAT3,
 	NOUVEAU_STATE_TEX_OBJ0,
 	NOUVEAU_STATE_TEX_OBJ1,
 	NOUVEAU_STATE_TEX_OBJ2,
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_texture.c b/src/mesa/drivers/dri/nouveau/nouveau_texture.c
index bf365bfca3..dbf9a5cc61 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_texture.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_texture.c
@@ -177,15 +177,15 @@ nouveau_choose_tex_format(GLcontext *ctx, GLint internalFormat,
 }
 
 static GLboolean
-teximage_fits(struct gl_texture_object *t, int level,
-	      struct gl_texture_image *ti)
+teximage_fits(struct gl_texture_object *t, int level)
 {
 	struct nouveau_surface *s = &to_nouveau_texture(t)->surfaces[level];
+	struct gl_texture_image *ti = t->Image[0][level];
 
-	return t->Target == GL_TEXTURE_RECTANGLE ||
-		(s->bo && s->width == ti->Width &&
-		 s->height == ti->Height &&
-		 s->format == ti->TexFormat);
+	return ti && (t->Target == GL_TEXTURE_RECTANGLE ||
+		      (s->bo && s->width == ti->Width &&
+		       s->height == ti->Height &&
+		       s->format == ti->TexFormat));
 }
 
 static GLboolean
@@ -195,7 +195,7 @@ validate_teximage(GLcontext *ctx, struct gl_texture_object *t,
 {
 	struct gl_texture_image *ti = t->Image[0][level];
 
-	if (ti && teximage_fits(t, level, ti)) {
+	if (teximage_fits(t, level)) {
 		struct nouveau_surface *ss = to_nouveau_texture(t)->surfaces;
 		struct nouveau_surface *s = &to_nouveau_teximage(ti)->surface;
 
@@ -283,7 +283,8 @@ nouveau_texture_validate(GLcontext *ctx, struct gl_texture_object *t)
 	struct nouveau_texture *nt = to_nouveau_texture(t);
 	int i, last = get_last_level(t);
 
-	if (!nt->surfaces[last].bo)
+	if (!teximage_fits(t, t->BaseLevel) ||
+	    !teximage_fits(t, last))
 		return GL_FALSE;
 
 	if (nt->dirty) {
@@ -296,6 +297,8 @@ nouveau_texture_validate(GLcontext *ctx, struct gl_texture_object *t)
 			validate_teximage(ctx, t, i, 0, 0, 0,
 					  s->width, s->height, 1);
 		}
+
+		FIRE_RING(context_chan(ctx));
 	}
 
 	return GL_TRUE;
@@ -304,9 +307,12 @@ nouveau_texture_validate(GLcontext *ctx, struct gl_texture_object *t)
 void
 nouveau_texture_reallocate(GLcontext *ctx, struct gl_texture_object *t)
 {
-	texture_dirty(t);
-	relayout_texture(ctx, t);
-	nouveau_texture_validate(ctx, t);
+	if (!teximage_fits(t, t->BaseLevel) ||
+	    !teximage_fits(t, get_last_level(t))) {
+		texture_dirty(t);
+		relayout_texture(ctx, t);
+		nouveau_texture_validate(ctx, t);
+	}
 }
 
 static unsigned
@@ -364,7 +370,7 @@ nouveau_teximage(GLcontext *ctx, GLint dims, GLenum target, GLint level,
 	}
 
 	if (level == t->BaseLevel) {
-		if (!teximage_fits(t, level, ti))
+		if (!teximage_fits(t, level))
 			relayout_texture(ctx, t);
 		nouveau_texture_validate(ctx, t);
 	}
@@ -416,6 +422,40 @@ nouveau_teximage_3d(GLcontext *ctx, GLenum target, GLint level,
 }
 
 static void
+nouveau_texsubimage(GLcontext *ctx, GLint dims, GLenum target, GLint level,
+		    GLint xoffset, GLint yoffset, GLint zoffset,
+		    GLint width, GLint height, GLint depth,
+		    GLenum format, GLenum type, const void *pixels,
+		    const struct gl_pixelstore_attrib *packing,
+		    struct gl_texture_object *t,
+		    struct gl_texture_image *ti)
+{
+	struct nouveau_surface *s = &to_nouveau_teximage(ti)->surface;
+	int ret;
+
+	pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, depth,
+					     format, type, pixels, packing,
+					     "glTexSubImage");
+	if (pixels) {
+		nouveau_teximage_map(ctx, ti);
+
+		ret = _mesa_texstore(ctx, 3, ti->_BaseFormat, ti->TexFormat,
+				     ti->Data, xoffset, yoffset, zoffset,
+				     s->pitch, ti->ImageOffsets,
+				     width, height, depth, format, type,
+				     pixels, packing);
+		assert(ret);
+
+		nouveau_teximage_unmap(ctx, ti);
+		_mesa_unmap_teximage_pbo(ctx, packing);
+	}
+
+	if (!to_nouveau_texture(t)->dirty)
+		validate_teximage(ctx, t, level, xoffset, yoffset, zoffset,
+				  width, height, depth);
+}
+
+static void
 nouveau_texsubimage_3d(GLcontext *ctx, GLenum target, GLint level,
 		       GLint xoffset, GLint yoffset, GLint zoffset,
 		       GLint width, GLint height, GLint depth,
@@ -424,15 +464,9 @@ nouveau_texsubimage_3d(GLcontext *ctx, GLenum target, GLint level,
 		       struct gl_texture_object *t,
 		       struct gl_texture_image *ti)
 {
-	nouveau_teximage_map(ctx, ti);
-	_mesa_store_texsubimage3d(ctx, target, level, xoffset, yoffset, zoffset,
-				  width, height, depth, format, type, pixels,
-				  packing, t, ti);
-	nouveau_teximage_unmap(ctx, ti);
-
-	if (!to_nouveau_texture(t)->dirty)
-		validate_teximage(ctx, t, level, xoffset, yoffset, zoffset,
-				  width, height, depth);
+	nouveau_texsubimage(ctx, 3, target, level, xoffset, yoffset, zoffset,
+			    width, height, depth, format, type, pixels,
+			    packing, t, ti);
 }
 
 static void
@@ -444,15 +478,9 @@ nouveau_texsubimage_2d(GLcontext *ctx, GLenum target, GLint level,
 		       struct gl_texture_object *t,
 		       struct gl_texture_image *ti)
 {
-	nouveau_teximage_map(ctx, ti);
-	_mesa_store_texsubimage2d(ctx, target, level, xoffset, yoffset,
-				  width, height, format, type, pixels,
-				  packing, t, ti);
-	nouveau_teximage_unmap(ctx, ti);
-
-	if (!to_nouveau_texture(t)->dirty)
-		validate_teximage(ctx, t, level, xoffset, yoffset, 0,
-				  width, height, 1);
+	nouveau_texsubimage(ctx, 2, target, level, xoffset, yoffset, 0,
+			    width, height, 1, format, type, pixels,
+			    packing, t, ti);
 }
 
 static void
@@ -463,15 +491,9 @@ nouveau_texsubimage_1d(GLcontext *ctx, GLenum target, GLint level,
 		       struct gl_texture_object *t,
 		       struct gl_texture_image *ti)
 {
-	nouveau_teximage_map(ctx, ti);
-	_mesa_store_texsubimage1d(ctx, target, level, xoffset,
-				  width, format, type, pixels,
-				  packing, t, ti);
-	nouveau_teximage_unmap(ctx, ti);
-
-	if (!to_nouveau_texture(t)->dirty)
-		validate_teximage(ctx, t, level, xoffset, 0, 0,
-				  width, 1, 1);
+	nouveau_texsubimage(ctx, 1, target, level, xoffset, 0, 0,
+			    width, 1, 1, format, type, pixels,
+			    packing, t, ti);
 }
 
 static void
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_texture.h b/src/mesa/drivers/dri/nouveau/nouveau_texture.h
index b91facbdeb..251f537bba 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_texture.h
+++ b/src/mesa/drivers/dri/nouveau/nouveau_texture.h
@@ -41,7 +41,7 @@ struct nouveau_texture {
 #define to_nouveau_texture(x) ((struct nouveau_texture *)(x))
 
 #define texture_dirty(t) \
-	to_nouveau_texture(t)->dirty = GL_TRUE;
+	to_nouveau_texture(t)->dirty = GL_TRUE
 
 void
 nouveau_set_texbuffer(__DRIcontext *dri_ctx,
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_util.h b/src/mesa/drivers/dri/nouveau/nouveau_util.h
index d6007aba2b..584cb80ef6 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_util.h
+++ b/src/mesa/drivers/dri/nouveau/nouveau_util.h
@@ -191,4 +191,22 @@ is_texture_source(int s)
 	return s == GL_TEXTURE || (s >= GL_TEXTURE0 && s <= GL_TEXTURE31);
 }
 
+static inline struct gl_texgen *
+get_texgen_coord(struct gl_texture_unit *u, int i)
+{
+	return ((struct gl_texgen *[])
+		{ &u->GenS, &u->GenT, &u->GenR, &u->GenQ }) [i];
+}
+
+static inline float *
+get_texgen_coeff(struct gl_texgen *c)
+{
+	if (c->Mode == GL_OBJECT_LINEAR)
+		return c->ObjectPlane;
+	else if (c->Mode == GL_EYE_LINEAR)
+		return c->EyePlane;
+	else
+		return NULL;
+}
+
 #endif
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_vbo_t.c b/src/mesa/drivers/dri/nouveau/nouveau_vbo_t.c
index a365b977f2..e5858f8268 100644
--- a/src/mesa/drivers/dri/nouveau/nouveau_vbo_t.c
+++ b/src/mesa/drivers/dri/nouveau/nouveau_vbo_t.c
@@ -85,6 +85,18 @@ vbo_deinit_array(struct nouveau_array_state *a)
 	a->fields = 0;
 }
 
+static int
+get_array_stride(GLcontext *ctx, const struct gl_client_array *a)
+{
+	struct nouveau_render_state *render = to_render_state(ctx);
+
+	if (render->mode == VBO && !_mesa_is_bufferobj(a->BufferObj))
+		/* Pack client buffers. */
+		return align(_mesa_sizeof_type(a->Type) * a->Size, 4);
+	else
+		return a->StrideB;
+}
+
 static void
 vbo_init_arrays(GLcontext *ctx, const struct _mesa_index_buffer *ib,
 		const struct gl_client_array **arrays)
@@ -101,18 +113,10 @@ vbo_init_arrays(GLcontext *ctx, const struct _mesa_index_buffer *ib,
 
 		if (attr >= 0) {
 			const struct gl_client_array *array = arrays[attr];
-			int stride;
-
-			if (render->mode == VBO &&
-			    !_mesa_is_bufferobj(array->BufferObj))
-				/* Pack client buffers. */
-				stride = align(_mesa_sizeof_type(array->Type)
-					       * array->Size, 4);
-			else
-				stride = array->StrideB;
 
 			vbo_init_array(&render->attrs[attr], attr,
-				       stride, array->Size, array->Type,
+				       get_array_stride(ctx, array),
+				       array->Size, array->Type,
 				       array->BufferObj, array->Ptr,
 				       render->mode == IMM);
 		}
@@ -224,9 +228,11 @@ vbo_choose_attrs(GLcontext *ctx, const struct gl_client_array **arrays)
 	if (ctx->Fog.Enabled && ctx->Fog.FogCoordinateSource == GL_FOG_COORD)
 		vbo_emit_attr(ctx, arrays, VERT_ATTRIB_FOG);
 
-	if (ctx->Light.Enabled) {
+	if (ctx->Light.Enabled ||
+	    (ctx->Texture._GenFlags & TEXGEN_NEED_NORMALS))
 		vbo_emit_attr(ctx, arrays, VERT_ATTRIB_NORMAL);
 
+	if (ctx->Light.Enabled) {
 		vbo_emit_attr(ctx, arrays, MAT(FRONT_AMBIENT));
 		vbo_emit_attr(ctx, arrays, MAT(FRONT_DIFFUSE));
 		vbo_emit_attr(ctx, arrays, MAT(FRONT_SPECULAR));
@@ -243,18 +249,21 @@ vbo_choose_attrs(GLcontext *ctx, const struct gl_client_array **arrays)
 	vbo_emit_attr(ctx, arrays, VERT_ATTRIB_POS);
 }
 
-static unsigned
-get_max_client_stride(GLcontext *ctx)
+static int
+get_max_client_stride(GLcontext *ctx, const struct gl_client_array **arrays)
 {
 	struct nouveau_render_state *render = to_render_state(ctx);
 	int i, s = 0;
 
 	for (i = 0; i < render->attr_count; i++) {
 		int attr = render->map[i];
-		struct nouveau_array_state *a = &render->attrs[attr];
 
-		if (attr >= 0 && !a->bo)
-			s = MAX2(a->stride, s);
+		if (attr >= 0) {
+			const struct gl_client_array *a = arrays[attr];
+
+			if (!_mesa_is_bufferobj(a->BufferObj))
+				s = MAX2(s, get_array_stride(ctx, a));
+		}
 	}
 
 	return s;
@@ -275,14 +284,15 @@ vbo_maybe_split(GLcontext *ctx, const struct gl_client_array **arrays,
 {
 	struct nouveau_context *nctx = to_nouveau_context(ctx);
 	struct nouveau_render_state *render = to_render_state(ctx);
-	unsigned pushbuf_avail = PUSHBUF_DWORDS - 2 * nctx->bo.count,
+	unsigned pushbuf_avail = PUSHBUF_DWORDS - 2 * (nctx->bo.count +
+						       render->attr_count),
 		vert_avail = get_max_vertices(ctx, NULL, pushbuf_avail),
 		idx_avail = get_max_vertices(ctx, ib, pushbuf_avail);
 	int stride;
 
 	/* Try to keep client buffers smaller than the scratch BOs. */
 	if (render->mode == VBO &&
-	    (stride = get_max_client_stride(ctx)))
+	    (stride = get_max_client_stride(ctx, arrays)))
 		    vert_avail = MIN2(vert_avail,
 				      RENDER_SCRATCH_SIZE / stride);
 
@@ -321,6 +331,7 @@ vbo_bind_vertices(GLcontext *ctx, const struct gl_client_array **arrays,
 				* array->StrideB;
 
 			if (a->bo) {
+				/* Array in a buffer obj. */
 				a->offset = (intptr_t)array->Ptr + delta;
 			} else {
 				int j, n = max_index - min_index + 1;
@@ -328,6 +339,8 @@ vbo_bind_vertices(GLcontext *ctx, const struct gl_client_array **arrays,
 				char *dp = get_scratch_vbo(ctx, n * a->stride,
 							   &a->bo, &a->offset);
 
+				/* Array in client memory, move it to
+				 * a scratch buffer obj. */
 				for (j = 0; j < n; j++)
 					memcpy(dp + j * a->stride,
 					       sp + j * array->StrideB,
@@ -371,8 +384,6 @@ vbo_draw_vbo(GLcontext *ctx, const struct gl_client_array **arrays,
 		dispatch(ctx, start, delta, count);
 		BATCH_END();
 	}
-
-	FIRE_RING(chan);
 }
 
 /* Immediate rendering path. */
@@ -416,8 +427,6 @@ vbo_draw_imm(GLcontext *ctx, const struct gl_client_array **arrays,
 
 		BATCH_END();
 	}
-
-	FIRE_RING(chan);
 }
 
 /* draw_prims entry point when we're doing hw-tnl. */
diff --git a/src/mesa/drivers/dri/nouveau/nv04_context.c b/src/mesa/drivers/dri/nouveau/nv04_context.c
index 3624b3af92..6834f7cd3d 100644
--- a/src/mesa/drivers/dri/nouveau/nv04_context.c
+++ b/src/mesa/drivers/dri/nouveau/nv04_context.c
@@ -276,6 +276,10 @@ const struct nouveau_driver nv04_driver = {
 		nouveau_emit_nothing,
 		nouveau_emit_nothing,
 		nouveau_emit_nothing,
+		nouveau_emit_nothing,
+		nouveau_emit_nothing,
+		nouveau_emit_nothing,
+		nouveau_emit_nothing,
 		nv04_emit_tex_obj,
 		nv04_emit_tex_obj,
 		nouveau_emit_nothing,
diff --git a/src/mesa/drivers/dri/nouveau/nv10_context.c b/src/mesa/drivers/dri/nouveau/nv10_context.c
index 860d0aeb8f..b6d10361de 100644
--- a/src/mesa/drivers/dri/nouveau/nv10_context.c
+++ b/src/mesa/drivers/dri/nouveau/nv10_context.c
@@ -212,7 +212,7 @@ nv10_hwctx_init(GLcontext *ctx)
 	OUT_RING(chan, 0);
 	BEGIN_RING(chan, celsius, NV10TCL_CULL_FACE_ENABLE, 1);
 	OUT_RING(chan, 0);
-	BEGIN_RING(chan, celsius, NV10TCL_TX_GEN_S(0), 8);
+	BEGIN_RING(chan, celsius, NV10TCL_TX_GEN_MODE_S(0), 8);
 	for (i = 0; i < 8; i++)
 		OUT_RING(chan, 0);
 
@@ -412,6 +412,10 @@ const struct nouveau_driver nv10_driver = {
 		nv10_emit_tex_gen,
 		nouveau_emit_nothing,
 		nouveau_emit_nothing,
+		nv10_emit_tex_mat,
+		nv10_emit_tex_mat,
+		nouveau_emit_nothing,
+		nouveau_emit_nothing,
 		nv10_emit_tex_obj,
 		nv10_emit_tex_obj,
 		nouveau_emit_nothing,
diff --git a/src/mesa/drivers/dri/nouveau/nv10_driver.h b/src/mesa/drivers/dri/nouveau/nv10_driver.h
index d662712533..cefd6c6fba 100644
--- a/src/mesa/drivers/dri/nouveau/nv10_driver.h
+++ b/src/mesa/drivers/dri/nouveau/nv10_driver.h
@@ -134,6 +134,9 @@ void
 nv10_emit_tex_gen(GLcontext *ctx, int emit);
 
 void
+nv10_emit_tex_mat(GLcontext *ctx, int emit);
+
+void
 nv10_emit_tex_obj(GLcontext *ctx, int emit);
 
 /* nv10_state_tnl.c */
diff --git a/src/mesa/drivers/dri/nouveau/nv10_state_fb.c b/src/mesa/drivers/dri/nouveau/nv10_state_fb.c
index 6bd383ebcd..a2fcb6b695 100644
--- a/src/mesa/drivers/dri/nouveau/nv10_state_fb.c
+++ b/src/mesa/drivers/dri/nouveau/nv10_state_fb.c
@@ -71,6 +71,7 @@ setup_lma_buffer(GLcontext *ctx)
 	nouveau_bo_markl(bctx, celsius, NV17TCL_LMA_DEPTH_BUFFER_OFFSET,
 			 nfb->lma_bo, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_RDWR);
 
+	WAIT_RING(chan, 9);
 	BEGIN_RING(chan, celsius, NV17TCL_LMA_DEPTH_WINDOW_X, 4);
 	OUT_RINGf(chan, - 1792);
 	OUT_RINGf(chan, - 2304 + fb->Height);
@@ -171,15 +172,13 @@ nv10_emit_viewport(GLcontext *ctx, int emit)
 	struct nouveau_grobj *celsius = context_eng3d(ctx);
 	struct gl_framebuffer *fb = ctx->DrawBuffer;
 	float a[4] = {};
-	int i;
 
 	get_viewport_translate(ctx, a);
 	a[0] -= 2048;
 	a[1] -= 2048;
 
 	BEGIN_RING(chan, celsius, NV10TCL_VIEWPORT_TRANSLATE_X, 4);
-	for (i = 0; i < 4; i++)
-		OUT_RINGf(chan, a[i]);
+	OUT_RINGp(chan, a, 4);
 
 	BEGIN_RING(chan, celsius, NV10TCL_VIEWPORT_CLIP_HORIZ(0), 1);
 	OUT_RING(chan, (fb->Width - 1) << 16 | 0x08000800);
diff --git a/src/mesa/drivers/dri/nouveau/nv10_state_tex.c b/src/mesa/drivers/dri/nouveau/nv10_state_tex.c
index 02a5ca797a..6dedb18c72 100644
--- a/src/mesa/drivers/dri/nouveau/nv10_state_tex.c
+++ b/src/mesa/drivers/dri/nouveau/nv10_state_tex.c
@@ -32,9 +32,64 @@
 #include "nouveau_util.h"
 #include "nv10_driver.h"
 
+#define TX_GEN_MODE(i, j) (NV10TCL_TX_GEN_MODE_S(i) + 4 * (j))
+#define TX_GEN_COEFF(i, j) (NV10TCL_TX_GEN_COEFF_S_A(i) + 16 * (j))
+#define TX_MATRIX(i) (NV10TCL_TX0_MATRIX(0) + 64 * (i))
+
 void
 nv10_emit_tex_gen(GLcontext *ctx, int emit)
 {
+	const int i = emit - NOUVEAU_STATE_TEX_GEN0;
+	struct nouveau_context *nctx = to_nouveau_context(ctx);
+	struct nouveau_channel *chan = context_chan(ctx);
+	struct nouveau_grobj *celsius = context_eng3d(ctx);
+	struct gl_texture_unit *unit = &ctx->Texture.Unit[i];
+	int j;
+
+	for (j = 0; j < 4; j++) {
+		if (nctx->fallback == HWTNL && (unit->TexGenEnabled & 1 << j)) {
+			struct gl_texgen *coord = get_texgen_coord(unit, j);
+			float *k = get_texgen_coeff(coord);
+
+			if (k) {
+				BEGIN_RING(chan, celsius,
+					   TX_GEN_COEFF(i, j), 4);
+				OUT_RINGp(chan, k, 4);
+			}
+
+			BEGIN_RING(chan, celsius, TX_GEN_MODE(i, j), 1);
+			OUT_RING(chan, nvgl_texgen_mode(coord->Mode));
+
+		} else {
+			BEGIN_RING(chan, celsius, TX_GEN_MODE(i, j), 1);
+			OUT_RING(chan, 0);
+		}
+	}
+
+	context_dirty_i(ctx, TEX_MAT, i);
+}
+
+void
+nv10_emit_tex_mat(GLcontext *ctx, int emit)
+{
+	const int i = emit - NOUVEAU_STATE_TEX_MAT0;
+	struct nouveau_context *nctx = to_nouveau_context(ctx);
+	struct nouveau_channel *chan = context_chan(ctx);
+	struct nouveau_grobj *celsius = context_eng3d(ctx);
+
+	if (nctx->fallback == HWTNL &&
+	    ((ctx->Texture._TexMatEnabled & 1 << i) ||
+	     ctx->Texture.Unit[i]._GenFlags)) {
+		BEGIN_RING(chan, celsius, NV10TCL_TX_MATRIX_ENABLE(i), 1);
+		OUT_RING(chan, 1);
+
+		BEGIN_RING(chan, celsius, TX_MATRIX(i), 16);
+		OUT_RINGm(chan, ctx->TextureMatrixStack[i].Top->m);
+
+	} else {
+		BEGIN_RING(chan, celsius, NV10TCL_TX_MATRIX_ENABLE(i), 1);
+		OUT_RING(chan, 0);
+	}
 }
 
 static uint32_t
diff --git a/src/mesa/drivers/dri/nouveau/nv10_state_tnl.c b/src/mesa/drivers/dri/nouveau/nv10_state_tnl.c
index 406e24c455..0e592a1629 100644
--- a/src/mesa/drivers/dri/nouveau/nv10_state_tnl.c
+++ b/src/mesa/drivers/dri/nouveau/nv10_state_tnl.c
@@ -140,9 +140,7 @@ nv10_emit_fog(GLcontext *ctx, int emit)
 	OUT_RING(chan, pack_rgba_f(MESA_FORMAT_RGBA8888_REV, f->Color));
 
 	BEGIN_RING(chan, celsius, NV10TCL_FOG_EQUATION_CONSTANT, 3);
-	OUT_RINGf(chan, k[0]);
-	OUT_RINGf(chan, k[1]);
-	OUT_RINGf(chan, k[2]);
+	OUT_RINGp(chan, k, 3);
 
 	context_dirty(ctx, FRAG);
 }
@@ -284,9 +282,7 @@ nv10_emit_light_source(GLcontext *ctx, int emit)
 
 	if (l->_Flags & LIGHT_POSITIONAL) {
 		BEGIN_RING(chan, celsius, NV10TCL_LIGHT_POSITION_X(i), 3);
-		OUT_RINGf(chan, l->_Position[0]);
-		OUT_RINGf(chan, l->_Position[1]);
-		OUT_RINGf(chan, l->_Position[2]);
+		OUT_RINGp(chan, l->_Position, 3);
 
 		BEGIN_RING(chan, celsius,
 			   NV10TCL_LIGHT_ATTENUATION_CONSTANT(i), 3);
@@ -296,14 +292,10 @@ nv10_emit_light_source(GLcontext *ctx, int emit)
 
 	} else {
 		BEGIN_RING(chan, celsius, NV10TCL_LIGHT_DIRECTION_X(i), 3);
-		OUT_RINGf(chan, l->_VP_inf_norm[0]);
-		OUT_RINGf(chan, l->_VP_inf_norm[1]);
-		OUT_RINGf(chan, l->_VP_inf_norm[2]);
+		OUT_RINGp(chan, l->_VP_inf_norm, 3);
 
 		BEGIN_RING(chan, celsius, NV10TCL_LIGHT_HALF_VECTOR_X(i), 3);
-		OUT_RINGf(chan, l->_h_inf_norm[0]);
-		OUT_RINGf(chan, l->_h_inf_norm[1]);
-		OUT_RINGf(chan, l->_h_inf_norm[2]);
+		OUT_RINGp(chan, l->_h_inf_norm, 3);
 	}
 
 	if (l->_Flags & LIGHT_SPOT) {
@@ -312,13 +304,7 @@ nv10_emit_light_source(GLcontext *ctx, int emit)
 		nv10_get_spot_coeff(l, k);
 
 		BEGIN_RING(chan, celsius, NV10TCL_LIGHT_SPOT_CUTOFF_A(i), 7);
-		OUT_RINGf(chan, k[0]);
-		OUT_RINGf(chan, k[1]);
-		OUT_RINGf(chan, k[2]);
-		OUT_RINGf(chan, k[3]);
-		OUT_RINGf(chan, k[4]);
-		OUT_RINGf(chan, k[5]);
-		OUT_RINGf(chan, k[6]);
+		OUT_RINGp(chan, k, 7);
 	}
 }
 
@@ -350,15 +336,11 @@ nv10_emit_material_ambient(GLcontext *ctx, int emit)
 	}
 
 	BEGIN_RING(chan, celsius, NV10TCL_LIGHT_MODEL_AMBIENT_R, 3);
-	OUT_RINGf(chan, c_scene[0]);
-	OUT_RINGf(chan, c_scene[1]);
-	OUT_RINGf(chan, c_scene[2]);
+	OUT_RINGp(chan, c_scene, 3);
 
 	if (ctx->Light.ColorMaterialEnabled) {
 		BEGIN_RING(chan, celsius, NV10TCL_MATERIAL_FACTOR_R, 3);
-		OUT_RINGf(chan, c_factor[0]);
-		OUT_RINGf(chan, c_factor[1]);
-		OUT_RINGf(chan, c_factor[2]);
+		OUT_RINGp(chan, c_factor, 3);
 	}
 
 	foreach(l, &ctx->Light.EnabledList) {
@@ -368,9 +350,7 @@ nv10_emit_material_ambient(GLcontext *ctx, int emit)
 				  l->_MatAmbient[0]);
 
 		BEGIN_RING(chan, celsius, NV10TCL_LIGHT_AMBIENT_R(i), 3);
-		OUT_RINGf(chan, c_light[0]);
-		OUT_RINGf(chan, c_light[1]);
-		OUT_RINGf(chan, c_light[2]);
+		OUT_RINGp(chan, c_light, 3);
 	}
 }
 
@@ -392,9 +372,7 @@ nv10_emit_material_diffuse(GLcontext *ctx, int emit)
 				  l->_MatDiffuse[0]);
 
 		BEGIN_RING(chan, celsius, NV10TCL_LIGHT_DIFFUSE_R(i), 3);
-		OUT_RINGf(chan, c_light[0]);
-		OUT_RINGf(chan, c_light[1]);
-		OUT_RINGf(chan, c_light[2]);
+		OUT_RINGp(chan, c_light, 3);
 	}
 }
 
@@ -412,9 +390,7 @@ nv10_emit_material_specular(GLcontext *ctx, int emit)
 				  l->_MatSpecular[0]);
 
 		BEGIN_RING(chan, celsius, NV10TCL_LIGHT_SPECULAR_R(i), 3);
-		OUT_RINGf(chan, c_light[0]);
-		OUT_RINGf(chan, c_light[1]);
-		OUT_RINGf(chan, c_light[2]);
+		OUT_RINGp(chan, c_light, 3);
 	}
 }
 
@@ -455,12 +431,7 @@ nv10_emit_material_shininess(GLcontext *ctx, int emit)
 		k);
 
 	BEGIN_RING(chan, celsius, NV10TCL_MATERIAL_SHININESS(0), 6);
-	OUT_RINGf(chan, k[0]);
-	OUT_RINGf(chan, k[1]);
-	OUT_RINGf(chan, k[2]);
-	OUT_RINGf(chan, k[3]);
-	OUT_RINGf(chan, k[4]);
-	OUT_RINGf(chan, k[5]);
+	OUT_RINGp(chan, k, 6);
 }
 
 void
@@ -474,12 +445,14 @@ nv10_emit_modelview(GLcontext *ctx, int emit)
 	if (nctx->fallback != HWTNL)
 		return;
 
-	if (ctx->Light._NeedEyeCoords || ctx->Fog.Enabled) {
+	if (ctx->Light._NeedEyeCoords || ctx->Fog.Enabled ||
+	    (ctx->Texture._GenFlags & TEXGEN_NEED_EYE_COORD)) {
 		BEGIN_RING(chan, celsius, NV10TCL_MODELVIEW0_MATRIX(0), 16);
 		OUT_RINGm(chan, m->m);
 	}
 
-	if (ctx->Light.Enabled) {
+	if (ctx->Light.Enabled ||
+	    (ctx->Texture._GenFlags & TEXGEN_NEED_EYE_COORD)) {
 		int i, j;
 
 		BEGIN_RING(chan, celsius,
diff --git a/src/mesa/drivers/dri/nouveau/nv20_context.c b/src/mesa/drivers/dri/nouveau/nv20_context.c
index db39ef7075..789dcaa6b4 100644
--- a/src/mesa/drivers/dri/nouveau/nv20_context.c
+++ b/src/mesa/drivers/dri/nouveau/nv20_context.c
@@ -297,9 +297,9 @@ nv20_hwctx_init(GLcontext *ctx)
 	BEGIN_RING(chan, kelvin, NV20TCL_POLYGON_STIPPLE_ENABLE, 1);
 	OUT_RING  (chan, 0);
 
-	BEGIN_RING(chan, kelvin, NV20TCL_TX_GEN_S(0),
-		   4 * NV20TCL_TX_GEN_S__SIZE);
-	for (i=0; i < 4 * NV20TCL_TX_GEN_S__SIZE; i++)
+	BEGIN_RING(chan, kelvin, NV20TCL_TX_GEN_MODE_S(0),
+		   4 * NV20TCL_TX_GEN_MODE_S__SIZE);
+	for (i=0; i < 4 * NV20TCL_TX_GEN_MODE_S__SIZE; i++)
 		OUT_RING(chan, 0);
 
 	BEGIN_RING(chan, kelvin, NV20TCL_FOG_EQUATION_CONSTANT, 3);
@@ -497,10 +497,14 @@ const struct nouveau_driver nv20_driver = {
 		nv20_emit_tex_env,
 		nv20_emit_tex_env,
 		nv20_emit_tex_env,
-		nv10_emit_tex_gen,
-		nv10_emit_tex_gen,
-		nv10_emit_tex_gen,
-		nv10_emit_tex_gen,
+		nv20_emit_tex_gen,
+		nv20_emit_tex_gen,
+		nv20_emit_tex_gen,
+		nv20_emit_tex_gen,
+		nv20_emit_tex_mat,
+		nv20_emit_tex_mat,
+		nv20_emit_tex_mat,
+		nv20_emit_tex_mat,
 		nv20_emit_tex_obj,
 		nv20_emit_tex_obj,
 		nv20_emit_tex_obj,
diff --git a/src/mesa/drivers/dri/nouveau/nv20_driver.h b/src/mesa/drivers/dri/nouveau/nv20_driver.h
index 18574e9be6..8adecef2c4 100644
--- a/src/mesa/drivers/dri/nouveau/nv20_driver.h
+++ b/src/mesa/drivers/dri/nouveau/nv20_driver.h
@@ -68,6 +68,12 @@ nv20_emit_frag(GLcontext *ctx, int emit);
 
 /* nv20_state_tex.c */
 void
+nv20_emit_tex_gen(GLcontext *ctx, int emit);
+
+void
+nv20_emit_tex_mat(GLcontext *ctx, int emit);
+
+void
 nv20_emit_tex_obj(GLcontext *ctx, int emit);
 
 void
diff --git a/src/mesa/drivers/dri/nouveau/nv20_state_fb.c b/src/mesa/drivers/dri/nouveau/nv20_state_fb.c
index d638541df9..21da4f7af1 100644
--- a/src/mesa/drivers/dri/nouveau/nv20_state_fb.c
+++ b/src/mesa/drivers/dri/nouveau/nv20_state_fb.c
@@ -106,13 +106,11 @@ nv20_emit_viewport(GLcontext *ctx, int emit)
 	struct nouveau_grobj *kelvin = context_eng3d(ctx);
 	struct gl_framebuffer *fb = ctx->DrawBuffer;
 	float a[4] = {};
-	int i;
 
 	get_viewport_translate(ctx, a);
 
 	BEGIN_RING(chan, kelvin, NV20TCL_VIEWPORT_TRANSLATE_X, 4);
-	for (i = 0; i < 4; i++)
-		OUT_RINGf(chan, a[i]);
+	OUT_RINGp(chan, a, 4);
 
 	BEGIN_RING(chan, kelvin, NV20TCL_VIEWPORT_CLIP_HORIZ(0), 1);
 	OUT_RING(chan, (fb->Width - 1) << 16);
diff --git a/src/mesa/drivers/dri/nouveau/nv20_state_tex.c b/src/mesa/drivers/dri/nouveau/nv20_state_tex.c
index 92870105f9..e46118e4fc 100644
--- a/src/mesa/drivers/dri/nouveau/nv20_state_tex.c
+++ b/src/mesa/drivers/dri/nouveau/nv20_state_tex.c
@@ -32,6 +32,62 @@
 #include "nouveau_util.h"
 #include "nv20_driver.h"
 
+#define TX_GEN_MODE(i, j) (NV20TCL_TX_GEN_MODE_S(i) + 4 * (j))
+#define TX_GEN_COEFF(i, j) (NV20TCL_TX_GEN_COEFF_S_A(i) + 16 * (j))
+#define TX_MATRIX(i) (NV20TCL_TX0_MATRIX(0) + 64 * (i))
+
+void
+nv20_emit_tex_gen(GLcontext *ctx, int emit)
+{
+	const int i = emit - NOUVEAU_STATE_TEX_GEN0;
+	struct nouveau_context *nctx = to_nouveau_context(ctx);
+	struct nouveau_channel *chan = context_chan(ctx);
+	struct nouveau_grobj *kelvin = context_eng3d(ctx);
+	struct gl_texture_unit *unit = &ctx->Texture.Unit[i];
+	int j;
+
+	for (j = 0; j < 4; j++) {
+		if (nctx->fallback == HWTNL && (unit->TexGenEnabled & 1 << j)) {
+			struct gl_texgen *coord = get_texgen_coord(unit, j);
+			float *k = get_texgen_coeff(coord);
+
+			if (k) {
+				BEGIN_RING(chan, kelvin, TX_GEN_COEFF(i, j), 4);
+				OUT_RINGp(chan, k, 4);
+			}
+
+			BEGIN_RING(chan, kelvin, TX_GEN_MODE(i, j), 1);
+			OUT_RING(chan, nvgl_texgen_mode(coord->Mode));
+
+		} else {
+			BEGIN_RING(chan, kelvin, TX_GEN_MODE(i, j), 1);
+			OUT_RING(chan, 0);
+		}
+	}
+}
+
+void
+nv20_emit_tex_mat(GLcontext *ctx, int emit)
+{
+	const int i = emit - NOUVEAU_STATE_TEX_MAT0;
+	struct nouveau_context *nctx = to_nouveau_context(ctx);
+	struct nouveau_channel *chan = context_chan(ctx);
+	struct nouveau_grobj *kelvin = context_eng3d(ctx);
+
+	if (nctx->fallback == HWTNL &&
+	    (ctx->Texture._TexMatEnabled & 1 << i)) {
+		BEGIN_RING(chan, kelvin, NV20TCL_TX_MATRIX_ENABLE(i), 1);
+		OUT_RING(chan, 1);
+
+		BEGIN_RING(chan, kelvin, TX_MATRIX(i), 16);
+		OUT_RINGm(chan, ctx->TextureMatrixStack[i].Top->m);
+
+	} else {
+		BEGIN_RING(chan, kelvin, NV20TCL_TX_MATRIX_ENABLE(i), 1);
+		OUT_RING(chan, 0);
+	}
+}
+
 static uint32_t
 get_tex_format_pot(struct gl_texture_image *ti)
 {
diff --git a/src/mesa/drivers/dri/nouveau/nv20_state_tnl.c b/src/mesa/drivers/dri/nouveau/nv20_state_tnl.c
index 43f8c72312..62efe80fe4 100644
--- a/src/mesa/drivers/dri/nouveau/nv20_state_tnl.c
+++ b/src/mesa/drivers/dri/nouveau/nv20_state_tnl.c
@@ -139,9 +139,7 @@ nv20_emit_fog(GLcontext *ctx, int emit)
 	OUT_RING(chan, pack_rgba_f(MESA_FORMAT_RGBA8888_REV, f->Color));
 
 	BEGIN_RING(chan, kelvin, NV20TCL_FOG_EQUATION_CONSTANT, 3);
-	OUT_RINGf(chan, k[0]);
-	OUT_RINGf(chan, k[1]);
-	OUT_RINGf(chan, k[2]);
+	OUT_RINGp(chan, k, 3);
 }
 
 void
@@ -176,9 +174,7 @@ nv20_emit_light_source(GLcontext *ctx, int emit)
 
 	if (l->_Flags & LIGHT_POSITIONAL) {
 		BEGIN_RING(chan, kelvin, NV20TCL_LIGHT_POSITION_X(i), 3);
-		OUT_RINGf(chan, l->_Position[0]);
-		OUT_RINGf(chan, l->_Position[1]);
-		OUT_RINGf(chan, l->_Position[2]);
+		OUT_RINGp(chan, l->_Position, 3);
 
 		BEGIN_RING(chan, kelvin, NV20TCL_LIGHT_ATTENUATION_CONSTANT(i), 3);
 		OUT_RINGf(chan, l->ConstantAttenuation);
@@ -187,14 +183,10 @@ nv20_emit_light_source(GLcontext *ctx, int emit)
 
 	} else {
 		BEGIN_RING(chan, kelvin, NV20TCL_LIGHT_DIRECTION_X(i), 3);
-		OUT_RINGf(chan, l->_VP_inf_norm[0]);
-		OUT_RINGf(chan, l->_VP_inf_norm[1]);
-		OUT_RINGf(chan, l->_VP_inf_norm[2]);
+		OUT_RINGp(chan, l->_VP_inf_norm, 3);
 
 		BEGIN_RING(chan, kelvin, NV20TCL_LIGHT_HALF_VECTOR_X(i), 3);
-		OUT_RINGf(chan, l->_h_inf_norm[0]);
-		OUT_RINGf(chan, l->_h_inf_norm[1]);
-		OUT_RINGf(chan, l->_h_inf_norm[2]);
+		OUT_RINGp(chan, l->_h_inf_norm, 3);
 	}
 
 	if (l->_Flags & LIGHT_SPOT) {
@@ -203,13 +195,7 @@ nv20_emit_light_source(GLcontext *ctx, int emit)
 		nv10_get_spot_coeff(l, k);
 
 		BEGIN_RING(chan, kelvin, NV20TCL_LIGHT_SPOT_CUTOFF_A(i), 7);
-		OUT_RINGf(chan, k[0]);
-		OUT_RINGf(chan, k[1]);
-		OUT_RINGf(chan, k[2]);
-		OUT_RINGf(chan, k[3]);
-		OUT_RINGf(chan, k[4]);
-		OUT_RINGf(chan, k[5]);
-		OUT_RINGf(chan, k[6]);
+		OUT_RINGp(chan, k, 7);
 	}
 }
 
@@ -246,15 +232,11 @@ nv20_emit_material_ambient(GLcontext *ctx, int emit)
 	}
 
 	BEGIN_RING(chan, kelvin, m_scene[side], 3);
-	OUT_RINGf(chan, c_scene[0]);
-	OUT_RINGf(chan, c_scene[1]);
-	OUT_RINGf(chan, c_scene[2]);
+	OUT_RINGp(chan, c_scene, 3);
 
 	if (ctx->Light.ColorMaterialEnabled) {
 		BEGIN_RING(chan, kelvin, m_factor[side], 3);
-		OUT_RINGf(chan, c_factor[0]);
-		OUT_RINGf(chan, c_factor[1]);
-		OUT_RINGf(chan, c_factor[2]);
+		OUT_RINGp(chan, c_factor, 3);
 	}
 
 	foreach(l, &ctx->Light.EnabledList) {
@@ -266,9 +248,7 @@ nv20_emit_material_ambient(GLcontext *ctx, int emit)
 				  l->_MatAmbient[side]);
 
 		BEGIN_RING(chan, kelvin, m_light[side], 3);
-		OUT_RINGf(chan, c_light[0]);
-		OUT_RINGf(chan, c_light[1]);
-		OUT_RINGf(chan, c_light[2]);
+		OUT_RINGp(chan, c_light, 3);
 	}
 }
 
@@ -295,9 +275,7 @@ nv20_emit_material_diffuse(GLcontext *ctx, int emit)
 				  l->_MatDiffuse[side]);
 
 		BEGIN_RING(chan, kelvin, m_light[side], 3);
-		OUT_RINGf(chan, c_light[0]);
-		OUT_RINGf(chan, c_light[1]);
-		OUT_RINGf(chan, c_light[2]);
+		OUT_RINGp(chan, c_light, 3);
 	}
 }
 
@@ -318,9 +296,7 @@ nv20_emit_material_specular(GLcontext *ctx, int emit)
 				  l->_MatSpecular[side]);
 
 		BEGIN_RING(chan, kelvin, m_light[side], 3);
-		OUT_RINGf(chan, c_light[0]);
-		OUT_RINGf(chan, c_light[1]);
-		OUT_RINGf(chan, c_light[2]);
+		OUT_RINGp(chan, c_light, 3);
 	}
 }
 
@@ -340,12 +316,7 @@ nv20_emit_material_shininess(GLcontext *ctx, int emit)
 		k);
 
 	BEGIN_RING(chan, kelvin, mthd[side], 6);
-	OUT_RINGf(chan, k[0]);
-	OUT_RINGf(chan, k[1]);
-	OUT_RINGf(chan, k[2]);
-	OUT_RINGf(chan, k[3]);
-	OUT_RINGf(chan, k[4]);
-	OUT_RINGf(chan, k[5]);
+	OUT_RINGp(chan, k, 6);
 }
 
 void
@@ -359,12 +330,14 @@ nv20_emit_modelview(GLcontext *ctx, int emit)
 	if (nctx->fallback != HWTNL)
 		return;
 
-	if (ctx->Light._NeedEyeCoords || ctx->Fog.Enabled) {
+	if (ctx->Light._NeedEyeCoords || ctx->Fog.Enabled ||
+	    (ctx->Texture._GenFlags & TEXGEN_NEED_EYE_COORD)) {
 		BEGIN_RING(chan, kelvin, NV20TCL_MODELVIEW0_MATRIX(0), 16);
 		OUT_RINGm(chan, m->m);
 	}
 
-	if (ctx->Light.Enabled) {
+	if (ctx->Light.Enabled ||
+	    (ctx->Texture._GenFlags & TEXGEN_NEED_EYE_COORD)) {
 		int i, j;
 
 		BEGIN_RING(chan, kelvin,
diff --git a/src/mesa/drivers/dri/r200/r200_cmdbuf.c b/src/mesa/drivers/dri/r200/r200_cmdbuf.c
index 2f2b8d94dc..ad43a8ca92 100644
--- a/src/mesa/drivers/dri/r200/r200_cmdbuf.c
+++ b/src/mesa/drivers/dri/r200/r200_cmdbuf.c
@@ -189,7 +189,8 @@ void r200FlushElts(GLcontext *ctx)
    if (R200_ELT_BUF_SZ > elt_used)
      radeonReturnDmaRegion(&rmesa->radeon, R200_ELT_BUF_SZ - elt_used);
 
-   if (radeon_is_debug_enabled(RADEON_SYNC, RADEON_CRITICAL)) {
+   if (radeon_is_debug_enabled(RADEON_SYNC, RADEON_CRITICAL)
+         && !rmesa->radeon.radeonScreen->kernel_mm) {
       radeon_print(RADEON_SYNC, RADEON_NORMAL, "%s: Syncing\n", __FUNCTION__);
       radeonFinish( rmesa->radeon.glCtx );
    }
diff --git a/src/mesa/drivers/dri/r200/r200_state.c b/src/mesa/drivers/dri/r200/r200_state.c
index 9c2ac05ad6..29d7bed8b6 100644
--- a/src/mesa/drivers/dri/r200/r200_state.c
+++ b/src/mesa/drivers/dri/r200/r200_state.c
@@ -2496,11 +2496,10 @@ void r200InitStateFuncs( radeonContextPtr radeon, struct dd_function_table *func
    functions->DrawBuffer		= radeonDrawBuffer;
    functions->ReadBuffer		= radeonReadBuffer;
 
-   if (radeon->radeonScreen->kernel_mm) {
-	   functions->CopyPixels                = _mesa_meta_CopyPixels;
-	   functions->DrawPixels                = _mesa_meta_DrawPixels;
+   functions->CopyPixels                = _mesa_meta_CopyPixels;
+   functions->DrawPixels                = _mesa_meta_DrawPixels;
+   if (radeon->radeonScreen->kernel_mm)
 	   functions->ReadPixels                = radeonReadPixels;
-   }
 
    functions->AlphaFunc			= r200AlphaFunc;
    functions->BlendColor		= r200BlendColor;
diff --git a/src/mesa/drivers/dri/r200/r200_tcl.c b/src/mesa/drivers/dri/r200/r200_tcl.c
index f3f558b7de..d43e14581e 100644
--- a/src/mesa/drivers/dri/r200/r200_tcl.c
+++ b/src/mesa/drivers/dri/r200/r200_tcl.c
@@ -404,8 +404,9 @@ static GLuint r200EnsureEmitSize( GLcontext * ctx , GLubyte* vimap_rev )
          rendering code may decide convert to elts.
 	 In that case we have to make pessimistic prediction.
 	 and use larger of 2 paths. */
-      const GLuint elts = ELTS_BUFSZ(nr_aos);
-      const GLuint index = INDEX_BUFSZ;
+      const GLuint elt_count =(VB->Primitive[i].count/GET_MAX_HW_ELTS() + 1);
+      const GLuint elts = ELTS_BUFSZ(nr_aos) * elt_count;
+      const GLuint index = INDEX_BUFSZ * elt_count;
       const GLuint vbuf = VBUF_BUFSZ;
       if ( (!VB->Elts && VB->Primitive[i].count >= MAX_CONVERSION_SIZE)
 	  || vbuf > index + elts)
@@ -687,25 +688,34 @@ static char *getFallbackString(GLuint bit)
 
 void r200TclFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 {
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLuint oldfallback = rmesa->radeon.TclFallback;
-
-   if (mode) {
-      rmesa->radeon.TclFallback |= bit;
-      if (oldfallback == 0) {
-	 if (R200_DEBUG & RADEON_FALLBACKS)
-	    fprintf(stderr, "R200 begin tcl fallback %s\n",
-		    getFallbackString( bit ));
-	 transition_to_swtnl( ctx );
-      }
-   }
-   else {
-      rmesa->radeon.TclFallback &= ~bit;
-      if (oldfallback == bit) {
-	 if (R200_DEBUG & RADEON_FALLBACKS)
-	    fprintf(stderr, "R200 end tcl fallback %s\n",
-		    getFallbackString( bit ));
-	 transition_to_hwtnl( ctx );
-      }
-   }
+	r200ContextPtr rmesa = R200_CONTEXT(ctx);
+	GLuint oldfallback = rmesa->radeon.TclFallback;
+
+	if (mode) {
+		if (oldfallback == 0) {
+			/* We have to flush before transition */
+			if ( rmesa->radeon.dma.flush )
+				rmesa->radeon.dma.flush( rmesa->radeon.glCtx );
+
+			if (R200_DEBUG & RADEON_FALLBACKS)
+				fprintf(stderr, "R200 begin tcl fallback %s\n",
+						getFallbackString( bit ));
+			rmesa->radeon.TclFallback |= bit;
+			transition_to_swtnl( ctx );
+		} else
+			rmesa->radeon.TclFallback |= bit;
+	} else {
+		if (oldfallback == bit) {
+			/* We have to flush before transition */
+			if ( rmesa->radeon.dma.flush )
+				rmesa->radeon.dma.flush( rmesa->radeon.glCtx );
+
+			if (R200_DEBUG & RADEON_FALLBACKS)
+				fprintf(stderr, "R200 end tcl fallback %s\n",
+						getFallbackString( bit ));
+			rmesa->radeon.TclFallback &= ~bit;
+			transition_to_hwtnl( ctx );
+		} else
+			rmesa->radeon.TclFallback &= ~bit;
+	}
 }
diff --git a/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c b/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c
index 710cae727a..b6dfe28def 100644
--- a/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c
+++ b/src/mesa/drivers/dri/r300/compiler/r500_fragprog_emit.c
@@ -190,6 +190,17 @@ static unsigned int use_source(struct r500_fragment_program_code* code, struct r
 	return 0;
 }
 
+/**
+ * NOP the specified instruction if it is not a texture lookup.
+ */
+static void alu_nop(struct r300_fragment_program_compiler *c, int ip)
+{
+	PROG_CODE;
+
+	if ((code->inst[ip].inst0 & 0x3) != R500_INST_TYPE_TEX) {
+		code->inst[ip].inst0 |= R500_INST_NOP;
+	}
+}
 
 /**
  * Emit a paired ALU instruction.
@@ -205,6 +216,14 @@ static void emit_paired(struct r300_fragment_program_compiler *c, struct rc_pair
 
 	int ip = ++code->inst_end;
 
+	/* Quirk: MDH/MDV (DDX/DDY) need a NOP on previous non-TEX instructions. */
+	if (inst->RGB.Opcode == RC_OPCODE_DDX || inst->Alpha.Opcode == RC_OPCODE_DDX ||
+		inst->RGB.Opcode == RC_OPCODE_DDY || inst->Alpha.Opcode == RC_OPCODE_DDY) {
+		if (ip > 0) {
+			alu_nop(c, ip - 1);
+		}
+	}
+
 	code->inst[ip].inst5 = translate_rgb_op(c, inst->RGB.Opcode);
 	code->inst[ip].inst4 = translate_alpha_op(c, inst->Alpha.Opcode);
 
@@ -252,8 +271,8 @@ static void emit_paired(struct r300_fragment_program_compiler *c, struct rc_pair
 	code->inst[ip].inst4 |= translate_arg_alpha(inst, 1) << R500_ALPHA_SEL_B_SHIFT;
 	code->inst[ip].inst5 |= translate_arg_alpha(inst, 2) << R500_ALU_RGBA_ALPHA_SEL_C_SHIFT;
 
-    code->inst[ip].inst3 |= R500_ALU_RGB_TARGET(inst->RGB.Target);
-    code->inst[ip].inst4 |= R500_ALPHA_TARGET(inst->Alpha.Target);
+	code->inst[ip].inst3 |= R500_ALU_RGB_TARGET(inst->RGB.Target);
+	code->inst[ip].inst4 |= R500_ALPHA_TARGET(inst->Alpha.Target);
 
 	if (inst->WriteALUResult) {
 		code->inst[ip].inst3 |= R500_ALU_RGB_WMASK;
@@ -469,9 +488,8 @@ void r500BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compi
 	if (compiler->Base.Error)
 		return;
 
-	assert(code->inst_end >= 0);
-
-	if ((code->inst[code->inst_end].inst0 & R500_INST_TYPE_MASK) != R500_INST_TYPE_OUT) {
+	if (code->inst_end == -1 ||
+	    (code->inst[code->inst_end].inst0 & R500_INST_TYPE_MASK) != R500_INST_TYPE_OUT) {
 		/* This may happen when dead-code elimination is disabled or
 		 * when most of the fragment program logic is leading to a KIL */
 		if (code->inst_end >= 511) {
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.c b/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.c
index c1c0181fac..9d289fce34 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.c
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_opcodes.c
@@ -75,14 +75,14 @@ struct rc_opcode_info rc_opcodes[MAX_RC_OPCODE] = {
 	{
 		.Opcode = RC_OPCODE_DDX,
 		.Name = "DDX",
-		.NumSrcRegs = 1,
+		.NumSrcRegs = 2,
 		.HasDstReg = 1,
 		.IsComponentwise = 1
 	},
 	{
 		.Opcode = RC_OPCODE_DDY,
 		.Name = "DDY",
-		.NumSrcRegs = 1,
+		.NumSrcRegs = 2,
 		.HasDstReg = 1,
 		.IsComponentwise = 1
 	},
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_pair_translate.c b/src/mesa/drivers/dri/r300/compiler/radeon_pair_translate.c
index fff5b0c217..3a26e7daaf 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_pair_translate.c
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_pair_translate.c
@@ -159,11 +159,6 @@ static void set_pair_instruction(struct r300_fragment_program_compiler *c,
 	int nargs = opcode->NumSrcRegs;
 	int i;
 
-	/* Special case for DDX/DDY (MDH/MDV). */
-	if (inst->Opcode == RC_OPCODE_DDX || inst->Opcode == RC_OPCODE_DDY) {
-		nargs++;
-	}
-
 	for(i = 0; i < opcode->NumSrcRegs; ++i) {
 		int source;
 		if (needrgb && !istranscendent) {
diff --git a/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c
index b5c08aea49..f5b7d57eab 100644
--- a/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c
+++ b/src/mesa/drivers/dri/r300/compiler/radeon_program_alu.c
@@ -506,6 +506,46 @@ static void transform_r300_vertex_ABS(struct radeon_compiler* c,
 	inst->U.I.SrcReg[1].Negate ^= RC_MASK_XYZW;
 }
 
+static void transform_r300_vertex_CMP(struct radeon_compiler* c,
+	struct rc_instruction* inst)
+{
+	/* There is no decent CMP available, so let's rig one up.
+	 * CMP is defined as dst = src0 < 0.0 ? src1 : src2
+	 * The following sequence consumes two temps and three extra slots,
+	 * but should be equivalent:
+	 *
+	 * SLT tmp0, src0, 0.0
+	 * SGE tmp1, src0, 0.0
+	 * MUL tmp0, tmp0, src1
+	 * MAD dst, src2, tmp1, tmp0
+	 *
+	 * Yes, I know, I'm a mad scientist. ~ C. */
+	int tempreg0 = rc_find_free_temporary(c);
+	int tempreg1 = rc_find_free_temporary(c);
+
+	/* SLT tmp0, src0, 0.0 */
+	emit2(c, inst->Prev, RC_OPCODE_SLT, 0,
+		dstreg(RC_FILE_TEMPORARY, tempreg0),
+		inst->U.I.SrcReg[0], builtin_zero);
+
+	/* SGE tmp1, src0, 0.0 */
+	emit2(c, inst->Prev, RC_OPCODE_SGE, 0,
+		dstreg(RC_FILE_TEMPORARY, tempreg1),
+		inst->U.I.SrcReg[0], builtin_zero);
+
+	/* MUL tmp0, tmp0, src1 */
+	emit2(c, inst->Prev, RC_OPCODE_MUL, 0,
+		dstreg(RC_FILE_TEMPORARY, tempreg0),
+		srcreg(RC_FILE_TEMPORARY, tempreg0), inst->U.I.SrcReg[1]);
+
+	/* MAD dst, src2, tmp1, tmp0 */
+	emit3(c, inst->Prev, RC_OPCODE_MAD, inst->U.I.SaturateMode,
+		inst->U.I.DstReg,
+		inst->U.I.SrcReg[2], srcreg(RC_FILE_TEMPORARY, tempreg1), srcreg(RC_FILE_TEMPORARY, tempreg0));
+
+	rc_remove_instruction(inst);
+}
+
 /**
  * For use with radeonLocalTransform, this transforms non-native ALU
  * instructions of the r300 up to r500 vertex engine.
@@ -517,6 +557,7 @@ int r300_transform_vertex_alu(
 {
 	switch(inst->U.I.Opcode) {
 	case RC_OPCODE_ABS: transform_r300_vertex_ABS(c, inst); return 1;
+	case RC_OPCODE_CMP: transform_r300_vertex_CMP(c, inst); return 1;
 	case RC_OPCODE_DP3: transform_DP3(c, inst); return 1;
 	case RC_OPCODE_DPH: transform_DPH(c, inst); return 1;
 	case RC_OPCODE_FLR: transform_FLR(c, inst); return 1;
diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
index e2dbb1dbf4..c40802aec6 100644
--- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c
+++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
@@ -77,12 +77,29 @@ static int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom)
 	cnt = vpu_count(atom->cmd);
 
 	if (r300->radeon.radeonScreen->kernel_mm) {
-		extra = 5;
+		extra = 3;
 	}
 
 	return cnt ? (cnt * 4) + extra : 0;
 }
 
+static int check_vpp(GLcontext *ctx, struct radeon_state_atom *atom)
+{
+    r300ContextPtr r300 = R300_CONTEXT(ctx);
+    int cnt;
+    int extra = 1;
+
+    if (r300->radeon.radeonScreen->kernel_mm) {
+        cnt = r300->selected_vp->code.constants.Count * 4;
+        extra = 3;
+    } else {
+        cnt = vpu_count(atom->cmd);
+        extra = 1;
+    }
+
+    return cnt ? (cnt * 4) + extra : 0;
+}
+
 void r300_emit_vpu(struct r300_context *r300,
                    uint32_t *data,
                    unsigned len,
@@ -101,15 +118,26 @@ static void emit_vpu_state(GLcontext *ctx, struct radeon_state_atom * atom)
 {
     r300ContextPtr r300 = R300_CONTEXT(ctx);
     drm_r300_cmd_header_t cmd;
-    uint32_t addr, ndw;
+    uint32_t addr;
 
     cmd.u = atom->cmd[0];
     addr = (cmd.vpu.adrhi << 8) | cmd.vpu.adrlo;
-    ndw = atom->check(ctx, atom);
 
     r300_emit_vpu(r300, &atom->cmd[1], vpu_count(atom->cmd) * 4, addr);
 }
 
+static void emit_vpp_state(GLcontext *ctx, struct radeon_state_atom * atom)
+{
+    r300ContextPtr r300 = R300_CONTEXT(ctx);
+    drm_r300_cmd_header_t cmd;
+    uint32_t addr;
+
+    cmd.u = atom->cmd[0];
+    addr = (cmd.vpu.adrhi << 8) | cmd.vpu.adrlo;
+
+    r300_emit_vpu(r300, &atom->cmd[1], r300->selected_vp->code.constants.Count * 4, addr);
+}
+
 void r500_emit_fp(struct r300_context *r300,
                   uint32_t *data,
                   unsigned len,
@@ -785,11 +813,11 @@ void r300InitCmdBuf(r300ContextPtr r300)
 			r300->hw.vpi.emit = emit_vpu_state;
 
 		if (is_r500) {
-			ALLOC_STATE(vpp, vpu, R300_VPP_CMDSIZE, 0);
+			ALLOC_STATE(vpp, vpp, R300_VPP_CMDSIZE, 0);
 			r300->hw.vpp.cmd[0] =
 				cmdvpu(r300->radeon.radeonScreen, R500_PVS_CONST_START, 0);
 			if (r300->radeon.radeonScreen->kernel_mm)
-				r300->hw.vpp.emit = emit_vpu_state;
+				r300->hw.vpp.emit = emit_vpp_state;
 
 			ALLOC_STATE(vps, vpu, R300_VPS_CMDSIZE, 0);
 			r300->hw.vps.cmd[0] =
@@ -806,11 +834,11 @@ void r300InitCmdBuf(r300ContextPtr r300)
 					r300->hw.vpucp[i].emit = emit_vpu_state;
 			}
 		} else {
-			ALLOC_STATE(vpp, vpu, R300_VPP_CMDSIZE, 0);
+			ALLOC_STATE(vpp, vpp, R300_VPP_CMDSIZE, 0);
 			r300->hw.vpp.cmd[0] =
 				cmdvpu(r300->radeon.radeonScreen, R300_PVS_CONST_START, 0);
 			if (r300->radeon.radeonScreen->kernel_mm)
-				r300->hw.vpp.emit = emit_vpu_state;
+				r300->hw.vpp.emit = emit_vpp_state;
 
 			ALLOC_STATE(vps, vpu, R300_VPS_CMDSIZE, 0);
 			r300->hw.vps.cmd[0] =
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_common.c b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
index 61ea5e4d9a..0646da4624 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog_common.c
+++ b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
@@ -256,6 +256,19 @@ static void translate_fragment_program(GLcontext *ctx, struct r300_fragment_prog
 
 	fp->InputsRead = compiler.Base.Program.InputsRead;
 
+	/* Clear the fog/wpos_attr if code accessing these
+	 * attributes has been removed during compilation
+	 */
+	if (fp->fog_attr != FRAG_ATTRIB_MAX) {
+		if (!(fp->InputsRead & (1 << fp->fog_attr)))
+			fp->fog_attr = FRAG_ATTRIB_MAX;
+	}
+
+	if (fp->wpos_attr != FRAG_ATTRIB_MAX) {
+		if (!(fp->InputsRead & (1 << fp->wpos_attr)))
+			fp->wpos_attr = FRAG_ATTRIB_MAX;
+	}
+
 	rc_destroy(&compiler.Base);
 }
 
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index 749a2464e7..e660b1fb3b 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -2396,11 +2396,10 @@ void r300InitStateFuncs(radeonContextPtr radeon, struct dd_function_table *funct
 	functions->DrawBuffer = radeonDrawBuffer;
 	functions->ReadBuffer = radeonReadBuffer;
 
-	if (radeon->radeonScreen->kernel_mm) {
-		functions->CopyPixels = _mesa_meta_CopyPixels;
-		functions->DrawPixels = _mesa_meta_DrawPixels;
+	functions->CopyPixels = _mesa_meta_CopyPixels;
+	functions->DrawPixels = _mesa_meta_DrawPixels;
+	if (radeon->radeonScreen->kernel_mm)
 		functions->ReadPixels = radeonReadPixels;
-	}
 }
 
 void r300InitShaderFunctions(r300ContextPtr r300)
diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index 129004fee7..a1fe378029 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -263,15 +263,25 @@ static struct r300_vertex_program *build_program(GLcontext *ctx,
 	rc_move_output(&compiler.Base, VERT_RESULT_PSIZ, VERT_RESULT_PSIZ, WRITEMASK_X);
 
 	if (vp->key.WPosAttr != FRAG_ATTRIB_MAX) {
-		rc_copy_output(&compiler.Base,
-			VERT_RESULT_HPOS,
-			vp->key.WPosAttr - FRAG_ATTRIB_TEX0 + VERT_RESULT_TEX0);
+		unsigned int vp_wpos_attr = vp->key.WPosAttr - FRAG_ATTRIB_TEX0 + VERT_RESULT_TEX0;
+
+		/* Set empty writemask for instructions writing to vp_wpos_attr
+		 * before moving the wpos attr there.
+		 * Such instructions will be removed by DCE.
+		 */
+		rc_move_output(&compiler.Base, vp_wpos_attr, vp->key.WPosAttr, 0);
+		rc_copy_output(&compiler.Base, VERT_RESULT_HPOS, vp_wpos_attr);
 	}
 
 	if (vp->key.FogAttr != FRAG_ATTRIB_MAX) {
-		rc_move_output(&compiler.Base,
-			VERT_RESULT_FOGC,
-			vp->key.FogAttr - FRAG_ATTRIB_TEX0 + VERT_RESULT_TEX0, WRITEMASK_X);
+		unsigned int vp_fog_attr = vp->key.FogAttr - FRAG_ATTRIB_TEX0 + VERT_RESULT_TEX0;
+
+		/* Set empty writemask for instructions writing to vp_fog_attr
+		 * before moving the fog attr there.
+		 * Such instructions will be removed by DCE.
+		 */
+		rc_move_output(&compiler.Base, vp_fog_attr, vp->key.FogAttr, 0);
+		rc_move_output(&compiler.Base, VERT_RESULT_FOGC, vp_fog_attr, WRITEMASK_X);
 	}
 
 	r3xx_compile_vertex_program(&compiler);
@@ -382,7 +392,11 @@ void r300SetupVertexProgram(r300ContextPtr rmesa)
 	R300_STATECHANGE(rmesa, vap_cntl);
 	R300_STATECHANGE(rmesa, vpp);
 	param_count = r300VertexProgUpdateParams(ctx, prog, (float *)&rmesa->hw.vpp.cmd[R300_VPP_PARAM_0]);
-	bump_vpu_count(rmesa->hw.vpp.cmd, param_count);
+	if (!rmesa->radeon.radeonScreen->kernel_mm && param_count > 255 * 4) {
+		WARN_ONCE("Too many VP params, expect rendering errors\n");
+	}
+	/* Prevent the overflow (vpu.count is u8) */
+	bump_vpu_count(rmesa->hw.vpp.cmd, MIN2(255 * 4, param_count));
 	param_count /= 4;
 
 	r300EmitVertexProgram(rmesa, R300_PVS_CODE_START, &(prog->code));
@@ -395,6 +409,6 @@ void r300SetupVertexProgram(r300ContextPtr rmesa)
 	rmesa->hw.pvs.cmd[R300_PVS_CNTL_1] = (0 << R300_PVS_FIRST_INST_SHIFT) | (inst_count << R300_PVS_XYZW_VALID_INST_SHIFT) |
 				(inst_count << R300_PVS_LAST_INST_SHIFT);
 
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_2] = (0 << R300_PVS_CONST_BASE_OFFSET_SHIFT) | (param_count << R300_PVS_MAX_CONST_ADDR_SHIFT);
+	rmesa->hw.pvs.cmd[R300_PVS_CNTL_2] = (0 << R300_PVS_CONST_BASE_OFFSET_SHIFT) | ((param_count - 1) << R300_PVS_MAX_CONST_ADDR_SHIFT);
 	rmesa->hw.pvs.cmd[R300_PVS_CNTL_3] = (inst_count << R300_PVS_LAST_VTX_SRC_INST_SHIFT);
 }
diff --git a/src/mesa/drivers/dri/r600/r700_state.c b/src/mesa/drivers/dri/r600/r700_state.c
index 2953ffd028..1da31e7b2b 100644
--- a/src/mesa/drivers/dri/r600/r700_state.c
+++ b/src/mesa/drivers/dri/r600/r700_state.c
@@ -615,7 +615,7 @@ static GLuint translate_logicop(GLenum logicop)
 	case GL_XOR:
 		return 0x66;
 	case GL_EQUIV:
-		return 0xaa;
+		return 0x99;
 	case GL_AND_REVERSE:
 		return 0x44;
 	case GL_AND_INVERTED:
@@ -1861,10 +1861,9 @@ void r700InitStateFuncs(radeonContextPtr radeon, struct dd_function_table *funct
 	functions->DrawBuffer = radeonDrawBuffer;
 	functions->ReadBuffer = radeonReadBuffer;
 
-	if (radeon->radeonScreen->kernel_mm) {
-		functions->CopyPixels = _mesa_meta_CopyPixels;
-		functions->DrawPixels = _mesa_meta_DrawPixels;
+	functions->CopyPixels = _mesa_meta_CopyPixels;
+	functions->DrawPixels = _mesa_meta_DrawPixels;
+	if (radeon->radeonScreen->kernel_mm)
 		functions->ReadPixels = radeonReadPixels;
-	}
 }
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c
index 22499bc38d..6b7690cf8b 100644
--- a/src/mesa/drivers/dri/radeon/radeon_dma.c
+++ b/src/mesa/drivers/dri/radeon/radeon_dma.c
@@ -184,9 +184,6 @@ void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size)
 	radeon_print(RADEON_DMA, RADEON_NORMAL, "%s size %d minimum_size %d\n",
 			__FUNCTION__, size, rmesa->dma.minimum_size);
 
-	if (!is_empty_list(&rmesa->dma.reserved))
-		radeon_bo_unmap(first_elem(&rmesa->dma.reserved)->bo);
-
 	if (is_empty_list(&rmesa->dma.free)
 	      || last_elem(&rmesa->dma.free)->bo->size < size) {
 		dma_bo = CALLOC_STRUCT(radeon_dma_bo);
@@ -336,9 +333,6 @@ void radeonReleaseDmaRegions(radeonContextPtr rmesa)
 		legacy_track_pending(rmesa->radeonScreen->bom, 0);
 	}
 
-	if (!is_empty_list(&rmesa->dma.reserved))
-		radeon_bo_unmap(first_elem(&rmesa->dma.reserved)->bo);
-
 	/* move waiting bos to free list.
 	   wait list provides gpu time to handle data before reuse */
 	foreach_s(dma_bo, temp, &rmesa->dma.wait) {
@@ -368,6 +362,7 @@ void radeonReleaseDmaRegions(radeonContextPtr rmesa)
 
 	/* move reserved to wait list */
 	foreach_s(dma_bo, temp, &rmesa->dma.reserved) {
+		radeon_bo_unmap(dma_bo->bo);
 		/* free objects that are too small to be used because of large request */
 		if (dma_bo->bo->size < rmesa->dma.minimum_size) {
 		   radeon_bo_unref(dma_bo->bo);
diff --git a/src/mesa/drivers/dri/radeon/radeon_state.c b/src/mesa/drivers/dri/radeon/radeon_state.c
index 0afbc19c12..539b067742 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state.c
+++ b/src/mesa/drivers/dri/radeon/radeon_state.c
@@ -2249,11 +2249,10 @@ void radeonInitStateFuncs( GLcontext *ctx , GLboolean dri2 )
 
    ctx->Driver.DrawBuffer		= radeonDrawBuffer;
    ctx->Driver.ReadBuffer		= radeonReadBuffer;
-   if (dri2) {
-	   ctx->Driver.CopyPixels               = _mesa_meta_CopyPixels;
-	   ctx->Driver.DrawPixels               = _mesa_meta_DrawPixels;
+   ctx->Driver.CopyPixels               = _mesa_meta_CopyPixels;
+   ctx->Driver.DrawPixels               = _mesa_meta_DrawPixels;
+   if (dri2)
 	   ctx->Driver.ReadPixels               = radeonReadPixels;
-   }
 
    ctx->Driver.AlphaFunc		= radeonAlphaFunc;
    ctx->Driver.BlendEquationSeparate	= radeonBlendEquationSeparate;
diff --git a/src/mesa/drivers/dri/swrast/Makefile b/src/mesa/drivers/dri/swrast/Makefile
index cc59eefdb2..d2cf6dbc55 100644
--- a/src/mesa/drivers/dri/swrast/Makefile
+++ b/src/mesa/drivers/dri/swrast/Makefile
@@ -5,6 +5,8 @@ include $(TOP)/configs/current
 
 LIBNAME = swrast_dri.so
 
+DRIVER_DEFINES = -D__NOT_HAVE_DRM_H
+
 DRIVER_SOURCES = \
 	swrast.c \
 	swrast_span.c
@@ -18,7 +20,7 @@ ASM_SOURCES =
 SWRAST_COMMON_SOURCES = \
 	../../common/driverfuncs.c \
 	../common/utils.c \
-	../common/dri_sw.c
+	../common/drisw_util.c
 
 include ../Makefile.template
 
diff --git a/src/mesa/drivers/dri/swrast/swrast.c b/src/mesa/drivers/dri/swrast/swrast.c
index e9ca99a86f..8b68281fab 100644
--- a/src/mesa/drivers/dri/swrast/swrast.c
+++ b/src/mesa/drivers/dri/swrast/swrast.c
@@ -206,12 +206,20 @@ swrast_delete_renderbuffer(struct gl_renderbuffer *rb)
     free(rb);
 }
 
+/* see bytes_per_line in libGL */
+static INLINE int
+bytes_per_line(unsigned pitch_bits, unsigned mul)
+{
+   unsigned mask = mul - 1;
+
+   return ((pitch_bits + mask) & ~mask) / 8;
+}
+
 static GLboolean
 swrast_alloc_front_storage(GLcontext *ctx, struct gl_renderbuffer *rb,
 			   GLenum internalFormat, GLuint width, GLuint height)
 {
     struct swrast_renderbuffer *xrb = swrast_renderbuffer(rb);
-    unsigned mask = PITCH_ALIGN_BITS - 1;
 
     TRACE;
 
@@ -219,8 +227,7 @@ swrast_alloc_front_storage(GLcontext *ctx, struct gl_renderbuffer *rb,
     rb->Width = width;
     rb->Height = height;
 
-    /* always pad to PITCH_ALIGN_BITS */
-    xrb->pitch = ((width * xrb->bpp + mask) & ~mask) / 8;
+    xrb->pitch = bytes_per_line(width * xrb->bpp, 32);
 
     return GL_TRUE;
 }
@@ -394,8 +401,10 @@ dri_swap_buffers(__DRIdrawable * dPriv)
 
     fb = &drawable->Base;
 
-    frontrb = swrast_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
-    backrb = swrast_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer);
+    frontrb =
+	swrast_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer);
+    backrb =
+	swrast_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer);
 
     /* check for signle-buffered */
     if (backrb == NULL)
diff --git a/src/mesa/drivers/dri/swrast/swrast_priv.h b/src/mesa/drivers/dri/swrast/swrast_priv.h
index 77670d89a5..6679061a98 100644
--- a/src/mesa/drivers/dri/swrast/swrast_priv.h
+++ b/src/mesa/drivers/dri/swrast/swrast_priv.h
@@ -30,7 +30,7 @@
 #include <GL/gl.h>
 #include <GL/internal/dri_interface.h>
 #include "main/mtypes.h"
-#include "dri_sw.h"
+#include "drisw_util.h"
 
 
 /**
@@ -124,14 +124,6 @@ swrast_renderbuffer(struct gl_renderbuffer *rb)
 #define PF_R3G3B2     3		/**<  8bpp TrueColor:  3-R, 3-G, 2-B bits */
 #define PF_X8R8G8B8   4		/**< 32bpp TrueColor:  8-R, 8-G, 8-B bits */
 
-/**
- * Renderbuffer pitch alignment (in bits).
- *
- * The xorg loader requires padding images to 32 bits. However, this should
- * become a screen/drawable parameter XXX
- */
-#define PITCH_ALIGN_BITS 32
-
 
 /* swrast_span.c */
 
diff --git a/src/mesa/drivers/dri/swrast/swrast_span.c b/src/mesa/drivers/dri/swrast/swrast_span.c
index c5681e34a9..c7d0bfdac7 100644
--- a/src/mesa/drivers/dri/swrast/swrast_span.c
+++ b/src/mesa/drivers/dri/swrast/swrast_span.c
@@ -236,7 +236,7 @@ static const GLubyte kernel[16] = {
    struct swrast_renderbuffer *xrb = swrast_renderbuffer(rb);
 #define INIT_PIXEL_PTR(P, X, Y) \
    GLushort *P = (GLushort *)row;
-#define INC_PIXEL_PTR(P) P += 2
+#define INC_PIXEL_PTR(P) P++
 #define STORE_PIXEL(DST, X, Y, VALUE) \
    STORE_PIXEL_R5G6B5(DST, X, Y, VALUE)
 #define FETCH_PIXEL(DST, SRC) \
diff --git a/src/mesa/drivers/dri/swrast/swrast_spantemp.h b/src/mesa/drivers/dri/swrast/swrast_spantemp.h
index 079726ae4a..1e9405eebf 100644
--- a/src/mesa/drivers/dri/swrast/swrast_spantemp.h
+++ b/src/mesa/drivers/dri/swrast/swrast_spantemp.h
@@ -37,7 +37,7 @@
 #define _SWRAST_SPANTEMP_ONCE
 
 static INLINE void
-PUT_PIXEL( GLcontext *glCtx, GLint x, GLint y, GLubyte *p )
+PUT_PIXEL( GLcontext *glCtx, GLint x, GLint y, GLvoid *p )
 {
     __DRIcontext *ctx = swrast_context(glCtx)->cPriv;
     __DRIdrawable *draw = swrast_drawable(glCtx->DrawBuffer)->dPriv;
@@ -168,7 +168,8 @@ NAME(put_row)( GLcontext *ctx, struct gl_renderbuffer *rb,
    if (mask) {
       for (i = 0; i < count; i++) {
          if (mask[i]) {
-            RB_TYPE pixel[4];
+            RB_TYPE row[4];
+            INIT_PIXEL_PTR(pixel, x, y);
             STORE_PIXEL(pixel, x + i, y, src[i]);
             PUT_PIXEL(ctx, x + i, YFLIP(xrb, y), pixel);
          }
@@ -200,7 +201,8 @@ NAME(put_row_rgb)( GLcontext *ctx, struct gl_renderbuffer *rb,
    if (mask) {
       for (i = 0; i < count; i++) {
          if (mask[i]) {
-            RB_TYPE pixel[4];
+            RB_TYPE row[4];
+            INIT_PIXEL_PTR(pixel, x, y);
 #ifdef STORE_PIXEL_RGB
             STORE_PIXEL_RGB(pixel, x + i, y, src[i]);
 #else
@@ -240,7 +242,8 @@ NAME(put_mono_row)( GLcontext *ctx, struct gl_renderbuffer *rb,
    if (mask) {
       for (i = 0; i < count; i++) {
          if (mask[i]) {
-            RB_TYPE pixel[4];
+            RB_TYPE row[4];
+            INIT_PIXEL_PTR(pixel, x, y);
             STORE_PIXEL(pixel, x + i, y, src);
             PUT_PIXEL(ctx, x + i, YFLIP(xrb, y), pixel);
          }
@@ -272,7 +275,8 @@ NAME(put_values)( GLcontext *ctx, struct gl_renderbuffer *rb,
    ASSERT(mask);
    for (i = 0; i < count; i++) {
       if (mask[i]) {
-         RB_TYPE pixel[4];
+         RB_TYPE row[4];
+         INIT_PIXEL_PTR(pixel, x, y);
          STORE_PIXEL(pixel, x[i], y[i], src[i]);
          PUT_PIXEL(ctx, x[i], YFLIP(xrb, y[i]), pixel);
       }
@@ -294,7 +298,8 @@ NAME(put_mono_values)( GLcontext *ctx, struct gl_renderbuffer *rb,
    ASSERT(mask);
    for (i = 0; i < count; i++) {
       if (mask[i]) {
-         RB_TYPE pixel[4];
+         RB_TYPE row[4];
+         INIT_PIXEL_PTR(pixel, x, y);
          STORE_PIXEL(pixel, x[i], y[i], src);
          PUT_PIXEL(ctx, x[i], YFLIP(xrb, y[i]), pixel);
       }
diff --git a/src/mesa/drivers/glslcompiler/Makefile b/src/mesa/drivers/glslcompiler/Makefile
index 080fe475c1..7dcf9a6541 100644
--- a/src/mesa/drivers/glslcompiler/Makefile
+++ b/src/mesa/drivers/glslcompiler/Makefile
@@ -9,12 +9,9 @@ PROGRAM = glslcompiler
 
 OBJECTS = \
 	glslcompiler.o \
-	../../glapi/glapi.o \
-	../../glapi/glapi_nop.o \
-	../../glapi/glthread.o \
-	../../main/dispatch.o \
 	../common/driverfuncs.o \
-	../../libmesa.a
+	../../libmesa.a \
+	../../libglapi.a
 
 INCLUDES = \
 	-I$(TOP)/include \
diff --git a/src/mesa/drivers/glslcompiler/glslcompiler.c b/src/mesa/drivers/glslcompiler/glslcompiler.c
index 66035a4a43..d58f32b293 100644
--- a/src/mesa/drivers/glslcompiler/glslcompiler.c
+++ b/src/mesa/drivers/glslcompiler/glslcompiler.c
@@ -68,6 +68,7 @@ static const char *Prog = "glslcompiler";
 
 struct options {
    GLboolean LineNumbers;
+   GLboolean Link;
    gl_prog_print_mode Mode;
    const char *VertFile;
    const char *FragFile;
@@ -207,23 +208,29 @@ ReadShader(GLuint shader, const char *filename)
 }
 
 
-#if 0
 static void
-CheckLink(GLuint prog)
+CheckLink(GLuint v_shader, GLuint f_shader)
 {
+   GLuint prog;
    GLint stat;
+
+   prog = _mesa_CreateProgram();
+
+   _mesa_AttachShader(prog, v_shader);
+   _mesa_AttachShader(prog, f_shader);
+
+   _mesa_LinkProgramARB(prog);
    _mesa_GetProgramiv(prog, GL_LINK_STATUS, &stat);
    if (!stat) {
       GLchar log[1000];
       GLsizei len;
       _mesa_GetProgramInfoLog(prog, 1000, &len, log);
-      fprintf(stderr, "%s: Linker error:\n%s\n", Prog, log);
+      fprintf(stderr, "Linker error:\n%s\n", log);
    }
    else {
-      fprintf(stderr, "%s: Link success!\n", Prog);
+      fprintf(stderr, "Link success!\n");
    }
 }
-#endif
 
 
 static void
@@ -262,6 +269,7 @@ Usage(void)
    printf("  --fs FILE          fragment shader input filename\n");
    printf("  --arb              emit ARB-style instructions\n");
    printf("  --nv               emit NV-style instructions\n");
+   printf("  --link             run linker\n");
    printf("  --debug            force #pragma debug(on)\n");
    printf("  --nodebug          force #pragma debug(off)\n");
    printf("  --opt              force #pragma optimize(on)\n");
@@ -309,6 +317,9 @@ ParseOptions(int argc, char *argv[])
       else if (strcmp(argv[i], "--nv") == 0) {
          Options.Mode = PROG_PRINT_NV;
       }
+      else if (strcmp(argv[i], "--link") == 0) {
+         Options.Link = GL_TRUE;
+      }
       else if (strcmp(argv[i], "--debug") == 0) {
          Options.Pragmas.IgnoreDebug = GL_TRUE;
          Options.Pragmas.Debug = GL_TRUE;
@@ -358,7 +369,7 @@ ParseOptions(int argc, char *argv[])
 int
 main(int argc, char *argv[])
 {
-   GLuint shader = 0;
+   GLuint v_shader = 0, f_shader = 0;
 
    ParseOptions(argc, argv);
 
@@ -368,24 +379,38 @@ main(int argc, char *argv[])
    }
 
    if (Options.VertFile) {
-      shader = CompileShader(Options.VertFile, GL_VERTEX_SHADER);
+      v_shader = CompileShader(Options.VertFile, GL_VERTEX_SHADER);
    }
-   else if (Options.FragFile) {
-      shader = CompileShader(Options.FragFile, GL_FRAGMENT_SHADER);
+
+   if (Options.FragFile) {
+      f_shader = CompileShader(Options.FragFile, GL_FRAGMENT_SHADER);
    }
 
-   if (shader) {
+   if (v_shader || f_shader) {
       if (Options.OutputFile) {
          fclose(stdout);
          /*stdout =*/ freopen(Options.OutputFile, "w", stdout);
       }
-      if (stdout) {
-         PrintShaderInstructions(shader, stdout);
+      if (stdout && v_shader) {
+         PrintShaderInstructions(v_shader, stdout);
+      }
+      if (stdout && f_shader) {
+         PrintShaderInstructions(f_shader, stdout);
       }
       if (Options.OutputFile) {
          fclose(stdout);
       }
    }
 
+   if (Options.Link) {
+      if (!v_shader || !f_shader) {
+         fprintf(stderr,
+                 "--link option requires both a vertex and fragment shader.\n");
+         exit(1);
+      }
+
+      CheckLink(v_shader, f_shader);
+   }
+
    return 0;
 }
diff --git a/src/mesa/drivers/x11/glxapi.c b/src/mesa/drivers/x11/glxapi.c
index e11aff1a84..955eba4e94 100644
--- a/src/mesa/drivers/x11/glxapi.c
+++ b/src/mesa/drivers/x11/glxapi.c
@@ -1113,7 +1113,7 @@ glXGetAGPOffsetMESA( const GLvoid *pointer )
 
 /*** GLX_MESA_allocate_memory */
 
-void *
+void PUBLIC *
 glXAllocateMemoryMESA(Display *dpy, int scrn, size_t size,
                       float readfreq, float writefreq, float priority)
 {
@@ -1121,14 +1121,14 @@ glXAllocateMemoryMESA(Display *dpy, int scrn, size_t size,
    return NULL;
 }
 
-void
+void PUBLIC
 glXFreeMemoryMESA(Display *dpy, int scrn, void *pointer)
 {
    /* dummy */
 }
 
 
-GLuint
+GLuint PUBLIC
 glXGetMemoryOffsetMESA(Display *dpy, int scrn, const void *pointer)
 {
    /* dummy */
@@ -1138,7 +1138,7 @@ glXGetMemoryOffsetMESA(Display *dpy, int scrn, const void *pointer)
 
 /*** GLX_EXT_texture_from_pixmap */
 
-void
+void PUBLIC
 glXBindTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer,
                    const int *attrib_list)
 {
@@ -1148,7 +1148,7 @@ glXBindTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer,
       t->BindTexImageEXT(dpy, drawable, buffer, attrib_list);
 }
 
-void
+void PUBLIC
 glXReleaseTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer)
 {
    struct _glxapi_table *t;
@@ -1426,7 +1426,7 @@ _glxapi_get_proc_address(const char *funcName)
  * This function does not get dispatched through the dispatch table
  * since it's really a "meta" function.
  */
-__GLXextFuncPtr
+__GLXextFuncPtr PUBLIC
 glXGetProcAddressARB(const GLubyte *procName)
 {
    __GLXextFuncPtr f;
@@ -1442,7 +1442,8 @@ glXGetProcAddressARB(const GLubyte *procName)
 
 
 /* GLX 1.4 */
-void (*glXGetProcAddress(const GLubyte *procName))()
+void PUBLIC
+(*glXGetProcAddress(const GLubyte *procName))()
 {
    return glXGetProcAddressARB(procName);
 }