Merge commit 'origin/mesa_7_7_branch'

Conflicts:
	src/gallium/drivers/svga/svga_screen_texture.c
	src/gallium/state_trackers/xorg/xorg_composite.c
	src/gallium/state_trackers/xorg/xorg_exa.c
	src/gallium/state_trackers/xorg/xorg_renderer.c
	src/gallium/state_trackers/xorg/xorg_xv.c
	src/mesa/main/texgetimage.c
	src/mesa/main/version.h

60 files changed, 1476 insertions, 1325 deletions

diff --git a/Makefile b/Makefile
index 7f073fd516c..bd793e7a06a 100644
--- a/Makefile
+++ b/Makefile
@@ -439,8 +439,7 @@ DEMO_FILES = \
 	$(DIRECTORY)/progs/windml/Makefile.ugl		\
 	$(DIRECTORY)/progs/windml/*.c			\
 	$(DIRECTORY)/progs/windml/*.bmp			\
-	$(DIRECTORY)/progs/ggi/*.c			\
-	$(DIRECTORY)/windows/VC7/progs/progs.sln
+	$(DIRECTORY)/progs/ggi/*.c
 
 GLUT_FILES = \
 	$(DIRECTORY)/include/GL/glut.h			\
@@ -463,9 +462,7 @@ GLUT_FILES = \
 	$(DIRECTORY)/src/glut/mini/glut.pc.in		\
 	$(DIRECTORY)/src/glut/directfb/Makefile		\
 	$(DIRECTORY)/src/glut/directfb/NOTES		\
-	$(DIRECTORY)/src/glut/directfb/*[ch]		\
-	$(DIRECTORY)/windows/VC6/progs/glut/glut.dsp	\
-	$(DIRECTORY)/windows/VC7/progs/glut/glut.vcproj
+	$(DIRECTORY)/src/glut/directfb/*[ch]
 
 DEPEND_FILES = \
 	$(TOP)/src/mesa/depend		\
diff --git a/configs/linux b/configs/linux
index 73a6b61b2f6..c60f0d8ac3b 100644
--- a/configs/linux
+++ b/configs/linux
@@ -23,11 +23,11 @@ DEFINES = -D_POSIX_SOURCE -D_POSIX_C_SOURCE=199309L -D_SVID_SOURCE \
 X11_INCLUDES = -I/usr/X11R6/include
 
 CFLAGS = -Wall -Wmissing-prototypes -Wdeclaration-after-statement \
-	$(OPT_FLAGS) $(PIC_FLAGS) $(ARCH_FLAGS) \
-	$(DEFINES) $(ASM_FLAGS) $(X11_INCLUDES) -std=c99 -ffast-math 
+	-Wpointer-arith $(OPT_FLAGS) $(PIC_FLAGS) $(ARCH_FLAGS) \
+	$(DEFINES) $(ASM_FLAGS) $(X11_INCLUDES) -std=c99 -ffast-math
 
-CXXFLAGS = -Wall $(OPT_FLAGS) $(PIC_FLAGS) $(ARCH_FLAGS) $(DEFINES) \
-	$(X11_INCLUDES) 
+CXXFLAGS = -Wall -Wpointer-arith $(OPT_FLAGS) $(PIC_FLAGS) $(ARCH_FLAGS) \
+	$(DEFINES) $(X11_INCLUDES)
 
 # Work around aliasing bugs - developers should comment this out
 CFLAGS += -fno-strict-aliasing
diff --git a/docs/README.WIN32 b/docs/README.WIN32
index 97e1ffb7a75..204b8e66041 100644
--- a/docs/README.WIN32
+++ b/docs/README.WIN32
@@ -27,12 +27,6 @@ using the SDK with Visual Studio Express can be found at
 
 http://msdn.microsoft.com/vstudio/express/visualc/usingpsdk/
 
-If you are stuck using VC6 or VC7, you may start with these project
-files, but you may need to modify them to reflect changes in the
-Mesa source code tree.  If you sucessfully update the project files,
-please submit them to the author of this document so that they may
-be included in the next distribution. 
-
 The project files to build the core Mesa library, Windows Mesa
 drivers, OSMesa, and GLU are in the mesa directory.  The project files
 to build GLUT and some demo programs are in the progs directory.
@@ -106,23 +100,6 @@ should build all the demos.
 Build System Notes
 ----- ------ -----
 
-VC6 (not actively supported)
----
-
-Visual Studio 6 does not recognize files with the .cc extension as C++
-language files, without a lot of unnatural tweaking.  So, the VC6
-build process uses custom build steps to compile these files in the
-GLU library.
-
-Two additional configurations are provided, Debug x86 and Release x86
-that activate the shader code compilation by defining SLANG_86.  It is
-unknown if and how this works.
-
-VC7 (not actively supported)
----
-
-The above-mentioned .cc problem does not exist in this version.
-
 VC8
 ---
 
diff --git a/docs/news.html b/docs/news.html
index 93da56a5c33..2abec2e6354 100644
--- a/docs/news.html
+++ b/docs/news.html
@@ -10,6 +10,13 @@
 
 <H1>News</H1>
 
+<h2>November XX, 2009</h2>
+<p>
+<a href="relnotes-7.6.1.html">Mesa 7.6.1</a> is released.  This is a bug-fix
+release fixing issues found in the 7.6 release.
+</p>
+
+
 <h2>September 28, 2009</h2>
 <p>
 <a href="relnotes-7.6.html">Mesa 7.6</a> is released.  This is a new feature
diff --git a/docs/relnotes-7.6.1.html b/docs/relnotes-7.6.1.html
index 9ee6babc9d8..d155cf5a673 100644
--- a/docs/relnotes-7.6.1.html
+++ b/docs/relnotes-7.6.1.html
@@ -34,6 +34,7 @@ tbd
 <ul>
 <li>Upgraded GL/glext.h to version 56, GL/glxext.h to version 25,
     GL/wglext.h to version 17
+<li>New 3D driver, r600, for Radeon R6xx, R7xx hardware
 </ul>
 
 
@@ -52,7 +53,16 @@ tbd
     it returns the actual compressed format chosen.
 <li>Fixed glBitmap bugs in Intel drivers.
 <li>Fixed a number of Microsoft Visual Studio compilation problems.
+<li>Fixed clipping / provoking vertex bugs in i965 driver.
+<li>Assorted build fixes for AIX.
+<li>Endianness fixes for the DRI swrast driver (bug 22767).</li>
 </ul>
 
+<h2>Changes</h2>
+<ul>
+<li>Removed old VC6, VC7 project files for Windows
+</ul>
+
+
 </body>
 </html>
diff --git a/include/GL/glx_mangle.h b/include/GL/glx_mangle.h
index 4439a96b5d2..fa664a7f59e 100644
--- a/include/GL/glx_mangle.h
+++ b/include/GL/glx_mangle.h
@@ -76,6 +76,7 @@
 
 /* GLX 1.4 */
 #define glXGetProcAddress mglXGetProcAddress
+#define glXGetProcAddressARB mglXGetProcAddressARB
 
 
 #endif
diff --git a/progs/Makefile b/progs/Makefile
index 3700707dfb1..d5852fa416c 100644
--- a/progs/Makefile
+++ b/progs/Makefile
@@ -4,7 +4,7 @@ TOP = ..
 
 include $(TOP)/configs/current
 
-SUBDIRS = $(PROGRAM_DIRS)
+SUBDIRS = "$(strip "$(PROGRAM_DIRS)")"
 
 
 default: message subdirs
diff --git a/progs/util/shaderutil.c b/progs/util/shaderutil.c
index 4fb3199a190..2f44c388d8a 100644
--- a/progs/util/shaderutil.c
+++ b/progs/util/shaderutil.c
@@ -25,7 +25,11 @@ GLboolean
 ShadersSupported(void)
 {
    const char *version = (const char *) glGetString(GL_VERSION);
-   if (version[0] == '2' && version[1] == '.') {
+
+   /* NVIDIA binary drivers will return "3.0.0", and they clearly support
+    * shaders.
+    */
+   if (version[0] >= '2' && version[1] == '.') {
       return GL_TRUE;
    }
    else if (glutExtensionSupported("GL_ARB_vertex_shader")
diff --git a/scons/gallium.py b/scons/gallium.py
index 483bd5d29eb..ca1ca5153e7 100644
--- a/scons/gallium.py
+++ b/scons/gallium.py
@@ -366,7 +366,7 @@ def generate(env):
         ccflags += [
             '-Wall',
             '-Wmissing-field-initializers',
-            '-Wpointer-arith',
+            '-Werror=pointer-arith',
             '-Wno-long-long',
             '-ffast-math',
             '-fmessage-length=0', # be nice to Eclipse
diff --git a/src/gallium/auxiliary/rbug/README b/src/gallium/auxiliary/rbug/README
index 33d76371de4..d984067893c 100644
--- a/src/gallium/auxiliary/rbug/README
+++ b/src/gallium/auxiliary/rbug/README
@@ -16,6 +16,10 @@ for information about applications look in:
 
 progs/rbug/README
 
+for a GUI see:
+
+  http://cgit.freedesktop.org/mesa/rbug-gui
+
 
 --
 Jakob Bornecrantz <[email protected]>
diff --git a/src/gallium/auxiliary/util/u_mm.c b/src/gallium/auxiliary/util/u_mm.c
index 4b75d4ba1d0..82f83702d1e 100644
--- a/src/gallium/auxiliary/util/u_mm.c
+++ b/src/gallium/auxiliary/util/u_mm.c
@@ -39,13 +39,20 @@ u_mmDumpMemInfo(const struct mem_block *heap)
    }
    else {
       const struct mem_block *p;
+      int total_used = 0, total_free = 0;
 
       for (p = heap->next; p != heap; p = p->next) {
 	 debug_printf("  Offset:%08x, Size:%08x, %c%c\n", p->ofs, p->size,
                       p->free ? 'F':'.',
                       p->reserved ? 'R':'.');
+         if (p->free)
+            total_free += p->size;
+         else
+            total_used += p->size;
       }
 
+      debug_printf("'\nMemory stats: total = %d, used = %d, free = %d\n",
+                   total_used + total_free, total_used, total_free);
       debug_printf("\nFree list:\n");
 
       for (p = heap->next_free; p != heap; p = p->next_free) {
diff --git a/src/gallium/drivers/svga/Makefile b/src/gallium/drivers/svga/Makefile
index 38b63394e38..f3619081875 100644
--- a/src/gallium/drivers/svga/Makefile
+++ b/src/gallium/drivers/svga/Makefile
@@ -50,6 +50,9 @@ C_SOURCES = \
 LIBRARY_INCLUDES = \
 	-I$(TOP)/src/gallium/drivers/svga/include
 
+# With linux-debug we get a lots of warnings, filter out the bad flags.
+CFLAGS := $(filter-out -pedantic, $(filter-out -ansi, $(CFLAGS)))
+
 LIBRARY_DEFINES = \
 	-std=gnu99 -fvisibility=hidden \
 	-DHAVE_STDINT_H -DHAVE_SYS_TYPES_H
diff --git a/src/gallium/drivers/svga/svga_context.c b/src/gallium/drivers/svga/svga_context.c
index 73233957f36..c3de12b4a39 100644
--- a/src/gallium/drivers/svga/svga_context.c
+++ b/src/gallium/drivers/svga/svga_context.c
@@ -230,7 +230,9 @@ void svga_context_flush( struct svga_context *svga,
                          struct pipe_fence_handle **pfence )
 {
    struct svga_screen *svgascreen = svga_screen(svga->pipe.screen);
-   
+
+   svga->curr.nr_fbs = 0;
+
    /* Unmap upload manager buffers: 
     */
    u_upload_flush(svga->upload_vb);
diff --git a/src/gallium/drivers/svga/svga_context.h b/src/gallium/drivers/svga/svga_context.h
index 9a3e92fd8d1..e650a251d19 100644
--- a/src/gallium/drivers/svga/svga_context.h
+++ b/src/gallium/drivers/svga/svga_context.h
@@ -191,6 +191,11 @@ struct svga_state
    struct pipe_framebuffer_state framebuffer;
    float depthscale;
 
+   /* Hack to limit the number of different render targets between
+    * flushes.  Helps avoid blowing out our surface cache in EXA.
+    */
+   int nr_fbs;
+
    struct pipe_poly_stipple poly_stipple;
    struct pipe_scissor_state scissor;
    struct pipe_blend_color blend_color;
diff --git a/src/gallium/drivers/svga/svga_debug.h b/src/gallium/drivers/svga/svga_debug.h
index b7bb5686ed3..3a3fcd8fae2 100644
--- a/src/gallium/drivers/svga/svga_debug.h
+++ b/src/gallium/drivers/svga/svga_debug.h
@@ -43,6 +43,7 @@
 #define DEBUG_FLUSH    0x1000   /* flush after every draw */
 #define DEBUG_SYNC     0x2000   /* sync after every flush */
 #define DEBUG_QUERY    0x4000
+#define DEBUG_CACHE    0x8000
 
 #ifdef DEBUG
 extern int SVGA_DEBUG;
diff --git a/src/gallium/drivers/svga/svga_draw.c b/src/gallium/drivers/svga/svga_draw.c
index 1b371cecc61..8db40d0fd57 100644
--- a/src/gallium/drivers/svga/svga_draw.c
+++ b/src/gallium/drivers/svga/svga_draw.c
@@ -29,10 +29,13 @@
 #include "util/u_memory.h"
 #include "util/u_math.h"
 
+#include "svga_context.h"
 #include "svga_draw.h"
 #include "svga_draw_private.h"
+#include "svga_debug.h"
 #include "svga_screen.h"
 #include "svga_screen_buffer.h"
+#include "svga_screen_texture.h"
 #include "svga_winsys.h"
 #include "svga_cmd.h"
 
@@ -160,6 +163,10 @@ svga_hwtnl_flush( struct svga_hwtnl *hwtnl )
          ib_handle[i] = handle;
       }
 
+      SVGA_DBG(DEBUG_DMA, "draw to sid %p, %d prims\n",
+               svga_surface(svga->curr.framebuffer.cbufs[0])->handle,
+               hwtnl->cmd.prim_count);
+
       ret = SVGA3D_BeginDrawPrimitives(swc, 
                                        &vdecl, 
                                        hwtnl->cmd.vdecl_count, 
diff --git a/src/gallium/drivers/svga/svga_pipe_blit.c b/src/gallium/drivers/svga/svga_pipe_blit.c
index 5a4a8c0f5f1..4f575b06e62 100644
--- a/src/gallium/drivers/svga/svga_pipe_blit.c
+++ b/src/gallium/drivers/svga/svga_pipe_blit.c
@@ -25,6 +25,7 @@
 
 #include "svga_screen_texture.h"
 #include "svga_context.h"
+#include "svga_debug.h"
 #include "svga_cmd.h"
 
 #define FILE_DEBUG_FLAG DEBUG_BLIT
@@ -43,6 +44,13 @@ static void svga_surface_copy(struct pipe_context *pipe,
 
    svga_hwtnl_flush_retry( svga );
 
+   SVGA_DBG(DEBUG_DMA, "blit to sid %p (%d,%d), from sid %p (%d,%d) sz %dx%d\n",
+            svga_surface(dest)->handle,
+            destx, desty,
+            svga_surface(src)->handle,
+            srcx, srcy,
+            width, height);
+
    ret = SVGA3D_BeginSurfaceCopy(svga->swc,
                                  src,
                                  dest,
diff --git a/src/gallium/drivers/svga/svga_pipe_clear.c b/src/gallium/drivers/svga/svga_pipe_clear.c
index 8977d26541c..6195c3897ed 100644
--- a/src/gallium/drivers/svga/svga_pipe_clear.c
+++ b/src/gallium/drivers/svga/svga_pipe_clear.c
@@ -24,12 +24,14 @@
  **********************************************************/
 
 #include "svga_cmd.h"
+#include "svga_debug.h"
 
 #include "pipe/p_defines.h"
 #include "util/u_pack_color.h"
 
 #include "svga_context.h"
 #include "svga_state.h"
+#include "svga_screen_texture.h"
 
 
 static enum pipe_error
@@ -98,6 +100,10 @@ svga_clear(struct pipe_context *pipe, unsigned buffers, const float *rgba,
 {
    struct svga_context *svga = svga_context( pipe );
    int ret;
+   
+   if (buffers & PIPE_CLEAR_COLOR)
+      SVGA_DBG(DEBUG_DMA, "clear sid %p\n",
+               svga_surface(svga->curr.framebuffer.cbufs[0])->handle);
 
    ret = try_clear( svga, buffers, rgba, depth, stencil );
 
diff --git a/src/gallium/drivers/svga/svga_pipe_flush.c b/src/gallium/drivers/svga/svga_pipe_flush.c
index 942366de721..0becb0765ac 100644
--- a/src/gallium/drivers/svga/svga_pipe_flush.c
+++ b/src/gallium/drivers/svga/svga_pipe_flush.c
@@ -59,6 +59,9 @@ static void svga_flush( struct pipe_context *pipe,
    /* Flush command queue.
     */
    svga_context_flush(svga, fence);
+
+   SVGA_DBG(DEBUG_DMA|DEBUG_PERF, "%s flags %x fence_ptr %p\n",
+            __FUNCTION__, flags, fence ? *fence : 0x0);
 }
 
 
diff --git a/src/gallium/drivers/svga/svga_screen.c b/src/gallium/drivers/svga/svga_screen.c
index 3afcaffff55..fc1b3c980ef 100644
--- a/src/gallium/drivers/svga/svga_screen.c
+++ b/src/gallium/drivers/svga/svga_screen.c
@@ -57,6 +57,7 @@ static const struct debug_named_value svga_debug_flags[] = {
    { "perf",     DEBUG_PERF },
    { "flush",    DEBUG_FLUSH },
    { "sync",     DEBUG_SYNC },
+   { "cache",    DEBUG_CACHE },
    {NULL, 0}
 };
 #endif
@@ -297,6 +298,10 @@ svga_fence_finish(struct pipe_screen *screen,
                   unsigned flag)
 {
    struct svga_winsys_screen *sws = svga_screen(screen)->sws;
+
+   SVGA_DBG(DEBUG_DMA|DEBUG_PERF, "%s fence_ptr %p\n",
+            __FUNCTION__, fence);
+
    return sws->fence_finish(sws, fence, flag);
 }
 
diff --git a/src/gallium/drivers/svga/svga_screen_buffer.c b/src/gallium/drivers/svga/svga_screen_buffer.c
index 3b7811734ed..1f8a8896723 100644
--- a/src/gallium/drivers/svga/svga_screen_buffer.c
+++ b/src/gallium/drivers/svga/svga_screen_buffer.c
@@ -71,7 +71,10 @@ svga_buffer_create_host_surface(struct svga_screen *ss,
       
       sbuf->key.numFaces = 1;
       sbuf->key.numMipLevels = 1;
+      sbuf->key.cachable = 1;
       
+      SVGA_DBG(DEBUG_DMA, "surface_create for buffer sz %d\n", sbuf->base.size);
+
       sbuf->handle = svga_screen_surface_create(ss, &sbuf->key);
       if(!sbuf->handle)
          return PIPE_ERROR_OUT_OF_MEMORY;
@@ -82,7 +85,7 @@ svga_buffer_create_host_surface(struct svga_screen *ss,
        */
       sbuf->hw.flags.discard = TRUE;
 
-      SVGA_DBG(DEBUG_DMA, "   grab sid %p sz %d\n", sbuf->handle, sbuf->base.size);
+      SVGA_DBG(DEBUG_DMA, "   --> got sid %p sz %d (buffer)\n", sbuf->handle, sbuf->base.size);
    }
    
    return PIPE_OK;
@@ -444,7 +447,7 @@ svga_buffer_map_range( struct pipe_screen *screen,
             enum pipe_error ret;
             struct pipe_fence_handle *fence = NULL;
             
-            SVGA_DBG(DEBUG_DMA|DEBUG_PERF, "dma from sid %p, bytes %u - %u\n", 
+            SVGA_DBG(DEBUG_DMA|DEBUG_PERF, "dma from sid %p (buffer), bytes %u - %u\n", 
                      sbuf->handle, 0, sbuf->base.size);
 
             memset(&flags, 0, sizeof flags);
@@ -776,12 +779,11 @@ svga_screen_buffer_wrap_surface(struct pipe_screen *screen,
 
    /*
     * We are not the creator of this surface and therefore we must not
-    * cache it for reuse. The caching code only caches SVGA3D_BUFFER surfaces
-    * so make sure this isn't one of those.
+    * cache it for reuse. Set the cacheable flag to zero in the key to
+    * prevent this.
     */
-
-   assert(format != SVGA3D_BUFFER);
    sbuf->key.format = format;
+   sbuf->key.cachable = 0;
    sws->surface_reference(sws, &sbuf->handle, srf);
 
    return buf;
@@ -794,6 +796,8 @@ svga_screen_buffer_get_winsys_surface(struct pipe_buffer *buffer)
    struct svga_winsys_screen *sws = svga_winsys_screen(buffer->screen);
    struct svga_winsys_surface *vsurf = NULL;
 
+   assert(svga_buffer(buffer)->key.cachable == 0);
+   svga_buffer(buffer)->key.cachable = 0;
    sws->surface_reference(sws, &vsurf, svga_buffer(buffer)->handle);
    return vsurf;
 }
diff --git a/src/gallium/drivers/svga/svga_screen_cache.c b/src/gallium/drivers/svga/svga_screen_cache.c
index 7360c1688bb..8a06383f61e 100644
--- a/src/gallium/drivers/svga/svga_screen_cache.c
+++ b/src/gallium/drivers/svga/svga_screen_cache.c
@@ -24,6 +24,7 @@
  **********************************************************/
 
 #include "util/u_memory.h"
+#include "util/u_hash.h"
 
 #include "svga_debug.h"
 #include "svga_winsys.h"
@@ -36,24 +37,11 @@
 
 /** 
  * Compute the bucket for this key. 
- * 
- * We simply compute log2(width) for now, but
  */
 static INLINE unsigned
 svga_screen_cache_bucket(const struct svga_host_surface_cache_key *key)
 {
-   unsigned bucket = 0;
-   unsigned size = key->size.width;
-   
-   while ((size >>= 1))
-      ++bucket;
-   
-   if(key->flags & SVGA3D_SURFACE_HINT_INDEXBUFFER)
-      bucket += 32;
-   
-   assert(bucket < SVGA_HOST_SURFACE_CACHE_BUCKETS);
-   
-   return bucket;
+   return util_hash_crc32( key, sizeof *key ) % SVGA_HOST_SURFACE_CACHE_BUCKETS;
 }
 
 
@@ -69,6 +57,8 @@ svga_screen_cache_lookup(struct svga_screen *svgascreen,
    unsigned bucket;
    unsigned tries = 0;
 
+   assert(key->cachable);
+
    bucket = svga_screen_cache_bucket(key);
 
    pipe_mutex_lock(cache->mutex);
@@ -104,11 +94,9 @@ svga_screen_cache_lookup(struct svga_screen *svgascreen,
 
    pipe_mutex_unlock(cache->mutex);
    
-#if 0
-   _debug_printf("%s: cache %s after %u tries\n", __FUNCTION__, handle ? "hit" : "miss", tries);
-#else
-   (void)tries;
-#endif
+   if (SVGA_DEBUG & DEBUG_DMA)
+      debug_printf("%s: cache %s after %u tries (bucket %d)\n", __FUNCTION__, 
+                   handle ? "hit" : "miss", tries, bucket);
    
    return handle;
 }
@@ -128,6 +116,7 @@ svga_screen_cache_add(struct svga_screen *svgascreen,
    struct svga_host_surface_cache_entry *entry = NULL;
    struct svga_winsys_surface *handle = *p_handle;
    
+   assert(key->cachable);
 
    assert(handle);
    if(!handle)
@@ -137,15 +126,16 @@ svga_screen_cache_add(struct svga_screen *svgascreen,
    pipe_mutex_lock(cache->mutex);
    
    if(!LIST_IS_EMPTY(&cache->empty)) {
-        /* use the first empty entry */
-        entry = LIST_ENTRY(struct svga_host_surface_cache_entry, cache->empty.next, head);
+      /* use the first empty entry */
+      entry = LIST_ENTRY(struct svga_host_surface_cache_entry, cache->empty.next, head);
         
-        LIST_DEL(&entry->head);
-     }
+      LIST_DEL(&entry->head);
+   }
    else if(!LIST_IS_EMPTY(&cache->unused)) {
       /* free the last used buffer and reuse its entry */
       entry = LIST_ENTRY(struct svga_host_surface_cache_entry, cache->unused.prev, head);
-      SVGA_DBG(DEBUG_DMA, "unref sid %p\n", entry->handle);
+      SVGA_DBG(DEBUG_CACHE|DEBUG_DMA,
+               "unref sid %p (make space)\n", entry->handle);
       sws->surface_reference(sws, &entry->handle, NULL);
 
       LIST_DEL(&entry->bucket_head);
@@ -157,11 +147,14 @@ svga_screen_cache_add(struct svga_screen *svgascreen,
       entry->handle = handle;
       memcpy(&entry->key, key, sizeof entry->key);
    
+      SVGA_DBG(DEBUG_CACHE|DEBUG_DMA,
+               "cache sid %p\n", entry->handle);
       LIST_ADD(&entry->head, &cache->validated);
    }
    else {
       /* Couldn't cache the buffer -- this really shouldn't happen */
-      SVGA_DBG(DEBUG_DMA, "unref sid %p\n", handle);
+      SVGA_DBG(DEBUG_CACHE|DEBUG_DMA,
+               "unref sid %p (couldn't find space)\n", handle);
       sws->surface_reference(sws, &handle, NULL);
    }
    
@@ -220,7 +213,8 @@ svga_screen_cache_cleanup(struct svga_screen *svgascreen)
    
    for(i = 0; i < SVGA_HOST_SURFACE_CACHE_SIZE; ++i) {
       if(cache->entries[i].handle) {
-	 SVGA_DBG(DEBUG_DMA, "unref sid %p\n", cache->entries[i].handle);
+	 SVGA_DBG(DEBUG_CACHE|DEBUG_DMA,
+                  "unref sid %p (shutdown)\n", cache->entries[i].handle);
 	 sws->surface_reference(sws, &cache->entries[i].handle, NULL);
       }
 
@@ -261,18 +255,45 @@ svga_screen_surface_create(struct svga_screen *svgascreen,
 {
    struct svga_winsys_screen *sws = svgascreen->sws;
    struct svga_winsys_surface *handle = NULL;
+   boolean cachable = SVGA_SURFACE_CACHE_ENABLED && key->cachable;
+
+   SVGA_DBG(DEBUG_CACHE|DEBUG_DMA,
+            "%s sz %dx%dx%d mips %d faces %d cachable %d\n", 
+            __FUNCTION__,
+            key->size.width,
+            key->size.height,
+            key->size.depth,
+            key->numMipLevels,
+            key->numFaces,
+            key->cachable);
+
+   if (cachable) {
+      if (key->format == SVGA3D_BUFFER) {
+         /* For buffers, round the buffer size up to the nearest power
+          * of two to increase the probability of cache hits.  Keep
+          * texture surface dimensions unchanged.
+          */
+         uint32_t size = 1;
+         while(size < key->size.width)
+            size <<= 1;
+         key->size.width = size;
+      }
 
-   if (SVGA_SURFACE_CACHE_ENABLED && key->format == SVGA3D_BUFFER) {
-      /* round the buffer size up to the nearest power of two to increase the
-       * probability of cache hits */
-      uint32_t size = 1;
-      while(size < key->size.width)
-         size <<= 1;
-      key->size.width = size;
-      
       handle = svga_screen_cache_lookup(svgascreen, key);
-      if (handle)
-         SVGA_DBG(DEBUG_DMA, "  reuse sid %p sz %d\n", handle, size);
+      if (handle) {
+         if (key->format == SVGA3D_BUFFER)
+            SVGA_DBG(DEBUG_CACHE|DEBUG_DMA,
+                     "reuse sid %p sz %d (buffer)\n", handle, 
+                     key->size.width);
+         else
+            SVGA_DBG(DEBUG_CACHE|DEBUG_DMA,
+                     "reuse sid %p sz %dx%dx%d mips %d faces %d\n", handle, 
+                     key->size.width,
+                     key->size.height,
+                     key->size.depth,
+                     key->numMipLevels,
+                     key->numFaces);
+      }
    }
 
    if (!handle) {
@@ -283,7 +304,12 @@ svga_screen_surface_create(struct svga_screen *svgascreen,
                                    key->numFaces, 
                                    key->numMipLevels);
       if (handle)
-         SVGA_DBG(DEBUG_DMA, "create sid %p sz %d\n", handle, key->size);
+         SVGA_DBG(DEBUG_CACHE|DEBUG_DMA,
+                  "  CREATE sid %p sz %dx%dx%d\n", 
+                  handle, 
+                  key->size.width,
+                  key->size.height,
+                  key->size.depth);
    }
 
    return handle;
@@ -297,11 +323,16 @@ svga_screen_surface_destroy(struct svga_screen *svgascreen,
 {
    struct svga_winsys_screen *sws = svgascreen->sws;
    
-   if(SVGA_SURFACE_CACHE_ENABLED && key->format == SVGA3D_BUFFER) {
+   /* We only set the cachable flag for surfaces of which we are the
+    * exclusive owner.  So just hold onto our existing reference in
+    * that case.
+    */
+   if(SVGA_SURFACE_CACHE_ENABLED && key->cachable) {
       svga_screen_cache_add(svgascreen, key, p_handle);
    }
    else {
-      SVGA_DBG(DEBUG_DMA, "unref sid %p\n", *p_handle);
+      SVGA_DBG(DEBUG_DMA,
+               "unref sid %p (uncachable)\n", *p_handle);
       sws->surface_reference(sws, p_handle, NULL);
    }
 }
diff --git a/src/gallium/drivers/svga/svga_screen_cache.h b/src/gallium/drivers/svga/svga_screen_cache.h
index 1bbe9877688..f5aa740d408 100644
--- a/src/gallium/drivers/svga/svga_screen_cache.h
+++ b/src/gallium/drivers/svga/svga_screen_cache.h
@@ -36,10 +36,18 @@
 #include "util/u_double_list.h"
 
 
-/* TODO: Reduce this once we don't allocate an index buffer per draw call */ 
+/* Guess the storage size of cached surfaces and try and keep it under
+ * this amount:
+ */ 
+#define SVGA_HOST_SURFACE_CACHE_BYTES 16*1024*1024
+
+/* Maximum number of discrete surfaces in the cache:
+ */
 #define SVGA_HOST_SURFACE_CACHE_SIZE 1024
 
-#define SVGA_HOST_SURFACE_CACHE_BUCKETS 64
+/* Number of hash buckets:
+ */
+#define SVGA_HOST_SURFACE_CACHE_BUCKETS 256
 
 
 struct svga_winsys_surface;
@@ -53,8 +61,9 @@ struct svga_host_surface_cache_key
    SVGA3dSurfaceFlags flags;
    SVGA3dSurfaceFormat format;
    SVGA3dSize size;
-   uint32_t numFaces;
-   uint32_t numMipLevels;
+   uint32_t numFaces:24;
+   uint32_t numMipLevels:7;
+   uint32_t cachable:1;         /* False if this is a shared surface */
 };
 
 
diff --git a/src/gallium/drivers/svga/svga_screen_texture.c b/src/gallium/drivers/svga/svga_screen_texture.c
index fb11b80dcf9..6e10d65a205 100644
--- a/src/gallium/drivers/svga/svga_screen_texture.c
+++ b/src/gallium/drivers/svga/svga_screen_texture.c
@@ -266,14 +266,8 @@ svga_texture_create(struct pipe_screen *screen,
                     const struct pipe_texture *templat)
 {
    struct svga_screen *svgascreen = svga_screen(screen);
-   struct svga_winsys_screen *sws = svgascreen->sws;
    struct svga_texture *tex = CALLOC_STRUCT(svga_texture);
    unsigned width, height, depth;
-   SVGA3dSurfaceFlags flags = 0;
-   SVGA3dSurfaceFormat format;
-   SVGA3dSize size;
-   uint32 numFaces;
-   uint32 numMipLevels;
    unsigned level;
    
    if (!tex)
@@ -298,23 +292,24 @@ svga_texture_create(struct pipe_screen *screen,
       depth = u_minify(depth, 1);
    }
    
-   size.width = templat->width0;
-   size.height = templat->height0;
-   size.depth = templat->depth0;
+   tex->key.flags = 0;
+   tex->key.size.width = templat->width0;
+   tex->key.size.height = templat->height0;
+   tex->key.size.depth = templat->depth0;
    
    if(templat->target == PIPE_TEXTURE_CUBE) {
-      flags |= SVGA3D_SURFACE_CUBEMAP;
-      numFaces = 6;
+      tex->key.flags |= SVGA3D_SURFACE_CUBEMAP;
+      tex->key.numFaces = 6;
    }
    else {
-      numFaces = 1;
+      tex->key.numFaces = 1;
    }
 
    if(templat->tex_usage & PIPE_TEXTURE_USAGE_SAMPLER)
-      flags |= SVGA3D_SURFACE_HINT_TEXTURE;
+      tex->key.flags |= SVGA3D_SURFACE_HINT_TEXTURE;
 
    if(templat->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY)
-      flags |= SVGA3D_SURFACE_HINT_SCANOUT;
+      tex->key.flags |= SVGA3D_SURFACE_HINT_SCANOUT;
    
    /* 
     * XXX: Never pass the SVGA3D_SURFACE_HINT_RENDERTARGET hint. Mesa cannot
@@ -325,21 +320,24 @@ svga_texture_create(struct pipe_screen *screen,
 #if 0
    if((templat->tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) &&
       !pf_is_compressed(templat->format))
-      flags |= SVGA3D_SURFACE_HINT_RENDERTARGET;
+      tex->key.flags |= SVGA3D_SURFACE_HINT_RENDERTARGET;
 #endif
    
    if(templat->tex_usage & PIPE_TEXTURE_USAGE_DEPTH_STENCIL)
-      flags |= SVGA3D_SURFACE_HINT_DEPTHSTENCIL;
+      tex->key.flags |= SVGA3D_SURFACE_HINT_DEPTHSTENCIL;
    
-   numMipLevels = templat->last_level + 1;
+   tex->key.numMipLevels = templat->last_level + 1;
    
-   format = svga_translate_format(templat->format);
-   if(format == SVGA3D_FORMAT_INVALID)
+   tex->key.format = svga_translate_format(templat->format);
+   if(tex->key.format == SVGA3D_FORMAT_INVALID)
       goto error2;
+
+   tex->key.cachable = 1;
    
-   tex->handle = sws->surface_create(sws, flags, format, size, numFaces, numMipLevels);
+   SVGA_DBG(DEBUG_DMA, "surface_create for texture\n", tex->handle);
+   tex->handle = svga_screen_surface_create(svgascreen, &tex->key);
    if (tex->handle)
-      SVGA_DBG(DEBUG_DMA, "create sid %p (texture)\n", tex->handle);
+      SVGA_DBG(DEBUG_DMA, "  --> got sid %p (texture)\n", tex->handle);
 
    return &tex->base;
 
@@ -395,6 +393,7 @@ svga_texture_blanket(struct pipe_screen * screen,
       return NULL;
 
    tex->base = *base;
+   
 
    if (sbuf->key.format == 1)
       tex->base.format = PIPE_FORMAT_X8R8G8B8_UNORM;
@@ -404,6 +403,12 @@ svga_texture_blanket(struct pipe_screen * screen,
    pipe_reference_init(&tex->base.reference, 1);
    tex->base.screen = screen;
 
+   SVGA_DBG(DEBUG_DMA, "blanket sid %p\n", sbuf->handle);
+
+   /* We don't own this storage, so don't try to cache it.
+    */
+   assert(sbuf->key.cachable == 0);
+   tex->key.cachable = 0;
    sws->surface_reference(sws, &tex->handle, sbuf->handle);
 
    return &tex->base;
@@ -424,7 +429,7 @@ svga_texture_destroy(struct pipe_texture *pt)
      DBG("%s deleting %p\n", __FUNCTION__, (void *) tex);
    */
    SVGA_DBG(DEBUG_DMA, "unref sid %p (texture)\n", tex->handle);
-   ss->sws->surface_reference(ss->sws, &tex->handle, NULL);
+   svga_screen_surface_destroy(ss, &tex->key, &tex->handle);
 
    FREE(tex);
 }
@@ -515,43 +520,47 @@ svga_texture_view_surface(struct pipe_context *pipe,
                           unsigned start_mip,
                           unsigned num_mip,
                           int face_pick,
-                          int zslice_pick)
+                          int zslice_pick,
+                          struct svga_host_surface_cache_key *key) /* OUT */
 {
    struct svga_screen *ss = svga_screen(tex->base.screen);
-   struct svga_winsys_screen *sws = ss->sws;
    struct svga_winsys_surface *handle;
    int i, j;
-   SVGA3dSurfaceFlags flags = 0;
-   SVGA3dSize size;
-   uint32 numFaces;
-   uint32 numMipLevels = num_mip;
    unsigned z_offset = 0;
 
    SVGA_DBG(DEBUG_PERF, 
             "svga: Create surface view: face %d zslice %d mips %d..%d\n",
             face_pick, zslice_pick, start_mip, start_mip+num_mip-1);
 
-   size.width = u_minify(tex->base.width0, start_mip);
-   size.height = u_minify(tex->base.height0, start_mip);
-   size.depth = zslice_pick < 0 ? u_minify(tex->base.depth0, start_mip) : 1;
-   assert(size.depth == 1);
+   key->flags = 0;
+   key->format = format;
+   key->numMipLevels = num_mip;
+   key->size.width = u_minify(tex->base.width0, start_mip);
+   key->size.height = u_minify(tex->base.height0, start_mip);
+   key->size.depth = zslice_pick < 0 ? u_minify(tex->base.depth0, start_mip) : 1;
+   key->cachable = 1;
+   assert(key->size.depth == 1);
    
    if(tex->base.target == PIPE_TEXTURE_CUBE && face_pick < 0) {
-      flags |= SVGA3D_SURFACE_CUBEMAP;
-      numFaces = 6;
+      key->flags |= SVGA3D_SURFACE_CUBEMAP;
+      key->numFaces = 6;
    } else {
-      numFaces = 1;
+      key->numFaces = 1;
    }
 
-   if(format == SVGA3D_FORMAT_INVALID)
+   if(key->format == SVGA3D_FORMAT_INVALID) {
+      key->cachable = 0;
       return NULL;
+   }
 
-   handle = sws->surface_create(sws, flags, format, size, numFaces, numMipLevels);
-
-   if (!handle)
+   SVGA_DBG(DEBUG_DMA, "surface_create for texture view\n");
+   handle = svga_screen_surface_create(ss, key);
+   if (!handle) {
+      key->cachable = 0;
       return NULL;
+   }
 
-   SVGA_DBG(DEBUG_DMA, "create sid %p (texture view)\n", handle);
+   SVGA_DBG(DEBUG_DMA, " --> got sid %p (texture view)\n", handle);
 
    if (face_pick < 0)
       face_pick = 0;
@@ -559,15 +568,23 @@ svga_texture_view_surface(struct pipe_context *pipe,
    if (zslice_pick >= 0)
        z_offset = zslice_pick;
 
-   for (i = 0; i < num_mip; i++) {
-      for (j = 0; j < numFaces; j++) {
+   for (i = 0; i < key->numMipLevels; i++) {
+      for (j = 0; j < key->numFaces; j++) {
          if(tex->defined[j + face_pick][i + start_mip]) {
-            unsigned depth = zslice_pick < 0 ? u_minify(tex->base.depth0, i + start_mip) : 1;
-            svga_texture_copy_handle(svga_context(pipe), ss,
-                                     tex->handle, 0, 0, z_offset, i + start_mip, j + face_pick,
+            unsigned depth = (zslice_pick < 0 ?
+                              u_minify(tex->base.depth0, i + start_mip) :
+                              1);
+
+            svga_texture_copy_handle(svga_context(pipe),
+                                     ss,
+                                     tex->handle, 
+                                     0, 0, z_offset, 
+                                     i + start_mip, 
+                                     j + face_pick,
                                      handle, 0, 0, 0, i, j,
                                      u_minify(tex->base.width0, i + start_mip),
-                                     u_minify(tex->base.height0, i + start_mip), depth);
+                                     u_minify(tex->base.height0, i + start_mip),
+                                     depth);
          }
       }
    }
@@ -584,25 +601,23 @@ svga_get_tex_surface(struct pipe_screen *screen,
 {
    struct svga_texture *tex = svga_texture(pt);
    struct svga_surface *s;
-   struct pipe_surface *ps;
    boolean render = flags & PIPE_BUFFER_USAGE_GPU_WRITE ? TRUE : FALSE;
    boolean view = FALSE;
    SVGA3dSurfaceFormat format;
 
    s = CALLOC_STRUCT(svga_surface);
-   ps = &s->base;
-   if (!ps)
+   if (!s)
       return NULL;
 
-   pipe_reference_init(&ps->reference, 1);
-   pipe_texture_reference(&ps->texture, pt);
-   ps->format = pt->format;
-   ps->width = u_minify(pt->width0, level);
-   ps->height = u_minify(pt->height0, level);
-   ps->usage = flags;
-   ps->level = level;
-   ps->face = face;
-   ps->zslice = zslice;
+   pipe_reference_init(&s->base.reference, 1);
+   pipe_texture_reference(&s->base.texture, pt);
+   s->base.format = pt->format;
+   s->base.width = u_minify(pt->width0, level);
+   s->base.height = u_minify(pt->height0, level);
+   s->base.usage = flags;
+   s->base.level = level;
+   s->base.face = face;
+   s->base.zslice = zslice;
 
    if (!render)
       format = svga_translate_format(pt->format);
@@ -617,11 +632,13 @@ svga_get_tex_surface(struct pipe_screen *screen,
       view = TRUE;
 
    /* Currently only used for compressed textures */
-   if (render && (format != svga_translate_format(pt->format))) {
+   if (render && 
+       format != svga_translate_format(pt->format)) {
       view = TRUE;
    }
 
-   if (level != 0 && svga_screen(screen)->debug.force_level_surface_view)
+   if (level != 0 && 
+       svga_screen(screen)->debug.force_level_surface_view)
       view = TRUE;
 
    if (pt->target == PIPE_TEXTURE_3D)
@@ -632,9 +649,10 @@ svga_get_tex_surface(struct pipe_screen *screen,
 
    if (view) {
       SVGA_DBG(DEBUG_VIEWS, "svga: Surface view: yes %p, level %u face %u z %u, %p\n",
-               pt, level, face, zslice, ps);
+               pt, level, face, zslice, s);
 
-      s->handle = svga_texture_view_surface(NULL, tex, format, level, 1, face, zslice);
+      s->handle = svga_texture_view_surface(NULL, tex, format, level, 1, face, zslice,
+                                            &s->key);
       s->real_face = 0;
       s->real_level = 0;
       s->real_zslice = 0;
@@ -642,15 +660,16 @@ svga_get_tex_surface(struct pipe_screen *screen,
       struct svga_winsys_screen *sws = svga_winsys_screen(screen);
 
       SVGA_DBG(DEBUG_VIEWS, "svga: Surface view: no %p, level %u, face %u, z %u, %p\n",
-               pt, level, face, zslice, ps);
+               pt, level, face, zslice, s);
 
+      memset(&s->key, 0, sizeof s->key);
       sws->surface_reference(sws, &s->handle, tex->handle);
       s->real_face = face;
       s->real_level = level;
       s->real_zslice = zslice;
    }
 
-   return ps;
+   return &s->base;
 }
 
 
@@ -661,7 +680,8 @@ svga_tex_surface_destroy(struct pipe_surface *surf)
    struct svga_screen *ss = svga_screen(surf->texture->screen);
 
    SVGA_DBG(DEBUG_DMA, "unref sid %p (tex surface)\n", s->handle);
-   ss->sws->surface_reference(ss->sws, &s->handle, NULL);
+   assert(s->key.cachable == 0);
+   svga_screen_surface_destroy(ss, &s->key, &s->handle);
    pipe_texture_reference(&surf->texture, NULL);
    FREE(surf);
 }
@@ -956,6 +976,7 @@ svga_get_tex_sampler_view(struct pipe_context *pipe, struct pipe_texture *pt,
                pt->height0,
                pt->depth0,
                pt->last_level);
+      sv->key.cachable = 0;
       sws->surface_reference(sws, &sv->handle, tex->handle);
       return sv;
    }
@@ -973,10 +994,12 @@ svga_get_tex_sampler_view(struct pipe_context *pipe, struct pipe_texture *pt,
    sv->handle = svga_texture_view_surface(pipe, tex, format,
                                           min_lod,
                                           max_lod - min_lod + 1,
-                                          -1, -1);
+                                          -1, -1,
+                                          &sv->key);
 
    if (!sv->handle) {
       assert(0);
+      sv->key.cachable = 0;
       sws->surface_reference(sws, &sv->handle, tex->handle);
       return sv;
    }
@@ -1029,7 +1052,7 @@ svga_destroy_sampler_view_priv(struct svga_sampler_view *v)
    struct svga_screen *ss = svga_screen(v->texture->base.screen);
 
    SVGA_DBG(DEBUG_DMA, "unref sid %p (sampler view)\n", v->handle);
-   ss->sws->surface_reference(ss->sws, &v->handle, NULL);
+   svga_screen_surface_destroy(ss, &v->key, &v->handle);
 
    FREE(v);
 }
@@ -1059,6 +1082,8 @@ svga_screen_texture_get_winsys_surface(struct pipe_texture *texture)
    struct svga_winsys_screen *sws = svga_winsys_screen(texture->screen);
    struct svga_winsys_surface *vsurf = NULL;
 
+   assert(svga_texture(texture)->key.cachable == 0);
+   svga_texture(texture)->key.cachable = 0;
    sws->surface_reference(sws, &vsurf, svga_texture(texture)->handle);
    return vsurf;
 }
diff --git a/src/gallium/drivers/svga/svga_screen_texture.h b/src/gallium/drivers/svga/svga_screen_texture.h
index 1e6fef59a39..1cc4063e653 100644
--- a/src/gallium/drivers/svga/svga_screen_texture.h
+++ b/src/gallium/drivers/svga/svga_screen_texture.h
@@ -29,7 +29,7 @@
 
 #include "pipe/p_compiler.h"
 #include "pipe/p_state.h"
-
+#include "svga_screen_cache.h"
 
 struct pipe_context;
 struct pipe_screen;
@@ -68,6 +68,7 @@ struct svga_sampler_view
 
    unsigned age;
 
+   struct svga_host_surface_cache_key key;
    struct svga_winsys_surface *handle;
 };
 
@@ -76,8 +77,6 @@ struct svga_texture
 {
    struct pipe_texture base;
 
-   struct svga_winsys_surface *handle;
-
    boolean defined[6][PIPE_MAX_TEXTURE_LEVELS];
    
    struct svga_sampler_view *cached_view;
@@ -86,6 +85,16 @@ struct svga_texture
    unsigned age;
 
    boolean views_modified;
+
+   /**
+    * Creation key for the host surface handle.
+    * 
+    * This structure describes all the host surface characteristics so that it 
+    * can be looked up in cache, since creating a host surface is often a slow
+    * operation.
+    */
+   struct svga_host_surface_cache_key key;
+   struct svga_winsys_surface *handle;
 };
 
 
@@ -93,6 +102,7 @@ struct svga_surface
 {
    struct pipe_surface base;
 
+   struct svga_host_surface_cache_key key;
    struct svga_winsys_surface *handle;
 
    unsigned real_face;
diff --git a/src/gallium/drivers/svga/svga_state_framebuffer.c b/src/gallium/drivers/svga/svga_state_framebuffer.c
index 7d7f93d8e3c..cfdcae4ee4a 100644
--- a/src/gallium/drivers/svga/svga_state_framebuffer.c
+++ b/src/gallium/drivers/svga/svga_state_framebuffer.c
@@ -54,6 +54,9 @@ static int emit_framebuffer( struct svga_context *svga,
    
    for(i = 0; i < PIPE_MAX_COLOR_BUFS; ++i) {
       if (curr->cbufs[i] != hw->cbufs[i]) {
+         if (svga->curr.nr_fbs++ > 8)
+            return PIPE_ERROR_OUT_OF_MEMORY;
+
          ret = SVGA3D_SetRenderTarget(svga->swc, SVGA3D_RT_COLOR0 + i, curr->cbufs[i]);
          if (ret != PIPE_OK)
             return ret;
diff --git a/src/gallium/drivers/svga/svga_tgsi.c b/src/gallium/drivers/svga/svga_tgsi.c
index 81eea1a145a..b8ef137c015 100644
--- a/src/gallium/drivers/svga/svga_tgsi.c
+++ b/src/gallium/drivers/svga/svga_tgsi.c
@@ -222,6 +222,20 @@ svga_tgsi_translate( const struct svga_shader *shader,
    result->nr_tokens = (emit.ptr - emit.buf) / sizeof(unsigned);
    memcpy(&result->key, &key, sizeof key);
 
+   if (SVGA_DEBUG & DEBUG_TGSI) 
+   {
+      debug_printf( "#####################################\n" );
+      debug_printf( "Shader %u below\n", shader->id );
+      tgsi_dump( shader->tokens, 0 );
+      if (SVGA_DEBUG & DEBUG_TGSI) {
+         debug_printf( "Shader %u compiled below\n", shader->id );
+         svga_shader_dump( result->tokens,
+                           result->nr_tokens ,
+                           FALSE );
+      }
+      debug_printf( "#####################################\n" );
+   }
+
    return result;
 
 fail:
diff --git a/src/gallium/state_trackers/xorg/Makefile b/src/gallium/state_trackers/xorg/Makefile
index 22c107370e8..cb2c3aea410 100644
--- a/src/gallium/state_trackers/xorg/Makefile
+++ b/src/gallium/state_trackers/xorg/Makefile
@@ -7,6 +7,9 @@ LIBRARY_INCLUDES = \
 	-DHAVE_CONFIG_H \
 	$(shell pkg-config xextproto --atleast-version=7.0.99.1 \
 				&& echo "-DHAVE_XEXTPROTO_71") \
+	$(shell pkg-config libkms --atleast-version=1.0 \
+				&& echo "-DHAVE_LIBKMS") \
+	$(shell pkg-config libkms --silence-errors --cflags-only-I) \
 	$(shell pkg-config --cflags-only-I pixman-1 xorg-server libdrm xproto) \
 	-I$(TOP)/src/gallium/include \
 	-I$(TOP)/src/gallium/auxiliary \
diff --git a/src/gallium/state_trackers/xorg/xorg_composite.c b/src/gallium/state_trackers/xorg/xorg_composite.c
index 69a83b15234..1c248a629e6 100644
--- a/src/gallium/state_trackers/xorg/xorg_composite.c
+++ b/src/gallium/state_trackers/xorg/xorg_composite.c
@@ -232,15 +232,25 @@ bind_blend_state(struct exa_context *exa, int op,
 }
 
 static unsigned
-picture_format_fixups(struct exa_pixmap_priv *pSrc, PicturePtr pSrcPicture, boolean mask)
+picture_format_fixups(struct exa_pixmap_priv *pSrc, PicturePtr pSrcPicture, boolean mask,
+                      PicturePtr pDstPicture)
 {
    boolean set_alpha = FALSE;
    boolean swizzle = FALSE;
    unsigned ret = 0;
 
    if (pSrc->picture_format == pSrcPicture->format) {
-      if (pSrc->picture_format == PICT_a8)
-         return mask ? FS_MASK_LUMINANCE : FS_SRC_LUMINANCE;
+      if (pSrc->picture_format == PICT_a8) {
+         if (mask)
+            return FS_MASK_LUMINANCE;
+         else if (pDstPicture->format != PICT_a8) {
+            /* if both dst and src are luminance then
+             * we don't want to swizzle the alpha (X) of the
+             * source into W component of the dst because
+             * it will break our destination */
+            return FS_SRC_LUMINANCE;
+         }
+      }
       return 0;
    }
 
@@ -285,7 +295,7 @@ picture_format_fixups(struct exa_pixmap_priv *pSrc, PicturePtr pSrcPicture, bool
 
 static void
 bind_shaders(struct exa_context *exa, int op,
-             PicturePtr pSrcPicture, PicturePtr pMaskPicture,
+             PicturePtr pSrcPicture, PicturePtr pMaskPicture, PicturePtr pDstPicture,
              struct exa_pixmap_priv *pSrc, struct exa_pixmap_priv *pMask)
 {
    unsigned vs_traits = 0, fs_traits = 0;
@@ -313,7 +323,7 @@ bind_shaders(struct exa_context *exa, int op,
          vs_traits |= VS_COMPOSITE;
       }
 
-      fs_traits |= picture_format_fixups(pSrc, pSrcPicture, FALSE);
+      fs_traits |= picture_format_fixups(pSrc, pSrcPicture, FALSE, pDstPicture);
    }
 
    if (pMaskPicture) {
@@ -331,7 +341,7 @@ bind_shaders(struct exa_context *exa, int op,
             fs_traits |= FS_CA_FULL;
       }
 
-      fs_traits |= picture_format_fixups(pMask, pMaskPicture, TRUE);
+      fs_traits |= picture_format_fixups(pMask, pMaskPicture, TRUE, pDstPicture);
    }
 
    shader = xorg_shaders_get(exa->renderer->shaders, vs_traits, fs_traits);
@@ -411,40 +421,7 @@ bind_samplers(struct exa_context *exa, int op,
                             exa->bound_textures);
 }
 
-static void
-setup_vs_constant_buffer(struct exa_context *exa,
-                         int width, int height)
-{
-   const int param_bytes = 8 * sizeof(float);
-   float vs_consts[8] = {
-      2.f/width, 2.f/height, 1, 1,
-      -1, -1, 0, 0
-   };
-   renderer_set_constants(exa->renderer, PIPE_SHADER_VERTEX,
-                          vs_consts, param_bytes);
-}
-
-
-static void
-setup_fs_constant_buffer(struct exa_context *exa)
-{
-   const int param_bytes = 4 * sizeof(float);
-   const float fs_consts[8] = {
-      0, 0, 0, 1,
-   };
-   renderer_set_constants(exa->renderer, PIPE_SHADER_FRAGMENT,
-                          fs_consts, param_bytes);
-}
-
-static void
-setup_constant_buffers(struct exa_context *exa, struct exa_pixmap_priv *pDst)
-{
-   int width = pDst->tex->width0;
-   int height = pDst->tex->height0;
 
-   setup_vs_constant_buffer(exa, width, height);
-   setup_fs_constant_buffer(exa);
-}
 
 static INLINE boolean matrix_from_pict_transform(PictTransform *trans, float *matrix)
 {
@@ -493,14 +470,16 @@ boolean xorg_composite_bind_state(struct exa_context *exa,
                                   struct exa_pixmap_priv *pMask,
                                   struct exa_pixmap_priv *pDst)
 {
-   renderer_bind_framebuffer(exa->renderer, pDst);
-   renderer_bind_viewport(exa->renderer, pDst);
+   struct pipe_surface *dst_surf = xorg_gpu_surface(exa->scrn, pDst);
+
+   renderer_bind_destination(exa->renderer, dst_surf,
+                             pDst->width,
+                             pDst->height);
+
    bind_blend_state(exa, op, pSrcPicture, pMaskPicture, pDstPicture);
-   renderer_bind_rasterizer(exa->renderer);
-   bind_shaders(exa, op, pSrcPicture, pMaskPicture, pSrc, pMask);
+   bind_shaders(exa, op, pSrcPicture, pMaskPicture, pDstPicture, pSrc, pMask);
    bind_samplers(exa, op, pSrcPicture, pMaskPicture,
                  pDstPicture, pSrc, pMask, pDst);
-   setup_constant_buffers(exa, pDst);
 
    setup_transforms(exa, pSrcPicture, pMaskPicture);
 
@@ -512,6 +491,8 @@ boolean xorg_composite_bind_state(struct exa_context *exa,
                               exa->num_bound_samplers);
    }
 
+
+   pipe_surface_reference(&dst_surf, NULL);
    return TRUE;
 }
 
@@ -546,6 +527,7 @@ boolean xorg_solid_bind_state(struct exa_context *exa,
                               struct exa_pixmap_priv *pixmap,
                               Pixel fg)
 {
+   struct pipe_surface *dst_surf = xorg_gpu_surface(exa->scrn, pixmap);
    unsigned vs_traits, fs_traits;
    struct xorg_shader shader;
 
@@ -563,13 +545,11 @@ boolean xorg_solid_bind_state(struct exa_context *exa,
    vs_traits = VS_SOLID_FILL;
    fs_traits = FS_SOLID_FILL;
 
-   renderer_bind_framebuffer(exa->renderer, pixmap);
-   renderer_bind_viewport(exa->renderer, pixmap);
-   renderer_bind_rasterizer(exa->renderer);
+   renderer_bind_destination(exa->renderer, dst_surf, 
+                             pixmap->width, pixmap->height);
    bind_blend_state(exa, PictOpSrc, NULL, NULL, NULL);
    cso_set_samplers(exa->renderer->cso, 0, NULL);
    cso_set_sampler_textures(exa->renderer->cso, 0, NULL);
-   setup_constant_buffers(exa, pixmap);
 
    shader = xorg_shaders_get(exa->renderer->shaders, vs_traits, fs_traits);
    cso_set_vertex_shader_handle(exa->renderer->cso, shader.vs);
@@ -577,6 +557,7 @@ boolean xorg_solid_bind_state(struct exa_context *exa,
 
    renderer_begin_solid(exa->renderer);
 
+   pipe_surface_reference(&dst_surf, NULL);
    return TRUE;
 }
 
@@ -588,3 +569,13 @@ void xorg_solid(struct exa_context *exa,
                   x0, y0, x1, y1, exa->solid_color);
 }
 
+void
+xorg_composite_done(struct exa_context *exa)
+{
+   renderer_draw_flush(exa->renderer);
+
+   exa->transform.has_src = FALSE;
+   exa->transform.has_mask = FALSE;
+   exa->has_solid_color = FALSE;
+   exa->num_bound_samplers = 0;
+}
diff --git a/src/gallium/state_trackers/xorg/xorg_composite.h b/src/gallium/state_trackers/xorg/xorg_composite.h
index 236addf1ce4..ec71ebfe0dc 100644
--- a/src/gallium/state_trackers/xorg/xorg_composite.h
+++ b/src/gallium/state_trackers/xorg/xorg_composite.h
@@ -29,4 +29,8 @@ void xorg_solid(struct exa_context *exa,
                 struct exa_pixmap_priv *pixmap,
                 int x0, int y0, int x1, int y1);
 
+
+void
+xorg_composite_done(struct exa_context *exa);
+
 #endif
diff --git a/src/gallium/state_trackers/xorg/xorg_crtc.c b/src/gallium/state_trackers/xorg/xorg_crtc.c
index c4751724c99..9e8c14d741c 100644
--- a/src/gallium/state_trackers/xorg/xorg_crtc.c
+++ b/src/gallium/state_trackers/xorg/xorg_crtc.c
@@ -52,12 +52,18 @@
 #include "pipe/p_inlines.h"
 #include "util/u_rect.h"
 
+#ifdef HAVE_LIBKMS
+#include "libkms.h"
+#endif
+
 struct crtc_private
 {
     drmModeCrtcPtr drm_crtc;
 
     /* hwcursor */
     struct pipe_texture *cursor_tex;
+    struct kms_bo *cursor_bo;
+
     unsigned cursor_handle;
 };
 
@@ -134,6 +140,7 @@ static void
 crtc_gamma_set(xf86CrtcPtr crtc, CARD16 * red, CARD16 * green, CARD16 * blue,
 	       int size)
 {
+    /* XXX: hockup */
 }
 
 static void *
@@ -160,6 +167,7 @@ crtc_shadow_destroy(xf86CrtcPtr crtc, PixmapPtr rotate_pixmap, void *data)
 static void
 crtc_set_cursor_colors(xf86CrtcPtr crtc, int bg, int fg)
 {
+    /* XXX: See if this one is needed, as we only support ARGB cursors */
 }
 
 static void
@@ -170,8 +178,9 @@ crtc_set_cursor_position(xf86CrtcPtr crtc, int x, int y)
 
     drmModeMoveCursor(ms->fd, crtcp->drm_crtc->crtc_id, x, y);
 }
+
 static void
-crtc_load_cursor_argb(xf86CrtcPtr crtc, CARD32 * image)
+crtc_load_cursor_argb_ga3d(xf86CrtcPtr crtc, CARD32 * image)
 {
     unsigned char *ptr;
     modesettingPtr ms = modesettingPTR(crtc->scrn);
@@ -214,13 +223,63 @@ crtc_load_cursor_argb(xf86CrtcPtr crtc, CARD32 * image)
     ms->screen->tex_transfer_destroy(transfer);
 }
 
+#if HAVE_LIBKMS
+static void
+crtc_load_cursor_argb_kms(xf86CrtcPtr crtc, CARD32 * image)
+{
+    modesettingPtr ms = modesettingPTR(crtc->scrn);
+    struct crtc_private *crtcp = crtc->driver_private;
+    unsigned char *ptr;
+
+    if (!crtcp->cursor_bo) {
+	unsigned attr[8];
+
+	attr[0] = KMS_BO_TYPE;
+	attr[1] = KMS_BO_TYPE_CURSOR;
+	attr[2] = KMS_WIDTH;
+	attr[3] = 64;
+	attr[4] = KMS_HEIGHT;
+	attr[5] = 64;
+	attr[6] = 0;
+
+        if (kms_bo_create(ms->kms, attr, &crtcp->cursor_bo))
+	   return;
+
+	if (kms_bo_get_prop(crtcp->cursor_bo, KMS_HANDLE,
+			    &crtcp->cursor_handle))
+	    goto err_bo_destroy;
+    }
+
+    kms_bo_map(crtcp->cursor_bo, (void**)&ptr);
+    memcpy(ptr, image, 64*64*4);
+    kms_bo_unmap(crtcp->cursor_bo);
+
+    return;
+
+err_bo_destroy:
+    kms_bo_destroy(crtcp->cursor_bo);
+}
+#endif
+
+static void
+crtc_load_cursor_argb(xf86CrtcPtr crtc, CARD32 * image)
+{
+    modesettingPtr ms = modesettingPTR(crtc->scrn);
+    if (ms->screen)
+	crtc_load_cursor_argb_ga3d(crtc, image);
+#ifdef HAVE_LIBKMS
+    else if (ms->kms)
+	crtc_load_cursor_argb_kms(crtc, image);
+#endif
+}
+
 static void
 crtc_show_cursor(xf86CrtcPtr crtc)
 {
     modesettingPtr ms = modesettingPTR(crtc->scrn);
     struct crtc_private *crtcp = crtc->driver_private;
 
-    if (crtcp->cursor_tex)
+    if (crtcp->cursor_tex || crtcp->cursor_bo)
 	drmModeSetCursor(ms->fd, crtcp->drm_crtc->crtc_id,
 			 crtcp->cursor_handle, 64, 64);
 }
@@ -234,14 +293,22 @@ crtc_hide_cursor(xf86CrtcPtr crtc)
     drmModeSetCursor(ms->fd, crtcp->drm_crtc->crtc_id, 0, 0, 0);
 }
 
+/**
+ * Called at vt leave
+ */
 void
-crtc_cursor_destroy(xf86CrtcPtr crtc)
+xorg_crtc_cursor_destroy(xf86CrtcPtr crtc)
 {
     struct crtc_private *crtcp = crtc->driver_private;
 
-    if (crtcp->cursor_tex) {
+    if (crtcp->cursor_tex)
 	pipe_texture_reference(&crtcp->cursor_tex, NULL);
-    }
+#ifdef HAVE_LIBKMS
+    if (crtcp->cursor_bo)
+	kms_bo_destroy(crtcp->cursor_bo);
+#endif
+
+    xfree(crtcp);
 }
 
 /*
@@ -279,7 +346,7 @@ static const xf86CrtcFuncsRec crtc_funcs = {
 };
 
 void
-crtc_init(ScrnInfoPtr pScrn)
+xorg_crtc_init(ScrnInfoPtr pScrn)
 {
     modesettingPtr ms = modesettingPTR(pScrn);
     xf86CrtcPtr crtc;
diff --git a/src/gallium/state_trackers/xorg/xorg_dri2.c b/src/gallium/state_trackers/xorg/xorg_dri2.c
index 406e0afff4b..2394f004d25 100644
--- a/src/gallium/state_trackers/xorg/xorg_dri2.c
+++ b/src/gallium/state_trackers/xorg/xorg_dri2.c
@@ -55,7 +55,7 @@ typedef struct {
 } *BufferPrivatePtr;
 
 static Bool
-driDoCreateBuffer(DrawablePtr pDraw, DRI2BufferPtr buffer, unsigned int format)
+dri2_do_create_buffer(DrawablePtr pDraw, DRI2BufferPtr buffer, unsigned int format)
 {
     struct pipe_texture *tex = NULL;
     ScreenPtr pScreen = pDraw->pScreen;
@@ -157,7 +157,7 @@ driDoCreateBuffer(DrawablePtr pDraw, DRI2BufferPtr buffer, unsigned int format)
 }
 
 static void
-driDoDestroyBuffer(DrawablePtr pDraw, DRI2BufferPtr buffer)
+dri2_do_destroy_buffer(DrawablePtr pDraw, DRI2BufferPtr buffer)
 {
     ScreenPtr pScreen = pDraw->pScreen;
     ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
@@ -174,7 +174,7 @@ driDoDestroyBuffer(DrawablePtr pDraw, DRI2BufferPtr buffer)
 #if DRI2INFOREC_VERSION >= 2
 
 static DRI2Buffer2Ptr
-driCreateBuffer(DrawablePtr pDraw, unsigned int attachment, unsigned int format)
+dri2_create_buffer(DrawablePtr pDraw, unsigned int attachment, unsigned int format)
 {
     DRI2Buffer2Ptr buffer;
     BufferPrivatePtr private;
@@ -192,7 +192,7 @@ driCreateBuffer(DrawablePtr pDraw, unsigned int attachment, unsigned int format)
     buffer->driverPrivate = private;
 
     /* So far it is safe to downcast a DRI2Buffer2Ptr to DRI2BufferPtr */
-    if (driDoCreateBuffer(pDraw, (DRI2BufferPtr)buffer, format))
+    if (dri2_do_create_buffer(pDraw, (DRI2BufferPtr)buffer, format))
 	return buffer;
 
     xfree(private);
@@ -202,10 +202,10 @@ fail:
 }
 
 static void
-driDestroyBuffer(DrawablePtr pDraw, DRI2Buffer2Ptr buffer)
+dri2_destroy_buffer(DrawablePtr pDraw, DRI2Buffer2Ptr buffer)
 {
     /* So far it is safe to downcast a DRI2Buffer2Ptr to DRI2BufferPtr */
-    driDoDestroyBuffer(pDraw, (DRI2BufferPtr)buffer);
+    dri2_do_destroy_buffer(pDraw, (DRI2BufferPtr)buffer);
 
     xfree(buffer->driverPrivate);
     xfree(buffer);
@@ -214,7 +214,7 @@ driDestroyBuffer(DrawablePtr pDraw, DRI2Buffer2Ptr buffer)
 #else /* DRI2INFOREC_VERSION < 2 */
 
 static DRI2BufferPtr
-driCreateBuffers(DrawablePtr pDraw, unsigned int *attachments, int count)
+dri2_create_buffers(DrawablePtr pDraw, unsigned int *attachments, int count)
 {
     BufferPrivatePtr privates;
     DRI2BufferPtr buffers;
@@ -232,7 +232,7 @@ driCreateBuffers(DrawablePtr pDraw, unsigned int *attachments, int count)
 	buffers[i].attachment = attachments[i];
 	buffers[i].driverPrivate = &privates[i];
 
-	if (!driDoCreateBuffer(pDraw, &buffers[i], 0))
+	if (!dri2_do_create_buffer(pDraw, &buffers[i], 0))
 	    goto fail;
     }
 
@@ -247,12 +247,12 @@ fail_buffers:
 }
 
 static void
-driDestroyBuffers(DrawablePtr pDraw, DRI2BufferPtr buffers, int count)
+dri2_destroy_buffers(DrawablePtr pDraw, DRI2BufferPtr buffers, int count)
 {
     int i;
 
     for (i = 0; i < count; i++) {
-	driDoDestroyBuffer(pDraw, &buffers[i]);
+	dri2_do_destroy_buffer(pDraw, &buffers[i]);
     }
 
     if (buffers) {
@@ -264,16 +264,16 @@ driDestroyBuffers(DrawablePtr pDraw, DRI2BufferPtr buffers, int count)
 #endif /* DRI2INFOREC_VERSION >= 2 */
 
 static void
-driCopyRegion(DrawablePtr pDraw, RegionPtr pRegion,
-              DRI2BufferPtr pDestBuffer, DRI2BufferPtr pSrcBuffer)
+dri2_copy_region(DrawablePtr pDraw, RegionPtr pRegion,
+                 DRI2BufferPtr pDestBuffer, DRI2BufferPtr pSrcBuffer)
 {
     ScreenPtr pScreen = pDraw->pScreen;
     ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
     BufferPrivatePtr dst_priv = pDestBuffer->driverPrivate;
     BufferPrivatePtr src_priv = pSrcBuffer->driverPrivate;
-    PixmapPtr src_pixmap;
-    PixmapPtr dst_pixmap;
+    DrawablePtr src_draw;
+    DrawablePtr dst_draw;
     GCPtr gc;
     RegionPtr copy_clip;
     Bool save_accel;
@@ -284,12 +284,10 @@ driCopyRegion(DrawablePtr pDraw, RegionPtr pRegion,
      * We need to use the real drawable in CopyArea
      * so that cliprects and offsets are correct.
      */
-    src_pixmap = src_priv->pPixmap;
-    dst_pixmap = dst_priv->pPixmap;
-    if (pSrcBuffer->attachment == DRI2BufferFrontLeft)
-	src_pixmap = (PixmapPtr)pDraw;
-    if (pDestBuffer->attachment == DRI2BufferFrontLeft)
-	dst_pixmap = (PixmapPtr)pDraw;
+    src_draw = (pSrcBuffer->attachment == DRI2BufferFrontLeft) ? pDraw :
+       &src_priv->pPixmap->drawable;
+    dst_draw = (pDestBuffer->attachment == DRI2BufferFrontLeft) ? pDraw :
+       &dst_priv->pPixmap->drawable;
 
     /*
      * The clients implements glXWaitX with a copy front to fake and then
@@ -308,7 +306,7 @@ driCopyRegion(DrawablePtr pDraw, RegionPtr pRegion,
      * must in the glXWaitGL case but we don't know if this is a glXWaitGL
      * or a glFlush/glFinish call.
      */
-    if (dst_pixmap == src_pixmap) {
+    if (dst_priv->pPixmap == src_priv->pPixmap) {
 	/* pixmap glXWaitX */
 	if (pSrcBuffer->attachment == DRI2BufferFrontLeft &&
 	    pDestBuffer->attachment == DRI2BufferFakeFrontLeft) {
@@ -329,7 +327,7 @@ driCopyRegion(DrawablePtr pDraw, RegionPtr pRegion,
     copy_clip = REGION_CREATE(pScreen, NULL, 0);
     REGION_COPY(pScreen, copy_clip, pRegion);
     (*gc->funcs->ChangeClip) (gc, CT_REGION, copy_clip, 0);
-    ValidateGC(&dst_pixmap->drawable, gc);
+    ValidateGC(dst_draw, gc);
 
     /* If this is a full buffer swap, throttle on the previous one */
     if (dst_priv->fence && REGION_NUM_RECTS(pRegion) == 1) {
@@ -342,9 +340,19 @@ driCopyRegion(DrawablePtr pDraw, RegionPtr pRegion,
 	}
     }
 
+    /* Try to make sure the blit will be accelerated */
     save_accel = ms->exa->accel;
     ms->exa->accel = TRUE;
-    (*gc->ops->CopyArea)(&src_pixmap->drawable, &dst_pixmap->drawable, gc,
+
+    /* In case it won't be though, make sure the GPU copy contents of the
+     * source pixmap will be used for the software fallback - presumably the
+     * client modified them before calling in here.
+     */
+    exaMoveInPixmap(src_priv->pPixmap);
+    DamageRegionAppend(src_draw, pRegion);
+    DamageRegionProcessPending(src_draw);
+
+    (*gc->ops->CopyArea)(src_draw, dst_draw, gc,
 			 0, 0, pDraw->width, pDraw->height, 0, 0);
     ms->exa->accel = save_accel;
 
@@ -356,7 +364,7 @@ driCopyRegion(DrawablePtr pDraw, RegionPtr pRegion,
 }
 
 Bool
-driScreenInit(ScreenPtr pScreen)
+xorg_dri2_init(ScreenPtr pScreen)
 {
     ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
@@ -373,13 +381,13 @@ driScreenInit(ScreenPtr pScreen)
     dri2info.deviceName = "/dev/dri/card0"; /* FIXME */
 
 #if DRI2INFOREC_VERSION >= 2
-    dri2info.CreateBuffer = driCreateBuffer;
-    dri2info.DestroyBuffer = driDestroyBuffer;
+    dri2info.CreateBuffer = dri2_create_buffer;
+    dri2info.DestroyBuffer = dri2_destroy_buffer;
 #else
-    dri2info.CreateBuffers = driCreateBuffers;
-    dri2info.DestroyBuffers = driDestroyBuffers;
+    dri2info.CreateBuffers = dri2_create_buffers;
+    dri2info.DestroyBuffers = dri2_destroy_buffers;
 #endif
-    dri2info.CopyRegion = driCopyRegion;
+    dri2info.CopyRegion = dri2_copy_region;
     dri2info.Wait = NULL;
 
     ms->d_depth_bits_last =
@@ -395,7 +403,7 @@ driScreenInit(ScreenPtr pScreen)
 }
 
 void
-driCloseScreen(ScreenPtr pScreen)
+xorg_dri2_close(ScreenPtr pScreen)
 {
     DRI2CloseScreen(pScreen);
 }
diff --git a/src/gallium/state_trackers/xorg/xorg_driver.c b/src/gallium/state_trackers/xorg/xorg_driver.c
index d949167adce..12915912986 100644
--- a/src/gallium/state_trackers/xorg/xorg_driver.c
+++ b/src/gallium/state_trackers/xorg/xorg_driver.c
@@ -56,34 +56,38 @@
 #include "xorg_tracker.h"
 #include "xorg_winsys.h"
 
-static void AdjustFrame(int scrnIndex, int x, int y, int flags);
-static Bool CloseScreen(int scrnIndex, ScreenPtr pScreen);
-static Bool EnterVT(int scrnIndex, int flags);
-static Bool SaveHWState(ScrnInfoPtr pScrn);
-static Bool RestoreHWState(ScrnInfoPtr pScrn);
-
-
-static ModeStatus ValidMode(int scrnIndex, DisplayModePtr mode, Bool verbose,
-			    int flags);
-static void FreeScreen(int scrnIndex, int flags);
-static void LeaveVT(int scrnIndex, int flags);
-static Bool SwitchMode(int scrnIndex, DisplayModePtr mode, int flags);
-static Bool ScreenInit(int scrnIndex, ScreenPtr pScreen, int argc,
-		       char **argv);
-static Bool PreInit(ScrnInfoPtr pScrn, int flags);
+#ifdef HAVE_LIBKMS
+#include "libkms.h"
+#endif
+
+/*
+ * Functions and symbols exported to Xorg via pointers.
+ */
+
+static Bool drv_pre_init(ScrnInfoPtr pScrn, int flags);
+static Bool drv_screen_init(int scrnIndex, ScreenPtr pScreen, int argc,
+			    char **argv);
+static Bool drv_switch_mode(int scrnIndex, DisplayModePtr mode, int flags);
+static void drv_adjust_frame(int scrnIndex, int x, int y, int flags);
+static Bool drv_enter_vt(int scrnIndex, int flags);
+static void drv_leave_vt(int scrnIndex, int flags);
+static void drv_free_screen(int scrnIndex, int flags);
+static ModeStatus drv_valid_mode(int scrnIndex, DisplayModePtr mode, Bool verbose,
+			         int flags);
 
 typedef enum
 {
     OPTION_SW_CURSOR,
     OPTION_2D_ACCEL,
-} modesettingOpts;
+} drv_option_enums;
 
-static const OptionInfoRec Options[] = {
+static const OptionInfoRec drv_options[] = {
     {OPTION_SW_CURSOR, "SWcursor", OPTV_BOOLEAN, {0}, FALSE},
     {OPTION_2D_ACCEL, "2DAccel", OPTV_BOOLEAN, {0}, FALSE},
     {-1, NULL, OPTV_NONE, {0}, FALSE}
 };
 
+
 /*
  * Exported Xorg driver functions to winsys
  */
@@ -91,28 +95,39 @@ static const OptionInfoRec Options[] = {
 const OptionInfoRec *
 xorg_tracker_available_options(int chipid, int busid)
 {
-    return Options;
+    return drv_options;
 }
 
 void
 xorg_tracker_set_functions(ScrnInfoPtr scrn)
 {
-    scrn->PreInit = PreInit;
-    scrn->ScreenInit = ScreenInit;
-    scrn->SwitchMode = SwitchMode;
-    scrn->AdjustFrame = AdjustFrame;
-    scrn->EnterVT = EnterVT;
-    scrn->LeaveVT = LeaveVT;
-    scrn->FreeScreen = FreeScreen;
-    scrn->ValidMode = ValidMode;
+    scrn->PreInit = drv_pre_init;
+    scrn->ScreenInit = drv_screen_init;
+    scrn->SwitchMode = drv_switch_mode;
+    scrn->AdjustFrame = drv_adjust_frame;
+    scrn->EnterVT = drv_enter_vt;
+    scrn->LeaveVT = drv_leave_vt;
+    scrn->FreeScreen = drv_free_screen;
+    scrn->ValidMode = drv_valid_mode;
 }
 
+
 /*
- * Static Xorg funtctions
+ * Internal function definitions
+ */
+
+static Bool drv_init_front_buffer_functions(ScrnInfoPtr pScrn);
+static Bool drv_close_screen(int scrnIndex, ScreenPtr pScreen);
+static Bool drv_save_hw_state(ScrnInfoPtr pScrn);
+static Bool drv_restore_hw_state(ScrnInfoPtr pScrn);
+
+
+/*
+ * Internal functions
  */
 
 static Bool
-GetRec(ScrnInfoPtr pScrn)
+drv_get_rec(ScrnInfoPtr pScrn)
 {
     if (pScrn->driverPrivate)
 	return TRUE;
@@ -123,7 +138,7 @@ GetRec(ScrnInfoPtr pScrn)
 }
 
 static void
-FreeRec(ScrnInfoPtr pScrn)
+drv_free_rec(ScrnInfoPtr pScrn)
 {
     if (!pScrn)
 	return;
@@ -137,88 +152,21 @@ FreeRec(ScrnInfoPtr pScrn)
 }
 
 static void
-ProbeDDC(ScrnInfoPtr pScrn, int index)
+drv_probe_ddc(ScrnInfoPtr pScrn, int index)
 {
     ConfiguredMonitor = NULL;
 }
 
 static Bool
-CreateFrontBuffer(ScrnInfoPtr pScrn)
-{
-    modesettingPtr ms = modesettingPTR(pScrn);
-    unsigned handle, stride;
-    struct pipe_texture *tex;
-
-    ms->noEvict = TRUE;
-
-    tex = xorg_exa_create_root_texture(pScrn, pScrn->virtualX, pScrn->virtualY,
-				       pScrn->depth, pScrn->bitsPerPixel);
-
-    if (!tex)
-	return FALSE;
-
-    if (!ms->api->local_handle_from_texture(ms->api, ms->screen,
-					    tex,
-					    &stride,
-					    &handle))
-	return FALSE;
-
-    drmModeAddFB(ms->fd,
-		 pScrn->virtualX,
-		 pScrn->virtualY,
-		 pScrn->depth,
-		 pScrn->bitsPerPixel,
-		 stride,
-		 handle,
-                 &ms->fb_id);
-
-    pScrn->frameX0 = 0;
-    pScrn->frameY0 = 0;
-    AdjustFrame(pScrn->scrnIndex, pScrn->frameX0, pScrn->frameY0, 0);
-
-    pipe_texture_reference(&ms->root_texture, tex);
-    pipe_texture_reference(&tex, NULL);
-    return TRUE;
-}
-
-static Bool
-BindTextureToRoot(ScrnInfoPtr pScrn)
-{
-    modesettingPtr ms = modesettingPTR(pScrn);
-    ScreenPtr pScreen = pScrn->pScreen;
-    struct pipe_texture *check;
-    PixmapPtr rootPixmap;
-
-    rootPixmap = pScreen->GetScreenPixmap(pScreen);
-
-    xorg_exa_set_displayed_usage(rootPixmap);
-    xorg_exa_set_shared_usage(rootPixmap);
-    xorg_exa_set_texture(rootPixmap, ms->root_texture);
-    if (!pScreen->ModifyPixmapHeader(rootPixmap, -1, -1, -1, -1, -1, NULL))
-	FatalError("Couldn't adjust screen pixmap\n");
-
-    check = xorg_exa_get_texture(rootPixmap);
-    if (ms->root_texture != check)
-	FatalError("Created new root texture\n");
-
-    pipe_texture_reference(&check, NULL);
-
-    return TRUE;
-}
-
-static Bool
-crtc_resize(ScrnInfoPtr pScrn, int width, int height)
+drv_crtc_resize(ScrnInfoPtr pScrn, int width, int height)
 {
     modesettingPtr ms = modesettingPTR(pScrn);
-    unsigned handle, stride;
     PixmapPtr rootPixmap;
     ScreenPtr pScreen = pScrn->pScreen;
 
     if (width == pScrn->virtualX && height == pScrn->virtualY)
 	return TRUE;
 
-    ErrorF("RESIZING TO %dx%d\n", width, height);
-
     pScrn->virtualX = width;
     pScrn->virtualY = height;
 
@@ -226,44 +174,26 @@ crtc_resize(ScrnInfoPtr pScrn, int width, int height)
      * Remove the old framebuffer & texture.
      */
     drmModeRmFB(ms->fd, ms->fb_id);
-    pipe_texture_reference(&ms->root_texture, NULL);
-
+    if (!ms->destroy_front_buffer(pScrn))
+	FatalError("failed to destroy front buffer\n");
 
     rootPixmap = pScreen->GetScreenPixmap(pScreen);
     if (!pScreen->ModifyPixmapHeader(rootPixmap, width, height, -1, -1, -1, NULL))
 	return FALSE;
 
-    /* takes one ref */
-    ms->root_texture = xorg_exa_get_texture(rootPixmap);
-
-    if (!ms->api->local_handle_from_texture(ms->api, ms->screen,
-					    ms->root_texture,
-					    &stride,
-					    &handle))
-	FatalError("Could not get handle and stride from texture\n");
-
-    drmModeAddFB(ms->fd,
-		 pScrn->virtualX,
-		 pScrn->virtualY,
-		 pScrn->depth,
-		 pScrn->bitsPerPixel,
-		 stride,
-		 handle,
-                 &ms->fb_id);
-
     /* HW dependent - FIXME */
     pScrn->displayWidth = pScrn->virtualX;
 
     /* now create new frontbuffer */
-    return CreateFrontBuffer(pScrn) && BindTextureToRoot(pScrn);
+    return ms->create_front_buffer(pScrn) && ms->bind_front_buffer(pScrn);
 }
 
 static const xf86CrtcConfigFuncsRec crtc_config_funcs = {
-    crtc_resize
+    .resize = drv_crtc_resize
 };
 
 static Bool
-InitDRM(ScrnInfoPtr pScrn)
+drv_init_drm(ScrnInfoPtr pScrn)
 {
     modesettingPtr ms = modesettingPTR(pScrn);
 
@@ -283,18 +213,62 @@ InitDRM(ScrnInfoPtr pScrn)
 	    return FALSE;
     }
 
-    if (!ms->api) {
-	ms->api = drm_api_create();
+    return TRUE;
+}
 
-	if (!ms->api)
-	    return FALSE;
+static Bool
+drv_init_resource_management(ScrnInfoPtr pScrn)
+{
+    modesettingPtr ms = modesettingPTR(pScrn);
+
+    if (ms->screen || ms->kms)
+	return TRUE;
+
+    ms->api = drm_api_create();
+    if (ms->api) {
+	ms->screen = ms->api->create_screen(ms->api, ms->fd, NULL);
+
+	if (ms->screen)
+	    return TRUE;
+
+	if (ms->api->destroy)
+	    ms->api->destroy(ms->api);
+
+	ms->api = NULL;
     }
 
+#ifdef HAVE_LIBKMS
+    if (!kms_create(ms->fd, &ms->kms))
+	return TRUE;
+#endif
+
+    return FALSE;
+}
+
+static Bool
+drv_close_resource_management(ScrnInfoPtr pScrn)
+{
+    modesettingPtr ms = modesettingPTR(pScrn);
+
+    if (ms->screen)
+	ms->screen->destroy(ms->screen);
+    ms->screen = NULL;
+
+    if (ms->api && ms->api->destroy)
+	ms->api->destroy(ms->api);
+    ms->api = NULL;
+
+#ifdef HAVE_LIBKMS
+    if (ms->kms)
+	kms_destroy(ms->kms);
+    ms->kms = NULL;
+#endif
+
     return TRUE;
 }
 
 static Bool
-PreInit(ScrnInfoPtr pScrn, int flags)
+drv_pre_init(ScrnInfoPtr pScrn, int flags)
 {
     xf86CrtcConfigPtr xf86_config;
     modesettingPtr ms;
@@ -309,12 +283,12 @@ PreInit(ScrnInfoPtr pScrn, int flags)
     pEnt = xf86GetEntityInfo(pScrn->entityList[0]);
 
     if (flags & PROBE_DETECT) {
-	ProbeDDC(pScrn, pEnt->index);
+	drv_probe_ddc(pScrn, pEnt->index);
 	return TRUE;
     }
 
     /* Allocate driverPrivate */
-    if (!GetRec(pScrn))
+    if (!drv_get_rec(pScrn))
 	return FALSE;
 
     ms = modesettingPTR(pScrn);
@@ -351,7 +325,7 @@ PreInit(ScrnInfoPtr pScrn, int flags)
 
     ms->fd = -1;
     ms->api = NULL;
-    if (!InitDRM(pScrn))
+    if (!drv_init_drm(pScrn))
 	return FALSE;
 
     pScrn->monitor = pScrn->confScreen->monitor;
@@ -383,9 +357,9 @@ PreInit(ScrnInfoPtr pScrn, int flags)
 
     /* Process the options */
     xf86CollectOptions(pScrn, NULL);
-    if (!(ms->Options = xalloc(sizeof(Options))))
+    if (!(ms->Options = xalloc(sizeof(drv_options))))
 	return FALSE;
-    memcpy(ms->Options, Options, sizeof(Options));
+    memcpy(ms->Options, drv_options, sizeof(drv_options));
     xf86ProcessOptions(pScrn->scrnIndex, pScrn->options, ms->Options);
 
     /* Allocate an xf86CrtcConfig */
@@ -400,18 +374,18 @@ PreInit(ScrnInfoPtr pScrn, int flags)
 	ms->SWCursor = TRUE;
     }
 
-    SaveHWState(pScrn);
+    drv_save_hw_state(pScrn);
 
-    crtc_init(pScrn);
-    output_init(pScrn);
+    xorg_crtc_init(pScrn);
+    xorg_output_init(pScrn);
 
     if (!xf86InitialConfiguration(pScrn, TRUE)) {
 	xf86DrvMsg(pScrn->scrnIndex, X_ERROR, "No valid modes.\n");
-	RestoreHWState(pScrn);
+	drv_restore_hw_state(pScrn);
 	return FALSE;
     }
 
-    RestoreHWState(pScrn);
+    drv_restore_hw_state(pScrn);
 
     /*
      * If the driver can do gamma correction, it should call xf86SetGamma() here.
@@ -435,21 +409,23 @@ PreInit(ScrnInfoPtr pScrn, int flags)
     xf86SetDpi(pScrn, 0, 0);
 
     /* Load the required sub modules */
-    if (!xf86LoadSubModule(pScrn, "fb")) {
+    if (!xf86LoadSubModule(pScrn, "fb"))
 	return FALSE;
-    }
 
-    xf86LoadSubModule(pScrn, "exa");
+    /* XXX: these aren't needed when we are using libkms */
+    if (!xf86LoadSubModule(pScrn, "exa"))
+	return FALSE;
 
 #ifdef DRI2
-    xf86LoadSubModule(pScrn, "dri2");
+    if (!xf86LoadSubModule(pScrn, "dri2"))
+	return FALSE;
 #endif
 
     return TRUE;
 }
 
 static Bool
-SaveHWState(ScrnInfoPtr pScrn)
+drv_save_hw_state(ScrnInfoPtr pScrn)
 {
     /*xf86CrtcConfigPtr xf86_config = XF86_CRTC_CONFIG_PTR(pScrn);*/
 
@@ -457,24 +433,45 @@ SaveHWState(ScrnInfoPtr pScrn)
 }
 
 static Bool
-RestoreHWState(ScrnInfoPtr pScrn)
+drv_restore_hw_state(ScrnInfoPtr pScrn)
 {
     /*xf86CrtcConfigPtr config = XF86_CRTC_CONFIG_PTR(pScrn);*/
 
     return TRUE;
 }
 
-static void xorgBlockHandler(int i, pointer blockData, pointer pTimeout,
-                             pointer pReadmask)
+static void drv_block_handler(int i, pointer blockData, pointer pTimeout,
+                              pointer pReadmask)
 {
     ScreenPtr pScreen = screenInfo.screens[i];
     modesettingPtr ms = modesettingPTR(xf86Screens[pScreen->myNum]);
 
     pScreen->BlockHandler = ms->blockHandler;
     pScreen->BlockHandler(i, blockData, pTimeout, pReadmask);
-    pScreen->BlockHandler = xorgBlockHandler;
-
-    ms->ctx->flush(ms->ctx, PIPE_FLUSH_RENDER_CACHE, NULL);
+    pScreen->BlockHandler = drv_block_handler;
+
+    if (ms->ctx) {
+       int j;
+
+       ms->ctx->flush(ms->ctx, PIPE_FLUSH_RENDER_CACHE, &ms->fence[XORG_NR_FENCES-1]);
+       
+       if (ms->fence[0])
+          ms->ctx->screen->fence_finish(ms->ctx->screen, ms->fence[0], 0);
+  
+       /* The amount of rendering generated by a block handler can be
+        * quite small.  Let us get a fair way ahead of hardware before
+        * throttling.
+        */
+       for (j = 0; j < XORG_NR_FENCES; j++)
+          ms->screen->fence_reference(ms->screen,
+                                      &ms->fence[j],
+                                      ms->fence[j+1]);
+
+       ms->screen->fence_reference(ms->screen,
+                                   &ms->fence[XORG_NR_FENCES-1],
+                                   NULL);
+    }
+        
 
 #ifdef DRM_MODE_FEATURE_DIRTYFB
     {
@@ -504,7 +501,7 @@ static void xorgBlockHandler(int i, pointer blockData, pointer pTimeout,
 }
 
 static Bool
-CreateScreenResources(ScreenPtr pScreen)
+drv_create_screen_resources(ScreenPtr pScreen)
 {
     ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
@@ -515,13 +512,13 @@ CreateScreenResources(ScreenPtr pScreen)
 
     pScreen->CreateScreenResources = ms->createScreenResources;
     ret = pScreen->CreateScreenResources(pScreen);
-    pScreen->CreateScreenResources = CreateScreenResources;
+    pScreen->CreateScreenResources = drv_create_screen_resources;
 
-    BindTextureToRoot(pScrn);
+    ms->bind_front_buffer(pScrn);
 
     ms->noEvict = FALSE;
 
-    AdjustFrame(pScrn->scrnIndex, pScrn->frameX0, pScrn->frameY0, 0);
+    drv_adjust_frame(pScrn->scrnIndex, pScrn->frameX0, pScrn->frameY0, 0);
 
 #ifdef DRM_MODE_FEATURE_DIRTYFB
     rootPixmap = pScreen->GetScreenPixmap(pScreen);
@@ -545,22 +542,25 @@ CreateScreenResources(ScreenPtr pScreen)
 }
 
 static Bool
-ScreenInit(int scrnIndex, ScreenPtr pScreen, int argc, char **argv)
+drv_screen_init(int scrnIndex, ScreenPtr pScreen, int argc, char **argv)
 {
     ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
     modesettingPtr ms = modesettingPTR(pScrn);
     VisualPtr visual;
 
-    if (!InitDRM(pScrn))
+    if (!drv_init_drm(pScrn)) {
+	FatalError("Could not init DRM");
 	return FALSE;
+    }
 
-    if (!ms->screen) {
-	ms->screen = ms->api->create_screen(ms->api, ms->fd, NULL);
+    if (!drv_init_resource_management(pScrn)) {
+	FatalError("Could not init resource management (!pipe_screen && !libkms)");
+	return FALSE;
+    }
 
-	if (!ms->screen) {
-	    FatalError("Could not init pipe_screen\n");
-	    return FALSE;
-	}
+    if (!drv_init_front_buffer_functions(pScrn)) {
+	FatalError("Could not init front buffer manager");
+	return FALSE;
     }
 
     pScrn->pScreen = pScreen;
@@ -605,17 +605,22 @@ ScreenInit(int scrnIndex, ScreenPtr pScreen, int argc, char **argv)
     fbPictureInit(pScreen, NULL, 0);
 
     ms->blockHandler = pScreen->BlockHandler;
-    pScreen->BlockHandler = xorgBlockHandler;
+    pScreen->BlockHandler = drv_block_handler;
     ms->createScreenResources = pScreen->CreateScreenResources;
-    pScreen->CreateScreenResources = CreateScreenResources;
+    pScreen->CreateScreenResources = drv_create_screen_resources;
 
     xf86SetBlackWhitePixels(pScreen);
 
-    ms->exa = xorg_exa_init(pScrn, xf86ReturnOptValBool(ms->Options,
-                                                        OPTION_2D_ACCEL, TRUE));
-    ms->debug_fallback = debug_get_bool_option("XORG_DEBUG_FALLBACK", TRUE);
+    if (ms->screen) {
+	ms->exa = xorg_exa_init(pScrn, xf86ReturnOptValBool(ms->Options,
+							    OPTION_2D_ACCEL, TRUE));
+	ms->debug_fallback = debug_get_bool_option("XORG_DEBUG_FALLBACK", TRUE);
 
-    xorg_init_video(pScreen);
+	xorg_xv_init(pScreen);
+#ifdef DRI2
+	xorg_dri2_init(pScreen);
+#endif
+    }
 
     miInitializeBackingStore(pScreen);
     xf86SetBackingStore(pScreen);
@@ -634,7 +639,7 @@ ScreenInit(int scrnIndex, ScreenPtr pScreen, int argc, char **argv)
 
     pScreen->SaveScreen = xf86SaveScreen;
     ms->CloseScreen = pScreen->CloseScreen;
-    pScreen->CloseScreen = CloseScreen;
+    pScreen->CloseScreen = drv_close_screen;
 
     if (!xf86CrtcScreenInit(pScreen))
 	return FALSE;
@@ -647,17 +652,11 @@ ScreenInit(int scrnIndex, ScreenPtr pScreen, int argc, char **argv)
     if (serverGeneration == 1)
 	xf86ShowUnusedOptions(pScrn->scrnIndex, pScrn->options);
 
-#if 1
-#ifdef DRI2
-    driScreenInit(pScreen);
-#endif
-#endif
-
-    return EnterVT(scrnIndex, 1);
+    return drv_enter_vt(scrnIndex, 1);
 }
 
 static void
-AdjustFrame(int scrnIndex, int x, int y, int flags)
+drv_adjust_frame(int scrnIndex, int x, int y, int flags)
 {
     ScrnInfoPtr pScrn = xf86Screens[scrnIndex];
     xf86CrtcConfigPtr config = XF86_CRTC_CONFIG_PTR(pScrn);
@@ -673,13 +672,13 @@ AdjustFrame(int scrnIndex, int x, int y, int flags)
 }
 
 static void
-FreeScreen(int scrnIndex, int flags)
+drv_free_screen(int scrnIndex, int flags)
 {
-    FreeRec(xf86Screens[scrnIndex]);
+    drv_free_rec(xf86Screens[scrnIndex]);
 }
 
 static void
-LeaveVT(int scrnIndex, int flags)
+drv_leave_vt(int scrnIndex, int flags)
 {
     ScrnInfoPtr pScrn = xf86Screens[scrnIndex];
     modesettingPtr ms = modesettingPTR(pScrn);
@@ -689,7 +688,7 @@ LeaveVT(int scrnIndex, int flags)
     for (o = 0; o < config->num_crtc; o++) {
 	xf86CrtcPtr crtc = config->crtc[o];
 
-	crtc_cursor_destroy(crtc);
+	xorg_crtc_cursor_destroy(crtc);
 
 	if (crtc->rotatedPixmap || crtc->rotatedData) {
 	    crtc->funcs->shadow_destroy(crtc, crtc->rotatedPixmap,
@@ -701,7 +700,7 @@ LeaveVT(int scrnIndex, int flags)
 
     drmModeRmFB(ms->fd, ms->fb_id);
 
-    RestoreHWState(pScrn);
+    drv_restore_hw_state(pScrn);
 
     if (drmDropMaster(ms->fd))
 	xf86DrvMsg(pScrn->scrnIndex, X_WARNING,
@@ -714,7 +713,7 @@ LeaveVT(int scrnIndex, int flags)
  * This gets called when gaining control of the VT, and from ScreenInit().
  */
 static Bool
-EnterVT(int scrnIndex, int flags)
+drv_enter_vt(int scrnIndex, int flags)
 {
     ScrnInfoPtr pScrn = xf86Screens[scrnIndex];
     modesettingPtr ms = modesettingPTR(pScrn);
@@ -736,13 +735,13 @@ EnterVT(int scrnIndex, int flags)
      */
     if (ms->SaveGeneration != serverGeneration) {
 	ms->SaveGeneration = serverGeneration;
-	SaveHWState(pScrn);
+	drv_save_hw_state(pScrn);
     }
 
-    if (!CreateFrontBuffer(pScrn))
+    if (!ms->create_front_buffer(pScrn))
 	return FALSE;
 
-    if (!flags && !BindTextureToRoot(pScrn))
+    if (!flags && !ms->bind_front_buffer(pScrn))
 	return FALSE;
 
     if (!xf86SetDesiredModes(pScrn))
@@ -752,7 +751,7 @@ EnterVT(int scrnIndex, int flags)
 }
 
 static Bool
-SwitchMode(int scrnIndex, DisplayModePtr mode, int flags)
+drv_switch_mode(int scrnIndex, DisplayModePtr mode, int flags)
 {
     ScrnInfoPtr pScrn = xf86Screens[scrnIndex];
 
@@ -760,16 +759,18 @@ SwitchMode(int scrnIndex, DisplayModePtr mode, int flags)
 }
 
 static Bool
-CloseScreen(int scrnIndex, ScreenPtr pScreen)
+drv_close_screen(int scrnIndex, ScreenPtr pScreen)
 {
     ScrnInfoPtr pScrn = xf86Screens[scrnIndex];
     modesettingPtr ms = modesettingPTR(pScrn);
 
     if (pScrn->vtSema) {
-	LeaveVT(scrnIndex, 0);
+	drv_leave_vt(scrnIndex, 0);
     }
+
 #ifdef DRI2
-    driCloseScreen(pScreen);
+    if (ms->screen)
+	xorg_dri2_close(pScreen);
 #endif
 
     pScreen->BlockHandler = ms->blockHandler;
@@ -783,14 +784,14 @@ CloseScreen(int scrnIndex, ScreenPtr pScreen)
     }
 #endif
 
-    pipe_texture_reference(&ms->root_texture, NULL);
+    drmModeRmFB(ms->fd, ms->fb_id);
+    ms->destroy_front_buffer(pScrn);
 
     if (ms->exa)
 	xorg_exa_close(pScrn);
+    ms->exa = NULL;
 
-    if (ms->api && ms->api->destroy)
-	ms->api->destroy(ms->api);
-    ms->api = NULL;
+    drv_close_resource_management(pScrn);
 
     drmClose(ms->fd);
     ms->fd = -1;
@@ -801,9 +802,190 @@ CloseScreen(int scrnIndex, ScreenPtr pScreen)
 }
 
 static ModeStatus
-ValidMode(int scrnIndex, DisplayModePtr mode, Bool verbose, int flags)
+drv_valid_mode(int scrnIndex, DisplayModePtr mode, Bool verbose, int flags)
 {
     return MODE_OK;
 }
 
+
+/*
+ * Front buffer backing store functions.
+ */
+
+static Bool
+drv_destroy_front_buffer_ga3d(ScrnInfoPtr pScrn)
+{
+    modesettingPtr ms = modesettingPTR(pScrn);
+    pipe_texture_reference(&ms->root_texture, NULL);
+    return TRUE;
+}
+
+static Bool
+drv_create_front_buffer_ga3d(ScrnInfoPtr pScrn)
+{
+    modesettingPtr ms = modesettingPTR(pScrn);
+    unsigned handle, stride;
+    struct pipe_texture *tex;
+
+    ms->noEvict = TRUE;
+
+    tex = xorg_exa_create_root_texture(pScrn, pScrn->virtualX, pScrn->virtualY,
+				       pScrn->depth, pScrn->bitsPerPixel);
+
+    if (!tex)
+	return FALSE;
+
+    if (!ms->api->local_handle_from_texture(ms->api, ms->screen,
+					    tex,
+					    &stride,
+					    &handle))
+	return FALSE;
+
+    drmModeAddFB(ms->fd,
+		 pScrn->virtualX,
+		 pScrn->virtualY,
+		 pScrn->depth,
+		 pScrn->bitsPerPixel,
+		 stride,
+		 handle,
+                 &ms->fb_id);
+
+    pScrn->frameX0 = 0;
+    pScrn->frameY0 = 0;
+    drv_adjust_frame(pScrn->scrnIndex, pScrn->frameX0, pScrn->frameY0, 0);
+
+    pipe_texture_reference(&ms->root_texture, tex);
+    pipe_texture_reference(&tex, NULL);
+
+    return TRUE;
+}
+
+static Bool
+drv_bind_front_buffer_ga3d(ScrnInfoPtr pScrn)
+{
+    modesettingPtr ms = modesettingPTR(pScrn);
+    ScreenPtr pScreen = pScrn->pScreen;
+    PixmapPtr rootPixmap = pScreen->GetScreenPixmap(pScreen);
+    struct pipe_texture *check;
+
+    xorg_exa_set_displayed_usage(rootPixmap);
+    xorg_exa_set_shared_usage(rootPixmap);
+    xorg_exa_set_texture(rootPixmap, ms->root_texture);
+    if (!pScreen->ModifyPixmapHeader(rootPixmap, -1, -1, -1, -1, -1, NULL))
+	FatalError("Couldn't adjust screen pixmap\n");
+
+    check = xorg_exa_get_texture(rootPixmap);
+    if (ms->root_texture != check)
+	FatalError("Created new root texture\n");
+
+    pipe_texture_reference(&check, NULL);
+    return TRUE;
+}
+
+#ifdef HAVE_LIBKMS
+static Bool
+drv_destroy_front_buffer_kms(ScrnInfoPtr pScrn)
+{
+    modesettingPtr ms = modesettingPTR(pScrn);
+
+    if (!ms->root_bo)
+	return TRUE;
+
+    kms_bo_unmap(ms->root_bo);
+    kms_bo_destroy(ms->root_bo);
+    ms->root_bo = NULL;
+    return TRUE;
+}
+
+static Bool
+drv_create_front_buffer_kms(ScrnInfoPtr pScrn)
+{
+    modesettingPtr ms = modesettingPTR(pScrn);
+    unsigned handle, stride;
+    struct kms_bo *bo;
+    unsigned attr[8];
+
+    attr[0] = KMS_BO_TYPE;
+    attr[1] = KMS_BO_TYPE_SCANOUT;
+    attr[2] = KMS_WIDTH;
+    attr[3] = pScrn->virtualX;
+    attr[4] = KMS_HEIGHT;
+    attr[5] = pScrn->virtualY;
+    attr[6] = 0;
+
+    if (kms_bo_create(ms->kms, attr, &bo))
+	return FALSE;
+
+    if (kms_bo_get_prop(bo, KMS_PITCH, &stride))
+	goto err_destroy;
+
+    if (kms_bo_get_prop(bo, KMS_HANDLE, &handle))
+	goto err_destroy;
+
+    drmModeAddFB(ms->fd,
+		 pScrn->virtualX,
+		 pScrn->virtualY,
+		 pScrn->depth,
+		 pScrn->bitsPerPixel,
+		 stride,
+		 handle,
+                 &ms->fb_id);
+
+    pScrn->frameX0 = 0;
+    pScrn->frameY0 = 0;
+    drv_adjust_frame(pScrn->scrnIndex, pScrn->frameX0, pScrn->frameY0, 0);
+    ms->root_bo = bo;
+
+    return TRUE;
+
+err_destroy:
+    kms_bo_destroy(bo);
+    return FALSE;
+}
+
+static Bool
+drv_bind_front_buffer_kms(ScrnInfoPtr pScrn)
+{
+    modesettingPtr ms = modesettingPTR(pScrn);
+    ScreenPtr pScreen = pScrn->pScreen;
+    PixmapPtr rootPixmap = pScreen->GetScreenPixmap(pScreen);
+    unsigned stride;
+    void *ptr;
+
+    if (kms_bo_get_prop(ms->root_bo, KMS_PITCH, &stride))
+	return FALSE;
+
+    if (kms_bo_map(ms->root_bo, &ptr))
+	return FALSE;
+
+    pScreen->ModifyPixmapHeader(rootPixmap,
+				pScreen->width,
+				pScreen->height,
+				pScreen->rootDepth,
+				pScrn->bitsPerPixel,
+				stride,
+				ptr);
+    return TRUE;
+}
+#endif /* HAVE_LIBKMS */
+
+static Bool drv_init_front_buffer_functions(ScrnInfoPtr pScrn)
+{
+    modesettingPtr ms = modesettingPTR(pScrn);
+    if (ms->screen) {
+	ms->destroy_front_buffer = drv_destroy_front_buffer_ga3d;
+	ms->create_front_buffer = drv_create_front_buffer_ga3d;
+	ms->bind_front_buffer = drv_bind_front_buffer_ga3d;
+#ifdef HAVE_LIBKMS
+    } else if (ms->kms) {
+	ms->destroy_front_buffer = drv_destroy_front_buffer_kms;
+	ms->create_front_buffer = drv_create_front_buffer_kms;
+	ms->bind_front_buffer = drv_bind_front_buffer_kms;
+#endif
+    } else
+	return FALSE;
+
+    return TRUE;
+}
+
 /* vim: set sw=4 ts=8 sts=4: */
diff --git a/src/gallium/state_trackers/xorg/xorg_exa.c b/src/gallium/state_trackers/xorg/xorg_exa.c
index a68a626fa48..16e683019c1 100644
--- a/src/gallium/state_trackers/xorg/xorg_exa.c
+++ b/src/gallium/state_trackers/xorg/xorg_exa.c
@@ -44,8 +44,11 @@
 #include "pipe/p_inlines.h"
 
 #include "util/u_rect.h"
+#include "util/u_math.h"
+#include "util/u_debug.h"
 
 #define DEBUG_PRINT 0
+#define ROUND_UP_TEXTURES 1
 
 /*
  * Helper functions
@@ -149,18 +152,6 @@ exa_get_pipe_format(int depth, enum pipe_format *format, int *bbp, int *picture_
     }
 }
 
-static void
-xorg_exa_common_done(struct exa_context *exa)
-{
-   renderer_draw_flush(exa->renderer);
-
-   exa->copy.src = NULL;
-   exa->copy.dst = NULL;
-   exa->transform.has_src = FALSE;
-   exa->transform.has_mask = FALSE;
-   exa->has_solid_color = FALSE;
-   exa->num_bound_samplers = 0;
-}
 
 /*
  * Static exported EXA functions
@@ -178,6 +169,11 @@ ExaMarkSync(ScreenPtr pScreen)
    return 1;
 }
 
+
+/***********************************************************************
+ * Screen upload/download
+ */
+
 static Bool
 ExaDownloadFromScreen(PixmapPtr pPix, int x,  int y, int w,  int h, char *dst,
 		      int dst_pitch)
@@ -278,13 +274,22 @@ ExaPrepareAccess(PixmapPtr pPix, int index)
 	    PIPE_REFERENCED_FOR_WRITE)
 	    exa->pipe->flush(exa->pipe, 0, NULL);
 
+        assert(pPix->drawable.width <= priv->tex->width0);
+        assert(pPix->drawable.height <= priv->tex->height0);
+
 	priv->map_transfer =
 	    exa->scrn->get_tex_transfer(exa->scrn, priv->tex, 0, 0, 0,
 #ifdef EXA_MIXED_PIXMAPS
 					PIPE_TRANSFER_MAP_DIRECTLY |
 #endif
 					PIPE_TRANSFER_READ_WRITE,
+<<<<<<< HEAD:src/gallium/state_trackers/xorg/xorg_exa.c
 					0, 0, priv->tex->width0, priv->tex->height0);
+=======
+					0, 0, 
+                                        pPix->drawable.width,
+                                        pPix->drawable.height );
+>>>>>>> origin/mesa_7_7_branch:src/gallium/state_trackers/xorg/xorg_exa.c
 	if (!priv->map_transfer)
 #ifdef EXA_MIXED_PIXMAPS
 	    return FALSE;
@@ -327,29 +332,9 @@ ExaFinishAccess(PixmapPtr pPix, int index)
     }
 }
 
-static void
-ExaDone(PixmapPtr pPixmap)
-{
-    ScrnInfoPtr pScrn = xf86Screens[pPixmap->drawable.pScreen->myNum];
-    modesettingPtr ms = modesettingPTR(pScrn);
-    struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap);
-    struct exa_context *exa = ms->exa;
-
-    if (!priv)
-	return;
-
-    xorg_exa_common_done(exa);
-}
-
-static void
-ExaDoneComposite(PixmapPtr pPixmap)
-{
-   ScrnInfoPtr pScrn = xf86Screens[pPixmap->drawable.pScreen->myNum];
-   modesettingPtr ms = modesettingPTR(pScrn);
-   struct exa_context *exa = ms->exa;
-
-   xorg_exa_common_done(exa);
-}
+/***********************************************************************
+ * Solid Fills
+ */
 
 static Bool
 ExaPrepareSolid(PixmapPtr pPixmap, int alu, Pixel planeMask, Pixel fg)
@@ -395,9 +380,34 @@ ExaSolid(PixmapPtr pPixmap, int x0, int y0, int x1, int y1)
     debug_printf("\tExaSolid(%d, %d, %d, %d)\n", x0, y0, x1, y1);
 #endif
 
+    if (x0 == 0 && y0 == 0 &&
+        x1 == pPixmap->drawable.width && y1 == pPixmap->drawable.height) {
+       exa->pipe->clear(exa->pipe, PIPE_CLEAR_COLOR, exa->solid_color, 0.0, 0);
+       return;
+    }
+
     xorg_solid(exa, priv, x0, y0, x1, y1) ;
 }
 
+
+static void
+ExaDoneSolid(PixmapPtr pPixmap)
+{
+    ScrnInfoPtr pScrn = xf86Screens[pPixmap->drawable.pScreen->myNum];
+    modesettingPtr ms = modesettingPTR(pScrn);
+    struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap);
+    struct exa_context *exa = ms->exa;
+
+    if (!priv)
+	return;
+   
+    xorg_composite_done(exa);
+}
+
+/***********************************************************************
+ * Copy Blits
+ */
+
 static Bool
 ExaPrepareCopy(PixmapPtr pSrcPixmap, PixmapPtr pDstPixmap, int xdir,
 	       int ydir, int alu, Pixel planeMask)
@@ -439,6 +449,51 @@ ExaPrepareCopy(PixmapPtr pSrcPixmap, PixmapPtr pDstPixmap, int xdir,
     exa->copy.src = src_priv;
     exa->copy.dst = priv;
 
+    /* For same-surface copies, the pipe->surface_copy path is clearly
+     * superior, providing it is implemented.  In other cases it's not
+     * clear what the better path would be, and eventually we'd
+     * probably want to gather timings and choose dynamically.
+     */
+    if (exa->pipe->surface_copy &&
+        exa->copy.src == exa->copy.dst) {
+
+       exa->copy.use_surface_copy = TRUE;
+       
+       exa->copy.src_surface =
+          exa->scrn->get_tex_surface( exa->scrn,
+                                      exa->copy.src->tex,
+                                      0, 0, 0,
+                                      PIPE_BUFFER_USAGE_GPU_READ);
+
+       exa->copy.dst_surface =
+          exa->scrn->get_tex_surface( exa->scrn, 
+                                      exa->copy.dst->tex,
+                                      0, 0, 0, 
+                                      PIPE_BUFFER_USAGE_GPU_WRITE );
+    }
+    else {
+       exa->copy.use_surface_copy = FALSE;
+
+       if (exa->copy.dst == exa->copy.src)
+          exa->copy.src_texture = renderer_clone_texture( exa->renderer,
+                                                          exa->copy.src->tex );
+       else
+          pipe_texture_reference(&exa->copy.src_texture,
+                                 exa->copy.src->tex);
+
+       exa->copy.dst_surface =
+          exa->scrn->get_tex_surface(exa->scrn,
+                                     exa->copy.dst->tex,
+                                     0, 0, 0,
+                                     PIPE_BUFFER_USAGE_GPU_WRITE);
+
+
+       renderer_copy_prepare(exa->renderer, 
+                             exa->copy.dst_surface,
+                             exa->copy.src_texture );
+    }
+
+
     return exa->accel;
 }
 
@@ -458,11 +513,48 @@ ExaCopy(PixmapPtr pDstPixmap, int srcX, int srcY, int dstX, int dstY,
 
    debug_assert(priv == exa->copy.dst);
 
-   renderer_copy_pixmap(exa->renderer, exa->copy.dst, dstX, dstY,
-                        exa->copy.src, srcX, srcY,
-                        width, height);
+   if (exa->copy.use_surface_copy) {
+      /* XXX: consider exposing >1 box in surface_copy interface.
+       */
+      exa->pipe->surface_copy( exa->pipe,
+                             exa->copy.dst_surface,
+                             dstX, dstY,
+                             exa->copy.src_surface,
+                             srcX, srcY,
+                             width, height );
+   }
+   else {
+      renderer_copy_pixmap(exa->renderer, 
+                           dstX, dstY,
+                           srcX, srcY,
+                           width, height,
+                           exa->copy.src_texture->width0,
+                           exa->copy.src_texture->height0);
+   }
 }
 
+static void
+ExaDoneCopy(PixmapPtr pPixmap)
+{
+    ScrnInfoPtr pScrn = xf86Screens[pPixmap->drawable.pScreen->myNum];
+    modesettingPtr ms = modesettingPTR(pScrn);
+    struct exa_pixmap_priv *priv = exaGetPixmapDriverPrivate(pPixmap);
+    struct exa_context *exa = ms->exa;
+
+    if (!priv)
+	return;
+
+   renderer_draw_flush(exa->renderer);
+
+   exa->copy.src = NULL;
+   exa->copy.dst = NULL;
+   pipe_surface_reference(&exa->copy.src_surface, NULL);
+   pipe_surface_reference(&exa->copy.dst_surface, NULL);
+   pipe_texture_reference(&exa->copy.src_texture, NULL);
+}
+
+
+
 static Bool
 picture_check_formats(struct exa_pixmap_priv *pSrc, PicturePtr pSrcPicture)
 {
@@ -498,6 +590,30 @@ picture_check_formats(struct exa_pixmap_priv *pSrc, PicturePtr pSrcPicture)
    return FALSE;
 }
 
+/***********************************************************************
+ * Composite entrypoints
+ */
+
+static Bool
+ExaCheckComposite(int op,
+		  PicturePtr pSrcPicture, PicturePtr pMaskPicture,
+		  PicturePtr pDstPicture)
+{
+   ScrnInfoPtr pScrn = xf86Screens[pDstPicture->pDrawable->pScreen->myNum];
+   modesettingPtr ms = modesettingPTR(pScrn);
+   struct exa_context *exa = ms->exa;
+   boolean accelerated = xorg_composite_accelerated(op,
+                                                    pSrcPicture,
+                                                    pMaskPicture,
+                                                    pDstPicture);
+#if DEBUG_PRINT
+   debug_printf("ExaCheckComposite(%d, %p, %p, %p) = %d\n",
+                op, pSrcPicture, pMaskPicture, pDstPicture, accelerated);
+#endif
+   return exa->accel && accelerated;
+}
+
+
 static Bool
 ExaPrepareComposite(int op, PicturePtr pSrcPicture,
 		    PicturePtr pMaskPicture, PicturePtr pDstPicture,
@@ -548,8 +664,6 @@ ExaPrepareComposite(int op, PicturePtr pSrcPicture,
                        render_format_name(priv->picture_format),
                        render_format_name(pSrcPicture->format));
 
-      if (priv->picture_format == PICT_a8)
-         XORG_FALLBACK("pSrc pic_format == PICT_a8");
    }
 
    if (pMask) {
@@ -596,25 +710,23 @@ ExaComposite(PixmapPtr pDst, int srcX, int srcY, int maskX, int maskY,
                   dstX, dstY, width, height);
 }
 
-static Bool
-ExaCheckComposite(int op,
-		  PicturePtr pSrcPicture, PicturePtr pMaskPicture,
-		  PicturePtr pDstPicture)
+
+
+static void
+ExaDoneComposite(PixmapPtr pPixmap)
 {
-   ScrnInfoPtr pScrn = xf86Screens[pDstPicture->pDrawable->pScreen->myNum];
+   ScrnInfoPtr pScrn = xf86Screens[pPixmap->drawable.pScreen->myNum];
    modesettingPtr ms = modesettingPTR(pScrn);
    struct exa_context *exa = ms->exa;
-   boolean accelerated = xorg_composite_accelerated(op,
-                                                    pSrcPicture,
-                                                    pMaskPicture,
-                                                    pDstPicture);
-#if DEBUG_PRINT
-   debug_printf("ExaCheckComposite(%d, %p, %p, %p) = %d\n",
-                op, pSrcPicture, pMaskPicture, pDstPicture, accelerated);
-#endif
-   return exa->accel && accelerated;
+
+   xorg_composite_done(exa);
 }
 
+
+/***********************************************************************
+ * Pixmaps
+ */
+
 static void *
 ExaCreatePixmap(ScreenPtr pScreen, int size, int align)
 {
@@ -718,6 +830,22 @@ xorg_exa_get_pixmap_handle(PixmapPtr pPixmap, unsigned *stride_out)
 }
 
 static Bool
+size_match( int width, int tex_width )
+{
+#if ROUND_UP_TEXTURES
+   if (width > tex_width)
+      return FALSE;
+
+   if (width * 2 < tex_width)
+      return FALSE;
+
+   return TRUE;
+#else
+   return width == tex_width;
+#endif
+}
+
+static Bool
 ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height,
 		      int depth, int bitsPerPixel, int devKind,
 		      pointer pPixData)
@@ -731,6 +859,17 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height,
     if (!priv || pPixData)
 	return FALSE;
 
+    if (0) {
+       debug_printf("%s pixmap %p sz %dx%dx%d devKind %d\n",
+                    __FUNCTION__, pPixmap, width, height, bitsPerPixel, devKind);
+       
+       if (priv->tex)
+          debug_printf("  ==> old texture %dx%d\n",
+                       priv->tex->width0, 
+                       priv->tex->height0);
+    }
+
+
     if (depth <= 0)
 	depth = pPixmap->drawable.depth;
 
@@ -749,12 +888,15 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height,
     miModifyPixmapHeader(pPixmap, width, height, depth,
 			     bitsPerPixel, devKind, NULL);
 
+    priv->width = width;
+    priv->height = height;
+
     /* Deal with screen resize */
     if ((exa->accel || priv->flags) &&
         (!priv->tex ||
-         (priv->tex->width0 != width ||
-          priv->tex->height0 != height ||
-          priv->tex_flags != priv->flags))) {
+         !size_match(width, priv->tex->width0) ||
+         !size_match(height, priv->tex->height0) ||
+         priv->tex_flags != priv->flags)) {
 	struct pipe_texture *texture = NULL;
 	struct pipe_texture template;
 
@@ -762,8 +904,15 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height,
 	template.target = PIPE_TEXTURE_2D;
 	exa_get_pipe_format(depth, &template.format, &bitsPerPixel, &priv->picture_format);
 	pf_get_block(template.format, &template.block);
-	template.width0 = width;
-	template.height0 = height;
+        if (ROUND_UP_TEXTURES && priv->flags == 0) {
+           template.width0 = util_next_power_of_two(width);
+           template.height0 = util_next_power_of_two(height);
+        }
+        else {
+           template.width0 = width;
+           template.height0 = height;
+        }
+
 	template.depth0 = 1;
 	template.last_level = 0;
 	template.tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET | priv->flags;
@@ -777,15 +926,15 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height,
 	    dst_surf = exa->scrn->get_tex_surface(
 		exa->scrn, texture, 0, 0, 0, PIPE_BUFFER_USAGE_GPU_WRITE);
 	    src_surf = xorg_gpu_surface(exa->pipe->screen, priv);
-        if (exa->pipe->surface_copy) {
-            exa->pipe->surface_copy(exa->pipe, dst_surf, 0, 0, src_surf,
-                        0, 0, min(width, texture->width0),
-                        min(height, texture->height0));
-        } else {
-            util_surface_copy(exa->pipe, FALSE, dst_surf, 0, 0, src_surf,
-                        0, 0, min(width, texture->width0),
-                        min(height, texture->height0));
-        }
+            if (exa->pipe->surface_copy) {
+               exa->pipe->surface_copy(exa->pipe, dst_surf, 0, 0, src_surf,
+                                       0, 0, min(width, texture->width0),
+                                       min(height, texture->height0));
+            } else {
+               util_surface_copy(exa->pipe, FALSE, dst_surf, 0, 0, src_surf,
+                                 0, 0, min(width, texture->width0),
+                                 min(height, texture->height0));
+            }
 	    exa->scrn->tex_surface_destroy(dst_surf);
 	    exa->scrn->tex_surface_destroy(src_surf);
 	}
@@ -907,10 +1056,10 @@ xorg_exa_init(ScrnInfoPtr pScrn, Bool accel)
    pExa->MarkSync           = ExaMarkSync;
    pExa->PrepareSolid       = ExaPrepareSolid;
    pExa->Solid              = ExaSolid;
-   pExa->DoneSolid          = ExaDone;
+   pExa->DoneSolid          = ExaDoneSolid;
    pExa->PrepareCopy        = ExaPrepareCopy;
    pExa->Copy               = ExaCopy;
-   pExa->DoneCopy           = ExaDone;
+   pExa->DoneCopy           = ExaDoneCopy;
    pExa->CheckComposite     = ExaCheckComposite;
    pExa->PrepareComposite   = ExaPrepareComposite;
    pExa->Composite          = ExaComposite;
diff --git a/src/gallium/state_trackers/xorg/xorg_exa.h b/src/gallium/state_trackers/xorg/xorg_exa.h
index 15cc29d6620..f2cefe23b99 100644
--- a/src/gallium/state_trackers/xorg/xorg_exa.h
+++ b/src/gallium/state_trackers/xorg/xorg_exa.h
@@ -35,13 +35,22 @@ struct exa_context
    } transform;
 
    struct {
+      boolean use_surface_copy;
+
       struct exa_pixmap_priv *src;
       struct exa_pixmap_priv *dst;
+
+      struct pipe_surface *src_surface;
+      struct pipe_surface *dst_surface;
+
+      struct pipe_texture *src_texture;
    } copy;
 };
 
 struct exa_pixmap_priv
 {
+   int width, height;
+
    int flags;
    int tex_flags;
 
diff --git a/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c
index 52b97af1635..89b794a09ac 100644
--- a/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c
+++ b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c
@@ -396,15 +396,11 @@ xrender_tex(struct ureg_program *ureg,
             struct ureg_dst dst,
             struct ureg_src coords,
             struct ureg_src sampler,
+            struct ureg_src imm0,
             boolean repeat_none,
             boolean swizzle,
             boolean set_alpha)
 {
-   struct ureg_src imm0 = { 0 };
-
-   if (repeat_none || set_alpha)
-      imm0 = ureg_imm4f(ureg, 0, 0, 0, 1);
-
    if (repeat_none) {
       struct ureg_dst tmp0 = ureg_DECL_temporary(ureg);
       struct ureg_dst tmp1 = ureg_DECL_temporary(ureg);
@@ -466,6 +462,7 @@ create_fs(struct pipe_context *pipe,
    struct ureg_src /*dst_pos,*/ src_input, mask_pos;
    struct ureg_dst src, mask;
    struct ureg_dst out;
+   struct ureg_src imm0 = { 0 };
    unsigned has_mask = (fs_traits & FS_MASK) != 0;
    unsigned is_fill = (fs_traits & FS_FILL) != 0;
    unsigned is_composite = (fs_traits & FS_COMPOSITE) != 0;
@@ -483,8 +480,6 @@ create_fs(struct pipe_context *pipe,
    unsigned src_luminance = (fs_traits & FS_SRC_LUMINANCE) != 0;
    unsigned mask_luminance = (fs_traits & FS_MASK_LUMINANCE) != 0;
 
-   if (src_luminance)
-      assert(!"src_luminance not supported");
 #if 0
    print_fs_traits(fs_traits);
 #else
@@ -502,6 +497,11 @@ create_fs(struct pipe_context *pipe,
                           TGSI_SEMANTIC_COLOR,
                           0);
 
+   if (src_repeat_none || mask_repeat_none ||
+       src_set_alpha || mask_set_alpha ||
+       src_luminance) {
+      imm0 = ureg_imm4f(ureg, 0, 0, 0, 1);
+   }
    if (is_composite) {
       src_sampler = ureg_DECL_sampler(ureg, 0);
       src_input = ureg_DECL_fs_input(ureg,
@@ -540,16 +540,17 @@ create_fs(struct pipe_context *pipe,
                                 TGSI_INTERPOLATE_PERSPECTIVE);
 #endif
 
+
    if (is_composite) {
-      if (has_mask)
+      if (has_mask || src_luminance)
          src = ureg_DECL_temporary(ureg);
       else
          src = out;
-      xrender_tex(ureg, src, src_input, src_sampler,
+      xrender_tex(ureg, src, src_input, src_sampler, imm0,
                   src_repeat_none, src_swizzle, src_set_alpha);
    } else if (is_fill) {
       if (is_solid) {
-         if (has_mask)
+         if (has_mask || src_luminance)
             src = ureg_dst(src_input);
          else
             ureg_MOV(ureg, out, src_input);
@@ -557,7 +558,7 @@ create_fs(struct pipe_context *pipe,
          struct ureg_src coords, const0124,
             matrow0, matrow1, matrow2;
 
-         if (has_mask)
+         if (has_mask || src_luminance)
             src = ureg_DECL_temporary(ureg);
          else
             src = out;
@@ -582,10 +583,18 @@ create_fs(struct pipe_context *pipe,
       } else
          debug_assert(!"Unknown fill type!");
    }
+   if (src_luminance) {
+      ureg_MOV(ureg, src,
+               ureg_scalar(ureg_src(src), TGSI_SWIZZLE_X));
+      ureg_MOV(ureg, ureg_writemask(src, TGSI_WRITEMASK_XYZ),
+               ureg_scalar(imm0, TGSI_SWIZZLE_X));
+      if (!has_mask)
+         ureg_MOV(ureg, out, ureg_src(src));
+   }
 
    if (has_mask) {
       mask = ureg_DECL_temporary(ureg);
-      xrender_tex(ureg, mask, mask_pos, mask_sampler,
+      xrender_tex(ureg, mask, mask_pos, mask_sampler, imm0,
                   mask_repeat_none, mask_swizzle, mask_set_alpha);
       /* src IN mask */
       src_in_mask(ureg, out, ureg_src(src), ureg_src(mask),
diff --git a/src/gallium/state_trackers/xorg/xorg_output.c b/src/gallium/state_trackers/xorg/xorg_output.c
index bfeddc5e114..251f331ea7a 100644
--- a/src/gallium/state_trackers/xorg/xorg_output.c
+++ b/src/gallium/state_trackers/xorg/xorg_output.c
@@ -53,7 +53,7 @@
 
 #include "xorg_tracker.h"
 
-static char *connector_enum_list[] = {
+static char *output_enum_list[] = {
     "Unknown",
     "VGA",
     "DVI",
@@ -70,19 +70,19 @@ static char *connector_enum_list[] = {
 };
 
 static void
-create_resources(xf86OutputPtr output)
+output_create_resources(xf86OutputPtr output)
 {
 #ifdef RANDR_12_INTERFACE
 #endif /* RANDR_12_INTERFACE */
 }
 
 static void
-dpms(xf86OutputPtr output, int mode)
+output_dpms(xf86OutputPtr output, int mode)
 {
 }
 
 static xf86OutputStatus
-detect(xf86OutputPtr output)
+output_detect(xf86OutputPtr output)
 {
     drmModeConnectorPtr drm_connector = output->driver_private;
 
@@ -97,7 +97,7 @@ detect(xf86OutputPtr output)
 }
 
 static DisplayModePtr
-get_modes(xf86OutputPtr output)
+output_get_modes(xf86OutputPtr output)
 {
     drmModeConnectorPtr drm_connector = output->driver_private;
     drmModeModeInfoPtr drm_mode = NULL;
@@ -110,7 +110,6 @@ get_modes(xf86OutputPtr output)
 	    mode = xcalloc(1, sizeof(DisplayModeRec));
 	    if (!mode)
 		continue;
-	    mode->type = 0;
 	    mode->Clock = drm_mode->clock;
 	    mode->HDisplay = drm_mode->hdisplay;
 	    mode->HSyncStart = drm_mode->hsync_start;
@@ -125,6 +124,11 @@ get_modes(xf86OutputPtr output)
 	    mode->VScan = drm_mode->vscan;
 	    mode->VRefresh = xf86ModeVRefresh(mode);
 	    mode->Private = (void *)drm_mode;
+	    mode->type = 0;
+	    if (drm_mode->type & DRM_MODE_TYPE_PREFERRED)
+		mode->type |= M_T_PREFERRED;
+	    if (drm_mode->type & DRM_MODE_TYPE_DRIVER)
+		mode->type |= M_T_DRIVER;
 	    xf86SetModeDefaultName(mode);
 	    modes = xf86ModesAdd(modes, mode);
 	    xf86PrintModeline(0, mode);
@@ -135,14 +139,14 @@ get_modes(xf86OutputPtr output)
 }
 
 static int
-mode_valid(xf86OutputPtr output, DisplayModePtr pMode)
+output_mode_valid(xf86OutputPtr output, DisplayModePtr pMode)
 {
     return MODE_OK;
 }
 
 #ifdef RANDR_12_INTERFACE
 static Bool
-set_property(xf86OutputPtr output, Atom property, RRPropertyValuePtr value)
+output_set_property(xf86OutputPtr output, Atom property, RRPropertyValuePtr value)
 {
     return TRUE;
 }
@@ -150,36 +154,36 @@ set_property(xf86OutputPtr output, Atom property, RRPropertyValuePtr value)
 
 #ifdef RANDR_13_INTERFACE
 static Bool
-get_property(xf86OutputPtr output, Atom property)
+output_get_property(xf86OutputPtr output, Atom property)
 {
     return TRUE;
 }
 #endif /* RANDR_13_INTERFACE */
 
 static void
-destroy(xf86OutputPtr output)
+output_destroy(xf86OutputPtr output)
 {
     drmModeFreeConnector(output->driver_private);
 }
 
 static const xf86OutputFuncsRec output_funcs = {
-    .create_resources = create_resources,
+    .create_resources = output_create_resources,
 #ifdef RANDR_12_INTERFACE
-    .set_property = set_property,
+    .set_property = output_set_property,
 #endif
 #ifdef RANDR_13_INTERFACE
-    .get_property = get_property,
+    .get_property = output_get_property,
 #endif
-    .dpms = dpms,
-    .detect = detect,
+    .dpms = output_dpms,
+    .detect = output_detect,
 
-    .get_modes = get_modes,
-    .mode_valid = mode_valid,
-    .destroy = destroy,
+    .get_modes = output_get_modes,
+    .mode_valid = output_mode_valid,
+    .destroy = output_destroy,
 };
 
 void
-output_init(ScrnInfoPtr pScrn)
+xorg_output_init(ScrnInfoPtr pScrn)
 {
     modesettingPtr ms = modesettingPTR(pScrn);
     xf86OutputPtr output;
@@ -220,7 +224,7 @@ output_init(ScrnInfoPtr pScrn)
 #endif
 
 	snprintf(name, 32, "%s%d",
-		 connector_enum_list[drm_connector->connector_type],
+		 output_enum_list[drm_connector->connector_type],
 		 drm_connector->connector_type_id);
 
 
diff --git a/src/gallium/state_trackers/xorg/xorg_renderer.c b/src/gallium/state_trackers/xorg/xorg_renderer.c
index 418a8dd88bd..8f73ec5fe11 100644
--- a/src/gallium/state_trackers/xorg/xorg_renderer.c
+++ b/src/gallium/state_trackers/xorg/xorg_renderer.c
@@ -13,11 +13,6 @@
 
 #include <math.h>
 
-enum AxisOrientation {
-   Y0_BOTTOM,
-   Y0_TOP
-};
-
 #define floatsEqual(x, y) (fabs(x - y) <= 0.00001f * MIN2(fabs(x), fabs(y)))
 #define floatIsZero(x) (floatsEqual((x) + 1, 1))
 
@@ -96,10 +91,18 @@ static void
 renderer_init_state(struct xorg_renderer *r)
 {
    struct pipe_depth_stencil_alpha_state dsa;
+   struct pipe_rasterizer_state raster;
 
    /* set common initial clip state */
    memset(&dsa, 0, sizeof(struct pipe_depth_stencil_alpha_state));
    cso_set_depth_stencil_alpha(r->cso, &dsa);
+
+
+   /* XXX: move to renderer_init_state? */
+   memset(&raster, 0, sizeof(struct pipe_rasterizer_state));
+   raster.gl_rasterization_rules = 1;
+   cso_set_rasterizer(r->cso, &raster);
+
 }
 
 
@@ -186,23 +189,6 @@ add_vertex_data1(struct xorg_renderer *r,
    add_vertex_1tex(r, dstX, dstY + height, s3, t3);
 }
 
-static struct pipe_buffer *
-setup_vertex_data_tex(struct xorg_renderer *r,
-                      float x0, float y0, float x1, float y1,
-                      float s0, float t0, float s1, float t1,
-                      float z)
-{
-   /* 1st vertex */
-   add_vertex_1tex(r, x0, y0, s0, t0);
-   /* 2nd vertex */
-   add_vertex_1tex(r, x1, y0, s1, t0);
-   /* 3rd vertex */
-   add_vertex_1tex(r, x1, y1, s1, t1);
-   /* 4th vertex */
-   add_vertex_1tex(r, x0, y1, s0, t1);
-
-   return renderer_buffer_create(r);
-}
 
 static INLINE void
 add_vertex_2tex(struct xorg_renderer *r,
@@ -322,15 +308,32 @@ setup_vertex_data_yuv(struct xorg_renderer *r,
 
 
 
-static void
-set_viewport(struct xorg_renderer *r, int width, int height,
-             enum AxisOrientation orientation)
+/* Set up framebuffer, viewport and vertex shader constant buffer
+ * state for a particular destinaton surface.  In all our rendering,
+ * these concepts are linked.
+ */
+void renderer_bind_destination(struct xorg_renderer *r,
+                               struct pipe_surface *surface,
+                               int width,
+                               int height )
 {
+
+   struct pipe_framebuffer_state fb;
    struct pipe_viewport_state viewport;
-   float y_scale = (orientation == Y0_BOTTOM) ? -2.f : 2.f;
 
+   /* Framebuffer uses actual surface width/height
+    */
+   memset(&fb, 0, sizeof fb);
+   fb.width  = surface->width;
+   fb.height = surface->height;
+   fb.nr_cbufs = 1;
+   fb.cbufs[0] = surface;
+   fb.zsbuf = 0;
+
+   /* Viewport just touches the bit we're interested in:
+    */
    viewport.scale[0] =  width / 2.f;
-   viewport.scale[1] =  height / y_scale;
+   viewport.scale[1] =  height / 2.f;
    viewport.scale[2] =  1.0;
    viewport.scale[3] =  1.0;
    viewport.translate[0] = width / 2.f;
@@ -338,11 +341,28 @@ set_viewport(struct xorg_renderer *r, int width, int height,
    viewport.translate[2] = 0.0;
    viewport.translate[3] = 0.0;
 
+   /* Constant buffer set up to match viewport dimensions:
+    */
+   if (r->fb_width != width ||
+       r->fb_height != height) 
+   {
+      float vs_consts[8] = {
+         2.f/width, 2.f/height, 1, 1,
+         -1, -1, 0, 0
+      };
+
+      r->fb_width = width;
+      r->fb_height = height;
+
+      renderer_set_constants(r, PIPE_SHADER_VERTEX,
+                             vs_consts, sizeof vs_consts);
+   }
+
+   cso_set_framebuffer(r->cso, &fb);
    cso_set_viewport(r->cso, &viewport);
 }
 
 
-
 struct xorg_renderer * renderer_create(struct pipe_context *pipe)
 {
    struct xorg_renderer *renderer = CALLOC_STRUCT(xorg_renderer);
@@ -379,51 +399,9 @@ void renderer_destroy(struct xorg_renderer *r)
    }
 }
 
-void renderer_bind_framebuffer(struct xorg_renderer *r,
-                               struct exa_pixmap_priv *priv)
-{
-   unsigned i;
-   struct pipe_framebuffer_state state;
-   struct pipe_surface *surface = xorg_gpu_surface(r->pipe->screen, priv);
-   memset(&state, 0, sizeof(struct pipe_framebuffer_state));
-
-   state.width  = priv->tex->width0;
-   state.height = priv->tex->height0;
-
-   state.nr_cbufs = 1;
-   state.cbufs[0] = surface;
-   for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i)
-      state.cbufs[i] = 0;
-
-   /* currently we don't use depth/stencil */
-   state.zsbuf = 0;
-
-   cso_set_framebuffer(r->cso, &state);
-
-   /* we do fire and forget for the framebuffer, this is the forget part */
-   pipe_surface_reference(&surface, NULL);
-}
-
-void renderer_bind_viewport(struct xorg_renderer *r,
-                            struct exa_pixmap_priv *dst)
-{
-   int width = dst->tex->width0;
-   int height = dst->tex->height0;
 
-   /*debug_printf("Bind viewport (%d, %d)\n", width, height);*/
 
-   set_viewport(r, width, height, Y0_TOP);
-}
 
-void renderer_bind_rasterizer(struct xorg_renderer *r)
-{
-   struct pipe_rasterizer_state raster;
-
-   /* XXX: move to renderer_init_state? */
-   memset(&raster, 0, sizeof(struct pipe_rasterizer_state));
-   raster.gl_rasterization_rules = 1;
-   cso_set_rasterizer(r->cso, &raster);
-}
 
 void renderer_set_constants(struct xorg_renderer *r,
                             int shader_type,
@@ -446,182 +424,20 @@ void renderer_set_constants(struct xorg_renderer *r,
    r->pipe->set_constant_buffer(r->pipe, shader_type, 0, cbuf);
 }
 
-static void
-setup_vs_constant_buffer(struct xorg_renderer *r,
-                         int width, int height)
-{
-   const int param_bytes = 8 * sizeof(float);
-   float vs_consts[8] = {
-      2.f/width, 2.f/height, 1, 1,
-      -1, -1, 0, 0
-   };
-   renderer_set_constants(r, PIPE_SHADER_VERTEX,
-                          vs_consts, param_bytes);
-}
 
-static void
-setup_fs_constant_buffer(struct xorg_renderer *r)
-{
-   const int param_bytes = 4 * sizeof(float);
-   const float fs_consts[8] = {
-      0, 0, 0, 1,
-   };
-   renderer_set_constants(r, PIPE_SHADER_FRAGMENT,
-                          fs_consts, param_bytes);
-}
-
-static INLINE void shift_rectx(float coords[4],
-                               const float *bounds,
-                               const float shift)
-{
-   coords[0] += shift;
-   coords[2] -= shift;
-   if (bounds) {
-      coords[2] = MIN2(coords[2], bounds[2]);
-      /* bound x/y + width/height */
-      if ((coords[0] + coords[2]) > (bounds[0] + bounds[2])) {
-         coords[2] = (bounds[0] + bounds[2]) - coords[0];
-      }
-   }
-}
-
-static INLINE void shift_recty(float coords[4],
-                               const float *bounds,
-                               const float shift)
-{
-   coords[1] += shift;
-   coords[3] -= shift;
-   if (bounds) {
-      coords[3] = MIN2(coords[3], bounds[3]);
-      if ((coords[1] + coords[3]) > (bounds[1] + bounds[3])) {
-         coords[3] = (bounds[1] + bounds[3]) - coords[1];
-      }
-   }
-}
-
-static INLINE void bound_rect(float coords[4],
-                              const float bounds[4],
-                              float shift[4])
-{
-   /* if outside the bounds */
-   if (coords[0] > (bounds[0] + bounds[2]) ||
-       coords[1] > (bounds[1] + bounds[3]) ||
-       (coords[0] + coords[2]) < bounds[0] ||
-       (coords[1] + coords[3]) < bounds[1]) {
-      coords[0] = 0.f;
-      coords[1] = 0.f;
-      coords[2] = 0.f;
-      coords[3] = 0.f;
-      shift[0] = 0.f;
-      shift[1] = 0.f;
-      return;
-   }
-
-   /* bound x */
-   if (coords[0] < bounds[0]) {
-      shift[0] = bounds[0] - coords[0];
-      coords[2] -= shift[0];
-      coords[0] = bounds[0];
-   } else
-      shift[0] = 0.f;
-
-   /* bound y */
-   if (coords[1] < bounds[1]) {
-      shift[1] = bounds[1] - coords[1];
-      coords[3] -= shift[1];
-      coords[1] = bounds[1];
-   } else
-      shift[1] = 0.f;
-
-   shift[2] = bounds[2] - coords[2];
-   shift[3] = bounds[3] - coords[3];
-   /* bound width/height */
-   coords[2] = MIN2(coords[2], bounds[2]);
-   coords[3] = MIN2(coords[3], bounds[3]);
-
-   /* bound x/y + width/height */
-   if ((coords[0] + coords[2]) > (bounds[0] + bounds[2])) {
-      coords[2] = (bounds[0] + bounds[2]) - coords[0];
-   }
-   if ((coords[1] + coords[3]) > (bounds[1] + bounds[3])) {
-      coords[3] = (bounds[1] + bounds[3]) - coords[1];
-   }
-
-   /* if outside the bounds */
-   if ((coords[0] + coords[2]) < bounds[0] ||
-       (coords[1] + coords[3]) < bounds[1]) {
-      coords[0] = 0.f;
-      coords[1] = 0.f;
-      coords[2] = 0.f;
-      coords[3] = 0.f;
-      return;
-   }
-}
-
-static INLINE void sync_size(float *src_loc, float *dst_loc)
-{
-   src_loc[2] = MIN2(src_loc[2], dst_loc[2]);
-   src_loc[3] = MIN2(src_loc[3], dst_loc[3]);
-   dst_loc[2] = src_loc[2];
-   dst_loc[3] = src_loc[3];
-}
-
-static void renderer_copy_texture(struct xorg_renderer *r,
-                                  struct pipe_texture *src,
-                                  float sx1, float sy1,
-                                  float sx2, float sy2,
-                                  struct pipe_texture *dst,
-                                  float dx1, float dy1,
-                                  float dx2, float dy2)
+void renderer_copy_prepare(struct xorg_renderer *r,
+                           struct pipe_surface *dst_surface,
+                           struct pipe_texture *src_texture)
 {
    struct pipe_context *pipe = r->pipe;
    struct pipe_screen *screen = pipe->screen;
-   struct pipe_buffer *buf;
-   struct pipe_surface *dst_surf = screen->get_tex_surface(
-      screen, dst, 0, 0, 0,
-      PIPE_BUFFER_USAGE_GPU_WRITE);
-   struct pipe_framebuffer_state fb;
-   float s0, t0, s1, t1;
    struct xorg_shader shader;
 
-   assert(src->width0 != 0);
-   assert(src->height0 != 0);
-   assert(dst->width0 != 0);
-   assert(dst->height0 != 0);
-
-#if 1
-   s0 = sx1 / src->width0;
-   s1 = sx2 / src->width0;
-   t0 = sy1 / src->height0;
-   t1 = sy2 / src->height0;
-#else
-   s0 = 0;
-   s1 = 1;
-   t0 = 0;
-   t1 = 1;
-#endif
-
-#if 0
-   debug_printf("copy texture src=[%f, %f, %f, %f], dst=[%f, %f, %f, %f], tex=[%f, %f, %f, %f]\n",
-                sx1, sy1, sx2, sy2, dx1, dy1, dx2, dy2,
-                s0, t0, s1, t1);
-#endif
-
-   assert(screen->is_format_supported(screen, dst_surf->format,
+   assert(screen->is_format_supported(screen, dst_surface->format,
                                       PIPE_TEXTURE_2D,
                                       PIPE_TEXTURE_USAGE_RENDER_TARGET,
                                       0));
 
-   /* save state (restored below) */
-   cso_save_blend(r->cso);
-   cso_save_samplers(r->cso);
-   cso_save_sampler_textures(r->cso);
-   cso_save_framebuffer(r->cso);
-   cso_save_fragment_shader(r->cso);
-   cso_save_vertex_shader(r->cso);
-
-   cso_save_viewport(r->cso);
-
 
    /* set misc state we care about */
    {
@@ -650,12 +466,12 @@ static void renderer_copy_texture(struct xorg_renderer *r,
       cso_single_sampler_done(r->cso);
    }
 
-   set_viewport(r, dst_surf->width, dst_surf->height, Y0_TOP);
+   renderer_bind_destination(r, dst_surface, 
+                             dst_surface->width,
+                             dst_surface->height);
 
    /* texture */
-   cso_set_sampler_textures(r->cso, 1, &src);
-
-   renderer_bind_rasterizer(r);
+   cso_set_sampler_textures(r->cso, 1, &src_texture);
 
    /* shaders */
    shader = xorg_shaders_get(r->shaders,
@@ -664,51 +480,12 @@ static void renderer_copy_texture(struct xorg_renderer *r,
    cso_set_vertex_shader_handle(r->cso, shader.vs);
    cso_set_fragment_shader_handle(r->cso, shader.fs);
 
-   /* drawing dest */
-   memset(&fb, 0, sizeof(fb));
-   fb.width = dst_surf->width;
-   fb.height = dst_surf->height;
-   fb.nr_cbufs = 1;
-   fb.cbufs[0] = dst_surf;
-   {
-      int i;
-      for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i)
-         fb.cbufs[i] = 0;
-   }
-   cso_set_framebuffer(r->cso, &fb);
-   setup_vs_constant_buffer(r, fb.width, fb.height);
-   setup_fs_constant_buffer(r);
-
-   /* draw quad */
-   buf = setup_vertex_data_tex(r,
-                               dx1, dy1,
-                               dx2, dy2,
-                               s0, t0, s1, t1,
-                               0.0f);
-
-   if (buf) {
-      util_draw_vertex_buffer(r->pipe, buf, 0,
-                              PIPE_PRIM_QUADS,
-                              4,  /* verts */
-                              2); /* attribs/vert */
-
-      pipe_buffer_reference(&buf, NULL);
-   }
-
-   /* restore state we changed */
-   cso_restore_blend(r->cso);
-   cso_restore_samplers(r->cso);
-   cso_restore_sampler_textures(r->cso);
-   cso_restore_framebuffer(r->cso);
-   cso_restore_vertex_shader(r->cso);
-   cso_restore_fragment_shader(r->cso);
-   cso_restore_viewport(r->cso);
-
-   pipe_surface_reference(&dst_surf, NULL);
+   r->buffer_size = 0;
+   r->attrs_per_vertex = 2;
 }
 
-static struct pipe_texture *
-create_sampler_texture(struct xorg_renderer *r,
+struct pipe_texture *
+renderer_clone_texture(struct xorg_renderer *r,
                        struct pipe_texture *src)
 {
    enum pipe_format format;
@@ -717,7 +494,9 @@ create_sampler_texture(struct xorg_renderer *r,
    struct pipe_texture *pt;
    struct pipe_texture templ;
 
-   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+   if (pipe->is_texture_referenced(pipe, src, 0, 0) &
+       PIPE_REFERENCED_FOR_WRITE)
+      pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
 
    /* the coming in texture should already have that invariance */
    debug_assert(screen->is_format_supported(screen, src->format,
@@ -771,16 +550,16 @@ create_sampler_texture(struct xorg_renderer *r,
 
 
 void renderer_copy_pixmap(struct xorg_renderer *r,
-                          struct exa_pixmap_priv *dst_priv, int dx, int dy,
-                          struct exa_pixmap_priv *src_priv, int sx, int sy,
-                          int width, int height)
+                          int dx, int dy,
+                          int sx, int sy,
+                          int width, int height,
+                          float src_width,
+                          float src_height)
 {
-   float dst_loc[4], src_loc[4];
-   float dst_bounds[4], src_bounds[4];
-   float src_shift[4], dst_shift[4], shift[4];
-   struct pipe_texture *dst = dst_priv->tex;
-   struct pipe_texture *src = src_priv->tex;
+   float s0, t0, s1, t1;
+   float x0, y0, x1, y1;
 
+<<<<<<< HEAD:src/gallium/state_trackers/xorg/xorg_renderer.c
    if (r->pipe->is_texture_referenced(r->pipe, src, 0, 0) &
        PIPE_REFERENCED_FOR_WRITE)
       r->pipe->flush(r->pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
@@ -842,8 +621,33 @@ void renderer_copy_pixmap(struct xorg_renderer *r,
       if (src == dst)
          pipe_texture_reference(&temp_src, NULL);
    }
+=======
+
+   /* XXX: could put the texcoord scaling calculation into the vertex
+    * shader.
+    */
+   s0 = sx            / src_width;
+   s1 = (sx + width)  / src_width;
+   t0 = sy            / src_height;
+   t1 = (sy + height) / src_height;
+
+   x0 = dx;
+   x1 = dx + width;
+   y0 = dy;
+   y1 = dy + height;
+
+   /* draw quad */
+   renderer_draw_conditional(r, 4*8);
+   add_vertex_1tex(r, x0, y0, s0, t0);
+   add_vertex_1tex(r, x1, y0, s1, t0);
+   add_vertex_1tex(r, x1, y1, s1, t1);
+   add_vertex_1tex(r, x0, y1, s0, t1);
+>>>>>>> origin/mesa_7_7_branch:src/gallium/state_trackers/xorg/xorg_renderer.c
 }
 
+
+
+
 void renderer_draw_yuv(struct xorg_renderer *r,
                        int src_x, int src_y, int src_w, int src_h,
                        int dst_x, int dst_y, int dst_w, int dst_h,
diff --git a/src/gallium/state_trackers/xorg/xorg_renderer.h b/src/gallium/state_trackers/xorg/xorg_renderer.h
index 2f0b865dbdc..5272cde2b3f 100644
--- a/src/gallium/state_trackers/xorg/xorg_renderer.h
+++ b/src/gallium/state_trackers/xorg/xorg_renderer.h
@@ -21,6 +21,8 @@ struct xorg_renderer {
    struct cso_context *cso;
    struct xorg_shaders *shaders;
 
+   int fb_width;
+   int fb_height;
    struct pipe_constant_buffer vs_const_buffer;
    struct pipe_constant_buffer fs_const_buffer;
 
@@ -35,19 +37,19 @@ struct xorg_renderer {
 struct xorg_renderer *renderer_create(struct pipe_context *pipe);
 void renderer_destroy(struct xorg_renderer *renderer);
 
+void renderer_bind_destination(struct xorg_renderer *r,
+                               struct pipe_surface *surface,
+                               int width,
+                               int height );
+
 void renderer_bind_framebuffer(struct xorg_renderer *r,
                                struct exa_pixmap_priv *priv);
 void renderer_bind_viewport(struct xorg_renderer *r,
                             struct exa_pixmap_priv *dst);
-void renderer_bind_rasterizer(struct xorg_renderer *r);
 void renderer_set_constants(struct xorg_renderer *r,
                             int shader_type,
                             const float *buffer,
                             int size);
-void renderer_copy_pixmap(struct xorg_renderer *r,
-                          struct exa_pixmap_priv *dst_priv, int dx, int dy,
-                          struct exa_pixmap_priv *src_priv, int sx, int sy,
-                          int width, int height);
 
 
 void renderer_draw_yuv(struct xorg_renderer *r,
@@ -74,5 +76,20 @@ void renderer_texture(struct xorg_renderer *r,
 
 void renderer_draw_flush(struct xorg_renderer *r);
 
+struct pipe_texture *
+renderer_clone_texture(struct xorg_renderer *r,
+                       struct pipe_texture *src);
+
+void renderer_copy_prepare(struct xorg_renderer *r,
+                           struct pipe_surface *dst_surface,
+                           struct pipe_texture *src_texture);
+
+void renderer_copy_pixmap(struct xorg_renderer *r,
+                          int dx, int dy,
+                          int sx, int sy,
+                          int width, int height,
+                          float src_width,
+                          float src_height);
+
 
 #endif
diff --git a/src/gallium/state_trackers/xorg/xorg_tracker.h b/src/gallium/state_trackers/xorg/xorg_tracker.h
index 20c9259c7bc..c6c7b2fe158 100644
--- a/src/gallium/state_trackers/xorg/xorg_tracker.h
+++ b/src/gallium/state_trackers/xorg/xorg_tracker.h
@@ -51,6 +51,8 @@
 
 #define DRV_ERROR(msg)	xf86DrvMsg(pScrn->scrnIndex, X_ERROR, msg);
 
+struct kms_bo;
+struct kms_driver;
 struct exa_context;
 
 typedef struct
@@ -61,6 +63,8 @@ typedef struct
     ScrnInfoPtr pScrn_2;
 } EntRec, *EntPtr;
 
+#define XORG_NR_FENCES 3
+
 typedef struct _modesettingRec
 {
     /* drm */
@@ -84,8 +88,19 @@ typedef struct _modesettingRec
     unsigned int SaveGeneration;
 
     void (*blockHandler)(int, pointer, pointer, pointer);
+    struct pipe_fence_handle *fence[XORG_NR_FENCES];
+
     CreateScreenResourcesProcPtr createScreenResources;
 
+    /* for frontbuffer backing store */
+    Bool (*destroy_front_buffer)(ScrnInfoPtr pScrn);
+    Bool (*create_front_buffer)(ScrnInfoPtr pScrn);
+    Bool (*bind_front_buffer)(ScrnInfoPtr pScrn);
+
+    /* kms */
+    struct kms_driver *kms;
+    struct kms_bo *root_bo;
+
     /* gallium */
     struct drm_api *api;
     struct pipe_screen *screen;
@@ -141,33 +156,34 @@ xorg_exa_close(ScrnInfoPtr pScrn);
  * xorg_dri2.c
  */
 Bool
-driScreenInit(ScreenPtr pScreen);
+xorg_dri2_init(ScreenPtr pScreen);
 
 void
-driCloseScreen(ScreenPtr pScreen);
+xorg_dri2_close(ScreenPtr pScreen);
 
 
 /***********************************************************************
  * xorg_crtc.c
  */
 void
-crtc_init(ScrnInfoPtr pScrn);
+xorg_crtc_init(ScrnInfoPtr pScrn);
 
 void
-crtc_cursor_destroy(xf86CrtcPtr crtc);
+xorg_crtc_cursor_destroy(xf86CrtcPtr crtc);
 
 
 /***********************************************************************
  * xorg_output.c
  */
 void
-output_init(ScrnInfoPtr pScrn);
+xorg_output_init(ScrnInfoPtr pScrn);
+
 
 /***********************************************************************
  * xorg_xv.c
  */
 void
-xorg_init_video(ScreenPtr pScreen);
+xorg_xv_init(ScreenPtr pScreen);
 
 
 #endif /* _XORG_TRACKER_H_ */
diff --git a/src/gallium/state_trackers/xorg/xorg_xv.c b/src/gallium/state_trackers/xorg/xorg_xv.c
index bb515a0f49b..b8eca8c8176 100644
--- a/src/gallium/state_trackers/xorg/xorg_xv.c
+++ b/src/gallium/state_trackers/xorg/xorg_xv.c
@@ -319,21 +319,6 @@ copy_packed_data(ScrnInfoPtr pScrn,
 
 
 static void
-setup_vs_video_constants(struct xorg_renderer *r, struct exa_pixmap_priv *dst)
-{
-   int width = dst->tex->width0;
-   int height = dst->tex->height0;
-   const int param_bytes = 8 * sizeof(float);
-   float vs_consts[8] = {
-      2.f/width, 2.f/height, 1, 1,
-      -1, -1, 0, 0
-   };
-
-   renderer_set_constants(r, PIPE_SHADER_VERTEX,
-                          vs_consts, param_bytes);
-}
-
-static void
 setup_fs_video_constants(struct xorg_renderer *r, boolean hdtv)
 {
    const int param_bytes = 12 * sizeof(float);
@@ -446,6 +431,7 @@ display_video(ScrnInfoPtr pScrn, struct xorg_xv_port_priv *pPriv, int id,
    Bool hdtv;
    int x, y, w, h;
    struct exa_pixmap_priv *dst = exaGetPixmapDriverPrivate(pPixmap);
+   struct pipe_surface *dst_surf = xorg_gpu_surface(pPriv->r->pipe->screen, dst);
 
    if (dst && !dst->tex) {
 	xorg_exa_set_shared_usage(pPixmap);
@@ -466,13 +452,12 @@ display_video(ScrnInfoPtr pScrn, struct xorg_xv_port_priv *pPriv, int id,
    pbox = REGION_RECTS(dstRegion);
    nbox = REGION_NUM_RECTS(dstRegion);
 
-   renderer_bind_framebuffer(pPriv->r, dst);
-   renderer_bind_viewport(pPriv->r, dst);
+   renderer_bind_destination(pPriv->r, dst_surf, 
+                             dst_surf->width, dst_surf->height);
+
    bind_blend_state(pPriv);
-   renderer_bind_rasterizer(pPriv->r);
    bind_shaders(pPriv);
    bind_samplers(pPriv);
-   setup_vs_video_constants(pPriv->r, dst);
    setup_fs_video_constants(pPriv->r, hdtv);
 
    exaMoveInPixmap(pPixmap);
@@ -506,6 +491,8 @@ display_video(ScrnInfoPtr pScrn, struct xorg_xv_port_priv *pPriv, int id,
    }
    DamageRegionProcessPending(&pPixmap->drawable);
 
+   pipe_surface_reference(&dst_surf, NULL);
+
    return TRUE;
 }
 
@@ -689,7 +676,7 @@ xorg_setup_textured_adapter(ScreenPtr pScreen)
 }
 
 void
-xorg_init_video(ScreenPtr pScreen)
+xorg_xv_init(ScreenPtr pScreen)
 {
    ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
    /*modesettingPtr ms = modesettingPTR(pScrn);*/
diff --git a/src/gallium/winsys/drm/vmware/core/vmw_surface.c b/src/gallium/winsys/drm/vmware/core/vmw_surface.c
index c19e556df9f..64eb32f8b94 100644
--- a/src/gallium/winsys/drm/vmware/core/vmw_surface.c
+++ b/src/gallium/winsys/drm/vmware/core/vmw_surface.c
@@ -37,11 +37,13 @@ vmw_svga_winsys_surface_reference(struct vmw_svga_winsys_surface **pdst,
 {
    struct pipe_reference *src_ref;
    struct pipe_reference *dst_ref;
-   struct vmw_svga_winsys_surface *dst = *pdst;
-   
+   struct vmw_svga_winsys_surface *dst;
+
    if(pdst == NULL || *pdst == src)
       return;
-   
+
+   dst = *pdst;
+
    src_ref = src ? &src->refcnt : NULL;
    dst_ref = dst ? &dst->refcnt : NULL;
 
diff --git a/src/gallium/winsys/drm/vmware/xorg/Makefile b/src/gallium/winsys/drm/vmware/xorg/Makefile
index e152263256a..48a9b08aa76 100644
--- a/src/gallium/winsys/drm/vmware/xorg/Makefile
+++ b/src/gallium/winsys/drm/vmware/xorg/Makefile
@@ -7,7 +7,6 @@ include $(TOP)/configs/current
 
 INCLUDES = \
 	$(shell pkg-config --cflags-only-I pixman-1 xorg-server libdrm xproto) \
-	-I../gem \
 	-I$(TOP)/src/gallium/include \
 	-I$(TOP)/src/gallium/drivers \
 	-I$(TOP)/src/gallium/auxiliary \
@@ -20,19 +19,29 @@ LIBS = \
 	$(TOP)/src/gallium/drivers/svga/libsvga.a \
 	$(GALLIUM_AUXILIARIES)
 
+LINKS = \
+	$(shell pkg-config --libs --silence-errors libkms) \
+	$(shell pkg-config --libs libdrm)
+
 DRIVER_DEFINES = \
 	-DHAVE_CONFIG_H
 
+TARGET_STAGING = $(TOP)/$(LIB_DIR)/gallium/$(TARGET)
 
 #############################################
 
 
 
-all default: $(TARGET)
+all default: $(TARGET) $(TARGET_STAGING)
+
+$(TARGET): $(OBJECTS) Makefile $(LIBS)
+	$(MKLIB) -noprefix -o $@ $(OBJECTS) $(LIBS) $(LINKS)
+
+$(TOP)/$(LIB_DIR)/gallium:
+	mkdir -p $@
 
-$(TARGET): $(OBJECTS) Makefile $(TOP)/src/gallium/state_trackers/xorg/libxorgtracker.a $(LIBS)
-	$(TOP)/bin/mklib -noprefix -o $@ \
-	$(OBJECTS) $(LIBS) $(shell pkg-config --libs libdrm) -ldrm_intel
+$(TARGET_STAGING): $(TARGET) $(TOP)/$(LIB_DIR)/gallium
+	$(INSTALL) $(TARGET) $(TOP)/$(LIB_DIR)/gallium
 
 clean:
 	rm -rf $(OBJECTS) $(TARGET)
diff --git a/src/glu/sgi/libutil/mipmap.c b/src/glu/sgi/libutil/mipmap.c
index c5faebd6a35..d1fd5a7d724 100644
--- a/src/glu/sgi/libutil/mipmap.c
+++ b/src/glu/sgi/libutil/mipmap.c
@@ -3608,6 +3608,7 @@ int gluBuild1DMipmapLevelsCore(GLenum target, GLint internalFormat,
 		    glPixelStorei(GL_UNPACK_SKIP_PIXELS,psm.unpack_skip_pixels);
 		    glPixelStorei(GL_UNPACK_ROW_LENGTH, psm.unpack_row_length);
 		    glPixelStorei(GL_UNPACK_SWAP_BYTES, psm.unpack_swap_bytes);
+		    free(newImage);
 		    return GLU_OUT_OF_MEMORY;
 		}
 	    }
@@ -4107,6 +4108,7 @@ static int gluBuild2DMipmapLevelsCore(GLenum target, GLint internalFormat,
 	  glPixelStorei(GL_UNPACK_SKIP_PIXELS, psm.unpack_skip_pixels);
 	  glPixelStorei(GL_UNPACK_ROW_LENGTH, psm.unpack_row_length);
 	  glPixelStorei(GL_UNPACK_SWAP_BYTES, psm.unpack_swap_bytes);
+	  free(srcImage);
 	  return GLU_OUT_OF_MEMORY;
 	}
 	/* level userLevel+1 is in srcImage; level userLevel already saved */
@@ -4349,6 +4351,7 @@ static int gluBuild2DMipmapLevelsCore(GLenum target, GLint internalFormat,
 	    glPixelStorei(GL_UNPACK_SKIP_PIXELS, psm.unpack_skip_pixels);
 	    glPixelStorei(GL_UNPACK_ROW_LENGTH, psm.unpack_row_length);
 	    glPixelStorei(GL_UNPACK_SWAP_BYTES, psm.unpack_swap_bytes);
+	    free(srcImage);
 	    return GLU_OUT_OF_MEMORY;
 	  }
 	}
@@ -8098,6 +8101,7 @@ static int gluBuild3DMipmapLevelsCore(GLenum target, GLint internalFormat,
 	 glPixelStorei(GL_UNPACK_SWAP_BYTES, psm.unpack_swap_bytes);
 	 glPixelStorei(GL_UNPACK_SKIP_IMAGES, psm.unpack_skip_images);
 	 glPixelStorei(GL_UNPACK_IMAGE_HEIGHT, psm.unpack_image_height);
+	 free(srcImage);
 	 return GLU_OUT_OF_MEMORY;
        }
        /* level userLevel+1 is in srcImage; level userLevel already saved */
@@ -8232,6 +8236,7 @@ static int gluBuild3DMipmapLevelsCore(GLenum target, GLint internalFormat,
 	   glPixelStorei(GL_UNPACK_SWAP_BYTES, psm.unpack_swap_bytes);
 	   glPixelStorei(GL_UNPACK_SKIP_IMAGES, psm.unpack_skip_images);
 	   glPixelStorei(GL_UNPACK_IMAGE_HEIGHT, psm.unpack_image_height);
+	   free(srcImage);
 	   return GLU_OUT_OF_MEMORY;
 	 }
        }
diff --git a/src/mesa/drivers/dri/common/dri_util.c b/src/mesa/drivers/dri/common/dri_util.c
index e48e10d7c06..da81ec9de51 100644
--- a/src/mesa/drivers/dri/common/dri_util.c
+++ b/src/mesa/drivers/dri/common/dri_util.c
@@ -167,11 +167,12 @@ static int driBindContext(__DRIcontext *pcp,
 			  __DRIdrawable *pdp,
 			  __DRIdrawable *prp)
 {
-    __DRIscreenPrivate *psp = pcp->driScreenPriv;
+    __DRIscreenPrivate *psp;
 
     /* Bind the drawable to the context */
 
     if (pcp) {
+	psp = pcp->driScreenPriv;
 	pcp->driDrawablePriv = pdp;
 	pcp->driReadablePriv = prp;
 	if (pdp) {
@@ -498,11 +499,11 @@ static void dri_put_drawable(__DRIdrawable *pdp)
 {
     __DRIscreenPrivate *psp;
 
-    pdp->refcount--;
-    if (pdp->refcount)
-	return;
-
     if (pdp) {
+	pdp->refcount--;
+	if (pdp->refcount)
+	    return;
+
 	psp = pdp->driScreenPriv;
         (*psp->DriverAPI.DestroyBuffer)(pdp);
 	if (pdp->pClipRects) {
diff --git a/src/mesa/drivers/dri/i915/i915_context.h b/src/mesa/drivers/dri/i915/i915_context.h
index 082d6144425..25418d5f7aa 100644
--- a/src/mesa/drivers/dri/i915/i915_context.h
+++ b/src/mesa/drivers/dri/i915/i915_context.h
@@ -39,6 +39,7 @@
 #define I915_FALLBACK_LOGICOP		 0x20000
 #define I915_FALLBACK_POLYGON_SMOOTH	 0x40000
 #define I915_FALLBACK_POINT_SMOOTH	 0x80000
+#define I915_FALLBACK_POINT_SPRITE_COORD_ORIGIN	 0x100000
 
 #define I915_UPLOAD_CTX              0x1
 #define I915_UPLOAD_BUFFERS          0x2
diff --git a/src/mesa/drivers/dri/i915/i915_state.c b/src/mesa/drivers/dri/i915/i915_state.c
index b60efea75bd..cc98d125dbc 100644
--- a/src/mesa/drivers/dri/i915/i915_state.c
+++ b/src/mesa/drivers/dri/i915/i915_state.c
@@ -585,7 +585,7 @@ i915PointSize(GLcontext * ctx, GLfloat size)
 {
    struct i915_context *i915 = I915_CONTEXT(ctx);
    int lis4 = i915->state.Ctx[I915_CTXREG_LIS4] & ~S4_POINT_WIDTH_MASK;
-   GLint point_size = (int) size;
+   GLint point_size = (int) round(size);
 
    DBG("%s\n", __FUNCTION__);
    
@@ -599,6 +599,24 @@ i915PointSize(GLcontext * ctx, GLfloat size)
 }
 
 
+static void
+i915PointParameterfv(GLcontext * ctx, GLenum pname, const GLfloat *params)
+{
+   struct i915_context *i915 = I915_CONTEXT(ctx);
+
+   switch (pname) {
+   case GL_POINT_SPRITE_COORD_ORIGIN:
+      /* This could be supported, but it would require modifying the fragment
+       * program to invert the y component of the texture coordinate by
+       * inserting a 'SUB tc.y, {1.0}.xxxx, tc' instruction.
+       */
+      FALLBACK(&i915->intel, I915_FALLBACK_POINT_SPRITE_COORD_ORIGIN,
+	       (params[0] != GL_UPPER_LEFT));
+      break;
+   }
+}
+
+
 /* =============================================================
  * Color masks
  */
@@ -939,6 +957,17 @@ i915Enable(GLcontext * ctx, GLenum cap, GLboolean state)
    case GL_POLYGON_SMOOTH:
       break;
 
+   case GL_POINT_SPRITE:
+      /* This state change is handled in i915_reduced_primitive_state because
+       * the hardware bit should only be set when rendering points.
+       */
+      I915_STATECHANGE(i915, I915_UPLOAD_CTX);
+      if (state)
+	 i915->state.Ctx[I915_CTXREG_LIS4] |= S4_SPRITE_POINT_ENABLE;
+      else
+	 i915->state.Ctx[I915_CTXREG_LIS4] &= ~S4_SPRITE_POINT_ENABLE;
+      break;
+
    case GL_POINT_SMOOTH:
       break;
 
@@ -1108,6 +1137,7 @@ i915InitStateFunctions(struct dd_function_table *functions)
    functions->LineWidth = i915LineWidth;
    functions->LogicOpcode = i915LogicOp;
    functions->PointSize = i915PointSize;
+   functions->PointParameterfv = i915PointParameterfv;
    functions->PolygonStipple = i915PolygonStipple;
    functions->Scissor = i915Scissor;
    functions->ShadeModel = i915ShadeModel;
diff --git a/src/mesa/drivers/dri/intel/intel_buffer_objects.c b/src/mesa/drivers/dri/intel/intel_buffer_objects.c
index 669becdab45..3b7015b5ad3 100644
--- a/src/mesa/drivers/dri/intel/intel_buffer_objects.c
+++ b/src/mesa/drivers/dri/intel/intel_buffer_objects.c
@@ -268,6 +268,8 @@ intel_bufferobj_map(GLcontext * ctx,
 
    if (intel_obj->sys_buffer) {
       obj->Pointer = intel_obj->sys_buffer;
+      obj->Length = obj->Size;
+      obj->Offset = 0;
       return obj->Pointer;
    }
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c
index 89a99974e29..b8c65f4ce62 100644
--- a/src/mesa/drivers/dri/radeon/radeon_dma.c
+++ b/src/mesa/drivers/dri/radeon/radeon_dma.c
@@ -205,6 +205,7 @@ again_alloc:
 		   counter on unused buffers for later freeing them from
 		   begin of list */
 		dma_bo = last_elem(&rmesa->dma.free);
+		assert(dma_bo->bo->cref == 1);
 		remove_from_list(dma_bo);
 		insert_at_head(&rmesa->dma.reserved, dma_bo);
 	}
@@ -301,10 +302,6 @@ static int radeon_bo_is_idle(struct radeon_bo* bo)
 		WARN_ONCE("Your libdrm or kernel doesn't have support for busy query.\n"
 			"This may cause small performance drop for you.\n");
 	}
-	/* Protect against bug in legacy bo handling that causes bos stay
-	 * referenced even after they should be freed */
-	if (bo->cref != 1)
-		return 0;
 	return ret != -EBUSY;
 }
 
@@ -341,9 +338,7 @@ void radeonReleaseDmaRegions(radeonContextPtr rmesa)
 	foreach_s(dma_bo, temp, &rmesa->dma.wait) {
 		if (dma_bo->expire_counter == time) {
 			WARN_ONCE("Leaking dma buffer object!\n");
-			/* force free of buffer so we don't realy start
-			 * leaking stuff now*/
-			while ((dma_bo->bo = radeon_bo_unref(dma_bo->bo))) {}
+			radeon_bo_unref(dma_bo->bo);
 			remove_from_list(dma_bo);
 			FREE(dma_bo);
 			continue;
diff --git a/src/mesa/drivers/dri/radeon/radeon_lock.c b/src/mesa/drivers/dri/radeon/radeon_lock.c
index 02de8e5fd1d..7ad781ba619 100644
--- a/src/mesa/drivers/dri/radeon/radeon_lock.c
+++ b/src/mesa/drivers/dri/radeon/radeon_lock.c
@@ -62,8 +62,6 @@ void radeonGetLock(radeonContextPtr rmesa, GLuint flags)
 	__DRIdrawablePrivate *const readable = radeon_get_readable(rmesa);
 	__DRIscreenPrivate *sPriv = rmesa->dri.screen;
 
-	assert(drawable != NULL);
-
 	drmGetLock(rmesa->dri.fd, rmesa->dri.hwContext, flags);
 
 	/* The window might have moved, so we might need to get new clip
@@ -74,12 +72,13 @@ void radeonGetLock(radeonContextPtr rmesa, GLuint flags)
 	 * Since the hardware state depends on having the latest drawable
 	 * clip rects, all state checking must be done _after_ this call.
 	 */
-	DRI_VALIDATE_DRAWABLE_INFO(sPriv, drawable);
-	if (drawable != readable) {
+	if (drawable)
+		DRI_VALIDATE_DRAWABLE_INFO(sPriv, drawable);
+	if (readable && drawable != readable) {
 		DRI_VALIDATE_DRAWABLE_INFO(sPriv, readable);
 	}
 
-	if (rmesa->lastStamp != drawable->lastStamp) {
+	if (drawable && (rmesa->lastStamp != drawable->lastStamp)) {
 		radeon_window_moved(rmesa);
 		rmesa->lastStamp = drawable->lastStamp;
 	}
diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c
index f2f7b2a9fdc..bdbb9460bf0 100644
--- a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c
+++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.c
@@ -140,20 +140,19 @@ static GLuint minify(GLuint size, GLuint levels)
 
 static void calculate_miptree_layout_r100(radeonContextPtr rmesa, radeon_mipmap_tree *mt)
 {
-	GLuint curOffset;
-	GLuint i;
-	GLuint face;
+	GLuint curOffset, i, face, level;
 
 	assert(mt->numLevels <= rmesa->glCtx->Const.MaxTextureLevels);
 
 	curOffset = 0;
 	for(face = 0; face < mt->faces; face++) {
 
-		for(i = 0; i < mt->numLevels; i++) {
-			mt->levels[i].width = minify(mt->width0, i);
-			mt->levels[i].height = minify(mt->height0, i);
-			mt->levels[i].depth = minify(mt->depth0, i);
-			compute_tex_image_offset(rmesa, mt, face, i, &curOffset);
+		for(i = 0, level = mt->baseLevel; i < mt->numLevels; i++, level++) {
+			mt->levels[level].valid = 1;
+			mt->levels[level].width = minify(mt->width0, i);
+			mt->levels[level].height = minify(mt->height0, i);
+			mt->levels[level].depth = minify(mt->depth0, i);
+			compute_tex_image_offset(rmesa, mt, face, level, &curOffset);
 		}
 	}
 
@@ -163,21 +162,21 @@ static void calculate_miptree_layout_r100(radeonContextPtr rmesa, radeon_mipmap_
 
 static void calculate_miptree_layout_r300(radeonContextPtr rmesa, radeon_mipmap_tree *mt)
 {
-	GLuint curOffset;
-	GLuint i;
+	GLuint curOffset, i, level;
 
 	assert(mt->numLevels <= rmesa->glCtx->Const.MaxTextureLevels);
 
 	curOffset = 0;
-	for(i = 0; i < mt->numLevels; i++) {
+	for(i = 0, level = mt->baseLevel; i < mt->numLevels; i++, level++) {
 		GLuint face;
 
-		mt->levels[i].width = minify(mt->width0, i);
-		mt->levels[i].height = minify(mt->height0, i);
-		mt->levels[i].depth = minify(mt->depth0, i);
+		mt->levels[level].valid = 1;
+		mt->levels[level].width = minify(mt->width0, i);
+		mt->levels[level].height = minify(mt->height0, i);
+		mt->levels[level].depth = minify(mt->depth0, i);
 
 		for(face = 0; face < mt->faces; face++)
-			compute_tex_image_offset(rmesa, mt, face, i, &curOffset);
+			compute_tex_image_offset(rmesa, mt, face, level, &curOffset);
 	}
 
 	/* Note the required size in memory */
@@ -293,18 +292,19 @@ static void calculate_min_max_lod(struct gl_texture_object *tObj,
  * given face and level.
  */
 GLboolean radeon_miptree_matches_image(radeon_mipmap_tree *mt,
-		struct gl_texture_image *texImage, GLuint face, GLuint mtLevel)
+		struct gl_texture_image *texImage, GLuint face, GLuint level)
 {
 	radeon_mipmap_level *lvl;
 
-	if (face >= mt->faces || mtLevel > mt->numLevels)
+	if (face >= mt->faces)
 		return GL_FALSE;
 
 	if (texImage->TexFormat != mt->mesaFormat)
 		return GL_FALSE;
 
-	lvl = &mt->levels[mtLevel];
-	if (lvl->width != texImage->Width ||
+	lvl = &mt->levels[level];
+	if (!lvl->valid ||
+	    lvl->width != texImage->Width ||
 	    lvl->height != texImage->Height ||
 	    lvl->depth != texImage->Depth)
 		return GL_FALSE;
@@ -410,38 +410,17 @@ radeon_miptree_image_offset(radeon_mipmap_tree *mt,
 }
 
 /**
- * Convert radeon miptree texture level to GL texture level
- * @param[in] tObj texture object whom level is to be converted
- * @param[in] level radeon miptree texture level
- * @return GL texture level
- */
-unsigned radeon_miptree_level_to_gl_level(struct gl_texture_object *tObj, unsigned level)
-{
-	return level + tObj->BaseLevel;
-}
-
-/**
- * Convert GL texture level to radeon miptree texture level
- * @param[in] tObj texture object whom level is to be converted
- * @param[in] level GL texture level
- * @return radeon miptree texture level
- */
-unsigned radeon_gl_level_to_miptree_level(struct gl_texture_object *tObj, unsigned level)
-{
-	return level - tObj->BaseLevel;
-}
-
-/**
  * Ensure that the given image is stored in the given miptree from now on.
  */
 static void migrate_image_to_miptree(radeon_mipmap_tree *mt,
 									 radeon_texture_image *image,
-									 int face, int mtLevel)
+									 int face, int level)
 {
-	radeon_mipmap_level *dstlvl = &mt->levels[mtLevel];
+	radeon_mipmap_level *dstlvl = &mt->levels[level];
 	unsigned char *dest;
 
 	assert(image->mt != mt);
+	assert(dstlvl->valid);
 	assert(dstlvl->width == image->base.Width);
 	assert(dstlvl->height == image->base.Height);
 	assert(dstlvl->depth == image->base.Depth);
@@ -458,6 +437,7 @@ static void migrate_image_to_miptree(radeon_mipmap_tree *mt,
 
 		radeon_mipmap_level *srclvl = &image->mt->levels[image->mtlevel];
 
+		assert(image->mtlevel == level);
 		assert(srclvl->size == dstlvl->size);
 		assert(srclvl->rowstride == dstlvl->rowstride);
 
@@ -495,7 +475,7 @@ static void migrate_image_to_miptree(radeon_mipmap_tree *mt,
 
 	radeon_miptree_reference(mt, &image->mt);
 	image->mtface = face;
-	image->mtlevel = mtLevel;
+	image->mtlevel = level;
 }
 
 /**
@@ -522,7 +502,7 @@ static radeon_mipmap_tree * get_biggest_matching_miptree(radeonTexObj *texObj,
 		if (!img)
 			break;
 
-		if (!img->mt || !radeon_miptree_matches_texture(img->mt, &texObj->base))
+		if (!img->mt)
 			continue;
 
 		for (int i = 0; i < mtCount; ++i) {
@@ -533,8 +513,8 @@ static radeon_mipmap_tree * get_biggest_matching_miptree(radeonTexObj *texObj,
 			}
 		}
 
-		if (!found) {
-			mtSizes[mtCount] += img->mt->levels[img->mtlevel].size;
+		if (!found && radeon_miptree_matches_texture(img->mt, &texObj->base)) {
+			mtSizes[mtCount] = img->mt->levels[img->mtlevel].size;
 			mts[mtCount] = img->mt;
 			mtCount++;
 		}
@@ -592,6 +572,11 @@ int radeon_validate_texture_miptree(GLcontext * ctx, struct gl_texture_object *t
 		radeon_miptree_unreference(&t->mt);
 		radeon_try_alloc_miptree(rmesa, t);
 		dst_miptree = t->mt;
+		if (RADEON_DEBUG & RADEON_TEXTURE) {
+			fprintf(stderr, "%s: No matching miptree found, allocated new one %p\n", __FUNCTION__, t->mt);
+		}
+	} else if (RADEON_DEBUG & RADEON_TEXTURE) {
+		fprintf(stderr, "%s: Using miptree %p\n", __FUNCTION__, t->mt);
 	}
 
 	const unsigned faces = texObj->Target == GL_TEXTURE_CUBE_MAP ? 6 : 1;
@@ -610,7 +595,11 @@ int radeon_validate_texture_miptree(GLcontext * ctx, struct gl_texture_object *t
 				if (RADEON_DEBUG & RADEON_TEXTURE) {
 					fprintf(stderr, "MIGRATING\n");
 				}
-				migrate_image_to_miptree(dst_miptree, img, face, radeon_gl_level_to_miptree_level(texObj, level));
+				struct radeon_bo *src_bo = (img->mt) ? img->mt->bo : img->bo;
+				if (src_bo && radeon_bo_is_referenced_by_cs(src_bo, rmesa->cmdbuf.cs)) {
+					radeon_firevertices(rmesa);
+				}
+				migrate_image_to_miptree(dst_miptree, img, face, level);
 			} else if (RADEON_DEBUG & RADEON_TEXTURE) {
 				fprintf(stderr, "OK\n");
 			}
diff --git a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h
index 28b84850954..a10649b5aea 100644
--- a/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h
+++ b/src/mesa/drivers/dri/radeon/radeon_mipmap_tree.h
@@ -44,6 +44,7 @@ struct _radeon_mipmap_level {
 	GLuint depth;
 	GLuint size; /** Size of each image, in bytes */
 	GLuint rowstride; /** in bytes */
+	GLuint valid;
 	radeon_mipmap_image faces[6];
 };
 
@@ -70,9 +71,9 @@ struct _radeon_mipmap_tree {
 	GLuint baseLevel; /** gl_texture_object->baseLevel it was created for */
 	GLuint numLevels; /** Number of mip levels stored in this mipmap tree */
 
-	GLuint width0; /** Width of firstLevel image */
-	GLuint height0; /** Height of firstLevel image */
-	GLuint depth0; /** Depth of firstLevel image */
+	GLuint width0; /** Width of baseLevel image */
+	GLuint height0; /** Height of baseLevel image */
+	GLuint depth0; /** Depth of baseLevel image */
 
 	GLuint tilebits; /** RADEON_TXO_xxx_TILE */
 
@@ -89,8 +90,5 @@ GLuint radeon_miptree_image_offset(radeon_mipmap_tree *mt,
 				   GLuint face, GLuint level);
 void radeon_miptree_depth_offsets(radeon_mipmap_tree *mt, GLuint level, GLuint *offsets);
 
-unsigned radeon_miptree_level_to_gl_level(struct gl_texture_object *tObj, unsigned level);
-unsigned radeon_gl_level_to_miptree_level(struct gl_texture_object *tObj, unsigned level);
-
 uint32_t get_base_teximage_offset(radeonTexObj *texObj);
 #endif /* __RADEON_MIPMAP_TREE_H_ */
diff --git a/src/mesa/drivers/dri/radeon/radeon_texture.c b/src/mesa/drivers/dri/radeon/radeon_texture.c
index 1ee9e2792a5..0390d376ba2 100644
--- a/src/mesa/drivers/dri/radeon/radeon_texture.c
+++ b/src/mesa/drivers/dri/radeon/radeon_texture.c
@@ -509,6 +509,27 @@ gl_format radeonChooseTextureFormat(GLcontext * ctx,
 	return MESA_FORMAT_NONE;		/* never get here */
 }
 
+/** Check if given image is valid within current texture object.
+ */
+static int image_matches_texture_obj(struct gl_texture_object *texObj,
+	struct gl_texture_image *texImage,
+	unsigned level)
+{
+	const struct gl_texture_image *baseImage = texObj->Image[0][level];
+
+	if (level < texObj->BaseLevel || level > texObj->MaxLevel)
+		return 0;
+
+	const unsigned levelDiff = level - texObj->BaseLevel;
+	const unsigned refWidth = baseImage->Width >> levelDiff;
+	const unsigned refHeight = baseImage->Height >> levelDiff;
+	const unsigned refDepth = baseImage->Depth >> levelDiff;
+
+	return (texImage->Width == refWidth &&
+			texImage->Height == refHeight &&
+			texImage->Depth == refDepth);
+}
+
 static void teximage_assign_miptree(radeonContextPtr rmesa,
 	struct gl_texture_object *texObj,
 	struct gl_texture_image *texImage,
@@ -518,18 +539,28 @@ static void teximage_assign_miptree(radeonContextPtr rmesa,
 	radeonTexObj *t = radeon_tex_obj(texObj);
 	radeon_texture_image* image = get_radeon_texture_image(texImage);
 
+	/* Since miptree holds only images for levels <BaseLevel..MaxLevel>
+	 * don't allocate the miptree if the teximage won't fit.
+	 */
+	if (!image_matches_texture_obj(texObj, texImage, level))
+		return;
+
 	/* Try using current miptree, or create new if there isn't any */
-	if (!t->mt || !radeon_miptree_matches_image(t->mt, texImage, face,
-					radeon_gl_level_to_miptree_level(texObj, level))) {
+	if (!t->mt || !radeon_miptree_matches_image(t->mt, texImage, face, level)) {
 		radeon_miptree_unreference(&t->mt);
 		radeon_try_alloc_miptree(rmesa, t);
+		if (RADEON_DEBUG & RADEON_TEXTURE) {
+			fprintf(stderr, "%s: texObj %p, texImage %p, face %d, level %d, "
+				"texObj miptree doesn't match, allocated new miptree %p\n",
+				__FUNCTION__, texObj, texImage, face, level, t->mt);
+		}
 	}
 
 	/* Miptree alocation may have failed,
 	 * when there was no image for baselevel specified */
 	if (t->mt) {
 		image->mtface = face;
-		image->mtlevel = radeon_gl_level_to_miptree_level(texObj, level);
+		image->mtlevel = level;
 		radeon_miptree_reference(t->mt, &image->mt);
 	}
 }
@@ -585,6 +616,8 @@ static void radeon_store_teximage(GLcontext* ctx, int dims,
 		dstRowStride = _mesa_format_row_stride(texImage->TexFormat, texImage->Width);
 	}
 
+	assert(dstRowStride);
+
 	if (dims == 3) {
 		unsigned alignedWidth = dstRowStride/_mesa_get_format_bytes(texImage->TexFormat);
 		dstImageOffsets = allocate_image_offsets(ctx, alignedWidth, texImage->Height, texImage->Depth);
@@ -670,6 +703,11 @@ static void radeon_teximage(
 		}
 	}
 
+	if (RADEON_DEBUG & RADEON_TEXTURE) {
+		fprintf(stderr, "radeon_teximage%dd: texObj %p, texImage %p, face %d, level %d\n",
+				dims, texObj, texImage, face, level);
+	}
+
 	t->validated = GL_FALSE;
 
 	if (ctx->_ImageTransferState & IMAGE_CONVOLUTION_BIT) {
@@ -694,12 +732,17 @@ static void radeon_teximage(
 
 	if (!t->bo) {
 		teximage_assign_miptree(rmesa, texObj, texImage, face, level);
-		if (!t->mt) {
+		if (!image->mt) {
 			int size = _mesa_format_image_size(texImage->TexFormat,
 								texImage->Width,
 								texImage->Height,
 								texImage->Depth);
 			texImage->Data = _mesa_alloc_texmemory(size);
+			if (RADEON_DEBUG & RADEON_TEXTURE) {
+				fprintf(stderr, "radeon_teximage%dd: texObj %p, texImage %p, "
+					" no miptree assigned, using local memory %p\n",
+					dims, texObj, texImage, texImage->Data);
+			}
 		}
 	}
 
@@ -801,6 +844,11 @@ static void radeon_texsubimage(GLcontext* ctx, int dims, GLenum target, int leve
 		}
 	}
 
+	if (RADEON_DEBUG & RADEON_TEXTURE) {
+		fprintf(stderr, "radeon_texsubimage%dd: texObj %p, texImage %p, face %d, level %d\n",
+				dims, texObj, texImage, radeon_face_for_target(target), level);
+	}
+
 	t->validated = GL_FALSE;
 	if (compressed) {
 		pixels = _mesa_validate_pbo_compressed_teximage(
diff --git a/src/mesa/drivers/dri/swrast/swrast_span.c b/src/mesa/drivers/dri/swrast/swrast_span.c
index 2d3c25dcbef..f8e503463fa 100644
--- a/src/mesa/drivers/dri/swrast/swrast_span.c
+++ b/src/mesa/drivers/dri/swrast/swrast_span.c
@@ -63,56 +63,42 @@ static const GLubyte kernel[16] = {
 
 /* 32-bit BGRA */
 #define STORE_PIXEL_A8R8G8B8(DST, X, Y, VALUE) \
-   DST[3] = VALUE[ACOMP]; \
-   DST[2] = VALUE[RCOMP]; \
-   DST[1] = VALUE[GCOMP]; \
-   DST[0] = VALUE[BCOMP]
+   *DST = VALUE[ACOMP] << 24 | VALUE[RCOMP] << 16 | VALUE[GCOMP] << 8 | VALUE[BCOMP]
 #define STORE_PIXEL_RGB_A8R8G8B8(DST, X, Y, VALUE) \
-   DST[3] = 0xff; \
-   DST[2] = VALUE[RCOMP]; \
-   DST[1] = VALUE[GCOMP]; \
-   DST[0] = VALUE[BCOMP]
+   *DST = 0xff << 24 | VALUE[RCOMP] << 16 | VALUE[GCOMP] << 8 | VALUE[BCOMP]
 #define FETCH_PIXEL_A8R8G8B8(DST, SRC) \
-   DST[ACOMP] = SRC[3]; \
-   DST[RCOMP] = SRC[2]; \
-   DST[GCOMP] = SRC[1]; \
-   DST[BCOMP] = SRC[0]
+   DST[ACOMP] = *SRC >> 24;            \
+   DST[RCOMP] = (*SRC >> 16) & 0xff;   \
+   DST[GCOMP] = (*SRC >> 8) & 0xff;    \
+   DST[BCOMP] = *SRC & 0xff
 
 
 /* 32-bit BGRX */
 #define STORE_PIXEL_X8R8G8B8(DST, X, Y, VALUE) \
-   DST[3] = 0xff; \
-   DST[2] = VALUE[RCOMP]; \
-   DST[1] = VALUE[GCOMP]; \
-   DST[0] = VALUE[BCOMP]
+   *DST = 0xff << 24 | VALUE[RCOMP] << 16 | VALUE[GCOMP] << 8 | VALUE[BCOMP]
 #define STORE_PIXEL_RGB_X8R8G8B8(DST, X, Y, VALUE) \
-   DST[3] = 0xff; \
-   DST[2] = VALUE[RCOMP]; \
-   DST[1] = VALUE[GCOMP]; \
-   DST[0] = VALUE[BCOMP]
+   *DST = 0xff << 24 | VALUE[RCOMP] << 16 | VALUE[GCOMP] << 8 | VALUE[BCOMP]
 #define FETCH_PIXEL_X8R8G8B8(DST, SRC) \
-   DST[ACOMP] = 0xff; \
-   DST[RCOMP] = SRC[2]; \
-   DST[GCOMP] = SRC[1]; \
-   DST[BCOMP] = SRC[0]
+   DST[ACOMP] = 0xff;                  \
+   DST[RCOMP] = (*SRC >> 16) & 0xff;   \
+   DST[GCOMP] = (*SRC >> 8) & 0xff;    \
+   DST[BCOMP] = *SRC & 0xff
 
 
 /* 16-bit BGR */
 #define STORE_PIXEL_R5G6B5(DST, X, Y, VALUE) \
    do { \
    int d = DITHER_COMP(X, Y) >> 6; \
-   GLushort *p = (GLushort *)DST; \
-   *p = ( ((DITHER_CLAMP((VALUE[RCOMP]) + d) & 0xf8) << 8) | \
-	  ((DITHER_CLAMP((VALUE[GCOMP]) + d) & 0xfc) << 3) | \
-	  ((DITHER_CLAMP((VALUE[BCOMP]) + d) & 0xf8) >> 3) ); \
+   *DST = ( ((DITHER_CLAMP((VALUE[RCOMP]) + d) & 0xf8) << 8) | \
+            ((DITHER_CLAMP((VALUE[GCOMP]) + d) & 0xfc) << 3) | \
+            ((DITHER_CLAMP((VALUE[BCOMP]) + d) & 0xf8) >> 3) ); \
    } while(0)
 #define FETCH_PIXEL_R5G6B5(DST, SRC) \
    do { \
-   GLushort p = *(GLushort *)SRC; \
    DST[ACOMP] = 0xff; \
-   DST[RCOMP] = ((p >> 8) & 0xf8) * 255 / 0xf8; \
-   DST[GCOMP] = ((p >> 3) & 0xfc) * 255 / 0xfc; \
-   DST[BCOMP] = ((p << 3) & 0xf8) * 255 / 0xf8; \
+   DST[RCOMP] = ((*SRC >> 8) & 0xf8) * 255 / 0xf8; \
+   DST[GCOMP] = ((*SRC >> 3) & 0xfc) * 255 / 0xfc; \
+   DST[BCOMP] = ((*SRC << 3) & 0xf8) * 255 / 0xf8; \
    } while(0)
 
 
@@ -145,8 +131,8 @@ static const GLubyte kernel[16] = {
 #define SPAN_VARS \
    struct swrast_renderbuffer *xrb = swrast_renderbuffer(rb);
 #define INIT_PIXEL_PTR(P, X, Y) \
-   GLubyte *P = (GLubyte *)xrb->Base.Data + YFLIP(xrb, Y) * xrb->pitch + (X) * 4;
-#define INC_PIXEL_PTR(P) P += 4
+   GLuint *P = (GLuint *)xrb->Base.Data + YFLIP(xrb, Y) * xrb->pitch / 4 + (X)
+#define INC_PIXEL_PTR(P) P++
 #define STORE_PIXEL(DST, X, Y, VALUE) \
    STORE_PIXEL_A8R8G8B8(DST, X, Y, VALUE)
 #define STORE_PIXEL_RGB(DST, X, Y, VALUE) \
@@ -163,8 +149,8 @@ static const GLubyte kernel[16] = {
 #define SPAN_VARS \
    struct swrast_renderbuffer *xrb = swrast_renderbuffer(rb);
 #define INIT_PIXEL_PTR(P, X, Y) \
-   GLubyte *P = (GLubyte *)xrb->Base.Data + YFLIP(xrb, Y) * xrb->pitch + (X) * 4;
-#define INC_PIXEL_PTR(P) P += 4
+   GLuint *P = (GLuint *)xrb->Base.Data + YFLIP(xrb, Y) * xrb->pitch / 4 + (X);
+#define INC_PIXEL_PTR(P) P++
 #define STORE_PIXEL(DST, X, Y, VALUE) \
    STORE_PIXEL_X8R8G8B8(DST, X, Y, VALUE)
 #define STORE_PIXEL_RGB(DST, X, Y, VALUE) \
@@ -181,8 +167,8 @@ static const GLubyte kernel[16] = {
 #define SPAN_VARS \
    struct swrast_renderbuffer *xrb = swrast_renderbuffer(rb);
 #define INIT_PIXEL_PTR(P, X, Y) \
-   GLubyte *P = (GLubyte *)xrb->Base.Data + YFLIP(xrb, Y) * xrb->pitch + (X) * 2;
-#define INC_PIXEL_PTR(P) P += 2
+   GLushort *P = (GLushort *)xrb->Base.Data + YFLIP(xrb, Y) * xrb->pitch / 2 + (X);
+#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) \
@@ -234,8 +220,8 @@ static const GLubyte kernel[16] = {
 #define SPAN_VARS \
    struct swrast_renderbuffer *xrb = swrast_renderbuffer(rb);
 #define INIT_PIXEL_PTR(P, X, Y) \
-   GLubyte *P = (GLubyte *)row;
-#define INC_PIXEL_PTR(P) P += 4
+   GLuint *P = (GLuint *)row;
+#define INC_PIXEL_PTR(P) P++
 #define STORE_PIXEL(DST, X, Y, VALUE) \
    STORE_PIXEL_A8R8G8B8(DST, X, Y, VALUE)
 #define STORE_PIXEL_RGB(DST, X, Y, VALUE) \
@@ -252,8 +238,8 @@ static const GLubyte kernel[16] = {
 #define SPAN_VARS \
    struct swrast_renderbuffer *xrb = swrast_renderbuffer(rb);
 #define INIT_PIXEL_PTR(P, X, Y) \
-   GLubyte *P = (GLubyte *)row;
-#define INC_PIXEL_PTR(P) P += 4
+   GLuint *P = (GLuint *)row;
+#define INC_PIXEL_PTR(P) P++
 #define STORE_PIXEL(DST, X, Y, VALUE) \
    STORE_PIXEL_X8R8G8B8(DST, X, Y, VALUE)
 #define STORE_PIXEL_RGB(DST, X, Y, VALUE) \
@@ -270,7 +256,7 @@ static const GLubyte kernel[16] = {
 #define SPAN_VARS \
    struct swrast_renderbuffer *xrb = swrast_renderbuffer(rb);
 #define INIT_PIXEL_PTR(P, X, Y) \
-   GLubyte *P = (GLubyte *)row;
+   GLushort *P = (GLushort *)row;
 #define INC_PIXEL_PTR(P) P += 2
 #define STORE_PIXEL(DST, X, Y, VALUE) \
    STORE_PIXEL_R5G6B5(DST, X, Y, VALUE)
diff --git a/src/mesa/main/compiler.h b/src/mesa/main/compiler.h
index 380663ec971..87c3b2e5621 100644
--- a/src/mesa/main/compiler.h
+++ b/src/mesa/main/compiler.h
@@ -234,7 +234,12 @@ extern "C" {
 #elif defined(__APPLE__)
 #include <CoreFoundation/CFByteOrder.h>
 #define CPU_TO_LE32( x )	CFSwapInt32HostToLittle( x )
-#else /*__linux__ __APPLE__*/
+#elif defined(_AIX)
+#define CPU_TO_LE32( x )        x = ((x & 0x000000ff) << 24) | \
+                                    ((x & 0x0000ff00) <<  8) | \
+                                    ((x & 0x00ff0000) >>  8) | \
+                                    ((x & 0xff000000) >> 24);
+#else /*__linux__ */
 #include <sys/endian.h>
 #define CPU_TO_LE32( x )	bswap32( x )
 #endif /*__linux__*/
diff --git a/src/mesa/main/points.c b/src/mesa/main/points.c
index 4c8fc1f72e0..b3305448904 100644
--- a/src/mesa/main/points.c
+++ b/src/mesa/main/points.c
@@ -200,7 +200,12 @@ _mesa_PointParameterfv( GLenum pname, const GLfloat *params)
          }
          break;
       case GL_POINT_SPRITE_COORD_ORIGIN:
-         if (ctx->Extensions.ARB_point_sprite || ctx->Extensions.NV_point_sprite) {
+	 /* This is not completely correct.  GL_POINT_SPRITE_COORD_ORIGIN was
+	  * added to point sprites when the extension was merged into OpenGL
+	  * 2.0.  It is expected that an implementation supporting OpenGL 1.4
+	  * and GL_ARB_point_sprite will generate an error here.
+	  */
+         if (ctx->Extensions.ARB_point_sprite) {
             GLenum value = (GLenum) params[0];
             if (value != GL_LOWER_LEFT && value != GL_UPPER_LEFT) {
                _mesa_error(ctx, GL_INVALID_VALUE,
diff --git a/src/mesa/main/texenv.c b/src/mesa/main/texenv.c
index 6d86a4275cc..4442fb8cf8e 100644
--- a/src/mesa/main/texenv.c
+++ b/src/mesa/main/texenv.c
@@ -598,7 +598,10 @@ _mesa_TexEnvfv( GLenum target, GLenum pname, const GLfloat *param )
 void GLAPIENTRY
 _mesa_TexEnvf( GLenum target, GLenum pname, GLfloat param )
 {
-   _mesa_TexEnvfv( target, pname, &param );
+   GLfloat p[4];
+   p[0] = param;
+   p[1] = p[2] = p[3] = 0.0;
+   _mesa_TexEnvfv( target, pname, p );
 }
 
 
diff --git a/src/mesa/main/texgen.c b/src/mesa/main/texgen.c
index 733e129fcfe..5abb1ff0ab6 100644
--- a/src/mesa/main/texgen.c
+++ b/src/mesa/main/texgen.c
@@ -211,14 +211,20 @@ _mesa_TexGendv(GLenum coord, GLenum pname, const GLdouble *params )
 static void GLAPIENTRY
 _mesa_TexGenf( GLenum coord, GLenum pname, GLfloat param )
 {
-   _mesa_TexGenfv(coord, pname, &param);
+   GLfloat p[4];
+   p[0] = param;
+   p[1] = p[2] = p[3] = 0.0F;
+   _mesa_TexGenfv(coord, pname, p);
 }
 
 
 void GLAPIENTRY
 _mesa_TexGeni( GLenum coord, GLenum pname, GLint param )
 {
-   _mesa_TexGeniv( coord, pname, &param );
+   GLint p[4];
+   p[0] = param;
+   p[1] = p[2] = p[3] = 0;
+   _mesa_TexGeniv( coord, pname, p );
 }
 
 
diff --git a/src/mesa/main/texparam.c b/src/mesa/main/texparam.c
index ab170cde35b..310d594cd55 100644
--- a/src/mesa/main/texparam.c
+++ b/src/mesa/main/texparam.c
@@ -547,13 +547,20 @@ _mesa_TexParameterf(GLenum target, GLenum pname, GLfloat param)
    case GL_DEPTH_TEXTURE_MODE_ARB:
       {
          /* convert float param to int */
-         GLint p = (GLint) param;
-         need_update = set_tex_parameteri(ctx, texObj, pname, &p);
+         GLint p[4];
+         p[0] = (GLint) param;
+         p[1] = p[2] = p[3] = 0;
+         need_update = set_tex_parameteri(ctx, texObj, pname, p);
       }
       break;
    default:
-      /* this will generate an error if pname is illegal */
-      need_update = set_tex_parameterf(ctx, texObj, pname, &param);
+      {
+         /* this will generate an error if pname is illegal */
+         GLfloat p[4];
+         p[0] = param;
+         p[1] = p[2] = p[3] = 0.0F;
+         need_update = set_tex_parameterf(ctx, texObj, pname, p);
+      }
    }
 
    if (ctx->Driver.TexParameter && need_update) {
diff --git a/windows/VC6/progs/glut/glut.dsp b/windows/VC6/progs/glut/glut.dsp
deleted file mode 100644
index 0a67689d786..00000000000
--- a/windows/VC6/progs/glut/glut.dsp
+++ /dev/null
@@ -1,333 +0,0 @@
-# Microsoft Developer Studio Project File - Name="glut" - Package Owner=<4>

-# Microsoft Developer Studio Generated Build File, Format Version 6.00

-# ** DO NOT EDIT **

-

-# TARGTYPE "Win32 (x86) Dynamic-Link Library" 0x0102

-

-CFG=glut - Win32 Debug

-!MESSAGE This is not a valid makefile. To build this project using NMAKE,

-!MESSAGE use the Export Makefile command and run

-!MESSAGE 

-!MESSAGE NMAKE /f "glut.mak".

-!MESSAGE 

-!MESSAGE You can specify a configuration when running NMAKE

-!MESSAGE by defining the macro CFG on the command line. For example:

-!MESSAGE 

-!MESSAGE NMAKE /f "glut.mak" CFG="glut - Win32 Debug"

-!MESSAGE 

-!MESSAGE Possible choices for configuration are:

-!MESSAGE 

-!MESSAGE "glut - Win32 Release" (based on "Win32 (x86) Dynamic-Link Library")

-!MESSAGE "glut - Win32 Debug" (based on "Win32 (x86) Dynamic-Link Library")

-!MESSAGE 

-

-# Begin Project

-# PROP AllowPerConfigDependencies 0

-# PROP Scc_ProjName ""

-# PROP Scc_LocalPath ""

-CPP=cl.exe

-MTL=midl.exe

-RSC=rc.exe

-

-!IF  "$(CFG)" == "glut - Win32 Release"

-

-# PROP BASE Use_MFC 0

-# PROP BASE Use_Debug_Libraries 0

-# PROP BASE Output_Dir "Release"

-# PROP BASE Intermediate_Dir "Release"

-# PROP BASE Target_Dir ""

-# PROP Use_MFC 0

-# PROP Use_Debug_Libraries 0

-# PROP Output_Dir "Release"

-# PROP Intermediate_Dir "Release"

-# PROP Ignore_Export_Lib 0

-# PROP Target_Dir ""

-# ADD BASE CPP /nologo /MT /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_WINDOWS" /D "_MBCS" /D "_USRDLL" /D "GLUT_EXPORTS" /YX /FD /c

-# ADD CPP /nologo /MT /W3 /GX /O2 /I "../../../../include" /D "NDEBUG" /D "WIN32" /D "_WINDOWS" /D "_MBCS" /D "_DLL" /D "_USRDLL" /D "GLUT_EXPORTS" /D "MESA" /D "BUILD_GL32" /FD /c

-# SUBTRACT CPP /YX

-# ADD BASE MTL /nologo /D "NDEBUG" /mktyplib203 /win32

-# ADD MTL /nologo /D "NDEBUG" /mktyplib203 /win32

-# ADD BASE RSC /l 0x409 /d "NDEBUG"

-# ADD RSC /l 0x409 /d "NDEBUG"

-BSC32=bscmake.exe

-# ADD BASE BSC32 /nologo

-# ADD BSC32 /nologo

-LINK32=link.exe

-# ADD BASE LINK32 kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib /nologo /dll /machine:I386

-# ADD LINK32 opengl32.lib glu32.lib winmm.lib msvcrt.lib oldnames.lib kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib /nologo /dll /machine:I386 /nodefaultlib /out:"Release/GLUT32.DLL" /libpath:"../../mesa/Release"

-# Begin Special Build Tool

-SOURCE="$(InputPath)"

-PostBuild_Cmds=if not exist ..\..\..\..\lib md ..\..\..\..\lib	copy Release\GLUT32.LIB ..\..\..\..\lib	copy Release\GLUT32.DLL ..\..\..\..\lib	if exist ..\..\..\..\progs\demos copy Release\GLUT32.DLL ..\..\..\..\progs\demos

-# End Special Build Tool

-

-!ELSEIF  "$(CFG)" == "glut - Win32 Debug"

-

-# PROP BASE Use_MFC 0

-# PROP BASE Use_Debug_Libraries 1

-# PROP BASE Output_Dir "Debug"

-# PROP BASE Intermediate_Dir "Debug"

-# PROP BASE Target_Dir ""

-# PROP Use_MFC 0

-# PROP Use_Debug_Libraries 1

-# PROP Output_Dir "Debug"

-# PROP Intermediate_Dir "Debug"

-# PROP Ignore_Export_Lib 0

-# PROP Target_Dir ""

-# ADD BASE CPP /nologo /MTd /W3 /Gm /GX /ZI /Od /D "WIN32" /D "_DEBUG" /D "_WINDOWS" /D "_MBCS" /D "_USRDLL" /D "GLUT_EXPORTS" /YX /FD /GZ /c

-# ADD CPP /nologo /MTd /W3 /Gm /GX /Zi /Od /I "../../../../include" /D "_DEBUG" /D "WIN32" /D "_WINDOWS" /D "_MBCS" /D "_DLL" /D "_USRDLL" /D "GLUT_EXPORTS" /D "MESA" /D "BUILD_GL32" /FD /GZ /c

-# SUBTRACT CPP /YX

-# ADD BASE MTL /nologo /D "_DEBUG" /mktyplib203 /win32

-# ADD MTL /nologo /D "_DEBUG" /mktyplib203 /win32

-# ADD BASE RSC /l 0x409 /d "_DEBUG"

-# ADD RSC /l 0x409 /d "_DEBUG"

-BSC32=bscmake.exe

-# ADD BASE BSC32 /nologo

-# ADD BSC32 /nologo

-LINK32=link.exe

-# ADD BASE LINK32 kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib /nologo /dll /debug /machine:I386 /pdbtype:sept

-# ADD LINK32 winmm.lib msvcrtd.lib oldnames.lib kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib opengl32.lib glu32.lib /nologo /dll /incremental:no /debug /machine:I386 /nodefaultlib /out:"Debug/GLUT32.DLL" /pdbtype:sept /libpath:"../../mesa/Debug"

-# Begin Special Build Tool

-SOURCE="$(InputPath)"

-PostBuild_Cmds=if not exist ..\..\..\..\lib md ..\..\..\..\lib	copy Debug\GLUT32.LIB ..\..\..\..\lib	copy Debug\GLUT32.DLL ..\..\..\..\lib	if exist ..\..\..\..\progs\demos copy Debug\GLUT32.DLL ..\..\..\..\progs\demos

-# End Special Build Tool

-

-!ENDIF 

-

-# Begin Target

-

-# Name "glut - Win32 Release"

-# Name "glut - Win32 Debug"

-# Begin Group "Source Files"

-

-# PROP Default_Filter "cpp;c;cxx;rc;def;r;odl;idl;hpj;bat"

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_8x13.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_9x15.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_bitmap.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_bwidth.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_cindex.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_cmap.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_cursor.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_dials.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_dstr.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_event.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_ext.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_fcb.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_fullscrn.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_gamemode.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_get.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_hel10.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_hel12.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_hel18.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_init.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_input.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_joy.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_key.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_keyctrl.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_keyup.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_mesa.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_modifier.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_mroman.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_overlay.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_roman.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_shapes.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_space.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_stroke.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_swap.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_swidth.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_tablet.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_teapot.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_tr10.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_tr24.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_util.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_vidresize.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_warp.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_win.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glut_winmisc.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\win32_glx.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\win32_menu.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\win32_util.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\win32_winproc.c

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\win32_x11.c

-# End Source File

-# End Group

-# Begin Group "Header Files"

-

-# PROP Default_Filter "h;hpp;hxx;hm;inl"

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glutbitmap.h

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glutint.h

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glutstroke.h

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\glutwin32.h

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\stroke.h

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\win32_glx.h

-# End Source File

-# Begin Source File

-

-SOURCE=..\..\..\..\src\glut\glx\win32_x11.h

-# End Source File

-# End Group

-# Begin Group "Resource Files"

-

-# PROP Default_Filter "ico;cur;bmp;dlg;rc2;rct;bin;rgs;gif;jpg;jpeg;jpe"

-# End Group

-# End Target

-# End Project