Merge branch 'master' of git+ssh://git.freedesktop.org/git/mesa/mesa into gallium-0.2

Conflicts:

	src/mesa/drivers/dri/i915/intel_ioctl.c
	src/mesa/main/texstore.c
	src/mesa/tnl/t_vp_build.c
	src/mesa/vbo/vbo_exec_draw.c

69 files changed, 1322 insertions, 1756 deletions

diff --git a/src/mesa/Makefile.mgw b/src/mesa/Makefile.mgw
index 6244ded8768..886b344ec9b 100644
--- a/src/mesa/Makefile.mgw
+++ b/src/mesa/Makefile.mgw
@@ -154,7 +154,7 @@ DRIVER_SOURCES = \
 endif
 endif
 
-SOURCES = $(CORE_SOURCES) $(X86_SOURCES) $(COMMON_DRIVER_SOURCES) $(DRIVER_SOURCES)
+SOURCES = $(MESA_SOURCES) $(GLAPI_SOURCES) $(X86_SOURCES) $(DRIVER_SOURCES)
 
 OBJECTS = $(addsuffix .o,$(basename $(SOURCES)))
 
diff --git a/src/mesa/drivers/dri/Makefile b/src/mesa/drivers/dri/Makefile
index 3b3fcf11df1..eef68825bc9 100644
--- a/src/mesa/drivers/dri/Makefile
+++ b/src/mesa/drivers/dri/Makefile
@@ -40,8 +40,6 @@ install: dri.pc
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_INC_DIR)/GL/internal
 	$(INSTALL) -m 0644 $(TOP)/include/GL/internal/dri_interface.h \
 	  $(DESTDIR)$(INSTALL_INC_DIR)/GL/internal
-	$(INSTALL) -m 0644 $(TOP)/include/GL/internal/dri_sarea.h \
-	  $(DESTDIR)$(INSTALL_INC_DIR)/GL/internal
 	$(INSTALL) -d $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 	$(INSTALL) -m 0644 dri.pc $(DESTDIR)$(INSTALL_LIB_DIR)/pkgconfig
 
diff --git a/src/mesa/drivers/dri/common/dri_util.c b/src/mesa/drivers/dri/common/dri_util.c
index e68faa3f0a0..66b3d619fa6 100644
--- a/src/mesa/drivers/dri/common/dri_util.c
+++ b/src/mesa/drivers/dri/common/dri_util.c
@@ -221,7 +221,6 @@ static int driUnbindContext(__DRIcontext *pcp)
     return GL_TRUE;
 }
 
-
 /**
  * This function takes both a read buffer and a draw buffer.  This is needed
  * for \c glXMakeCurrentReadSGI or GLX 1.3's \c glXMakeContextCurrent
@@ -255,10 +254,7 @@ static int driBindContext(__DRIcontext *pcp,
     ** initialize the drawable information if has not been done before.
     */
 
-    if (psp->dri2.enabled) {
-       __driParseEvents(pcp, pdp);
-       __driParseEvents(pcp, prp);
-    } else {
+    if (!psp->dri2.enabled) {
 	if (!pdp->pStamp || *pdp->pStamp != pdp->lastStamp) {
 	    DRM_SPINLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID);
 	    __driUtilUpdateDrawableInfo(pdp);
@@ -348,139 +344,6 @@ __driUtilUpdateDrawableInfo(__DRIdrawablePrivate *pdp)
     DRM_SPINLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID);
 }
 
-
-int
-__driParseEvents(__DRIcontextPrivate *pcp, __DRIdrawablePrivate *pdp)
-{
-    __DRIscreenPrivate *psp = pdp->driScreenPriv;
-    __DRIDrawableConfigEvent *dc, *last_dc;
-    __DRIBufferAttachEvent *ba, *last_ba;
-    unsigned int tail, mask, *p, end, total, size, changed;
-    unsigned char *data;
-    size_t rect_size;
-
-    /* Check for wraparound. */
-    if (pcp && psp->dri2.buffer->prealloc - pdp->dri2.tail > psp->dri2.buffer->size) {
-       /* If prealloc overlaps into what we just parsed, the
-	* server overwrote it and we have to reset our tail
-	* pointer. */
-	DRM_UNLOCK(psp->fd, psp->lock, pcp->hHWContext);
-	(*psp->dri2.loader->reemitDrawableInfo)(pdp, &pdp->dri2.tail,
-						pdp->loaderPrivate);
-	DRM_LIGHT_LOCK(psp->fd, psp->lock, pcp->hHWContext);
-    }
-
-    total = psp->dri2.buffer->head - pdp->dri2.tail;
-    mask = psp->dri2.buffer->size - 1;
-    end = psp->dri2.buffer->head;
-    data = psp->dri2.buffer->data;
-
-    changed = 0;
-    last_dc = NULL;
-    last_ba = NULL;
-
-    for (tail = pdp->dri2.tail; tail != end; tail += size) {
-       p = (unsigned int *) (data + (tail & mask));
-       size = DRI2_EVENT_SIZE(*p);
-       if (size > total || (tail & mask) + size > psp->dri2.buffer->size) {
-	  /* illegal data, bail out. */
-	  fprintf(stderr, "illegal event size\n");
-	  break;
-       }
-
-       switch (DRI2_EVENT_TYPE(*p)) {
-       case DRI2_EVENT_DRAWABLE_CONFIG:
-	  dc = (__DRIDrawableConfigEvent *) p;
-	  if (dc->drawable == pdp->dri2.drawable_id)
-	     last_dc = dc;
-	  break;
-
-       case DRI2_EVENT_BUFFER_ATTACH:
-	  ba = (__DRIBufferAttachEvent *) p;
-	  if (ba->drawable == pdp->dri2.drawable_id && 
-	      ba->buffer.attachment == DRI_DRAWABLE_BUFFER_FRONT_LEFT)
-	     last_ba = ba;
-	  break;
-       }
-    }
-	  
-    if (last_dc) {
-       if (pdp->w != last_dc->width || pdp->h != last_dc->height)
-	  changed = 1;
-
-       pdp->x = last_dc->x;
-       pdp->y = last_dc->y;
-       pdp->w = last_dc->width;
-       pdp->h = last_dc->height;
-
-       pdp->backX = 0;
-       pdp->backY = 0;
-       pdp->numBackClipRects = 1;
-       pdp->pBackClipRects[0].x1 = 0;
-       pdp->pBackClipRects[0].y1 = 0;
-       pdp->pBackClipRects[0].x2 = pdp->w;
-       pdp->pBackClipRects[0].y2 = pdp->h;
-
-       pdp->numClipRects = last_dc->num_rects;
-       _mesa_free(pdp->pClipRects);
-       rect_size = last_dc->num_rects * sizeof last_dc->rects[0];
-       pdp->pClipRects = _mesa_malloc(rect_size);
-       memcpy(pdp->pClipRects, last_dc->rects, rect_size);
-    }
-
-    /* We only care about the most recent drawable config. */
-    if (last_dc && changed)
-       (*psp->DriverAPI.HandleDrawableConfig)(pdp, pcp, last_dc);
-
-    /* Front buffer attachments are special, they typically mean that
-     * we're rendering to a redirected window (or a child window of a
-     * redirected window) and that it got resized.  Resizing the root
-     * window on randr events is a special case of this.  Other causes
-     * may be a window transitioning between redirected and
-     * non-redirected, or a window getting reparented between parents
-     * with different window pixmaps (eg two redirected windows).
-     * These events are special in that the X server allocates the
-     * buffer and that the buffer may be shared by other child
-     * windows.  When our window share the window pixmap with its
-     * parent, drawable config events doesn't affect the front buffer.
-     * We only care about the last such event in the buffer; in fact,
-     * older events will refer to invalid buffer objects.*/
-    if (last_ba)
-       (*psp->DriverAPI.HandleBufferAttach)(pdp, pcp, last_ba);
-
-    /* If there was a drawable config event in the buffer and it
-     * changed the size of the window, all buffer auxillary buffer
-     * attachments prior to that are invalid (as opposed to the front
-     * buffer case discussed above).  In that case we can start
-     * looking for buffer attachment after the last drawable config
-     * event.  If there is no drawable config event in this batch of
-     * events, we have to assume that the last batch might have had
-     * one and process all buffer attach events.*/
-    if (last_dc && changed)
-       tail = (unsigned char *) last_dc - data;
-    else
-       tail = pdp->dri2.tail;
-
-    for ( ; tail != end; tail += size) {
-       ba = (__DRIBufferAttachEvent *) (data + (tail & mask));
-       size = DRI2_EVENT_SIZE(ba->event_header);
-
-       if (DRI2_EVENT_TYPE(ba->event_header) != DRI2_EVENT_BUFFER_ATTACH)
-	  continue;
-       if (ba->drawable != pdp->dri2.drawable_id)
-	  continue;
-       if (last_ba == ba)
-	  continue;
-
-       (*psp->DriverAPI.HandleBufferAttach)(pdp, pcp, ba);
-       changed = 1;
-    }
-
-    pdp->dri2.tail = tail;
-
-    return changed || last_ba;
-}
-
 /*@}*/
 
 /*****************************************************************/
@@ -494,12 +357,7 @@ static void driReportDamage(__DRIdrawable *pdp,
     __DRIscreen *psp = pdp->driScreenPriv;
 
     /* Check that we actually have the new damage report method */
-    if (psp->dri2.enabled) {
-	(*psp->dri2.loader->postDamage)(pdp,
-					pClipRects,
-					numClipRects,
-					pdp->loaderPrivate);
-    } else if (psp->damage) {
+    if (psp->damage) {
 	/* Report the damage.  Currently, all our drivers draw
 	 * directly to the front buffer, so we report the damage there
 	 * rather than to the backing storein (if any).
@@ -529,9 +387,6 @@ static void driSwapBuffers(__DRIdrawable *dPriv)
     if (!dPriv->numClipRects)
         return;
 
-    if (psp->dri2.enabled)
-       __driParseEvents(NULL, dPriv);
-
     psp->DriverAPI.SwapBuffers(dPriv);
 
     driReportDamage(dPriv, dPriv->pClipRects, dPriv->numClipRects);
@@ -671,17 +526,17 @@ driCreateNewDrawable(__DRIscreen *psp, const __DRIconfig *config,
 
 
 static __DRIdrawable *
-dri2CreateNewDrawable(__DRIscreen *screen, const __DRIconfig *config,
-		      unsigned int drawable_id, unsigned int head, void *data)
+dri2CreateNewDrawable(__DRIscreen *screen,
+		      const __DRIconfig *config,
+		      void *loaderPrivate)
 {
     __DRIdrawable *pdraw;
 
-    pdraw = driCreateNewDrawable(screen, config, 0, 0, NULL, data);
+    pdraw = driCreateNewDrawable(screen, config, 0, 0, NULL, loaderPrivate);
     if (!pdraw)
     	return NULL;
 
-    pdraw->dri2.drawable_id = drawable_id;
-    pdraw->dri2.tail = head;
+    pdraw->pClipRects = _mesa_malloc(sizeof *pdraw->pBackClipRects);
     pdraw->pBackClipRects = _mesa_malloc(sizeof *pdraw->pBackClipRects);
 
     return pdraw;
@@ -792,18 +647,7 @@ static __DRIcontext *
 dri2CreateNewContext(__DRIscreen *screen, const __DRIconfig *config,
 		      __DRIcontext *shared, void *data)
 {
-    drm_context_t hwContext;
-    DRM_CAS_RESULT(ret);
-
-    /* DRI2 doesn't use kernel with context IDs, we just need an ID that's
-     * different from the kernel context ID to make drmLock() happy. */
-
-    do {
-	hwContext = screen->dri2.lock->next_id;
-	DRM_CAS(&screen->dri2.lock->next_id, hwContext, hwContext + 1, ret);
-    } while (ret);
-
-    return driCreateNewContext(screen, config, 0, shared, hwContext, data);
+    return driCreateNewContext(screen, config, 0, shared, 0, data);
 }
 
 
@@ -839,12 +683,7 @@ static void driDestroyScreen(__DRIscreen *psp)
 	if (psp->DriverAPI.DestroyScreen)
 	    (*psp->DriverAPI.DestroyScreen)(psp);
 
-	if (psp->dri2.enabled) {
-#ifdef TTM_API
-	    drmBOUnmap(psp->fd, &psp->dri2.sareaBO);
-	    drmBOUnreference(psp->fd, &psp->dri2.sareaBO);
-#endif
-	} else {
+	if (!psp->dri2.enabled) {
 	   (void)drmUnmap((drmAddress)psp->pSAREA, SAREA_MAX);
 	   (void)drmUnmap((drmAddress)psp->pFB, psp->fbSize);
 	   (void)drmCloseOnce(psp->fd);
@@ -867,8 +706,8 @@ setupLoaderExtensions(__DRIscreen *psp,
 	    psp->damage = (__DRIdamageExtension *) extensions[i];
 	if (strcmp(extensions[i]->name, __DRI_SYSTEM_TIME) == 0)
 	    psp->systemTime = (__DRIsystemTimeExtension *) extensions[i];
-	if (strcmp(extensions[i]->name, __DRI_LOADER) == 0)
-	    psp->dri2.loader = (__DRIloaderExtension *) extensions[i];
+	if (strcmp(extensions[i]->name, __DRI_DRI2_LOADER) == 0)
+	    psp->dri2.loader = (__DRIdri2LoaderExtension *) extensions[i];
     }
 }
 
@@ -964,19 +803,16 @@ driCreateNewScreen(int scrn,
     return psp;
 }
 
-
 /**
  * DRI2
  */
 static __DRIscreen *
-dri2CreateNewScreen(int scrn, int fd, unsigned int sarea_handle,
+dri2CreateNewScreen(int scrn, int fd,
 		    const __DRIextension **extensions,
 		    const __DRIconfig ***driver_configs, void *data)
 {
-#ifdef TTM_API
     static const __DRIextension *emptyExtensionList[] = { NULL };
     __DRIscreen *psp;
-    unsigned int *p;
     drmVersionPtr version;
 
     if (driDriverAPI.InitScreen2 == NULL)
@@ -1001,39 +837,9 @@ dri2CreateNewScreen(int scrn, int fd, unsigned int sarea_handle,
     psp->myNum = scrn;
     psp->dri2.enabled = GL_TRUE;
 
-    if (drmBOReference(psp->fd, sarea_handle, &psp->dri2.sareaBO)) {
-	fprintf(stderr, "Failed to reference DRI2 sarea BO\n");
-	_mesa_free(psp);
-	return NULL;
-    }
-
-    if (drmBOMap(psp->fd, &psp->dri2.sareaBO,
-		 DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0, &psp->dri2.sarea)) {
-	drmBOUnreference(psp->fd, &psp->dri2.sareaBO);
-	_mesa_free(psp);
-	return NULL;
-    }
-
-    p = psp->dri2.sarea;
-    while (DRI2_SAREA_BLOCK_TYPE(*p)) {
-	switch (DRI2_SAREA_BLOCK_TYPE(*p)) {
-	case DRI2_SAREA_BLOCK_LOCK:
-	    psp->dri2.lock = (__DRILock *) p;
-	    break;
-	case DRI2_SAREA_BLOCK_EVENT_BUFFER:
-	    psp->dri2.buffer = (__DRIEventBuffer *) p;
-	    break;
-	}
-	p = DRI2_SAREA_BLOCK_NEXT(p);
-    }
-
-    psp->lock = (drmLock *) &psp->dri2.lock->lock;
-
     psp->DriverAPI = driDriverAPI;
     *driver_configs = driDriverAPI.InitScreen2(psp);
     if (*driver_configs == NULL) {
-	drmBOUnmap(psp->fd, &psp->dri2.sareaBO);
-	drmBOUnreference(psp->fd, &psp->dri2.sareaBO);
 	_mesa_free(psp);
 	return NULL;
     }
@@ -1041,9 +847,6 @@ dri2CreateNewScreen(int scrn, int fd, unsigned int sarea_handle,
     psp->DriverAPI = driDriverAPI;
 
     return psp;
-#else
-    return NULL;
-#endif
 }
 
 static const __DRIextension **driGetExtensions(__DRIscreen *psp)
@@ -1051,36 +854,45 @@ static const __DRIextension **driGetExtensions(__DRIscreen *psp)
     return psp->extensions;
 }
 
-/** Legacy DRI interface */
-const __DRIlegacyExtension driLegacyExtension = {
-    { __DRI_LEGACY, __DRI_LEGACY_VERSION },
-    driCreateNewScreen,
-    driCreateNewDrawable,
-    driCreateNewContext
-};
-
-/** DRI2 interface */
+/** Core interface */
 const __DRIcoreExtension driCoreExtension = {
     { __DRI_CORE, __DRI_CORE_VERSION },
-    dri2CreateNewScreen,
+    NULL,
     driDestroyScreen,
     driGetExtensions,
     driGetConfigAttrib,
     driIndexConfigAttrib,
-    dri2CreateNewDrawable,
+    NULL,
     driDestroyDrawable,
     driSwapBuffers,
-    dri2CreateNewContext,
+    NULL,
     driCopyContext,
     driDestroyContext,
     driBindContext,
     driUnbindContext
 };
 
+/** Legacy DRI interface */
+const __DRIlegacyExtension driLegacyExtension = {
+    { __DRI_LEGACY, __DRI_LEGACY_VERSION },
+    driCreateNewScreen,
+    driCreateNewDrawable,
+    driCreateNewContext,
+};
+
+/** Legacy DRI interface */
+const __DRIdri2Extension driDRI2Extension = {
+    { __DRI_DRI2, __DRI_DRI2_VERSION },
+    dri2CreateNewScreen,
+    dri2CreateNewDrawable,
+    dri2CreateNewContext,
+};
+
 /* This is the table of extensions that the loader will dlsym() for. */
 PUBLIC const __DRIextension *__driDriverExtensions[] = {
     &driCoreExtension.base,
     &driLegacyExtension.base,
+    &driDRI2Extension.base,
     NULL
 };
 
diff --git a/src/mesa/drivers/dri/common/dri_util.h b/src/mesa/drivers/dri/common/dri_util.h
index abd1482e913..1a8f020c778 100644
--- a/src/mesa/drivers/dri/common/dri_util.h
+++ b/src/mesa/drivers/dri/common/dri_util.h
@@ -54,7 +54,6 @@
 #include "glheader.h"
 #include "GL/internal/glcore.h"
 #include "GL/internal/dri_interface.h"
-#include "GL/internal/dri_sarea.h"
 
 #define GLX_BAD_CONTEXT                    5
 
@@ -202,16 +201,8 @@ struct __DriverAPIRec {
 
 
 
-    /* DRI2 Entry points */
+    /* DRI2 Entry point */
     const __DRIconfig **(*InitScreen2) (__DRIscreen * priv);
-    void (*HandleDrawableConfig)(__DRIdrawable *dPriv,
-				__DRIcontext *pcp,
-				__DRIDrawableConfigEvent *event);
-
-    void (*HandleBufferAttach)(__DRIdrawable *dPriv,
-			       __DRIcontext *pcp,
-			       __DRIBufferAttachEvent *ba);
-
 };
 
 extern const struct __DriverAPIRec driDriverAPI;
@@ -367,10 +358,6 @@ struct __DRIdrawableRec {
      * GLX_MESA_swap_control.
      */
     unsigned int swap_interval;
-    struct {
-	unsigned int tail;
-	unsigned int drawable_id;
-    } dri2;
 };
 
 /**
@@ -522,13 +509,7 @@ struct __DRIscreenRec {
 	/* Flag to indicate that this is a DRI2 screen.  Many of the above
 	 * fields will not be valid or initializaed in that case. */
 	int enabled;
-#ifdef TTM_API
-	drmBO sareaBO;
-#endif
-	void *sarea;
-	__DRIEventBuffer *buffer;
-	__DRILock *lock;
-	__DRIloaderExtension *loader;
+	__DRIdri2LoaderExtension *loader;
     } dri2;
 
     /* The lock actually in use, old sarea or DRI2 */
@@ -542,9 +523,6 @@ __driUtilMessage(const char *f, ...);
 extern void
 __driUtilUpdateDrawableInfo(__DRIdrawable *pdp);
 
-extern int
-__driParseEvents(__DRIcontext *psp, __DRIdrawable *pdp);
-
 extern float
 driCalculateSwapUsage( __DRIdrawable *dPriv,
 		       int64_t last_swap_ust, int64_t current_ust );
diff --git a/src/mesa/drivers/dri/i915/Makefile b/src/mesa/drivers/dri/i915/Makefile
index fd77980ebf8..5858e0ee9fd 100644
--- a/src/mesa/drivers/dri/i915/Makefile
+++ b/src/mesa/drivers/dri/i915/Makefile
@@ -46,7 +46,6 @@ DRIVER_SOURCES = \
 	i915_vtbl.c \
 	intel_context.c \
 	intel_decode.c \
-	intel_ioctl.c \
 	intel_screen.c \
 	intel_span.c \
 	intel_state.c \
@@ -64,6 +63,8 @@ DRIVER_DEFINES = -I../intel -I../intel/server -DI915 \
 	$(shell pkg-config libdrm --atleast-version=2.3.1 \
 				&& echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP")
 
+DRI_LIB_DEPS += -ldrm_intel
+
 include ../Makefile.template
 
 intel_decode.o: ../intel/intel_decode.c
diff --git a/src/mesa/drivers/dri/i915/i830_metaops.c b/src/mesa/drivers/dri/i915/i830_metaops.c
index 13e4ab3aaca..acfb3af9102 100644
--- a/src/mesa/drivers/dri/i915/i830_metaops.c
+++ b/src/mesa/drivers/dri/i915/i830_metaops.c
@@ -33,7 +33,6 @@
 
 #include "intel_screen.h"
 #include "intel_batchbuffer.h"
-#include "intel_ioctl.h"
 #include "intel_regions.h"
 
 #include "i830_context.h"
diff --git a/src/mesa/drivers/dri/i915/i830_tex.c b/src/mesa/drivers/dri/i915/i830_tex.c
index 79b0fcf596f..c0126ba19a4 100644
--- a/src/mesa/drivers/dri/i915/i830_tex.c
+++ b/src/mesa/drivers/dri/i915/i830_tex.c
@@ -38,8 +38,6 @@
 
 #include "mm.h"
 
-#include "intel_ioctl.h"
-
 #include "i830_context.h"
 #include "i830_reg.h"
 
diff --git a/src/mesa/drivers/dri/i915/i830_texblend.c b/src/mesa/drivers/dri/i915/i830_texblend.c
index 58f220eb7ce..dca72adecad 100644
--- a/src/mesa/drivers/dri/i915/i830_texblend.c
+++ b/src/mesa/drivers/dri/i915/i830_texblend.c
@@ -36,7 +36,6 @@
 #include "mm.h"
 
 #include "intel_screen.h"
-#include "intel_ioctl.h"
 #include "intel_tex.h"
 
 #include "i830_context.h"
diff --git a/src/mesa/drivers/dri/i915/i830_texstate.c b/src/mesa/drivers/dri/i915/i830_texstate.c
index 4e9b022eaeb..314bbe32b65 100644
--- a/src/mesa/drivers/dri/i915/i830_texstate.c
+++ b/src/mesa/drivers/dri/i915/i830_texstate.c
@@ -28,7 +28,6 @@
 #include "mtypes.h"
 #include "enums.h"
 #include "texformat.h"
-#include "dri_bufmgr.h"
 
 #include "intel_mipmap_tree.h"
 #include "intel_tex.h"
diff --git a/src/mesa/drivers/dri/i915/i915_metaops.c b/src/mesa/drivers/dri/i915/i915_metaops.c
index 73aa63451e0..bfe97f13e57 100644
--- a/src/mesa/drivers/dri/i915/i915_metaops.c
+++ b/src/mesa/drivers/dri/i915/i915_metaops.c
@@ -33,7 +33,6 @@
 
 #include "intel_screen.h"
 #include "intel_batchbuffer.h"
-#include "intel_ioctl.h"
 #include "intel_regions.h"
 
 #include "i915_context.h"
diff --git a/src/mesa/drivers/dri/i915/i915_tex.c b/src/mesa/drivers/dri/i915/i915_tex.c
index 386617a1ff1..cb55bf61253 100644
--- a/src/mesa/drivers/dri/i915/i915_tex.c
+++ b/src/mesa/drivers/dri/i915/i915_tex.c
@@ -38,8 +38,6 @@
 
 #include "mm.h"
 
-#include "intel_ioctl.h"
-
 #include "i915_context.h"
 #include "i915_reg.h"
 
diff --git a/src/mesa/drivers/dri/i915/i915_texstate.c b/src/mesa/drivers/dri/i915/i915_texstate.c
index 69c1e0755ef..48137f025f0 100644
--- a/src/mesa/drivers/dri/i915/i915_texstate.c
+++ b/src/mesa/drivers/dri/i915/i915_texstate.c
@@ -28,7 +28,6 @@
 #include "mtypes.h"
 #include "enums.h"
 #include "texformat.h"
-#include "dri_bufmgr.h"
 
 #include "intel_mipmap_tree.h"
 #include "intel_tex.h"
@@ -78,7 +77,7 @@ translate_texture_format(GLuint mesa_format, GLenum DepthMode)
       return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT2_3);
    case MESA_FORMAT_RGBA_DXT5:
       return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT4_5);
-   case MESA_FORMAT_Z24_S8:
+   case MESA_FORMAT_S8_Z24:
       return (MAPSURF_32BIT | MT_32BIT_xI824);
    default:
       fprintf(stderr, "%s: bad image format %x\n", __FUNCTION__, mesa_format);
diff --git a/src/mesa/drivers/dri/i915/intel_ioctl.c b/src/mesa/drivers/dri/i915/intel_ioctl.c
deleted file mode 100644
index aa7c6d106ce..00000000000
--- a/src/mesa/drivers/dri/i915/intel_ioctl.c
+++ /dev/null
@@ -1,659 +0,0 @@
-/**************************************************************************
- * 
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
- * All Rights Reserved.
- * 
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- * 
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- * 
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- * 
- **************************************************************************/
-
-
-#include <stdio.h>
-#include <unistd.h>
-#include <errno.h>
-#include <sched.h>
-
-#include "mtypes.h"
-#include "context.h"
-#include "swrast/swrast.h"
-
-#include "intel_context.h"
-#include "intel_ioctl.h"
-#include "intel_batchbuffer.h"
-#include "drm.h"
-
-uint32_t intelGetLastFrame (intelContextPtr intel) 
-{
-   int ret;
-   uint32_t frame;
-   drm_i915_getparam_t gp;
-   
-   gp.param = I915_PARAM_LAST_DISPATCH;
-   gp.value = (int *)&frame;
-   ret = drmCommandWriteRead( intel->driFd, DRM_I915_GETPARAM,
-			      &gp, sizeof(gp) );
-   return frame;
-}
-
-int intelEmitIrqLocked( intelContextPtr intel )
-{
-   drmI830IrqEmit ie;
-   int ret, seq;
-      
-   assert(((*(int *)intel->driHwLock) & ~DRM_LOCK_CONT) == 
-	  (DRM_LOCK_HELD|intel->hHWContext));
-
-   ie.irq_seq = &seq;
-	 
-   ret = drmCommandWriteRead( intel->driFd, DRM_I830_IRQ_EMIT, 
-			      &ie, sizeof(ie) );
-   if ( ret ) {
-      fprintf( stderr, "%s: drmI830IrqEmit: %d\n", __FUNCTION__, ret );
-      exit(1);
-   }
-   
-   if (0)
-      fprintf(stderr, "%s -->  %d\n", __FUNCTION__, seq );
-
-   return seq;
-}
-
-void intelWaitIrq( intelContextPtr intel, int seq )
-{
-   int ret;
-      
-   if (0)
-      fprintf(stderr, "%s %d\n", __FUNCTION__, seq );
-
-   intel->iw.irq_seq = seq;
-	 
-   do {
-     ret = drmCommandWrite( intel->driFd, DRM_I830_IRQ_WAIT, &intel->iw, sizeof(intel->iw) );
-   } while (ret == -EAGAIN || ret == -EINTR);
-
-   if ( ret ) {
-      fprintf( stderr, "%s: drmI830IrqWait: %d\n", __FUNCTION__, ret );
-      if (0)
-	 intel_dump_batchbuffer( intel->alloc.offset,
-				 intel->alloc.ptr,
-				 intel->alloc.size );
-      exit(1);
-   }
-}
-
-
-
-static void age_intel( intelContextPtr intel, int age )
-{
-   GLuint i;
-
-   for (i = 0 ; i < MAX_TEXTURE_UNITS ; i++)
-      if (intel->CurrentTexObj[i]) 
-	 intel->CurrentTexObj[i]->age = age;
-}
-
-void intel_dump_batchbuffer( long offset,
-			     int *ptr,
-			     int count )
-{
-   int i;
-   fprintf(stderr, "\n\n\nSTART BATCH (%d dwords):\n", count);
-   for (i = 0; i < count/4; i += 4) 
-      fprintf(stderr, "\t0x%x: 0x%08x 0x%08x 0x%08x 0x%08x\n", 
-	      (unsigned int)offset + i*4, ptr[i], ptr[i+1], ptr[i+2], ptr[i+3]);
-   fprintf(stderr, "END BATCH\n\n\n");
-}
-
-void intelRefillBatchLocked( intelContextPtr intel, GLboolean allow_unlock )
-{
-   GLuint last_irq = intel->alloc.irq_emitted;
-   GLuint half = intel->alloc.size / 2;
-   GLuint buf = (intel->alloc.active_buf ^= 1);
-
-   intel->alloc.irq_emitted = intelEmitIrqLocked( intel );
-
-   if (last_irq) {
-      if (allow_unlock) UNLOCK_HARDWARE( intel ); 
-      intelWaitIrq( intel, last_irq );
-      if (allow_unlock) LOCK_HARDWARE( intel ); 
-   }
-
-   if (0)
-      fprintf(stderr, "%s: now using half %d\n", __FUNCTION__, buf);
-
-   intel->batch.start_offset = intel->alloc.offset + buf * half;
-   intel->batch.ptr = (unsigned char *)intel->alloc.ptr + buf * half;
-   intel->batch.size = half - 8;
-   intel->batch.space = half - 8;
-   assert(intel->batch.space >= 0);
-}
-
-#define MI_BATCH_BUFFER_END 	(0xA<<23)
-
-
-void intelFlushBatchLocked( intelContextPtr intel, 
-			    GLboolean ignore_cliprects,
-			    GLboolean refill,
-			    GLboolean allow_unlock)
-{
-   drmI830BatchBuffer batch;
-
-   assert(intel->locked);
-
-   if (0)
-      fprintf(stderr, "%s used %d of %d offset %x..%x refill %d (started in %s)\n",
-	      __FUNCTION__, 
-	      (intel->batch.size - intel->batch.space), 
-	      intel->batch.size,
-	      intel->batch.start_offset,
-	      intel->batch.start_offset + 
-	      (intel->batch.size - intel->batch.space), 
-	      refill,
-	      intel->batch.func);
-
-   /* Throw away non-effective packets.  Won't work once we have
-    * hardware contexts which would preserve statechanges beyond a
-    * single buffer.
-    */
-   if (intel->numClipRects == 0 && !ignore_cliprects) {
-      
-      /* Without this yeild, an application with no cliprects can hog
-       * the hardware.  Without unlocking, the effect is much worse -
-       * effectively a lock-out of other contexts.
-       */
-      if (allow_unlock) {
-	 UNLOCK_HARDWARE( intel );
-	 sched_yield();
-	 LOCK_HARDWARE( intel );
-      }
-
-      /* Note that any state thought to have been emitted actually
-       * hasn't:
-       */
-      intel->batch.ptr -= (intel->batch.size - intel->batch.space);
-      intel->batch.space = intel->batch.size;
-      intel->vtbl.lost_hardware( intel ); 
-   }
-
-   if (intel->batch.space != intel->batch.size) {
-
-      if (intel->sarea->ctxOwner != intel->hHWContext) {
-	 intel->perf_boxes |= I830_BOX_LOST_CONTEXT;
-	 intel->sarea->ctxOwner = intel->hHWContext;
-      }
-
-      batch.start = intel->batch.start_offset;
-      batch.used = intel->batch.size - intel->batch.space;
-      batch.cliprects = intel->pClipRects;
-      batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects;
-      batch.DR1 = 0;
-      batch.DR4 = ((((GLuint)intel->drawX) & 0xffff) | 
-		   (((GLuint)intel->drawY) << 16));
-      
-      if (intel->alloc.offset) {
-	 if ((batch.used & 0x4) == 0) {
-	    ((int *)intel->batch.ptr)[0] = 0;
-	    ((int *)intel->batch.ptr)[1] = MI_BATCH_BUFFER_END;
-	    batch.used += 0x8;
-	    intel->batch.ptr += 0x8;
-	 }
-	 else {
-	    ((int *)intel->batch.ptr)[0] = MI_BATCH_BUFFER_END;
-	    batch.used += 0x4;
-	    intel->batch.ptr += 0x4;
-	 }      
-      }
-
-      if (0)
- 	 intel_dump_batchbuffer( batch.start,
-				 (int *)(intel->batch.ptr - batch.used),
-				 batch.used );
-
-      intel->batch.start_offset += batch.used;
-      intel->batch.size -= batch.used;
-
-      if (intel->batch.size < 8) {
-	 refill = GL_TRUE;
-	 intel->batch.space = intel->batch.size = 0;
-      }
-      else {
-	 intel->batch.size -= 8;
-	 intel->batch.space = intel->batch.size;
-      }
-
-
-      assert(intel->batch.space >= 0);
-      assert(batch.start >= intel->alloc.offset);
-      assert(batch.start < intel->alloc.offset + intel->alloc.size);
-      assert(batch.start + batch.used > intel->alloc.offset);
-      assert(batch.start + batch.used <= 
-	     intel->alloc.offset + intel->alloc.size);
-
-
-      if (intel->alloc.offset) {
-	 if (drmCommandWrite (intel->driFd, DRM_I830_BATCHBUFFER, &batch, 
-			      sizeof(batch))) {
-	    fprintf(stderr, "DRM_I830_BATCHBUFFER: %d\n",  -errno);
-	    UNLOCK_HARDWARE(intel);
-	    exit(1);
-	 }
-      } else {
-	 drmI830CmdBuffer cmd;
-	 cmd.buf = (char *)intel->alloc.ptr + batch.start;
-	 cmd.sz = batch.used;
-	 cmd.DR1 = batch.DR1;
-	 cmd.DR4 = batch.DR4;
-	 cmd.num_cliprects = batch.num_cliprects;
-	 cmd.cliprects = batch.cliprects;
-	 
-	 if (drmCommandWrite (intel->driFd, DRM_I830_CMDBUFFER, &cmd, 
-			      sizeof(cmd))) {
-	    fprintf(stderr, "DRM_I830_CMDBUFFER: %d\n",  -errno);
-	    UNLOCK_HARDWARE(intel);
-	    exit(1);
-	 }
-      }	 
-
-      
-      age_intel(intel, intel->sarea->last_enqueue);
-
-      /* FIXME: use hardware contexts to avoid 'losing' hardware after
-       * each buffer flush.
-       */
-      if (intel->batch.contains_geometry) 
-	 assert(intel->batch.last_emit_state == intel->batch.counter);
-
-      intel->batch.counter++;
-      intel->batch.contains_geometry = 0;
-      intel->batch.func = 0;
-      intel->vtbl.lost_hardware( intel );
-   }
-
-   if (refill)
-      intelRefillBatchLocked( intel, allow_unlock );
-}
-
-void intelFlushBatch( intelContextPtr intel, GLboolean refill )
-{
-   if (intel->locked) {
-      intelFlushBatchLocked( intel, GL_FALSE, refill, GL_FALSE );
-   } 
-   else {
-      LOCK_HARDWARE(intel);
-      intelFlushBatchLocked( intel, GL_FALSE, refill, GL_TRUE );
-      UNLOCK_HARDWARE(intel);
-   }
-}
-
-
-void intelWaitForIdle( intelContextPtr intel )
-{   
-   if (0)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   intel->vtbl.emit_flush( intel );
-   intelFlushBatch( intel, GL_TRUE );
-
-   /* Use an irq to wait for dma idle -- Need to track lost contexts
-    * to shortcircuit consecutive calls to this function:
-    */
-   intelWaitIrq( intel, intel->alloc.irq_emitted );
-   intel->alloc.irq_emitted = 0;
-}
-
-
-/**
- * Check if we need to rotate/warp the front color buffer to the
- * rotated screen.  We generally need to do this when we get a glFlush
- * or glFinish after drawing to the front color buffer.
- */
-static void
-intelCheckFrontRotate(GLcontext *ctx)
-{
-   intelContextPtr intel = INTEL_CONTEXT( ctx );
-   if (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT) {
-      intelScreenPrivate *screen = intel->intelScreen;
-      if (screen->current_rotation != 0) {
-         __DRIdrawablePrivate *dPriv = intel->driDrawable;
-         intelRotateWindow(intel, dPriv, BUFFER_BIT_FRONT_LEFT);
-      }
-   }
-}
-
-
-/**
- * NOT directly called via glFlush.
- */
-void intelFlush( GLcontext *ctx )
-{
-   intelContextPtr intel = INTEL_CONTEXT( ctx );
-
-   if (intel->Fallback)
-      _swrast_flush( ctx );
-
-   INTEL_FIREVERTICES( intel );
-
-   if (intel->batch.size != intel->batch.space)
-      intelFlushBatch( intel, GL_FALSE );
-}
-
-
-/**
- * Called via glFlush.
- */
-void intelglFlush( GLcontext *ctx )
-{
-   intelFlush(ctx);
-   intelCheckFrontRotate(ctx);
-}
-
-
-void intelFinish( GLcontext *ctx  ) 
-{
-   intelContextPtr intel = INTEL_CONTEXT( ctx );
-   intelFlush( ctx );
-   intelWaitForIdle( intel );
-   intelCheckFrontRotate(ctx);
-}
-
-
-void intelClear(GLcontext *ctx, GLbitfield mask)
-{
-   intelContextPtr intel = INTEL_CONTEXT( ctx );
-   const GLuint colorMask = *((GLuint *) &ctx->Color.ColorMask);
-   GLbitfield tri_mask = 0;
-   GLbitfield blit_mask = 0;
-   GLbitfield swrast_mask = 0;
-
-   if (0)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   /* Take care of cliprects, which are handled differently for
-    * clears, etc.
-    */
-   intelFlush( &intel->ctx );
-
-   if (mask & BUFFER_BIT_FRONT_LEFT) {
-      if (colorMask == ~0) {
-	 blit_mask |= BUFFER_BIT_FRONT_LEFT;
-      } 
-      else {
-	 tri_mask |= BUFFER_BIT_FRONT_LEFT;
-      }
-   }
-
-   if (mask & BUFFER_BIT_BACK_LEFT) {
-      if (colorMask == ~0) {
-	 blit_mask |= BUFFER_BIT_BACK_LEFT;
-      } 
-      else {
-	 tri_mask |= BUFFER_BIT_BACK_LEFT;
-      }
-   }
-
-   if (mask & BUFFER_BIT_DEPTH) {
-      blit_mask |= BUFFER_BIT_DEPTH;
-   }
-
-   if (mask & BUFFER_BIT_STENCIL) {
-      if (!intel->hw_stencil) {
-	 swrast_mask |= BUFFER_BIT_STENCIL;
-      }
-      else if ((ctx->Stencil.WriteMask[0] & 0xff) != 0xff) {
-	 tri_mask |= BUFFER_BIT_STENCIL;
-      } 
-      else {
-	 blit_mask |= BUFFER_BIT_STENCIL;
-      }
-   }
-
-   swrast_mask |= (mask & BUFFER_BIT_ACCUM);
-
-   if (blit_mask) 
-      intelClearWithBlit( ctx, blit_mask, 0, 0, 0, 0, 0);
-
-   if (tri_mask) 
-      intel->vtbl.clear_with_tris( intel, tri_mask, 0, 0, 0, 0, 0);
-
-   if (swrast_mask)
-      _swrast_Clear( ctx, swrast_mask );
-}
-
-
-void
-intelRotateWindow(intelContextPtr intel, __DRIdrawablePrivate *dPriv,
-                  GLuint srcBuffer)
-{
-   if (intel->vtbl.rotate_window) {
-      intel->vtbl.rotate_window(intel, dPriv, srcBuffer);
-   }
-}
-
-
-void *intelAllocateAGP( intelContextPtr intel, GLsizei size )
-{
-   int region_offset;
-   drmI830MemAlloc alloc;
-   int ret;
-
-   if (0)
-      fprintf(stderr, "%s: %d bytes\n", __FUNCTION__, size);
-
-   alloc.region = I830_MEM_REGION_AGP;
-   alloc.alignment = 0;
-   alloc.size = size;
-   alloc.region_offset = &region_offset;
-
-   LOCK_HARDWARE(intel);
-
-   /* Make sure the global heap is initialized
-    */
-   if (intel->texture_heaps[0])
-      driAgeTextures( intel->texture_heaps[0] );
-
-
-   ret = drmCommandWriteRead( intel->driFd,
-			      DRM_I830_ALLOC,
-			      &alloc, sizeof(alloc));
-   
-   if (ret) {
-      fprintf(stderr, "%s: DRM_I830_ALLOC ret %d\n", __FUNCTION__, ret);
-      UNLOCK_HARDWARE(intel);
-      return NULL;
-   }
-   
-   if (0)
-      fprintf(stderr, "%s: allocated %d bytes\n", __FUNCTION__, size);
-
-   /* Need to propogate this information (agp memory in use) to our
-    * local texture lru.  The kernel has already updated the global
-    * lru.  An alternative would have been to allocate memory the
-    * usual way and then notify the kernel to pin the allocation.
-    */
-   if (intel->texture_heaps[0])
-      driAgeTextures( intel->texture_heaps[0] );
-
-   UNLOCK_HARDWARE(intel);   
-
-   return (void *)((char *)intel->intelScreen->tex.map + region_offset);
-}
-
-void intelFreeAGP( intelContextPtr intel, void *pointer )
-{
-   int region_offset;
-   drmI830MemFree memfree;
-   int ret;
-
-   region_offset = (char *)pointer - (char *)intel->intelScreen->tex.map;
-
-   if (region_offset < 0 || 
-       region_offset > intel->intelScreen->tex.size) {
-      fprintf(stderr, "offset %d outside range 0..%d\n", region_offset,
-	      intel->intelScreen->tex.size);
-      return;
-   }
-
-   memfree.region = I830_MEM_REGION_AGP;
-   memfree.region_offset = region_offset;
-   
-   ret = drmCommandWrite( intel->driFd,
-			  DRM_I830_FREE,
-			  &memfree, sizeof(memfree));
-   
-   if (ret) 
-      fprintf(stderr, "%s: DRM_I830_FREE ret %d\n", __FUNCTION__, ret);
-}
-
-/* This version of AllocateMemoryMESA allocates only agp memory, and
- * only does so after the point at which the driver has been
- * initialized.
- *
- * Theoretically a valid context isn't required.  However, in this
- * implementation, it is, as I'm using the hardware lock to protect
- * the kernel data structures, and the current context to get the
- * device fd.
- */
-void *intelAllocateMemoryMESA(__DRInativeDisplay *dpy, int scrn,
-			      GLsizei size, GLfloat readfreq,
-			      GLfloat writefreq, GLfloat priority)
-{
-   GET_CURRENT_CONTEXT(ctx);
-
-   if (INTEL_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s sz %d %f/%f/%f\n", __FUNCTION__, size, readfreq, 
-	      writefreq, priority);
-
-   if (getenv("INTEL_NO_ALLOC"))
-      return NULL;
-   
-   if (!ctx || INTEL_CONTEXT(ctx) == 0) 
-      return NULL;
-   
-   return intelAllocateAGP( INTEL_CONTEXT(ctx), size );
-}
-
-
-/* Called via glXFreeMemoryMESA() */
-void intelFreeMemoryMESA(__DRInativeDisplay *dpy, int scrn, GLvoid *pointer)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   if (INTEL_DEBUG & DEBUG_IOCTL) 
-      fprintf(stderr, "%s %p\n", __FUNCTION__, pointer);
-
-   if (!ctx || INTEL_CONTEXT(ctx) == 0) {
-      fprintf(stderr, "%s: no context\n", __FUNCTION__);
-      return;
-   }
-
-   intelFreeAGP( INTEL_CONTEXT(ctx), pointer );
-}
-
-/* Called via glXGetMemoryOffsetMESA() 
- *
- * Returns offset of pointer from the start of agp aperture.
- */
-GLuint intelGetMemoryOffsetMESA(__DRInativeDisplay *dpy, int scrn, 
-				const GLvoid *pointer)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   intelContextPtr intel;
-
-   if (!ctx || !(intel = INTEL_CONTEXT(ctx)) ) {
-      fprintf(stderr, "%s: no context\n", __FUNCTION__);
-      return ~0;
-   }
-
-   if (!intelIsAgpMemory( intel, pointer, 0 ))
-      return ~0;
-
-   return intelAgpOffsetFromVirtual( intel, pointer );
-}
-
-
-GLboolean intelIsAgpMemory( intelContextPtr intel, const GLvoid *pointer,
-			   GLint size )
-{
-   int offset = (char *)pointer - (char *)intel->intelScreen->tex.map;
-   int valid = (size >= 0 &&
-		offset >= 0 &&
-		offset + size < intel->intelScreen->tex.size);
-
-   if (INTEL_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "intelIsAgpMemory( %p ) : %d\n", pointer, valid );
-   
-   return valid;
-}
-
-
-GLuint intelAgpOffsetFromVirtual( intelContextPtr intel, const GLvoid *pointer )
-{
-   int offset = (char *)pointer - (char *)intel->intelScreen->tex.map;
-
-   if (offset < 0 || offset > intel->intelScreen->tex.size)
-      return ~0;
-   else
-      return intel->intelScreen->tex.offset + offset;
-}
-
-
-
-
-
-/* Flip the front & back buffes
- */
-void intelPageFlip( const __DRIdrawablePrivate *dPriv )
-{
-#if 0
-   intelContextPtr intel;
-   int tmp, ret;
-
-   if (INTEL_DEBUG & DEBUG_IOCTL)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   assert(dPriv);
-   assert(dPriv->driContextPriv);
-   assert(dPriv->driContextPriv->driverPrivate);
-
-   intel = (intelContextPtr) dPriv->driContextPriv->driverPrivate;
-
-   intelFlush( &intel->ctx );
-   LOCK_HARDWARE( intel );
-
-   if (dPriv->pClipRects) {
-      *(drm_clip_rect_t *)intel->sarea->boxes = dPriv->pClipRects[0];
-      intel->sarea->nbox = 1;
-   }
-
-   ret = drmCommandNone(intel->driFd, DRM_I830_FLIP); 
-   if (ret) {
-      fprintf(stderr, "%s: %d\n", __FUNCTION__, ret);
-      UNLOCK_HARDWARE( intel );
-      exit(1);
-   }
-
-   tmp = intel->sarea->last_enqueue;
-   intelRefillBatchLocked( intel );
-   UNLOCK_HARDWARE( intel );
-
-
-   intelSetDrawBuffer( &intel->ctx, intel->ctx.Color.DriverDrawBuffer );
-#endif
-}
diff --git a/src/mesa/drivers/dri/i915/intel_pixel_read.c b/src/mesa/drivers/dri/i915/intel_pixel_read.c
index d009590a4bc..120c887abd6 100644
--- a/src/mesa/drivers/dri/i915/intel_pixel_read.c
+++ b/src/mesa/drivers/dri/i915/intel_pixel_read.c
@@ -35,7 +35,6 @@
 
 #include "intel_screen.h"
 #include "intel_context.h"
-#include "intel_ioctl.h"
 #include "intel_batchbuffer.h"
 #include "intel_blit.h"
 #include "intel_buffers.h"
diff --git a/src/mesa/drivers/dri/i965/Makefile b/src/mesa/drivers/dri/i965/Makefile
index 8fa205e79ff..c2d555cd0cf 100644
--- a/src/mesa/drivers/dri/i965/Makefile
+++ b/src/mesa/drivers/dri/i965/Makefile
@@ -13,7 +13,6 @@ DRIVER_SOURCES = \
 	intel_decode.c \
 	intel_depthstencil.c \
 	intel_fbo.c \
-	intel_ioctl.c \
 	intel_mipmap_tree.c \
 	intel_regions.c \
 	intel_screen.c \
@@ -91,6 +90,8 @@ ASM_SOURCES =
 
 DRIVER_DEFINES = -I../intel -I../intel/server
 
+DRI_LIB_DEPS += -ldrm_intel
+
 include ../Makefile.template
 
 symlinks:
diff --git a/src/mesa/drivers/dri/i965/brw_cc.c b/src/mesa/drivers/dri/i965/brw_cc.c
index d662cf75211..0458bf59832 100644
--- a/src/mesa/drivers/dri/i965/brw_cc.c
+++ b/src/mesa/drivers/dri/i965/brw_cc.c
@@ -255,12 +255,12 @@ cc_unit_create_from_key(struct brw_context *brw, struct brw_cc_unit_key *key)
 			 NULL, NULL);
 
    /* Emit CC viewport relocation */
-   intel_bo_emit_reloc(bo,
-		       I915_GEM_DOMAIN_INSTRUCTION,
-		       0,
-		       0,
-		       offsetof(struct brw_cc_unit_state, cc4),
-		       brw->cc.vp_bo);
+   dri_bo_emit_reloc(bo,
+		     I915_GEM_DOMAIN_INSTRUCTION,
+		     0,
+		     0,
+		     offsetof(struct brw_cc_unit_state, cc4),
+		     brw->cc.vp_bo);
 
    return bo;
 }
diff --git a/src/mesa/drivers/dri/i965/brw_clip_state.c b/src/mesa/drivers/dri/i965/brw_clip_state.c
index ae904c6253d..51045f91880 100644
--- a/src/mesa/drivers/dri/i965/brw_clip_state.c
+++ b/src/mesa/drivers/dri/i965/brw_clip_state.c
@@ -119,12 +119,12 @@ clip_unit_create_from_key(struct brw_context *brw,
 
    /* Emit clip program relocation */
    assert(brw->clip.prog_bo);
-   intel_bo_emit_reloc(bo,
-		       I915_GEM_DOMAIN_INSTRUCTION,
-		       0,
-		       clip.thread0.grf_reg_count << 1,
-		       offsetof(struct brw_clip_unit_state, thread0),
-		       brw->clip.prog_bo);
+   dri_bo_emit_reloc(bo,
+		     I915_GEM_DOMAIN_INSTRUCTION,
+		     0,
+		     clip.thread0.grf_reg_count << 1,
+		     offsetof(struct brw_clip_unit_state, thread0),
+		     brw->clip.prog_bo);
 
    return bo;
 }
diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c
index d43b52c2c7c..95d798238d9 100644
--- a/src/mesa/drivers/dri/i965/brw_draw.c
+++ b/src/mesa/drivers/dri/i965/brw_draw.c
@@ -39,7 +39,6 @@
 #include "brw_state.h"
 #include "brw_fallback.h"
 
-#include "intel_ioctl.h"
 #include "intel_batchbuffer.h"
 #include "intel_buffer_objects.h"
 
diff --git a/src/mesa/drivers/dri/i965/brw_draw.h b/src/mesa/drivers/dri/i965/brw_draw.h
index 2a3e0c1c5bf..e0bba589d5a 100644
--- a/src/mesa/drivers/dri/i965/brw_draw.h
+++ b/src/mesa/drivers/dri/i965/brw_draw.h
@@ -31,7 +31,6 @@
 #include "mtypes.h"		/* for GLcontext... */
 #include "vbo/vbo.h"
 
-#include "dri_bufmgr.h"
 struct brw_context;
 
 
diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c
index 0122d18f82b..465239f195f 100644
--- a/src/mesa/drivers/dri/i965/brw_draw_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c
@@ -39,7 +39,6 @@
 #include "brw_state.h"
 #include "brw_fallback.h"
 
-#include "intel_ioctl.h"
 #include "intel_batchbuffer.h"
 #include "intel_buffer_objects.h"
 #include "intel_tex.h"
diff --git a/src/mesa/drivers/dri/i965/brw_eu_emit.c b/src/mesa/drivers/dri/i965/brw_eu_emit.c
index 6b97f8b170c..0bfbec9d140 100644
--- a/src/mesa/drivers/dri/i965/brw_eu_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_eu_emit.c
@@ -513,6 +513,8 @@ struct brw_instruction *brw_IF(struct brw_compile *p, GLuint execute_size)
    insn->header.compression_control = BRW_COMPRESSION_NONE;
    insn->header.predicate_control = BRW_PREDICATE_NORMAL;
    insn->header.mask_control = BRW_MASK_ENABLE;
+   if (!p->single_program_flow)
+       insn->header.thread_control = BRW_THREAD_SWITCH;
 
    p->current->header.predicate_control = BRW_PREDICATE_NONE;
 
@@ -538,6 +540,8 @@ struct brw_instruction *brw_ELSE(struct brw_compile *p,
    insn->header.compression_control = BRW_COMPRESSION_NONE;
    insn->header.execution_size = if_insn->header.execution_size;
    insn->header.mask_control = BRW_MASK_ENABLE;
+   if (!p->single_program_flow)
+       insn->header.thread_control = BRW_THREAD_SWITCH;
 
    /* Patch the if instruction to point at this instruction.
     */
@@ -579,6 +583,7 @@ void brw_ENDIF(struct brw_compile *p,
       insn->header.compression_control = BRW_COMPRESSION_NONE;
       insn->header.execution_size = patch_insn->header.execution_size;
       insn->header.mask_control = BRW_MASK_ENABLE;
+      insn->header.thread_control = BRW_THREAD_SWITCH;
 
       assert(patch_insn->bits3.if_else.jump_count == 0);
 
@@ -617,7 +622,7 @@ struct brw_instruction *brw_BREAK(struct brw_compile *p)
    brw_set_src1(insn, brw_imm_d(0x0));
    insn->header.compression_control = BRW_COMPRESSION_NONE;
    insn->header.execution_size = BRW_EXECUTE_8;
-   insn->header.mask_control = BRW_MASK_DISABLE;
+   /* insn->header.mask_control = BRW_MASK_DISABLE; */
    insn->bits3.if_else.pad0 = 0;
    return insn;
 }
@@ -631,7 +636,7 @@ struct brw_instruction *brw_CONT(struct brw_compile *p)
    brw_set_src1(insn, brw_imm_d(0x0));
    insn->header.compression_control = BRW_COMPRESSION_NONE;
    insn->header.execution_size = BRW_EXECUTE_8;
-   insn->header.mask_control = BRW_MASK_DISABLE;
+   /* insn->header.mask_control = BRW_MASK_DISABLE; */
    insn->bits3.if_else.pad0 = 0;
    return insn;
 }
@@ -655,7 +660,7 @@ struct brw_instruction *brw_DO(struct brw_compile *p, GLuint execute_size)
       insn->header.execution_size = execute_size;
       insn->header.predicate_control = BRW_PREDICATE_NONE;
       /* insn->header.mask_control = BRW_MASK_ENABLE; */
-      insn->header.mask_control = BRW_MASK_DISABLE;
+      /* insn->header.mask_control = BRW_MASK_DISABLE; */
 
       return insn;
    }
@@ -694,7 +699,7 @@ struct brw_instruction *brw_WHILE(struct brw_compile *p,
 
 /*    insn->header.mask_control = BRW_MASK_ENABLE; */
 
-   insn->header.mask_control = BRW_MASK_DISABLE;
+   /* insn->header.mask_control = BRW_MASK_DISABLE; */
    p->current->header.predicate_control = BRW_PREDICATE_NONE;   
    return insn;
 }
diff --git a/src/mesa/drivers/dri/i965/brw_gs_state.c b/src/mesa/drivers/dri/i965/brw_gs_state.c
index ff2e3ab0598..633e17cc9d2 100644
--- a/src/mesa/drivers/dri/i965/brw_gs_state.c
+++ b/src/mesa/drivers/dri/i965/brw_gs_state.c
@@ -106,11 +106,11 @@ gs_unit_create_from_key(struct brw_context *brw, struct brw_gs_unit_key *key)
 
    if (key->prog_active) {
       /* Emit GS program relocation */
-      intel_bo_emit_reloc(bo,
-			  I915_GEM_DOMAIN_INSTRUCTION, 0,
-			  gs.thread0.grf_reg_count << 1,
-			  offsetof(struct brw_gs_unit_state, thread0),
-			  brw->gs.prog_bo);
+      dri_bo_emit_reloc(bo,
+			I915_GEM_DOMAIN_INSTRUCTION, 0,
+			gs.thread0.grf_reg_count << 1,
+			offsetof(struct brw_gs_unit_state, thread0),
+			brw->gs.prog_bo);
    }
 
    return bo;
diff --git a/src/mesa/drivers/dri/i965/brw_sf_state.c b/src/mesa/drivers/dri/i965/brw_sf_state.c
index 2478872b827..07d0b4c798a 100644
--- a/src/mesa/drivers/dri/i965/brw_sf_state.c
+++ b/src/mesa/drivers/dri/i965/brw_sf_state.c
@@ -251,18 +251,18 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key,
 			 NULL, NULL);
 
    /* Emit SF program relocation */
-   intel_bo_emit_reloc(bo,
-		       I915_GEM_DOMAIN_INSTRUCTION, 0,
-		       sf.thread0.grf_reg_count << 1,
-		       offsetof(struct brw_sf_unit_state, thread0),
-		       brw->sf.prog_bo);
+   dri_bo_emit_reloc(bo,
+		     I915_GEM_DOMAIN_INSTRUCTION, 0,
+		     sf.thread0.grf_reg_count << 1,
+		     offsetof(struct brw_sf_unit_state, thread0),
+		     brw->sf.prog_bo);
 
    /* Emit SF viewport relocation */
-   intel_bo_emit_reloc(bo,
-		       I915_GEM_DOMAIN_INSTRUCTION, 0,
-		       sf.sf5.front_winding | (sf.sf5.viewport_transform << 1),
-		       offsetof(struct brw_sf_unit_state, sf5),
-		       brw->sf.vp_bo);
+   dri_bo_emit_reloc(bo,
+		     I915_GEM_DOMAIN_INSTRUCTION, 0,
+		     sf.sf5.front_winding | (sf.sf5.viewport_transform << 1),
+		     offsetof(struct brw_sf_unit_state, sf5),
+		     brw->sf.vp_bo);
 
    return bo;
 }
diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c b/src/mesa/drivers/dri/i965/brw_state_upload.c
index d1d319d92e6..7d4fd467b1c 100644
--- a/src/mesa/drivers/dri/i965/brw_state_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_state_upload.c
@@ -33,7 +33,6 @@
 
 #include "brw_context.h"
 #include "brw_state.h"
-#include "dri_bufmgr.h"
 #include "intel_batchbuffer.h"
 
 /* This is used to initialize brw->state.atoms[].  We could use this
diff --git a/src/mesa/drivers/dri/i965/brw_tex.c b/src/mesa/drivers/dri/i965/brw_tex.c
index 258c6260fb9..ab6393420a8 100644
--- a/src/mesa/drivers/dri/i965/brw_tex.c
+++ b/src/mesa/drivers/dri/i965/brw_tex.c
@@ -42,7 +42,6 @@
 #include "texmem.h"
 
 #include "intel_context.h"
-#include "intel_ioctl.h"
 #include "intel_regions.h"
 #include "intel_tex.h"
 #include "brw_context.h"
diff --git a/src/mesa/drivers/dri/i965/brw_vs_state.c b/src/mesa/drivers/dri/i965/brw_vs_state.c
index 909b942610d..24bf5567bd0 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_state.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_state.c
@@ -115,11 +115,11 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key)
 			 NULL, NULL);
 
    /* Emit VS program relocation */
-   intel_bo_emit_reloc(bo,
-		       I915_GEM_DOMAIN_INSTRUCTION, 0,
-		       vs.thread0.grf_reg_count << 1,
-		       offsetof(struct brw_vs_unit_state, thread0),
-		       brw->vs.prog_bo);
+   dri_bo_emit_reloc(bo,
+		     I915_GEM_DOMAIN_INSTRUCTION, 0,
+		     vs.thread0.grf_reg_count << 1,
+		     offsetof(struct brw_vs_unit_state, thread0),
+		     brw->vs.prog_bo);
 
    return bo;
 }
diff --git a/src/mesa/drivers/dri/i965/brw_wm_glsl.c b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
index 305100f8833..8dce40f25b4 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_glsl.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_glsl.c
@@ -850,20 +850,20 @@ static void emit_sop(struct brw_wm_compile *c,
     struct brw_reg dst, src0, src1;
     int i;
 
-    brw_push_insn_state(p);
     for (i = 0; i < 4; i++) {
 	if (mask & (1<<i)) {
 	    dst = get_dst_reg(c, inst, i, 1);
 	    src0 = get_src_reg(c, &inst->SrcReg[0], i, 1);
 	    src1 = get_src_reg(c, &inst->SrcReg[1], i, 1);
+	    brw_push_insn_state(p);
 	    brw_CMP(p, brw_null_reg(), cond, src0, src1);
 	    brw_set_predicate_control(p, BRW_PREDICATE_NONE);
 	    brw_MOV(p, dst, brw_imm_f(0.0));
 	    brw_set_predicate_control(p, BRW_PREDICATE_NORMAL);
 	    brw_MOV(p, dst, brw_imm_f(1.0));
+	    brw_pop_insn_state(p);
 	}
     }
-    brw_pop_insn_state(p);
 }
 
 static void emit_slt(struct brw_wm_compile *c,
diff --git a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c
index 08d01823dee..e47dbe8b101 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c
@@ -303,12 +303,12 @@ static void upload_wm_samplers( struct brw_context *brw )
 	 if (!brw->attribs.Texture->Unit[i]._ReallyEnabled)
 	    continue;
 
-	 intel_bo_emit_reloc(brw->wm.sampler_bo,
-			     I915_GEM_DOMAIN_INSTRUCTION, 0,
-			     0,
-			     i * sizeof(struct brw_sampler_state) +
-			     offsetof(struct brw_sampler_state, ss2),
-			     brw->wm.sdc_bo[i]);
+	 dri_bo_emit_reloc(brw->wm.sampler_bo,
+			   I915_GEM_DOMAIN_INSTRUCTION, 0,
+			   0,
+			   i * sizeof(struct brw_sampler_state) +
+			   offsetof(struct brw_sampler_state, ss2),
+			   brw->wm.sdc_bo[i]);
       }
    }
 }
diff --git a/src/mesa/drivers/dri/i965/brw_wm_state.c b/src/mesa/drivers/dri/i965/brw_wm_state.c
index f97d0dc2854..61fe97a4634 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_state.c
@@ -34,7 +34,6 @@
 #include "brw_context.h"
 #include "brw_state.h"
 #include "brw_defines.h"
-#include "dri_bufmgr.h"
 #include "brw_wm.h"
 
 /***********************************************************************
@@ -199,28 +198,28 @@ wm_unit_create_from_key(struct brw_context *brw, struct brw_wm_unit_key *key,
 			 NULL, NULL);
 
    /* Emit WM program relocation */
-   intel_bo_emit_reloc(bo,
-		       I915_GEM_DOMAIN_INSTRUCTION, 0,
-		       wm.thread0.grf_reg_count << 1,
-		       offsetof(struct brw_wm_unit_state, thread0),
-		       brw->wm.prog_bo);
+   dri_bo_emit_reloc(bo,
+		     I915_GEM_DOMAIN_INSTRUCTION, 0,
+		     wm.thread0.grf_reg_count << 1,
+		     offsetof(struct brw_wm_unit_state, thread0),
+		     brw->wm.prog_bo);
 
    /* Emit scratch space relocation */
    if (key->total_scratch != 0) {
-      intel_bo_emit_reloc(bo,
-			  0, 0,
-			  wm.thread2.per_thread_scratch_space,
-			  offsetof(struct brw_wm_unit_state, thread2),
-			  brw->wm.scratch_buffer);
+      dri_bo_emit_reloc(bo,
+			0, 0,
+			wm.thread2.per_thread_scratch_space,
+			offsetof(struct brw_wm_unit_state, thread2),
+			brw->wm.scratch_buffer);
    }
 
    /* Emit sampler state relocation */
    if (key->sampler_count != 0) {
-      intel_bo_emit_reloc(bo,
-			  I915_GEM_DOMAIN_INSTRUCTION, 0,
-			  wm.wm4.stats_enable | (wm.wm4.sampler_count << 2),
-			  offsetof(struct brw_wm_unit_state, wm4),
-			  brw->wm.sampler_bo);
+      dri_bo_emit_reloc(bo,
+			I915_GEM_DOMAIN_INSTRUCTION, 0,
+			wm.wm4.stats_enable | (wm.wm4.sampler_count << 2),
+			offsetof(struct brw_wm_unit_state, wm4),
+			brw->wm.sampler_bo);
    }
 
    return bo;
diff --git a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
index af1401b34b7..8fcd50924f3 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
@@ -138,7 +138,7 @@ static GLuint translate_tex_format( GLuint mesa_format, GLenum depth_mode )
    case MESA_FORMAT_SRGB_DXT1:
       return BRW_SURFACEFORMAT_BC1_UNORM_SRGB;
 
-   case MESA_FORMAT_Z24_S8:
+   case MESA_FORMAT_S8_Z24:
       return BRW_SURFACEFORMAT_I24X8_UNORM;
 
    default:
@@ -234,11 +234,11 @@ brw_create_texture_surface( struct brw_context *brw,
 
    if (key->bo) {
       /* Emit relocation to surface contents */
-      intel_bo_emit_reloc(bo,
-			  I915_GEM_DOMAIN_SAMPLER, 0,
-			  0,
-			  offsetof(struct brw_surface_state, ss1),
-			  key->bo);
+      dri_bo_emit_reloc(bo,
+			I915_GEM_DOMAIN_SAMPLER, 0,
+			0,
+			offsetof(struct brw_surface_state, ss1),
+			key->bo);
    }
    return bo;
 }
@@ -374,13 +374,13 @@ brw_update_region_surface(struct brw_context *brw, struct intel_region *region,
 	  * them both.  We might be able to figure out from other state
 	  * a more restrictive relocation to emit.
 	  */
-	 intel_bo_emit_reloc(brw->wm.surf_bo[unit],
-			     I915_GEM_DOMAIN_RENDER |
-			     I915_GEM_DOMAIN_SAMPLER,
-			     I915_GEM_DOMAIN_RENDER,
-			     0,
-			     offsetof(struct brw_surface_state, ss1),
-			     region_bo);
+	 dri_bo_emit_reloc(brw->wm.surf_bo[unit],
+			   I915_GEM_DOMAIN_RENDER |
+			   I915_GEM_DOMAIN_SAMPLER,
+			   I915_GEM_DOMAIN_RENDER,
+			   0,
+			   offsetof(struct brw_surface_state, ss1),
+			   region_bo);
       }
    }
 }
@@ -420,11 +420,11 @@ brw_wm_get_binding_table(struct brw_context *brw)
       /* Emit binding table relocations to surface state */
       for (i = 0; i < BRW_WM_MAX_SURF; i++) {
 	 if (brw->wm.surf_bo[i] != NULL) {
-	    intel_bo_emit_reloc(bind_bo,
-				I915_GEM_DOMAIN_INSTRUCTION, 0,
-				0,
-				i * sizeof(GLuint),
-				brw->wm.surf_bo[i]);
+	    dri_bo_emit_reloc(bind_bo,
+			      I915_GEM_DOMAIN_INSTRUCTION, 0,
+			      0,
+			      i * sizeof(GLuint),
+			      brw->wm.surf_bo[i]);
 	 }
       }
 
diff --git a/src/mesa/drivers/dri/i965/intel_ioctl.c b/src/mesa/drivers/dri/i965/intel_ioctl.c
deleted file mode 120000
index effd0f378d2..00000000000
--- a/src/mesa/drivers/dri/i965/intel_ioctl.c
+++ /dev/null
@@ -1 +0,0 @@
-../intel/intel_ioctl.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/intel/intel_batchbuffer.c b/src/mesa/drivers/dri/intel/intel_batchbuffer.c
index 5afaad070cf..7dbc646370d 100644
--- a/src/mesa/drivers/dri/intel/intel_batchbuffer.c
+++ b/src/mesa/drivers/dri/intel/intel_batchbuffer.c
@@ -25,8 +25,8 @@
  * 
  **************************************************************************/
 
+#include "intel_context.h"
 #include "intel_batchbuffer.h"
-#include "intel_ioctl.h"
 #include "intel_decode.h"
 #include "intel_reg.h"
 #include "intel_bufmgr.h"
@@ -148,28 +148,15 @@ do_flush_locked(struct intel_batchbuffer *batch,
     */
 
    if (!(intel->numClipRects == 0 &&
-	 batch->cliprect_mode == LOOP_CLIPRECTS)) {
-      if (intel->ttm == GL_TRUE) {
-	 struct drm_i915_gem_execbuffer *execbuf;
-
-	 execbuf = dri_process_relocs(batch->buf);
-	 ret = intel_exec_ioctl(batch->intel,
-				used,
-				batch->cliprect_mode != LOOP_CLIPRECTS,
-				allow_unlock,
-				execbuf);
-      } else {
-	 dri_process_relocs(batch->buf);
-	 ret = intel_batch_ioctl(batch->intel,
-				 batch->buf->offset,
-				 used,
-				 batch->cliprect_mode != LOOP_CLIPRECTS,
-				 allow_unlock);
-      }
+	 batch->cliprect_mode == LOOP_CLIPRECTS) || intel->no_hw) {
+      dri_bo_exec(batch->buf, used,
+		  intel->pClipRects,
+		  batch->cliprect_mode != LOOP_CLIPRECTS ?
+		  0 : intel->numClipRects,
+		  (((GLuint) intel->drawX) & 0xffff) |
+		  (((GLuint) intel->drawY) << 16));
    }
 
-   dri_post_submit(batch->buf);
-
    if (intel->numClipRects == 0 &&
        batch->cliprect_mode == LOOP_CLIPRECTS) {
       if (allow_unlock) {
@@ -261,13 +248,9 @@ _intel_batchbuffer_flush(struct intel_batchbuffer *batch, const char *file,
       UNLOCK_HARDWARE(intel);
 
    if (INTEL_DEBUG & DEBUG_SYNC) {
-      int irq;
-
       fprintf(stderr, "waiting for idle\n");
-      LOCK_HARDWARE(intel);
-      irq = intelEmitIrqLocked(intel);
-      UNLOCK_HARDWARE(intel);
-      intelWaitIrq(intel, irq);
+      dri_bo_map(batch->buf, GL_TRUE);
+      dri_bo_unmap(batch->buf);
    }
 
    /* Reset the buffer:
@@ -289,8 +272,8 @@ intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch,
    if (batch->ptr - batch->map > batch->buf->size)
     _mesa_printf ("bad relocation ptr %p map %p offset %d size %d\n",
 		  batch->ptr, batch->map, batch->ptr - batch->map, batch->buf->size);
-   ret = intel_bo_emit_reloc(batch->buf, read_domains, write_domain,
-			     delta, batch->ptr - batch->map, buffer);
+   ret = dri_bo_emit_reloc(batch->buf, read_domains, write_domain,
+			   delta, batch->ptr - batch->map, buffer);
 
    /*
     * Using the old buffer offset, write in what the right data would be, in case
diff --git a/src/mesa/drivers/dri/intel/intel_batchbuffer.h b/src/mesa/drivers/dri/intel/intel_batchbuffer.h
index 52d6ecc2238..cf7d4a06336 100644
--- a/src/mesa/drivers/dri/intel/intel_batchbuffer.h
+++ b/src/mesa/drivers/dri/intel/intel_batchbuffer.h
@@ -3,11 +3,10 @@
 
 #include "mtypes.h"
 
-#include "dri_bufmgr.h"
+#include "intel_context.h"
+#include "intel_bufmgr.h"
 #include "intel_reg.h"
 
-struct intel_context;
-
 #define BATCH_SZ 16384
 #define BATCH_RESERVED 16
 
diff --git a/src/mesa/drivers/dri/intel/intel_blit.c b/src/mesa/drivers/dri/intel/intel_blit.c
index 7129a4ba917..653f485dd09 100644
--- a/src/mesa/drivers/dri/intel/intel_blit.c
+++ b/src/mesa/drivers/dri/intel/intel_blit.c
@@ -33,13 +33,13 @@
 #include "context.h"
 #include "enums.h"
 
-#include "intel_batchbuffer.h"
 #include "intel_blit.h"
 #include "intel_buffers.h"
 #include "intel_context.h"
 #include "intel_fbo.h"
 #include "intel_reg.h"
 #include "intel_regions.h"
+#include "intel_batchbuffer.h"
 
 #define FILE_DEBUG_FLAG DEBUG_BLIT
 
diff --git a/src/mesa/drivers/dri/intel/intel_blit.h b/src/mesa/drivers/dri/intel/intel_blit.h
index 0881cc4fdc7..52065b13ed7 100644
--- a/src/mesa/drivers/dri/intel/intel_blit.h
+++ b/src/mesa/drivers/dri/intel/intel_blit.h
@@ -29,8 +29,6 @@
 #define INTEL_BLIT_H
 
 #include "intel_context.h"
-#include "intel_ioctl.h"
-#include "dri_bufmgr.h"
 
 extern void intelCopyBuffer(const __DRIdrawablePrivate * dpriv,
                             const drm_clip_rect_t * rect);
diff --git a/src/mesa/drivers/dri/intel/intel_buffer_objects.c b/src/mesa/drivers/dri/intel/intel_buffer_objects.c
index 1923a21516c..88484a07af6 100644
--- a/src/mesa/drivers/dri/intel/intel_buffer_objects.c
+++ b/src/mesa/drivers/dri/intel/intel_buffer_objects.c
@@ -34,7 +34,6 @@
 #include "intel_buffer_objects.h"
 #include "intel_batchbuffer.h"
 #include "intel_regions.h"
-#include "dri_bufmgr.h"
 
 static GLboolean intel_bufferobj_unmap(GLcontext * ctx,
 				       GLenum target,
diff --git a/src/mesa/drivers/dri/intel/intel_buffers.c b/src/mesa/drivers/dri/intel/intel_buffers.c
index 77352e6a531..3fe67462d58 100644
--- a/src/mesa/drivers/dri/intel/intel_buffers.c
+++ b/src/mesa/drivers/dri/intel/intel_buffers.c
@@ -618,6 +618,9 @@ intel_wait_flips(struct intel_context *intel)
 			     BUFFER_FRONT_LEFT ? BUFFER_FRONT_LEFT :
 			     BUFFER_BACK_LEFT);
 
+   if (intel->intelScreen->driScrnPriv->dri2.enabled)
+      return;
+
    if (intel_fb->Base.Name == 0 && intel_rb &&
        intel_rb->pf_pending == intel_fb->pf_seq) {
       GLint pf_planes = intel_fb->pf_planes;
diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c
index 50c1964d878..4d1a742698c 100644
--- a/src/mesa/drivers/dri/intel/intel_context.c
+++ b/src/mesa/drivers/dri/intel/intel_context.c
@@ -51,7 +51,6 @@
 #include "intel_chipset.h"
 #include "intel_buffers.h"
 #include "intel_tex.h"
-#include "intel_ioctl.h"
 #include "intel_batchbuffer.h"
 #include "intel_blit.h"
 #include "intel_pixel.h"
@@ -195,6 +194,142 @@ intelGetString(GLcontext * ctx, GLenum name)
    }
 }
 
+void
+intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
+{
+   struct intel_framebuffer *intel_fb = drawable->driverPrivate;
+   struct intel_renderbuffer *rb;
+   struct intel_region *region, *depth_region;
+   struct intel_context *intel = context->driverPrivate;
+   __DRIbuffer *buffers;
+   __DRIscreen *screen;
+   int i, count;
+   unsigned int attachments[10];
+   uint32_t name;
+   const char *region_name;
+
+   if (INTEL_DEBUG & DEBUG_DRI)
+      fprintf(stderr, "enter %s, drawable %p\n", __func__, drawable);
+
+   screen = intel->intelScreen->driScrnPriv;
+
+   i = 0;
+   if (intel_fb->color_rb[0])
+      attachments[i++] = __DRI_BUFFER_FRONT_LEFT;
+   if (intel_fb->color_rb[1])
+      attachments[i++] = __DRI_BUFFER_BACK_LEFT;
+   if (intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH))
+      attachments[i++] = __DRI_BUFFER_DEPTH;
+   if (intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL))
+      attachments[i++] = __DRI_BUFFER_STENCIL;
+
+   buffers = (*screen->dri2.loader->getBuffers)(drawable,
+						&drawable->w,
+						&drawable->h,
+						attachments, i,
+						&count,
+						drawable->loaderPrivate);
+
+   drawable->x = 0;
+   drawable->y = 0;
+   drawable->backX = 0;
+   drawable->backY = 0;
+   drawable->numClipRects = 1;
+   drawable->pClipRects[0].x1 = 0;
+   drawable->pClipRects[0].y1 = 0;
+   drawable->pClipRects[0].x2 = drawable->w;
+   drawable->pClipRects[0].y2 = drawable->h;
+   drawable->numBackClipRects = 1;
+   drawable->pBackClipRects[0].x1 = 0;
+   drawable->pBackClipRects[0].y1 = 0;
+   drawable->pBackClipRects[0].x2 = drawable->w;
+   drawable->pBackClipRects[0].y2 = drawable->h;
+
+   depth_region = NULL;
+   for (i = 0; i < count; i++) {
+       switch (buffers[i].attachment) {
+       case __DRI_BUFFER_FRONT_LEFT:
+	   rb = intel_fb->color_rb[0];
+	   region_name = "dri2 front buffer";
+	   break;
+
+       case __DRI_BUFFER_BACK_LEFT:
+	   rb = intel_fb->color_rb[1];
+	   region_name = "dri2 back buffer";
+	   break;
+
+       case __DRI_BUFFER_DEPTH:
+	   rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH);
+	   region_name = "dri2 depth buffer";
+	   break;
+
+       case __DRI_BUFFER_STENCIL:
+	   rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL);
+	   region_name = "dri2 stencil buffer";
+	   break;
+
+       case __DRI_BUFFER_ACCUM:
+       default:
+	   fprintf(stderr,
+		   "unhandled buffer attach event, attacment type %d\n",
+		   buffers[i].attachment);
+	   return;
+       }
+
+       if (rb->region) {
+	  dri_bo_flink(rb->region->buffer, &name);
+	  if (name == buffers[i].name)
+	     continue;
+       }
+
+       if (INTEL_DEBUG & DEBUG_DRI)
+	  fprintf(stderr,
+		  "attaching buffer %d, at %d, cpp %d, pitch %d\n",
+		  buffers[i].name, buffers[i].attachment,
+		  buffers[i].cpp, buffers[i].pitch);
+       
+       if (buffers[i].attachment == __DRI_BUFFER_STENCIL && depth_region) {
+	  if (INTEL_DEBUG & DEBUG_DRI)
+	     fprintf(stderr, "(reusing depth buffer as stencil)\n");
+	  intel_region_reference(&region, depth_region);
+       }
+       else
+          region = intel_region_alloc_for_handle(intel, buffers[i].cpp,
+						 buffers[i].pitch / buffers[i].cpp,
+						 drawable->h,
+						 buffers[i].name,
+						 region_name);
+
+       if (buffers[i].attachment == __DRI_BUFFER_DEPTH)
+	  depth_region = region;
+
+       intel_renderbuffer_set_region(rb, region);
+       intel_region_release(&region);
+   }
+
+   driUpdateFramebufferSize(&intel->ctx, drawable);
+}
+
+static void
+intel_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h)
+{
+    struct intel_context *intel = intel_context(ctx);
+    __DRIcontext *driContext = intel->driContext;
+
+    if (!driContext->driScreenPriv->dri2.enabled)
+	return;
+
+    intel_update_renderbuffers(driContext, driContext->driDrawablePriv);
+    if (driContext->driDrawablePriv != driContext->driReadablePriv)
+	intel_update_renderbuffers(driContext, driContext->driReadablePriv);
+
+    ctx->Driver.Viewport = NULL;
+    intel->driDrawable = driContext->driDrawablePriv;
+    intelWindowMoved(intel);
+    intel_draw_buffer(ctx, intel->ctx.DrawBuffer);
+    ctx->Driver.Viewport = intel_viewport;
+}
+
 /**
  * Extension strings exported by the intel driver.
  *
@@ -273,10 +408,12 @@ static const struct dri_extension brw_extensions[] = {
    { NULL,                                NULL }
 };
 
+#ifdef I915_MMIO_READ
 static const struct dri_extension arb_oc_extensions[] = {
    {"GL_ARB_occlusion_query",            GL_ARB_occlusion_query_functions},
    {NULL, NULL}
 };
+#endif
 
 static const struct dri_extension ttm_extensions[] = {
    {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions},
@@ -301,10 +438,12 @@ void intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging)
    if (intel == NULL || intel->ttm)
       driInitExtensions(ctx, ttm_extensions, GL_FALSE);
 
+#ifdef I915_MMIO_READ
    if (intel == NULL || 
        (IS_965(intel->intelScreen->deviceID) && 
 	intel->intelScreen->drmMinor >= 8))
       driInitExtensions(ctx, arb_oc_extensions, GL_FALSE);
+#endif
 
    if (intel == NULL || IS_965(intel->intelScreen->deviceID))
       driInitExtensions(ctx, brw_extensions, GL_FALSE);
@@ -402,6 +541,7 @@ intelFinish(GLcontext * ctx)
    }
 }
 
+#ifdef I915_MMIO_READ
 static void
 intelBeginQuery(GLcontext *ctx, GLenum target, struct gl_query_object *q)
 {
@@ -432,105 +572,7 @@ intelEndQuery(GLcontext *ctx, GLenum target, struct gl_query_object *q)
 	q->Ready = GL_TRUE;
 	intel->stats_wm--;
 }
-
-/** Driver-specific fence emit implementation for the fake memory manager. */
-static unsigned int
-intel_fence_emit(void *private)
-{
-   struct intel_context *intel = (struct intel_context *)private;
-   unsigned int fence;
-
-   /* XXX: Need to emit a flush, if we haven't already (at least with the
-    * current batchbuffer implementation, we have).
-    */
-
-   fence = intelEmitIrqLocked(intel);
-
-   return fence;
-}
-
-/** Driver-specific fence wait implementation for the fake memory manager. */
-static int
-intel_fence_wait(void *private, unsigned int cookie)
-{
-   struct intel_context *intel = (struct intel_context *)private;
-
-   intelWaitIrq(intel, cookie);
-
-   return 0;
-}
-
-static GLboolean
-intel_init_bufmgr(struct intel_context *intel)
-{
-   intelScreenPrivate *intelScreen = intel->intelScreen;
-   GLboolean gem_disable = getenv("INTEL_NO_GEM") != NULL;
-   int gem_kernel = 0;
-   GLboolean gem_supported;
-   struct drm_i915_getparam gp;
-
-   gp.param = I915_PARAM_HAS_GEM;
-   gp.value = &gem_kernel;
-
-   (void) drmCommandWriteRead(intel->driFd, DRM_I915_GETPARAM, &gp, sizeof(gp));
-
-   /* If we've got a new enough DDX that's initializing GEM and giving us
-    * object handles for the shared buffers, use that.
-    */
-   intel->ttm = GL_FALSE;
-   if (intel->intelScreen->driScrnPriv->dri2.enabled)
-       gem_supported = GL_TRUE;
-   else if (intel->intelScreen->driScrnPriv->ddx_version.minor >= 9 &&
-	    gem_kernel &&
-	    intel->intelScreen->front.bo_handle != -1)
-       gem_supported = GL_TRUE;
-   else
-       gem_supported = GL_FALSE;
-
-   if (!gem_disable && gem_supported) {
-      int bo_reuse_mode;
-      intel->bufmgr = intel_bufmgr_gem_init(intel->driFd,
-					    BATCH_SZ);
-      if (intel->bufmgr != NULL)
-	 intel->ttm = GL_TRUE;
-
-      bo_reuse_mode = driQueryOptioni(&intel->optionCache, "bo_reuse");
-      switch (bo_reuse_mode) {
-      case DRI_CONF_BO_REUSE_DISABLED:
-	 break;
-      case DRI_CONF_BO_REUSE_ALL:
-	 intel_bufmgr_gem_enable_reuse(intel->bufmgr);
-	 break;
-      }
-   }
-   /* Otherwise, use the classic buffer manager. */
-   if (intel->bufmgr == NULL) {
-      if (gem_disable) {
-	 fprintf(stderr, "GEM disabled.  Using classic.\n");
-      } else {
-	 fprintf(stderr, "Failed to initialize GEM.  "
-		 "Falling back to classic.\n");
-      }
-
-      if (intelScreen->tex.size == 0) {
-	 fprintf(stderr, "[%s:%u] Error initializing buffer manager.\n",
-		 __func__, __LINE__);
-	 return GL_FALSE;
-      }
-
-      intel->bufmgr = intel_bufmgr_fake_init(intelScreen->tex.offset,
-					     intelScreen->tex.map,
-					     intelScreen->tex.size,
-					     intel_fence_emit,
-					     intel_fence_wait,
-					     intel);
-   }
-
-   /* XXX bufmgr should be per-screen, not per-context */
-   intelScreen->ttm = intel->ttm;
-
-   return GL_TRUE;
-}
+#endif
 
 void
 intelInitDriverFunctions(struct dd_function_table *functions)
@@ -541,14 +583,17 @@ intelInitDriverFunctions(struct dd_function_table *functions)
    functions->Finish = intelFinish;
    functions->GetString = intelGetString;
    functions->UpdateState = intelInvalidateState;
+   functions->Viewport = intel_viewport;
 
    functions->CopyColorTable = _swrast_CopyColorTable;
    functions->CopyColorSubTable = _swrast_CopyColorSubTable;
    functions->CopyConvolutionFilter1D = _swrast_CopyConvolutionFilter1D;
    functions->CopyConvolutionFilter2D = _swrast_CopyConvolutionFilter2D;
 
+#ifdef I915_MMIO_READ
    functions->BeginQuery = intelBeginQuery;
    functions->EndQuery = intelEndQuery;
+#endif
 
    intelInitTextureFuncs(functions);
    intelInitStateFuncs(functions);
@@ -568,8 +613,6 @@ intelInitContext(struct intel_context *intel,
    GLcontext *shareCtx = (GLcontext *) sharedContextPrivate;
    __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
    intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private;
-   volatile struct drm_i915_sarea *saPriv = (struct drm_i915_sarea *)
-      (((GLubyte *) sPriv->pSAREA) + intelScreen->sarea_priv_offset);
    int fthrottle_mode;
 
    if (!_mesa_initialize_context(&intel->ctx, mesaVis, shareCtx,
@@ -581,7 +624,8 @@ intelInitContext(struct intel_context *intel,
    driContextPriv->driverPrivate = intel;
    intel->intelScreen = intelScreen;
    intel->driScreen = sPriv;
-   intel->sarea = saPriv;
+   intel->sarea = intelScreen->sarea;
+   intel->driContext = driContextPriv;
 
    /* Dri stuff */
    intel->hHWContext = driContextPriv->hHWContext;
@@ -599,8 +643,20 @@ intelInitContext(struct intel_context *intel,
    else
       intel->maxBatchSize = BATCH_SZ;
 
-   if (!intel_init_bufmgr(intel))
-      return GL_FALSE;
+   intel->bufmgr = intelScreen->bufmgr;
+   intel->ttm = intelScreen->ttm;
+   if (intel->ttm) {
+      int bo_reuse_mode;
+
+      bo_reuse_mode = driQueryOptioni(&intel->optionCache, "bo_reuse");
+      switch (bo_reuse_mode) {
+      case DRI_CONF_BO_REUSE_DISABLED:
+	 break;
+      case DRI_CONF_BO_REUSE_ALL:
+	 intel_bufmgr_gem_enable_reuse(intel->bufmgr);
+	 break;
+      }
+   }
 
    ctx->Const.MaxTextureMaxAnisotropy = 2.0;
 
@@ -783,11 +839,14 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv,
       struct intel_framebuffer *intel_fb =
 	 (struct intel_framebuffer *) driDrawPriv->driverPrivate;
       GLframebuffer *readFb = (GLframebuffer *) driReadPriv->driverPrivate;
-
-
-      /* XXX FBO temporary fix-ups! */
-      /* if the renderbuffers don't have regions, init them from the context */
-      if (!driContextPriv->driScreenPriv->dri2.enabled) {
+ 
+      if (driContextPriv->driScreenPriv->dri2.enabled) {     
+          intel_update_renderbuffers(driContextPriv, driDrawPriv);
+          if (driDrawPriv != driReadPriv)
+              intel_update_renderbuffers(driContextPriv, driReadPriv);
+      } else {
+          /* XXX FBO temporary fix-ups! */
+          /* if the renderbuffers don't have regions, init them from the context */
          struct intel_renderbuffer *irbDepth
             = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH);
          struct intel_renderbuffer *irbStencil
@@ -989,18 +1048,12 @@ void LOCK_HARDWARE( struct intel_context *intel )
 	intel_fb->vbl_waited = vbl.reply.sequence;
     }
 
-    DRM_CAS(intel->driHwLock, intel->hHWContext,
-        (DRM_LOCK_HELD|intel->hHWContext), __ret);
+    if (!sPriv->dri2.enabled) {
+	DRM_CAS(intel->driHwLock, intel->hHWContext,
+		(DRM_LOCK_HELD|intel->hHWContext), __ret);
 
-    if (sPriv->dri2.enabled) {
 	if (__ret)
-	    drmGetLock(intel->driFd, intel->hHWContext, 0);
-	if (__driParseEvents(dPriv->driContextPriv, dPriv)) {
-	    intelWindowMoved(intel);
-	    intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer);
-	}
-    } else if (__ret) {
-        intelContendedLock( intel, 0 );
+	    intelContendedLock( intel, 0 );
     }
 
 
@@ -1013,10 +1066,13 @@ void LOCK_HARDWARE( struct intel_context *intel )
  */
 void UNLOCK_HARDWARE( struct intel_context *intel )
 {
+    __DRIscreen *sPriv = intel->driScreen;
+
    intel->vtbl.note_unlock( intel );
    intel->locked = 0;
 
-   DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext);
+   if (!sPriv->dri2.enabled)
+      DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext);
 
    _glthread_UNLOCK_MUTEX(lockMutex);
 
diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h
index f9a373cf748..8bd75753ea6 100644
--- a/src/mesa/drivers/dri/intel/intel_context.h
+++ b/src/mesa/drivers/dri/intel/intel_context.h
@@ -34,7 +34,6 @@
 #include "drm.h"
 #include "mm.h"
 #include "texmem.h"
-#include "dri_bufmgr.h"
 #include "intel_bufmgr.h"
 
 #include "intel_screen.h"
@@ -257,6 +256,7 @@ struct intel_context
    drmLock *driHwLock;
    int driFd;
 
+   __DRIcontextPrivate *driContext;
    __DRIdrawablePrivate *driDrawable;
    __DRIdrawablePrivate *driReadDrawable;
    __DRIscreenPrivate *driScreen;
@@ -492,6 +492,8 @@ extern int intel_translate_stencil_op(GLenum op);
 extern int intel_translate_blend_factor(GLenum factor);
 extern int intel_translate_logic_op(GLenum opcode);
 
+void intel_update_renderbuffers(__DRIcontext *context,
+				__DRIdrawable *drawable);
 
 /*======================================================================
  * Inline conversion functions.  
diff --git a/src/mesa/drivers/dri/intel/intel_ioctl.c b/src/mesa/drivers/dri/intel/intel_ioctl.c
deleted file mode 100644
index 58c81766cdd..00000000000
--- a/src/mesa/drivers/dri/intel/intel_ioctl.c
+++ /dev/null
@@ -1,185 +0,0 @@
-/**************************************************************************
- * 
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
- * All Rights Reserved.
- * 
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- * 
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- * 
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- * 
- **************************************************************************/
-
-
-#include <stdio.h>
-#include <unistd.h>
-#include <errno.h>
-#include <sched.h>
-#include <sys/types.h>
-#include <sys/ioctl.h>
-
-#include "mtypes.h"
-#include "context.h"
-#include "swrast/swrast.h"
-
-#include "intel_context.h"
-#include "intel_ioctl.h"
-#include "intel_batchbuffer.h"
-#include "intel_blit.h"
-#include "intel_regions.h"
-#include "drm.h"
-#include "i915_drm.h"
-
-#include "intel_bufmgr.h"
-
-#define FILE_DEBUG_FLAG DEBUG_IOCTL
-
-int
-intelEmitIrqLocked(struct intel_context *intel)
-{
-   struct drm_i915_irq_emit ie;
-   int ret, seq = 1;
-
-   if (intel->no_hw)
-      return 1;
-
-   /*
-     assert(((*(int *)intel->driHwLock) & ~DRM_LOCK_CONT) ==
-     (DRM_LOCK_HELD|intel->hHWContext));
-   */
-
-   ie.irq_seq = &seq;
-
-   ret = drmCommandWriteRead(intel->driFd, DRM_I915_IRQ_EMIT, &ie, sizeof(ie));
-   if (ret) {
-      fprintf(stderr, "%s: drm_i915_irq_emit: %d\n", __FUNCTION__, ret);
-      exit(1);
-   }
-
-   DBG("%s -->  %d\n", __FUNCTION__, seq);
-
-   return seq;
-}
-
-void
-intelWaitIrq(struct intel_context *intel, int seq)
-{
-   struct drm_i915_irq_wait iw;
-   int ret, lastdispatch;
-   volatile struct drm_i915_sarea *sarea = intel->sarea;
-
-   if (intel->no_hw)
-      return;
-
-   DBG("%s %d\n", __FUNCTION__, seq);
-
-   iw.irq_seq = seq;
-
-   do {
-      lastdispatch = sarea->last_dispatch;
-      ret = drmCommandWrite(intel->driFd, DRM_I915_IRQ_WAIT, &iw, sizeof(iw));
-   } while (ret == -EAGAIN ||
-	    ret == -EINTR ||
-	    (ret == -EBUSY && lastdispatch != sarea->last_dispatch) ||
-	    (ret == 0 && seq > sarea->last_dispatch) ||
-	    (ret == 0 && sarea->last_dispatch - seq >= (1 << 24)));
-
-   if (ret) {
-      fprintf(stderr, "%s: drm_i915_irq_wait: %d\n", __FUNCTION__, ret);
-      exit(1);
-   }
-}
-
-
-int
-intel_batch_ioctl(struct intel_context *intel,
-                  GLuint start_offset,
-                  GLuint used,
-                  GLboolean ignore_cliprects, GLboolean allow_unlock)
-{
-   struct drm_i915_batchbuffer batch;
-
-   if (intel->no_hw)
-      return 0;
-
-   assert(intel->locked);
-   assert(used);
-
-   DBG("%s used %d offset %x..%x ignore_cliprects %d\n",
-       __FUNCTION__,
-       used, start_offset, start_offset + used, ignore_cliprects);
-
-   /* Throw away non-effective packets.  Won't work once we have
-    * hardware contexts which would preserve statechanges beyond a
-    * single buffer.
-    */
-   batch.start = start_offset;
-   batch.used = used;
-   batch.cliprects = intel->pClipRects;
-   batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects;
-   batch.DR1 = 0;
-   batch.DR4 = ((((GLuint) intel->drawX) & 0xffff) |
-                (((GLuint) intel->drawY) << 16));
-
-   DBG("%s: 0x%x..0x%x DR4: %x cliprects: %d\n",
-       __FUNCTION__,
-       batch.start,
-       batch.start + batch.used * 4, batch.DR4, batch.num_cliprects);
-
-   if (drmCommandWrite(intel->driFd, DRM_I915_BATCHBUFFER, &batch,
-                       sizeof(batch))) {
-      fprintf(stderr, "DRM_I915_BATCHBUFFER: %d\n", -errno);
-      return -errno;
-   }
-
-   return 0;
-}
-
-int
-intel_exec_ioctl(struct intel_context *intel,
-		 GLuint used,
-		 GLboolean ignore_cliprects, GLboolean allow_unlock,
-		 struct drm_i915_gem_execbuffer *execbuf)
-{
-   int ret;
-
-   assert(intel->locked);
-   assert(used);
-
-   if (intel->no_hw)
-      return 0;
-
-   execbuf->batch_start_offset = 0;
-   execbuf->batch_len = used;
-   execbuf->cliprects_ptr = (uintptr_t)intel->pClipRects;
-   execbuf->num_cliprects = ignore_cliprects ? 0 : intel->numClipRects;
-   execbuf->DR1 = 0;
-   execbuf->DR4 = ((((GLuint) intel->drawX) & 0xffff) |
-		   (((GLuint) intel->drawY) << 16));
-
-   do {
-      ret = ioctl(intel->driFd, DRM_IOCTL_I915_GEM_EXECBUFFER, execbuf);
-   } while (ret == -EAGAIN);
-
-   if (ret != 0) {
-      fprintf(stderr, "DRM_I915_GEM_EXECBUFFER: %d\n", -errno);
-      return -errno;
-   }
-
-   return 0;
-}
diff --git a/src/mesa/drivers/dri/intel/intel_ioctl.h b/src/mesa/drivers/dri/intel/intel_ioctl.h
deleted file mode 100644
index 526e38358cc..00000000000
--- a/src/mesa/drivers/dri/intel/intel_ioctl.h
+++ /dev/null
@@ -1,46 +0,0 @@
-/**************************************************************************
- * 
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
- * All Rights Reserved.
- * 
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, and/or sell copies of the Software, and to
- * permit persons to whom the Software is furnished to do so, subject to
- * the following conditions:
- * 
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- * 
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
- * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
- * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
- * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- * 
- **************************************************************************/
-
-#ifndef INTEL_IOCTL_H
-#define INTEL_IOCTL_H
-
-#include "intel_context.h"
-
-void intelWaitIrq( struct intel_context *intel, int seq );
-int intelEmitIrqLocked( struct intel_context *intel );
-
-int intel_batch_ioctl(struct intel_context *intel,
-		      GLuint start_offset,
-		      GLuint used,
-		      GLboolean ignore_cliprects,
-		      GLboolean allow_unlock);
-int intel_exec_ioctl(struct intel_context *intel,
-		     GLuint used,
-		     GLboolean ignore_cliprects, GLboolean allow_unlock,
-		     struct drm_i915_gem_execbuffer *execbuf);
-
-#endif
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
index 7e0d20e6810..daa6d4be172 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
@@ -36,7 +36,6 @@
 
 #include "intel_screen.h"
 #include "intel_context.h"
-#include "intel_ioctl.h"
 #include "intel_batchbuffer.h"
 #include "intel_blit.h"
 #include "intel_regions.h"
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_copy.c b/src/mesa/drivers/dri/intel/intel_pixel_copy.c
index 3093ccf7c65..1c72d6179dd 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_copy.c
@@ -35,7 +35,6 @@
 
 #include "intel_screen.h"
 #include "intel_context.h"
-#include "intel_ioctl.h"
 #include "intel_batchbuffer.h"
 #include "intel_buffers.h"
 #include "intel_blit.h"
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_draw.c b/src/mesa/drivers/dri/intel/intel_pixel_draw.c
index 5675084da52..47460da9aad 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_draw.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_draw.c
@@ -35,7 +35,6 @@
 
 #include "intel_screen.h"
 #include "intel_context.h"
-#include "intel_ioctl.h"
 #include "intel_batchbuffer.h"
 #include "intel_blit.h"
 #include "intel_buffers.h"
diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c
index ddfdce38356..1cfc8ddd64c 100644
--- a/src/mesa/drivers/dri/intel/intel_regions.c
+++ b/src/mesa/drivers/dri/intel/intel_regions.c
@@ -46,7 +46,6 @@
 #include "intel_regions.h"
 #include "intel_blit.h"
 #include "intel_buffer_objects.h"
-#include "dri_bufmgr.h"
 #include "intel_bufmgr.h"
 #include "intel_batchbuffer.h"
 #include "intel_chipset.h"
@@ -145,12 +144,12 @@ intel_region_alloc(struct intel_context *intel,
 struct intel_region *
 intel_region_alloc_for_handle(struct intel_context *intel,
 			      GLuint cpp, GLuint pitch, GLuint height,
-			      GLuint handle)
+			      GLuint handle, const char *name)
 {
    struct intel_region *region;
    dri_bo *buffer;
 
-   buffer = intel_bo_gem_create_from_name(intel->bufmgr, "dri2 region", handle);
+   buffer = intel_bo_gem_create_from_name(intel->bufmgr, name, handle);
 
    region = intel_region_alloc_internal(intel, cpp, pitch, height, buffer);
    if (region == NULL)
@@ -164,6 +163,9 @@ intel_region_alloc_for_handle(struct intel_context *intel,
 void
 intel_region_reference(struct intel_region **dst, struct intel_region *src)
 {
+   if (src)
+      DBG("%s %d\n", __FUNCTION__, src->refcount);
+
    assert(*dst == NULL);
    if (src) {
       src->refcount++;
diff --git a/src/mesa/drivers/dri/intel/intel_regions.h b/src/mesa/drivers/dri/intel/intel_regions.h
index e5f19fbb452..a561de485e2 100644
--- a/src/mesa/drivers/dri/intel/intel_regions.h
+++ b/src/mesa/drivers/dri/intel/intel_regions.h
@@ -34,8 +34,10 @@
  * the basic structure for rectangular collections of pixels stored in a dri_bo.
  */
 
+#include <xf86drm.h>
+
 #include "mtypes.h"
-#include "dri_bufmgr.h"
+#include "intel_bufmgr.h"
 
 struct intel_context;
 struct intel_buffer_object;
@@ -76,7 +78,7 @@ struct intel_region *intel_region_alloc(struct intel_context *intel,
 struct intel_region *
 intel_region_alloc_for_handle(struct intel_context *intel,
 			      GLuint cpp, GLuint pitch, GLuint height,
-			      unsigned int handle);
+			      unsigned int handle, const char *name);
 
 void intel_region_reference(struct intel_region **dst,
                             struct intel_region *src);
diff --git a/src/mesa/drivers/dri/intel/intel_screen.c b/src/mesa/drivers/dri/intel/intel_screen.c
index 9ed89906d5c..1de1ac49ea5 100644
--- a/src/mesa/drivers/dri/intel/intel_screen.c
+++ b/src/mesa/drivers/dri/intel/intel_screen.c
@@ -41,7 +41,6 @@
 #include "intel_buffers.h"
 #include "intel_tex.h"
 #include "intel_span.h"
-#include "intel_ioctl.h"
 #include "intel_fbo.h"
 #include "intel_chipset.h"
 
@@ -211,102 +210,6 @@ intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen,
       intelPrintSAREA(sarea);
 }
 
-
-/**
- * DRI2 entrypoint
- */
-static void
-intelHandleDrawableConfig(__DRIdrawablePrivate *dPriv,
-			  __DRIcontextPrivate *pcp,
-			  __DRIDrawableConfigEvent *event)
-{
-   struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
-   struct intel_region *region = NULL;
-   struct intel_renderbuffer *rb, *depth_rb, *stencil_rb;
-   struct intel_context *intel = pcp->driverPrivate;
-   int cpp, pitch;
-
-   cpp = intel->ctx.Visual.rgbBits / 8;
-   pitch = ((cpp * dPriv->w + 63) & ~63) / cpp;
-
-   rb = intel_fb->color_rb[1];
-   if (rb) {
-      region = intel_region_alloc(intel, cpp, pitch, dPriv->h);
-      intel_renderbuffer_set_region(rb, region);
-   }
-
-   rb = intel_fb->color_rb[2];
-   if (rb) {
-      region = intel_region_alloc(intel, cpp, pitch, dPriv->h);
-      intel_renderbuffer_set_region(rb, region);
-   }
-
-   depth_rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH);
-   stencil_rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL);
-   if (depth_rb || stencil_rb)
-      region = intel_region_alloc(intel, cpp, pitch, dPriv->h);
-   if (depth_rb)
-      intel_renderbuffer_set_region(depth_rb, region);
-   if (stencil_rb)
-      intel_renderbuffer_set_region(stencil_rb, region);
-
-   /* FIXME: Tell the X server about the regions we just allocated and
-    * attached. */
-}
-
-/**
- * DRI2 entrypoint
- */
-static void
-intelHandleBufferAttach(__DRIdrawablePrivate *dPriv,
-			__DRIcontextPrivate *pcp,
-			__DRIBufferAttachEvent *ba)
-{
-   struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
-   struct intel_renderbuffer *rb;
-   struct intel_region *region;
-   struct intel_context *intel = pcp->driverPrivate;
-
-   switch (ba->buffer.attachment) {
-   case DRI_DRAWABLE_BUFFER_FRONT_LEFT:
-      rb = intel_fb->color_rb[0];
-      break;
-
-   case DRI_DRAWABLE_BUFFER_BACK_LEFT:
-      rb = intel_fb->color_rb[0];
-      break;
-
-   case DRI_DRAWABLE_BUFFER_DEPTH:
-     rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH);
-     break;
-
-   case DRI_DRAWABLE_BUFFER_STENCIL:
-     rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL);
-     break;
-
-   case DRI_DRAWABLE_BUFFER_ACCUM:
-   default:
-      fprintf(stderr, "unhandled buffer attach event, attacment type %d\n",
-	      ba->buffer.attachment);
-      return;
-   }
-
-#if 0
-   /* FIXME: Add this so we can filter out when the X server sends us
-    * attachment events for the buffers we just allocated.  Need to
-    * get the BO handle for a render buffer. */
-   if (intel_renderbuffer_get_region_handle(rb) == ba->buffer.handle)
-      return;
-#endif
-
-   region = intel_region_alloc_for_handle(intel, ba->buffer.cpp,
-					  ba->buffer.pitch / ba->buffer.cpp,
-					  dPriv->h,
-					  ba->buffer.handle);
-
-   intel_renderbuffer_set_region(rb, region);
-}
-
 static const __DRItexOffsetExtension intelTexOffsetExtension = {
    { __DRI_TEX_OFFSET },
    intelSetTexOffset,
@@ -370,9 +273,9 @@ static GLboolean intelInitDriver(__DRIscreenPrivate *sPriv)
 
    intelScreen->driScrnPriv = sPriv;
    sPriv->private = (void *) intelScreen;
-   intelScreen->sarea_priv_offset = gDRIPriv->sarea_priv_offset;
    sarea = (struct drm_i915_sarea *)
-      (((GLubyte *) sPriv->pSAREA) + intelScreen->sarea_priv_offset);
+      (((GLubyte *) sPriv->pSAREA) + gDRIPriv->sarea_priv_offset);
+   intelScreen->sarea = sarea;
 
    intelScreen->deviceID = gDRIPriv->deviceID;
 
@@ -385,8 +288,6 @@ static GLboolean intelInitDriver(__DRIscreenPrivate *sPriv)
       return GL_FALSE;
    }
 
-   intelScreen->sarea_priv_offset = gDRIPriv->sarea_priv_offset;
-
    if (0)
       intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv);
 
@@ -647,6 +548,69 @@ intelFillInModes(__DRIscreenPrivate *psp,
    return configs;
 }
 
+static GLboolean
+intel_init_bufmgr(intelScreenPrivate *intelScreen)
+{
+   GLboolean gem_disable = getenv("INTEL_NO_GEM") != NULL;
+   int gem_kernel = 0;
+   GLboolean gem_supported;
+   struct drm_i915_getparam gp;
+   __DRIscreenPrivate *spriv = intelScreen->driScrnPriv;
+
+   intelScreen->no_hw = getenv("INTEL_NO_HW") != NULL;
+
+   gp.param = I915_PARAM_HAS_GEM;
+   gp.value = &gem_kernel;
+
+   (void) drmCommandWriteRead(spriv->fd, DRM_I915_GETPARAM, &gp, sizeof(gp));
+
+   /* If we've got a new enough DDX that's initializing GEM and giving us
+    * object handles for the shared buffers, use that.
+    */
+   intelScreen->ttm = GL_FALSE;
+   if (intelScreen->driScrnPriv->dri2.enabled)
+       gem_supported = GL_TRUE;
+   else if (intelScreen->driScrnPriv->ddx_version.minor >= 9 &&
+	    gem_kernel &&
+	    intelScreen->front.bo_handle != -1)
+       gem_supported = GL_TRUE;
+   else
+       gem_supported = GL_FALSE;
+
+   if (!gem_disable && gem_supported) {
+      intelScreen->bufmgr = intel_bufmgr_gem_init(spriv->fd, BATCH_SZ);
+      if (intelScreen->bufmgr != NULL)
+	 intelScreen->ttm = GL_TRUE;
+   }
+   /* Otherwise, use the classic buffer manager. */
+   if (intelScreen->bufmgr == NULL) {
+      if (gem_disable) {
+	 fprintf(stderr, "GEM disabled.  Using classic.\n");
+      } else {
+	 fprintf(stderr, "Failed to initialize GEM.  "
+		 "Falling back to classic.\n");
+      }
+
+      if (intelScreen->tex.size == 0) {
+	 fprintf(stderr, "[%s:%u] Error initializing buffer manager.\n",
+		 __func__, __LINE__);
+	 return GL_FALSE;
+      }
+
+      intelScreen->bufmgr =
+	 intel_bufmgr_fake_init(spriv->fd,
+				intelScreen->tex.offset,
+				intelScreen->tex.map,
+				intelScreen->tex.size,
+				(unsigned int * volatile)
+				&intelScreen->sarea->last_dispatch);
+   }
+
+   /* XXX bufmgr should be per-screen, not per-context */
+   intelScreen->ttm = intelScreen->ttm;
+
+   return GL_TRUE;
+}
 
 /**
  * This is the driver specific part of the createNewScreen entry point.
@@ -658,6 +622,7 @@ intelFillInModes(__DRIscreenPrivate *psp,
  */
 static const __DRIconfig **intelInitScreen(__DRIscreenPrivate *psp)
 {
+   intelScreenPrivate *intelScreen;
 #ifdef I915
    static const __DRIversion ddx_expected = { 1, 5, 0 };
 #else
@@ -691,6 +656,10 @@ static const __DRIconfig **intelInitScreen(__DRIscreenPrivate *psp)
 
    psp->extensions = intelScreenExtensions;
 
+   intelScreen = psp->private;
+   if (!intel_init_bufmgr(intelScreen))
+       return GL_FALSE;
+
    return (const __DRIconfig **)
        intelFillInModes(psp, dri_priv->cpp * 8,
 			(dri_priv->cpp == 2) ? 16 : 24,
@@ -750,26 +719,15 @@ __DRIconfig **intelInitScreen2(__DRIscreenPrivate *psp)
 
    intelScreen->drmMinor = psp->drm_version.minor;
 
-   /* Determine chipset ID? */
+   /* Determine chipset ID */
    if (!intel_get_param(psp, I915_PARAM_CHIPSET_ID,
 			&intelScreen->deviceID))
       return GL_FALSE;
 
-   /* Determine if IRQs are active? */
-   if (!intel_get_param(psp, I915_PARAM_IRQ_ACTIVE,
-			&intelScreen->irq_active))
-      return GL_FALSE;
-
-   /* Determine if batchbuffers are allowed */
-   if (!intel_get_param(psp, I915_PARAM_ALLOW_BATCHBUFFER,
-			&intelScreen->allow_batchbuffer))
-      return GL_FALSE;
-
-   if (!intelScreen->allow_batchbuffer) {
-      fprintf(stderr, "batch buffer not allowed\n");
-      return GL_FALSE;
-   }
+   if (!intel_init_bufmgr(intelScreen))
+       return GL_FALSE;
 
+   intelScreen->irq_active = 1;
    psp->extensions = intelScreenExtensions;
 
    return driConcatConfigs(intelFillInModes(psp, 16, 16, 0, 1),
@@ -792,6 +750,4 @@ const struct __DriverAPIRec driDriverAPI = {
    .CopySubBuffer	 = intelCopySubBuffer,
 
    .InitScreen2		 = intelInitScreen2,
-   .HandleDrawableConfig = intelHandleDrawableConfig,
-   .HandleBufferAttach	 = intelHandleBufferAttach,
 };
diff --git a/src/mesa/drivers/dri/intel/intel_screen.h b/src/mesa/drivers/dri/intel/intel_screen.h
index 9a73b139517..daf2ab4c582 100644
--- a/src/mesa/drivers/dri/intel/intel_screen.h
+++ b/src/mesa/drivers/dri/intel/intel_screen.h
@@ -30,6 +30,7 @@
 
 #include <sys/time.h>
 #include "dri_util.h"
+#include "intel_bufmgr.h"
 #include "i915_drm.h"
 #include "xmlconfig.h"
 
@@ -67,14 +68,18 @@ typedef struct
    int logTextureGranularity;
 
    __DRIscreenPrivate *driScrnPriv;
-   unsigned int sarea_priv_offset;
+
+   volatile struct drm_i915_sarea *sarea;
 
    int drmMinor;
 
    int irq_active;
    int allow_batchbuffer;
 
+   GLboolean no_hw;
+
    int ttm;
+   dri_bufmgr *bufmgr;
 
    /**
    * Configuration cache with default values for all contexts
diff --git a/src/mesa/drivers/dri/intel/intel_span.c b/src/mesa/drivers/dri/intel/intel_span.c
index edede3a74b4..0ea8576a027 100644
--- a/src/mesa/drivers/dri/intel/intel_span.c
+++ b/src/mesa/drivers/dri/intel/intel_span.c
@@ -34,7 +34,6 @@
 #include "intel_screen.h"
 #include "intel_span.h"
 #include "intel_regions.h"
-#include "intel_ioctl.h"
 #include "intel_tex.h"
 
 #include "swrast/swrast.h"
diff --git a/src/mesa/drivers/dri/intel/intel_tex_format.c b/src/mesa/drivers/dri/intel/intel_tex_format.c
index 349a29bd714..8ae80e16014 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_format.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_format.c
@@ -138,7 +138,7 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat,
 
    case GL_DEPTH_STENCIL_EXT:
    case GL_DEPTH24_STENCIL8_EXT:
-      return &_mesa_texformat_z24_s8;
+      return &_mesa_texformat_s8_z24;
 
 #ifndef I915
    case GL_SRGB_EXT:
diff --git a/src/mesa/drivers/dri/intel/intel_tex_image.c b/src/mesa/drivers/dri/intel/intel_tex_image.c
index c11687a2cf6..d1f540b0f17 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_image.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_image.c
@@ -21,7 +21,6 @@
 #include "intel_buffer_objects.h"
 #include "intel_batchbuffer.h"
 #include "intel_tex.h"
-#include "intel_ioctl.h"
 #include "intel_blit.h"
 #include "intel_fbo.h"
 
@@ -731,7 +730,7 @@ intelSetTexBuffer(__DRIcontext *pDRICtx, GLint target, __DRIdrawable *dPriv)
    if (!intelObj)
       return;
 
-   __driParseEvents(pDRICtx, dPriv);
+   intel_update_renderbuffers(pDRICtx, dPriv);
 
    rb = intel_fb->color_rb[0];
    type = GL_BGRA;
diff --git a/src/mesa/drivers/windows/gdi/wmesa.c b/src/mesa/drivers/windows/gdi/wmesa.c
index 1c179eb3bb3..9dafb747231 100644
--- a/src/mesa/drivers/windows/gdi/wmesa.c
+++ b/src/mesa/drivers/windows/gdi/wmesa.c
@@ -118,6 +118,7 @@ static void wmSetPixelFormat(WMesaFramebuffer pwfb, HDC hDC)
     /* Only 16 and 32 bit targets are supported now */
     assert(pwfb->cColorBits == 0 ||
 	   pwfb->cColorBits == 16 || 
+	   pwfb->cColorBits == 24 || 
 	   pwfb->cColorBits == 32);
 
     switch(pwfb->cColorBits){
@@ -127,6 +128,7 @@ static void wmSetPixelFormat(WMesaFramebuffer pwfb, HDC hDC)
     case 16:
 	pwfb->pixelformat = PF_5R6G5B;
 	break;
+    case 24:
     case 32:
 	pwfb->pixelformat = PF_8R8G8B;
 	break;
@@ -856,6 +858,195 @@ static void read_rgba_pixels_32(const GLcontext *ctx,
 
 /*********************************************************************/
 
+/* DOUBLE BUFFER 24-bit */
+
+#define WMSETPIXEL24(pwc, y, x, r, g, b) { \
+LPBYTE lpb = ((LPBYTE)((pwc)->pbPixels + (pwc)->ScanWidth * (y)) + (3 * x)); \
+lpb[0] = (b); \
+lpb[1] = (g); \
+lpb[2] = (r); }
+
+/* Write a horizontal span of RGBA color pixels with a boolean mask. */
+static void write_rgba_span_24(const GLcontext *ctx, 
+			       struct gl_renderbuffer *rb, 
+			       GLuint n, GLint x, GLint y,
+			       const GLubyte rgba[][4], 
+			       const GLubyte mask[] )
+{
+    WMesaContext pwc = wmesa_context(ctx);
+    WMesaFramebuffer pwfb = wmesa_framebuffer(ctx->DrawBuffer);
+    GLuint i;
+    LPBYTE lpb;
+
+    (void) ctx;
+    
+    y=FLIP(y);
+    lpb = ((LPBYTE)(pwfb->pbPixels + pwfb->ScanWidth * y)) + (3 * x);
+    if (mask) {
+	for (i=0; i<n; i++)
+	    if (mask[i]) {
+                lpb[3*i] = rgba[i][BCOMP];
+                lpb[3*i+1] = rgba[i][GCOMP];
+                lpb[3*i+2] = rgba[i][RCOMP];
+	    }
+    }
+    else {
+	    for (i=0; i<n; i++) {
+            *lpb++ = rgba[i][BCOMP];
+            *lpb++ = rgba[i][GCOMP];
+            *lpb++ = rgba[i][RCOMP];
+	    }
+    }
+}
+
+
+/* Write a horizontal span of RGB color pixels with a boolean mask. */
+static void write_rgb_span_24(const GLcontext *ctx, 
+			      struct gl_renderbuffer *rb, 
+			      GLuint n, GLint x, GLint y,
+			      const GLubyte rgb[][3], 
+			      const GLubyte mask[] )
+{
+    WMesaContext pwc = wmesa_context(ctx);
+    WMesaFramebuffer pwfb = wmesa_framebuffer(ctx->DrawBuffer);
+    GLuint i;
+    LPBYTE lpb;
+
+    (void) ctx;
+    
+    y=FLIP(y);
+    lpb = ((LPBYTE)(pwfb->pbPixels + pwfb->ScanWidth * y)) + (3 * x);
+    if (mask) {
+	for (i=0; i<n; i++)
+	    if (mask[i]) {
+            lpb[3*i] = rgb[i][BCOMP];
+            lpb[3*i+1] = rgb[i][GCOMP];
+            lpb[3*i+2] = rgb[i][RCOMP];
+	    }
+    }
+    else {
+    	for (i=0; i<n; i++) {
+    		*lpb++ = rgb[i][BCOMP];
+    		*lpb++ = rgb[i][GCOMP];
+    		*lpb++ = rgb[i][RCOMP];
+    	}
+    }
+}
+
+/*
+ * Write a horizontal span of pixels with a boolean mask.  The current color
+ * is used for all pixels.
+ */
+static void write_mono_rgba_span_24(const GLcontext *ctx, 
+				    struct gl_renderbuffer *rb,
+				    GLuint n, GLint x, GLint y,
+				    const GLchan color[4], 
+				    const GLubyte mask[])
+{
+    LPBYTE lpb;
+    GLuint i;
+    WMesaContext pwc = wmesa_context(ctx);
+    WMesaFramebuffer pwfb = wmesa_framebuffer(ctx->DrawBuffer);
+    lpb = ((LPBYTE)(pwfb->pbPixels + pwfb->ScanWidth * y)) + (3 * x);
+    y=FLIP(y);
+    if (mask) {
+	for (i=0; i<n; i++)
+	    if (mask[i]) {
+	    	lpb[3*i] = color[BCOMP];
+	    	lpb[3*i+1] = color[GCOMP];
+	    	lpb[3*i+2] = color[RCOMP];
+	    }
+    }
+    else
+	for (i=0; i<n; i++) {
+		*lpb++ = color[BCOMP];
+		*lpb++ = color[GCOMP];
+		*lpb++ = color[RCOMP];		
+	}
+}
+
+/* Write an array of RGBA pixels with a boolean mask. */
+static void write_rgba_pixels_24(const GLcontext *ctx, 
+				 struct gl_renderbuffer *rb,
+				 GLuint n, const GLint x[], const GLint y[],
+				 const GLubyte rgba[][4], 
+				 const GLubyte mask[])
+{
+    GLuint i;
+    WMesaContext pwc = wmesa_context(ctx);
+    WMesaFramebuffer pwfb = wmesa_framebuffer(ctx->DrawBuffer);
+    for (i=0; i<n; i++)
+	if (mask[i])
+	    WMSETPIXEL24(pwfb, FLIP(y[i]), x[i],
+			 rgba[i][RCOMP], rgba[i][GCOMP], rgba[i][BCOMP]);
+}
+
+/*
+ * Write an array of pixels with a boolean mask.  The current color
+ * is used for all pixels.
+ */
+static void write_mono_rgba_pixels_24(const GLcontext *ctx, 
+				      struct gl_renderbuffer *rb,
+				      GLuint n,
+				      const GLint x[], const GLint y[],
+				      const GLchan color[4],
+				      const GLubyte mask[])
+{
+    GLuint i;
+    WMesaContext pwc = wmesa_context(ctx);
+    WMesaFramebuffer pwfb = wmesa_framebuffer(ctx->DrawBuffer);
+    for (i=0; i<n; i++)
+	if (mask[i])
+	    WMSETPIXEL24(pwfb, FLIP(y[i]),x[i],color[RCOMP],
+			 color[GCOMP], color[BCOMP]);
+}
+
+/* Read a horizontal span of color pixels. */
+static void read_rgba_span_24(const GLcontext *ctx, 
+			      struct gl_renderbuffer *rb,
+			      GLuint n, GLint x, GLint y,
+			      GLubyte rgba[][4] )
+{
+    GLuint i;
+    LPBYTE lpb;
+    WMesaContext pwc = wmesa_context(ctx);
+    WMesaFramebuffer pwfb = wmesa_framebuffer(ctx->DrawBuffer);
+    
+    y = FLIP(y);
+    lpb = ((LPBYTE)(pwfb->pbPixels + pwfb->ScanWidth * y)) + (3 * x);
+    for (i=0; i<n; i++) {
+	rgba[i][RCOMP] = lpb[3*i+2];
+	rgba[i][GCOMP] = lpb[3*i+1];
+	rgba[i][BCOMP] = lpb[3*i];
+	rgba[i][ACOMP] = 255;
+    }
+}
+
+
+/* Read an array of color pixels. */
+static void read_rgba_pixels_24(const GLcontext *ctx, 
+				struct gl_renderbuffer *rb,
+				GLuint n, const GLint x[], const GLint y[],
+				GLubyte rgba[][4])
+{
+    GLuint i;
+    LPBYTE lpb;
+    WMesaContext pwc = wmesa_context(ctx);
+    WMesaFramebuffer pwfb = wmesa_framebuffer(ctx->DrawBuffer);
+
+    for (i=0; i<n; i++) {
+	GLint y2 = FLIP(y[i]);
+	lpb = ((LPBYTE)(pwfb->pbPixels + pwfb->ScanWidth * y2)) + (3 * x[i]);
+	rgba[i][RCOMP] = lpb[3*i+2];
+	rgba[i][GCOMP] = lpb[3*i+1];
+	rgba[i][BCOMP] = lpb[3*i];
+	rgba[i][ACOMP] = 255;
+  }
+}
+
+
+/*********************************************************************/
+
 /* DOUBLE BUFFER 16-bit */
 
 #define WMSETPIXEL16(pwc, y, x, r, g, b) { \
@@ -1077,7 +1268,7 @@ wmesa_renderbuffer_storage(GLcontext *ctx,
  * on if we're drawing to the front or back color buffer.
  */
 void wmesa_set_renderbuffer_funcs(struct gl_renderbuffer *rb, int pixelformat,
-                                  int double_buffer)
+                                  BYTE cColorBits, int double_buffer)
 {
     if (double_buffer) {
         /* back buffer */
@@ -1097,16 +1288,32 @@ void wmesa_set_renderbuffer_funcs(struct gl_renderbuffer *rb, int pixelformat,
             rb->BlueBits = 5;
 	    break;
 	case PF_8R8G8B:
-	    rb->PutRow = write_rgba_span_32;
-	    rb->PutRowRGB = write_rgb_span_32;
-	    rb->PutMonoRow = write_mono_rgba_span_32;
-	    rb->PutValues = write_rgba_pixels_32;
-	    rb->PutMonoValues = write_mono_rgba_pixels_32;
-	    rb->GetRow = read_rgba_span_32;
-	    rb->GetValues = read_rgba_pixels_32;
+		if (cColorBits == 24)
+		{
+		    rb->PutRow = write_rgba_span_24;
+		    rb->PutRowRGB = write_rgb_span_24;
+		    rb->PutMonoRow = write_mono_rgba_span_24;
+		    rb->PutValues = write_rgba_pixels_24;
+		    rb->PutMonoValues = write_mono_rgba_pixels_24;
+		    rb->GetRow = read_rgba_span_24;
+		    rb->GetValues = read_rgba_pixels_24;
+	        rb->RedBits = 8;
+	        rb->GreenBits = 8;
+	        rb->BlueBits = 8;		
+		}
+		else
+		{
+	        rb->PutRow = write_rgba_span_32;
+	        rb->PutRowRGB = write_rgb_span_32;
+	        rb->PutMonoRow = write_mono_rgba_span_32;
+	        rb->PutValues = write_rgba_pixels_32;
+	        rb->PutMonoValues = write_mono_rgba_pixels_32;
+	        rb->GetRow = read_rgba_span_32;
+	        rb->GetValues = read_rgba_pixels_32;
             rb->RedBits = 8;
             rb->GreenBits = 8;
             rb->BlueBits = 8;
+		}
 	    break;
 	default:
 	    break;
@@ -1413,11 +1620,11 @@ void WMesaMakeCurrent(WMesaContext c, HDC hdc)
         if (visual->doubleBufferMode == 1) {
             rb = wmesa_new_renderbuffer();
             _mesa_add_renderbuffer(&pwfb->Base, BUFFER_BACK_LEFT, rb);
-            wmesa_set_renderbuffer_funcs(rb, pwfb->pixelformat, 1);
+            wmesa_set_renderbuffer_funcs(rb, pwfb->pixelformat, pwfb->cColorBits, 1);
 	}
         rb = wmesa_new_renderbuffer();
         _mesa_add_renderbuffer(&pwfb->Base, BUFFER_FRONT_LEFT, rb);
-        wmesa_set_renderbuffer_funcs(rb, pwfb->pixelformat, 0);
+        wmesa_set_renderbuffer_funcs(rb, pwfb->pixelformat, pwfb->cColorBits, 0);
 
 	/* Let Mesa own the Depth, Stencil, and Accum buffers */
         _mesa_add_soft_renderbuffers(&pwfb->Base,
diff --git a/src/mesa/main/arrayobj.c b/src/mesa/main/arrayobj.c
index d62661e2b52..14612393176 100644
--- a/src/mesa/main/arrayobj.c
+++ b/src/mesa/main/arrayobj.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5
+ * Version:  7.2
  *
- * Copyright (C) 1999-2004  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  * (C) Copyright IBM Corporation 2006
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -77,7 +77,7 @@ lookup_arrayobj(GLcontext *ctx, GLuint id)
 struct gl_array_object *
 _mesa_new_array_object( GLcontext *ctx, GLuint name )
 {
-   struct gl_array_object *obj = MALLOC_STRUCT(gl_array_object);
+   struct gl_array_object *obj = CALLOC_STRUCT(gl_array_object);
    if (obj)
       _mesa_initialize_array_object(ctx, obj, name);
    return obj;
@@ -219,6 +219,15 @@ _mesa_remove_array_object( GLcontext *ctx, struct gl_array_object *obj )
 }
 
 
+static void
+unbind_buffer_object( GLcontext *ctx, struct gl_buffer_object *bufObj )
+{
+   if (bufObj != ctx->Array.NullBufferObj) {
+      _mesa_reference_buffer_object(ctx, &bufObj, NULL);
+   }
+}
+
+
 /**********************************************************************/
 /* API Functions                                                      */
 /**********************************************************************/
@@ -320,18 +329,18 @@ _mesa_DeleteVertexArraysAPPLE(GLsizei n, const GLuint *ids)
 	 /* Unbind any buffer objects that might be bound to arrays in
 	  * this array object.
 	  */
-	 _mesa_unbind_buffer_object( ctx, obj->Vertex.BufferObj );
-	 _mesa_unbind_buffer_object( ctx, obj->Normal.BufferObj );
-	 _mesa_unbind_buffer_object( ctx, obj->Color.BufferObj );
-	 _mesa_unbind_buffer_object( ctx, obj->SecondaryColor.BufferObj );
-	 _mesa_unbind_buffer_object( ctx, obj->FogCoord.BufferObj );
-	 _mesa_unbind_buffer_object( ctx, obj->Index.BufferObj );
+	 unbind_buffer_object( ctx, obj->Vertex.BufferObj );
+	 unbind_buffer_object( ctx, obj->Normal.BufferObj );
+	 unbind_buffer_object( ctx, obj->Color.BufferObj );
+	 unbind_buffer_object( ctx, obj->SecondaryColor.BufferObj );
+	 unbind_buffer_object( ctx, obj->FogCoord.BufferObj );
+	 unbind_buffer_object( ctx, obj->Index.BufferObj );
 	 for (i = 0; i < MAX_TEXTURE_UNITS; i++) {
-	    _mesa_unbind_buffer_object( ctx, obj->TexCoord[i].BufferObj );
+	    unbind_buffer_object( ctx, obj->TexCoord[i].BufferObj );
 	 }
-	 _mesa_unbind_buffer_object( ctx, obj->EdgeFlag.BufferObj );
+	 unbind_buffer_object( ctx, obj->EdgeFlag.BufferObj );
 	 for (i = 0; i < VERT_ATTRIB_MAX; i++) {
-	    _mesa_unbind_buffer_object( ctx, obj->VertexAttrib[i].BufferObj );
+	    unbind_buffer_object( ctx, obj->VertexAttrib[i].BufferObj );
 	 }
 #endif
 
diff --git a/src/mesa/main/attrib.c b/src/mesa/main/attrib.c
index ce0733855cc..1a2c34b7109 100644
--- a/src/mesa/main/attrib.c
+++ b/src/mesa/main/attrib.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
+ * Version:  7.2
  *
- * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -1280,6 +1280,29 @@ adjust_buffer_object_ref_counts(struct gl_array_attrib *array, GLint step)
 }
 
 
+/**
+ * Copy gl_pixelstore_attrib from src to dst, updating buffer
+ * object refcounts.
+ */
+static void
+copy_pixelstore(GLcontext *ctx,
+                struct gl_pixelstore_attrib *dst,
+                const struct gl_pixelstore_attrib *src)
+{
+   dst->Alignment = src->Alignment;
+   dst->RowLength = src->RowLength;
+   dst->SkipPixels = src->SkipPixels;
+   dst->SkipRows = src->SkipRows;
+   dst->ImageHeight = src->ImageHeight;
+   dst->SkipImages = src->SkipImages;
+   dst->SwapBytes = src->SwapBytes;
+   dst->LsbFirst = src->LsbFirst;
+   dst->ClientStorage = src->ClientStorage;
+   dst->Invert = src->Invert;
+   _mesa_reference_buffer_object(ctx, &dst->BufferObj, src->BufferObj);
+}
+
+
 #define GL_CLIENT_PACK_BIT (1<<20)
 #define GL_CLIENT_UNPACK_BIT (1<<21)
 
@@ -1298,31 +1321,29 @@ _mesa_PushClientAttrib(GLbitfield mask)
       return;
    }
 
-   /* Build linked list of attribute nodes which save all attribute */
-   /* groups specified by the mask. */
+   /* Build linked list of attribute nodes which save all attribute
+    * groups specified by the mask.
+    */
    head = NULL;
 
    if (mask & GL_CLIENT_PIXEL_STORE_BIT) {
       struct gl_pixelstore_attrib *attr;
-#if FEATURE_EXT_pixel_buffer_object
-      ctx->Pack.BufferObj->RefCount++;
-      ctx->Unpack.BufferObj->RefCount++;
-#endif
       /* packing attribs */
-      attr = MALLOC_STRUCT( gl_pixelstore_attrib );
-      MEMCPY( attr, &ctx->Pack, sizeof(struct gl_pixelstore_attrib) );
+      attr = CALLOC_STRUCT( gl_pixelstore_attrib );
+      copy_pixelstore(ctx, attr, &ctx->Pack);
       newnode = new_attrib_node( GL_CLIENT_PACK_BIT );
       newnode->data = attr;
       newnode->next = head;
       head = newnode;
       /* unpacking attribs */
-      attr = MALLOC_STRUCT( gl_pixelstore_attrib );
-      MEMCPY( attr, &ctx->Unpack, sizeof(struct gl_pixelstore_attrib) );
+      attr = CALLOC_STRUCT( gl_pixelstore_attrib );
+      copy_pixelstore(ctx, attr, &ctx->Unpack);
       newnode = new_attrib_node( GL_CLIENT_UNPACK_BIT );
       newnode->data = attr;
       newnode->next = head;
       head = newnode;
    }
+
    if (mask & GL_CLIENT_VERTEX_ARRAY_BIT) {
       struct gl_array_attrib *attr;
       struct gl_array_object *obj;
@@ -1359,7 +1380,7 @@ _mesa_PushClientAttrib(GLbitfield mask)
 void GLAPIENTRY
 _mesa_PopClientAttrib(void)
 {
-   struct gl_attrib_node *attr, *next;
+   struct gl_attrib_node *node, *next;
 
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx);
@@ -1370,37 +1391,31 @@ _mesa_PopClientAttrib(void)
    }
 
    ctx->ClientAttribStackDepth--;
-   attr = ctx->ClientAttribStack[ctx->ClientAttribStackDepth];
+   node = ctx->ClientAttribStack[ctx->ClientAttribStackDepth];
 
-   while (attr) {
-      switch (attr->kind) {
+   while (node) {
+      switch (node->kind) {
          case GL_CLIENT_PACK_BIT:
-#if FEATURE_EXT_pixel_buffer_object
-            ctx->Pack.BufferObj->RefCount--;
-            if (ctx->Pack.BufferObj->RefCount <= 0) {
-               _mesa_remove_buffer_object( ctx, ctx->Pack.BufferObj );
-               (*ctx->Driver.DeleteBuffer)( ctx, ctx->Pack.BufferObj );
+            {
+               struct gl_pixelstore_attrib *store =
+                  (struct gl_pixelstore_attrib *) node->data;
+               copy_pixelstore(ctx, &ctx->Pack, store);
+               _mesa_reference_buffer_object(ctx, &store->BufferObj, NULL);
             }
-#endif
-            MEMCPY( &ctx->Pack, attr->data,
-                    sizeof(struct gl_pixelstore_attrib) );
 	    ctx->NewState |= _NEW_PACKUNPACK;
             break;
          case GL_CLIENT_UNPACK_BIT:
-#if FEATURE_EXT_pixel_buffer_object
-            ctx->Unpack.BufferObj->RefCount--;
-            if (ctx->Unpack.BufferObj->RefCount <= 0) {
-               _mesa_remove_buffer_object( ctx, ctx->Unpack.BufferObj );
-               (*ctx->Driver.DeleteBuffer)( ctx, ctx->Unpack.BufferObj );
+            {
+               struct gl_pixelstore_attrib *store =
+                  (struct gl_pixelstore_attrib *) node->data;
+               copy_pixelstore(ctx, &ctx->Unpack, store);
+               _mesa_reference_buffer_object(ctx, &store->BufferObj, NULL);
             }
-#endif
-            MEMCPY( &ctx->Unpack, attr->data,
-                    sizeof(struct gl_pixelstore_attrib) );
 	    ctx->NewState |= _NEW_PACKUNPACK;
             break;
          case GL_CLIENT_VERTEX_ARRAY_BIT: {
 	    struct gl_array_attrib * data =
-	      (struct gl_array_attrib *) attr->data;
+	      (struct gl_array_attrib *) node->data;
 
             adjust_buffer_object_ref_counts(&ctx->Array, -1);
 	 
@@ -1437,10 +1452,10 @@ _mesa_PopClientAttrib(void)
             break;
       }
 
-      next = attr->next;
-      FREE( attr->data );
-      FREE( attr );
-      attr = next;
+      next = node->next;
+      FREE( node->data );
+      FREE( node );
+      node = next;
    }
 }
 
diff --git a/src/mesa/main/bufferobj.c b/src/mesa/main/bufferobj.c
index 8d3ddaf60f0..11e1fd6b2b2 100644
--- a/src/mesa/main/bufferobj.c
+++ b/src/mesa/main/bufferobj.c
@@ -1,6 +1,6 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
+ * Version:  7.2
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
@@ -166,22 +166,75 @@ _mesa_delete_buffer_object( GLcontext *ctx, struct gl_buffer_object *bufObj )
 
    if (bufObj->Data)
       _mesa_free(bufObj->Data);
+
+   /* assign strange values here to help w/ debugging */
+   bufObj->RefCount = -1000;
+   bufObj->Name = ~0;
+
    _mesa_free(bufObj);
 }
 
 
+
+/**
+ * Set ptr to bufObj w/ reference counting.
+ */
 void
-_mesa_unbind_buffer_object( GLcontext *ctx, struct gl_buffer_object *bufObj )
+_mesa_reference_buffer_object(GLcontext *ctx,
+                              struct gl_buffer_object **ptr,
+                              struct gl_buffer_object *bufObj)
 {
-   if (bufObj != ctx->Array.NullBufferObj) {
-      bufObj->RefCount--;
-      if (bufObj->RefCount <= 0) {
+   if (*ptr == bufObj)
+      return;
+
+   if (*ptr) {
+      /* Unreference the old texture */
+      GLboolean deleteFlag = GL_FALSE;
+      struct gl_buffer_object *oldObj = *ptr;
+
+      /*_glthread_LOCK_MUTEX(oldObj->Mutex);*/
+      ASSERT(oldObj->RefCount > 0);
+      oldObj->RefCount--;
+#if 0
+      printf("BufferObj %p %d DECR to %d\n",
+             (void *) oldObj, oldObj->Name, oldObj->RefCount);
+#endif
+      deleteFlag = (oldObj->RefCount == 0);
+      /*_glthread_UNLOCK_MUTEX(oldObj->Mutex);*/
+
+      if (deleteFlag) {
+
+         /* some sanity checking: don't delete a buffer still in use */
 	 ASSERT(ctx->Array.ArrayBufferObj != bufObj);
 	 ASSERT(ctx->Array.ElementArrayBufferObj != bufObj);
 	 ASSERT(ctx->Array.ArrayObj->Vertex.BufferObj != bufObj);
 	 ASSERT(ctx->Driver.DeleteBuffer);
-	 ctx->Driver.DeleteBuffer(ctx, bufObj);
+
+         ctx->Driver.DeleteBuffer(ctx, oldObj);
       }
+
+      *ptr = NULL;
+   }
+   ASSERT(!*ptr);
+
+   if (bufObj) {
+      /* reference new texture */
+      /*_glthread_LOCK_MUTEX(tex->Mutex);*/
+      if (bufObj->RefCount == 0) {
+         /* this buffer's being deleted (look just above) */
+         /* Not sure this can every really happen.  Warn if it does. */
+         _mesa_problem(NULL, "referencing deleted buffer object");
+         *ptr = NULL;
+      }
+      else {
+         bufObj->RefCount++;
+#if 0
+         printf("BufferObj %p %d INCR to %d\n",
+                (void *) bufObj, bufObj->Name, bufObj->RefCount);
+#endif
+         *ptr = bufObj;
+      }
+      /*_glthread_UNLOCK_MUTEX(tex->Mutex);*/
    }
 }
 
@@ -204,33 +257,6 @@ _mesa_initialize_buffer_object( struct gl_buffer_object *obj,
 
 
 /**
- * Add the given buffer object to the buffer object pool.
- */
-void
-_mesa_save_buffer_object( GLcontext *ctx, struct gl_buffer_object *obj )
-{
-   if (obj->Name > 0) {
-      /* insert into hash table */
-      _mesa_HashInsert(ctx->Shared->BufferObjects, obj->Name, obj);
-   }
-}
-
-
-/**
- * Remove the given buffer object from the buffer object pool.
- * Do not deallocate the buffer object though.
- */
-void
-_mesa_remove_buffer_object( GLcontext *ctx, struct gl_buffer_object *bufObj )
-{
-   if (bufObj->Name > 0) {
-      /* remove from hash table */
-      _mesa_HashRemove(ctx->Shared->BufferObjects, bufObj->Name);
-   }
-}
-
-
-/**
  * Allocate space for and store data in a buffer object.  Any data that was
  * previously stored in the buffer object is lost.  If \c data is \c NULL,
  * memory will be allocated, but no copy will occur.
@@ -400,6 +426,7 @@ _mesa_init_buffer_objects( GLcontext *ctx )
 {
    /* Allocate the default buffer object and set refcount so high that
     * it never gets deleted.
+    * XXX with recent/improved refcounting this may not longer be needed.
     */
    ctx->Array.NullBufferObj = _mesa_new_buffer_object(ctx, 0, 0);
    if (ctx->Array.NullBufferObj)
@@ -462,28 +489,16 @@ bind_buffer_object(GLcontext *ctx, GLenum target, GLuint buffer)
             _mesa_error(ctx, GL_OUT_OF_MEMORY, "glBindBufferARB");
             return;
          }
-         _mesa_save_buffer_object(ctx, newBufObj);
+         _mesa_HashInsert(ctx->Shared->BufferObjects, buffer, newBufObj);
       }
    }
    
-   /* Make new binding */
-   *bindTarget = newBufObj;
-   newBufObj->RefCount++;
+   /* bind new buffer */
+   _mesa_reference_buffer_object(ctx, bindTarget, newBufObj);
 
    /* Pass BindBuffer call to device driver */
    if (ctx->Driver.BindBuffer && newBufObj)
       ctx->Driver.BindBuffer( ctx, target, newBufObj );
-
-   /* decr ref count on old buffer obj, delete if needed */
-   if (oldBufObj) {
-      oldBufObj->RefCount--;
-      assert(oldBufObj->RefCount >= 0);
-      if (oldBufObj->RefCount == 0) {
-         assert(oldBufObj->Name != 0);
-         ASSERT(ctx->Driver.DeleteBuffer);
-         ctx->Driver.DeleteBuffer( ctx, oldBufObj );
-      }
-   }
 }
 
 
@@ -716,6 +731,23 @@ _mesa_lookup_bufferobj(GLcontext *ctx, GLuint buffer)
 }
 
 
+/**
+ * If *ptr points to obj, set ptr = the Null/default buffer object.
+ * This is a helper for buffer object deletion.
+ * The GL spec says that deleting a buffer object causes it to get
+ * unbound from all arrays in the current context.
+ */
+static void
+unbind(GLcontext *ctx,
+       struct gl_buffer_object **ptr,
+       struct gl_buffer_object *obj)
+{
+   if (*ptr == obj) {
+      _mesa_reference_buffer_object(ctx, ptr, ctx->Array.NullBufferObj);
+   }
+}
+
+
 
 /**********************************************************************/
 /* API Functions                                                      */
@@ -759,54 +791,18 @@ _mesa_DeleteBuffersARB(GLsizei n, const GLuint *ids)
 
          ASSERT(bufObj->Name == ids[i]);
 
-         if (ctx->Array.ArrayObj->Vertex.BufferObj == bufObj) {
-            bufObj->RefCount--;
-            ctx->Array.ArrayObj->Vertex.BufferObj = ctx->Array.NullBufferObj;
-            ctx->Array.NullBufferObj->RefCount++;
-         }
-         if (ctx->Array.ArrayObj->Normal.BufferObj == bufObj) {
-            bufObj->RefCount--;
-            ctx->Array.ArrayObj->Normal.BufferObj = ctx->Array.NullBufferObj;
-            ctx->Array.NullBufferObj->RefCount++;
-         }
-         if (ctx->Array.ArrayObj->Color.BufferObj == bufObj) {
-            bufObj->RefCount--;
-            ctx->Array.ArrayObj->Color.BufferObj = ctx->Array.NullBufferObj;
-            ctx->Array.NullBufferObj->RefCount++;
-         }
-         if (ctx->Array.ArrayObj->SecondaryColor.BufferObj == bufObj) {
-            bufObj->RefCount--;
-            ctx->Array.ArrayObj->SecondaryColor.BufferObj = ctx->Array.NullBufferObj;
-            ctx->Array.NullBufferObj->RefCount++;
-         }
-         if (ctx->Array.ArrayObj->FogCoord.BufferObj == bufObj) {
-            bufObj->RefCount--;
-            ctx->Array.ArrayObj->FogCoord.BufferObj = ctx->Array.NullBufferObj;
-            ctx->Array.NullBufferObj->RefCount++;
-         }
-         if (ctx->Array.ArrayObj->Index.BufferObj == bufObj) {
-            bufObj->RefCount--;
-            ctx->Array.ArrayObj->Index.BufferObj = ctx->Array.NullBufferObj;
-            ctx->Array.NullBufferObj->RefCount++;
-         }
-         if (ctx->Array.ArrayObj->EdgeFlag.BufferObj == bufObj) {
-            bufObj->RefCount--;
-            ctx->Array.ArrayObj->EdgeFlag.BufferObj = ctx->Array.NullBufferObj;
-            ctx->Array.NullBufferObj->RefCount++;
-         }
+         unbind(ctx, &ctx->Array.ArrayObj->Vertex.BufferObj, bufObj);
+         unbind(ctx, &ctx->Array.ArrayObj->Normal.BufferObj, bufObj);
+         unbind(ctx, &ctx->Array.ArrayObj->Color.BufferObj, bufObj);
+         unbind(ctx, &ctx->Array.ArrayObj->SecondaryColor.BufferObj, bufObj);
+         unbind(ctx, &ctx->Array.ArrayObj->FogCoord.BufferObj, bufObj);
+         unbind(ctx, &ctx->Array.ArrayObj->Index.BufferObj, bufObj);
+         unbind(ctx, &ctx->Array.ArrayObj->EdgeFlag.BufferObj, bufObj);
          for (j = 0; j < MAX_TEXTURE_UNITS; j++) {
-            if (ctx->Array.ArrayObj->TexCoord[j].BufferObj == bufObj) {
-               bufObj->RefCount--;
-               ctx->Array.ArrayObj->TexCoord[j].BufferObj = ctx->Array.NullBufferObj;
-               ctx->Array.NullBufferObj->RefCount++;
-            }
+            unbind(ctx, &ctx->Array.ArrayObj->TexCoord[j].BufferObj, bufObj);
          }
          for (j = 0; j < VERT_ATTRIB_MAX; j++) {
-            if (ctx->Array.ArrayObj->VertexAttrib[j].BufferObj == bufObj) {
-               bufObj->RefCount--;
-               ctx->Array.ArrayObj->VertexAttrib[j].BufferObj = ctx->Array.NullBufferObj;
-               ctx->Array.NullBufferObj->RefCount++;
-            }
+            unbind(ctx, &ctx->Array.ArrayObj->VertexAttrib[j].BufferObj, bufObj);
          }
 
          if (ctx->Array.ArrayBufferObj == bufObj) {
@@ -824,8 +820,8 @@ _mesa_DeleteBuffersARB(GLsizei n, const GLuint *ids)
          }
 
          /* The ID is immediately freed for re-use */
-         _mesa_remove_buffer_object(ctx, bufObj);
-         _mesa_unbind_buffer_object(ctx, bufObj);
+         _mesa_HashRemove(ctx->Shared->BufferObjects, bufObj->Name);
+         _mesa_reference_buffer_object(ctx, &bufObj, NULL);
       }
    }
 
@@ -874,7 +870,7 @@ _mesa_GenBuffersARB(GLsizei n, GLuint *buffer)
          _mesa_error(ctx, GL_OUT_OF_MEMORY, "glGenBuffersARB");
          return;
       }
-      _mesa_save_buffer_object(ctx, bufObj);
+      _mesa_HashInsert(ctx->Shared->BufferObjects, first + i, bufObj);
       buffer[i] = first + i;
    }
 
diff --git a/src/mesa/main/bufferobj.h b/src/mesa/main/bufferobj.h
index 537b2404740..3c08f0083cf 100644
--- a/src/mesa/main/bufferobj.h
+++ b/src/mesa/main/bufferobj.h
@@ -1,6 +1,6 @@
 /*
  * Mesa 3-D graphics library
- * Version:  7.1
+ * Version:  7.2
  *
  * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
@@ -55,10 +55,9 @@ _mesa_initialize_buffer_object( struct gl_buffer_object *obj,
 				GLuint name, GLenum target );
 
 extern void
-_mesa_save_buffer_object( GLcontext *ctx, struct gl_buffer_object *obj );
-
-extern void
-_mesa_remove_buffer_object( GLcontext *ctx, struct gl_buffer_object *bufObj );
+_mesa_reference_buffer_object(GLcontext *ctx,
+                              struct gl_buffer_object **ptr,
+                              struct gl_buffer_object *bufObj);
 
 extern void
 _mesa_buffer_data( GLcontext *ctx, GLenum target, GLsizeiptrARB size,
@@ -117,8 +116,6 @@ _mesa_unmap_readpix_pbo(GLcontext *ctx,
                         const struct gl_pixelstore_attrib *pack);
 
 
-extern void
-_mesa_unbind_buffer_object( GLcontext *ctx, struct gl_buffer_object *bufObj );
 
 /*
  * API functions
diff --git a/src/mesa/main/extensions.c b/src/mesa/main/extensions.c
index 40a2c1b0a86..8dfbfeb62e4 100644
--- a/src/mesa/main/extensions.c
+++ b/src/mesa/main/extensions.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.1
+ * Version:  7.1
  *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
diff --git a/src/mesa/main/framebuffer.c b/src/mesa/main/framebuffer.c
index aba887a796f..af78363ad3a 100644
--- a/src/mesa/main/framebuffer.c
+++ b/src/mesa/main/framebuffer.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.1
+ * Version:  7.2
  *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -569,13 +569,13 @@ _mesa_update_framebuffer_visual(struct gl_framebuffer *fb)
    if (fb->Attachment[BUFFER_ACCUM].Renderbuffer) {
       fb->Visual.haveAccumBuffer = GL_TRUE;
       fb->Visual.accumRedBits
-         = fb->Attachment[BUFFER_DEPTH].Renderbuffer->RedBits;
+         = fb->Attachment[BUFFER_ACCUM].Renderbuffer->RedBits;
       fb->Visual.accumGreenBits
-         = fb->Attachment[BUFFER_DEPTH].Renderbuffer->GreenBits;
+         = fb->Attachment[BUFFER_ACCUM].Renderbuffer->GreenBits;
       fb->Visual.accumBlueBits
-         = fb->Attachment[BUFFER_DEPTH].Renderbuffer->BlueBits;
+         = fb->Attachment[BUFFER_ACCUM].Renderbuffer->BlueBits;
       fb->Visual.accumAlphaBits
-         = fb->Attachment[BUFFER_DEPTH].Renderbuffer->AlphaBits;
+         = fb->Attachment[BUFFER_ACCUM].Renderbuffer->AlphaBits;
    }
 
    compute_depth_max(fb);
diff --git a/src/mesa/main/getstring.c b/src/mesa/main/getstring.c
index 48c815772d8..1a82ccce59a 100644
--- a/src/mesa/main/getstring.c
+++ b/src/mesa/main/getstring.c
@@ -85,7 +85,7 @@ compute_version(const GLcontext *ctx)
                               ctx->Extensions.ARB_texture_non_power_of_two &&
                               ctx->Extensions.EXT_blend_equation_separate);
    const GLboolean ver_2_1 = (ver_2_0 &&
-                              ctx->Extensions.ARB_shading_language_120 &&
+                              /*ctx->Extensions.ARB_shading_language_120 &&*/
                               ctx->Extensions.EXT_pixel_buffer_object &&
                               ctx->Extensions.EXT_texture_sRGB);
    if (ver_2_1)
@@ -121,12 +121,6 @@ _mesa_GetString( GLenum name )
    static const char *vendor = "Brian Paul";
    static const char *renderer = "Mesa";
 
-#if FEATURE_ARB_shading_language_120_foo /* support not complete! */
-   static const char *sl_version = "1.20";
-#elif FEATURE_ARB_shading_language_100
-   static const char *sl_version = "1.10";
-#endif
-
    if (!ctx)
       return NULL;
 
@@ -154,8 +148,10 @@ _mesa_GetString( GLenum name )
          return (const GLubyte *) ctx->Extensions.String;
 #if FEATURE_ARB_shading_language_100
       case GL_SHADING_LANGUAGE_VERSION_ARB:
-         if (ctx->Extensions.ARB_shading_language_100)
-            return (const GLubyte *) sl_version;
+         if (ctx->Extensions.ARB_shading_language_120)
+            return (const GLubyte *) "1.20";
+         else if (ctx->Extensions.ARB_shading_language_100)
+            return (const GLubyte *) "1.10";
          goto error;
 #endif
 #if FEATURE_NV_fragment_program || FEATURE_ARB_fragment_program || \
diff --git a/src/mesa/main/imports.c b/src/mesa/main/imports.c
index 1307bae0d60..6cfd7ccc723 100644
--- a/src/mesa/main/imports.c
+++ b/src/mesa/main/imports.c
@@ -559,7 +559,7 @@ _mesa_pow(double x, double y)
 int
 _mesa_ffs(int32_t i)
 {
-#if (defined(_WIN32) && !defined(__MINGW32__) ) || defined(__IBMC__) || defined(__IBMCPP__)
+#if (defined(_WIN32) ) || defined(__IBMC__) || defined(__IBMCPP__)
    register int bit = 0;
    if (i != 0) {
       if ((i & 0xffff) == 0) {
diff --git a/src/mesa/main/light.c b/src/mesa/main/light.c
index 30e1908771f..d4db960f1b9 100644
--- a/src/mesa/main/light.c
+++ b/src/mesa/main/light.c
@@ -184,6 +184,7 @@ _mesa_Lightfv( GLenum light, GLenum pname, const GLfloat *params )
    GET_CURRENT_CONTEXT(ctx);
    GLint i = (GLint) (light - GL_LIGHT0);
    GLfloat temp[4];
+   ASSERT_OUTSIDE_BEGIN_END(ctx);
 
    if (i < 0 || i >= (GLint) ctx->Const.MaxLights) {
       _mesa_error( ctx, GL_INVALID_ENUM, "glLight(light=0x%x)", light );
diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c
index 39f9d1ddb08..57419aa2f31 100644
--- a/src/mesa/main/texstore.c
+++ b/src/mesa/main/texstore.c
@@ -2507,8 +2507,7 @@ _mesa_texstore_s8_z24(TEXSTORE_PARAMS)
             dstRow += dstRowStride / sizeof(GLuint);
          }
       }
-   }
-   else {
+   } else {
       for (img = 0; img < srcDepth; img++) {
          GLuint *dstRow = (GLuint *) dstAddr
             + dstImageOffsets[dstZoffset + img]
@@ -2522,7 +2521,7 @@ _mesa_texstore_s8_z24(TEXSTORE_PARAMS)
          for (row = 0; row < srcHeight; row++) {
             GLubyte stencil[MAX_WIDTH];
             GLint i;
-            /* the 24 depth bits will be in the high position: */
+            /* the 24 depth bits will be in the low position: */
             _mesa_unpack_depth_span(ctx, srcWidth,
                                     GL_UNSIGNED_INT, /* dst type */
                                     dstRow, /* dst addr */
@@ -2536,7 +2535,7 @@ _mesa_texstore_s8_z24(TEXSTORE_PARAMS)
                                       ctx->_ImageTransferState);
             /* merge stencil values into depth values */
             for (i = 0; i < srcWidth; i++)
-               dstRow[i] = stencil[i] << 24;
+               dstRow[i] |= stencil[i] << 24;
 
             src += srcRowStride;
             dstRow += dstRowStride / sizeof(GLuint);
diff --git a/src/mesa/main/varray.c b/src/mesa/main/varray.c
index 234b712f7f2..9d9b28b5189 100644
--- a/src/mesa/main/varray.c
+++ b/src/mesa/main/varray.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.1
+ * Version:  7.2
  *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -62,14 +62,9 @@ update_array(GLcontext *ctx, struct gl_client_array *array,
    array->Normalized = normalized;
    array->Ptr = (const GLubyte *) ptr;
 #if FEATURE_ARB_vertex_buffer_object
-   array->BufferObj->RefCount--;
-   if (array->BufferObj->RefCount <= 0) {
-      ASSERT(array->BufferObj->Name);
-      _mesa_remove_buffer_object( ctx, array->BufferObj );
-      (*ctx->Driver.DeleteBuffer)( ctx, array->BufferObj );
-   }
-   array->BufferObj = ctx->Array.ArrayBufferObj;
-   array->BufferObj->RefCount++;
+   _mesa_reference_buffer_object(ctx, &array->BufferObj,
+                                 ctx->Array.ArrayBufferObj);
+
    /* Compute the index of the last array element that's inside the buffer.
     * Later in glDrawArrays we'll check if start + count > _MaxElement to
     * be sure we won't go out of bounds.
diff --git a/src/mesa/tnl/t_vp_build.c b/src/mesa/tnl/t_vp_build.c
index 5ef0f7e7056..ca1df04868a 100644
--- a/src/mesa/tnl/t_vp_build.c
+++ b/src/mesa/tnl/t_vp_build.c
@@ -35,6 +35,437 @@
 #include "t_vp_build.h"
 
 
+struct state_key {
+   unsigned light_global_enabled:1;
+   unsigned light_local_viewer:1;
+   unsigned light_twoside:1;
+   unsigned light_color_material:1;
+   unsigned light_color_material_mask:12;
+   unsigned light_material_mask:12;
+
+   unsigned normalize:1;
+   unsigned rescale_normals:1;
+   unsigned fog_source_is_depth:1;
+   unsigned tnl_do_vertex_fog:1;
+   unsigned separate_specular:1;
+   unsigned fog_mode:2;
+   unsigned point_attenuated:1;
+   unsigned texture_enabled_global:1;
+   unsigned fragprog_inputs_read:12;
+
+   struct {
+      unsigned light_enabled:1;
+      unsigned light_eyepos3_is_zero:1;
+      unsigned light_spotcutoff_is_180:1;
+      unsigned light_attenuated:1;      
+      unsigned texunit_really_enabled:1;
+      unsigned texmat_enabled:1;
+      unsigned texgen_enabled:4;
+      unsigned texgen_mode0:4;
+      unsigned texgen_mode1:4;
+      unsigned texgen_mode2:4;
+      unsigned texgen_mode3:4;
+   } unit[8];
+};
+
+
+
+#define FOG_NONE   0
+#define FOG_LINEAR 1
+#define FOG_EXP    2
+#define FOG_EXP2   3
+
+static GLuint translate_fog_mode( GLenum mode )
+{
+   switch (mode) {
+   case GL_LINEAR: return FOG_LINEAR;
+   case GL_EXP: return FOG_EXP;
+   case GL_EXP2: return FOG_EXP2;
+   default: return FOG_NONE;
+   }
+}
+
+#define TXG_NONE           0
+#define TXG_OBJ_LINEAR     1
+#define TXG_EYE_LINEAR     2
+#define TXG_SPHERE_MAP     3
+#define TXG_REFLECTION_MAP 4
+#define TXG_NORMAL_MAP     5
+
+static GLuint translate_texgen( GLboolean enabled, GLenum mode )
+{
+   if (!enabled)
+      return TXG_NONE;
+
+   switch (mode) {
+   case GL_OBJECT_LINEAR: return TXG_OBJ_LINEAR;
+   case GL_EYE_LINEAR: return TXG_EYE_LINEAR;
+   case GL_SPHERE_MAP: return TXG_SPHERE_MAP;
+   case GL_REFLECTION_MAP_NV: return TXG_REFLECTION_MAP;
+   case GL_NORMAL_MAP_NV: return TXG_NORMAL_MAP;
+   default: return TXG_NONE;
+   }
+}
+
+static struct state_key *make_state_key( GLcontext *ctx )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   const struct gl_fragment_program *fp = ctx->FragmentProgram._Current;
+   struct state_key *key = CALLOC_STRUCT(state_key);
+   GLuint i;
+
+   /* This now relies on texenvprogram.c being active:
+    */
+   assert(fp);
+
+   key->fragprog_inputs_read = fp->Base.InputsRead;
+
+   key->separate_specular = (ctx->Light.Model.ColorControl ==
+			     GL_SEPARATE_SPECULAR_COLOR);
+
+   if (ctx->Light.Enabled) {
+      key->light_global_enabled = 1;
+
+      if (ctx->Light.Model.LocalViewer)
+	 key->light_local_viewer = 1;
+
+      if (ctx->Light.Model.TwoSide)
+	 key->light_twoside = 1;
+
+      if (ctx->Light.ColorMaterialEnabled) {
+	 key->light_color_material = 1;
+	 key->light_color_material_mask = ctx->Light.ColorMaterialBitmask;
+      }
+
+      for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) 
+	 if (VB->AttribPtr[i]->stride) 
+	    key->light_material_mask |= 1<<(i-_TNL_ATTRIB_MAT_FRONT_AMBIENT);
+
+      for (i = 0; i < MAX_LIGHTS; i++) {
+	 struct gl_light *light = &ctx->Light.Light[i];
+
+	 if (light->Enabled) {
+	    key->unit[i].light_enabled = 1;
+
+	    if (light->EyePosition[3] == 0.0)
+	       key->unit[i].light_eyepos3_is_zero = 1;
+	    
+	    if (light->SpotCutoff == 180.0)
+	       key->unit[i].light_spotcutoff_is_180 = 1;
+
+	    if (light->ConstantAttenuation != 1.0 ||
+		light->LinearAttenuation != 0.0 ||
+		light->QuadraticAttenuation != 0.0)
+	       key->unit[i].light_attenuated = 1;
+	 }
+      }
+   }
+
+   if (ctx->Transform.Normalize)
+      key->normalize = 1;
+
+   if (ctx->Transform.RescaleNormals)
+      key->rescale_normals = 1;
+
+   key->fog_mode = translate_fog_mode(fp->FogOption);
+   
+   if (ctx->Fog.FogCoordinateSource == GL_FRAGMENT_DEPTH_EXT)
+      key->fog_source_is_depth = 1;
+   
+   if (tnl->_DoVertexFog)
+      key->tnl_do_vertex_fog = 1;
+
+   if (ctx->Point._Attenuated)
+      key->point_attenuated = 1;
+
+   if (ctx->Texture._TexGenEnabled ||
+       ctx->Texture._TexMatEnabled ||
+       ctx->Texture._EnabledUnits)
+      key->texture_enabled_global = 1;
+      
+   for (i = 0; i < MAX_TEXTURE_UNITS; i++) {
+      struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i];
+
+      if (texUnit->_ReallyEnabled)
+	 key->unit[i].texunit_really_enabled = 1;
+
+      if (ctx->Texture._TexMatEnabled & ENABLE_TEXMAT(i))      
+	 key->unit[i].texmat_enabled = 1;
+      
+      if (texUnit->TexGenEnabled) {
+	 key->unit[i].texgen_enabled = 1;
+      
+	 key->unit[i].texgen_mode0 = 
+	    translate_texgen( texUnit->TexGenEnabled & (1<<0),
+			      texUnit->GenModeS );
+	 key->unit[i].texgen_mode1 = 
+	    translate_texgen( texUnit->TexGenEnabled & (1<<1),
+			      texUnit->GenModeT );
+	 key->unit[i].texgen_mode2 = 
+	    translate_texgen( texUnit->TexGenEnabled & (1<<2),
+			      texUnit->GenModeR );
+	 key->unit[i].texgen_mode3 = 
+	    translate_texgen( texUnit->TexGenEnabled & (1<<3),
+			      texUnit->GenModeQ );
+      }
+   }
+   
+   return key;
+}
+
+
+   
+/* Very useful debugging tool - produces annotated listing of
+ * generated program with line/function references for each
+ * instruction back into this file:
+ */
+#define DISASSEM (MESA_VERBOSE&VERBOSE_DISASSEM)
+
+/* Should be tunable by the driver - do we want to do matrix
+ * multiplications with DP4's or with MUL/MAD's?  SSE works better
+ * with the latter, drivers may differ.
+ */
+#define PREFER_DP4 0
+
+#define MAX_INSN 350
+
+/* Use uregs to represent registers internally, translate to Mesa's
+ * expected formats on emit.  
+ *
+ * NOTE: These are passed by value extensively in this file rather
+ * than as usual by pointer reference.  If this disturbs you, try
+ * remembering they are just 32bits in size.
+ *
+ * GCC is smart enough to deal with these dword-sized structures in
+ * much the same way as if I had defined them as dwords and was using
+ * macros to access and set the fields.  This is much nicer and easier
+ * to evolve.
+ */
+struct ureg {
+   GLuint file:4;
+   GLint idx:8;      /* relative addressing may be negative */
+   GLuint negate:1;
+   GLuint swz:12;
+   GLuint pad:7;
+};
+
+
+struct tnl_program {
+   const struct state_key *state;
+   struct gl_vertex_program *program;
+   
+   GLuint temp_in_use;
+   GLuint temp_reserved;
+   
+   struct ureg eye_position;
+   struct ureg eye_position_normalized;
+   struct ureg eye_normal;
+   struct ureg identity;
+
+   GLuint materials;
+   GLuint color_materials;
+};
+
+
+static const struct ureg undef = { 
+   PROGRAM_UNDEFINED,
+   ~0,
+   0,
+   0,
+   0
+};
+
+/* Local shorthand:
+ */
+#define X    SWIZZLE_X
+#define Y    SWIZZLE_Y
+#define Z    SWIZZLE_Z
+#define W    SWIZZLE_W
+
+
+/* Construct a ureg:
+ */
+static struct ureg make_ureg(GLuint file, GLint idx)
+{
+   struct ureg reg;
+   reg.file = file;
+   reg.idx = idx;
+   reg.negate = 0;
+   reg.swz = SWIZZLE_NOOP;
+   reg.pad = 0;
+   return reg;
+}
+
+
+
+static struct ureg negate( struct ureg reg )
+{
+   reg.negate ^= 1;
+   return reg;
+} 
+
+
+static struct ureg swizzle( struct ureg reg, int x, int y, int z, int w )
+{
+   reg.swz = MAKE_SWIZZLE4(GET_SWZ(reg.swz, x),
+			   GET_SWZ(reg.swz, y),
+			   GET_SWZ(reg.swz, z),
+			   GET_SWZ(reg.swz, w));
+
+   return reg;
+}
+
+static struct ureg swizzle1( struct ureg reg, int x )
+{
+   return swizzle(reg, x, x, x, x);
+}
+
+static struct ureg get_temp( struct tnl_program *p )
+{
+   int bit = _mesa_ffs( ~p->temp_in_use );
+   if (!bit) {
+      _mesa_problem(NULL, "%s: out of temporaries\n", __FILE__);
+      _mesa_exit(1);
+   }
+
+   if ((GLuint) bit > p->program->Base.NumTemporaries)
+      p->program->Base.NumTemporaries = bit;
+
+   p->temp_in_use |= 1<<(bit-1);
+   return make_ureg(PROGRAM_TEMPORARY, bit-1);
+}
+
+static struct ureg reserve_temp( struct tnl_program *p )
+{
+   struct ureg temp = get_temp( p );
+   p->temp_reserved |= 1<<temp.idx;
+   return temp;
+}
+
+static void release_temp( struct tnl_program *p, struct ureg reg )
+{
+   if (reg.file == PROGRAM_TEMPORARY) {
+      p->temp_in_use &= ~(1<<reg.idx);
+      p->temp_in_use |= p->temp_reserved; /* can't release reserved temps */
+   }
+}
+
+static void release_temps( struct tnl_program *p )
+{
+   p->temp_in_use = p->temp_reserved;
+}
+
+
+
+static struct ureg register_input( struct tnl_program *p, GLuint input )
+{
+   p->program->Base.InputsRead |= (1<<input);
+   return make_ureg(PROGRAM_INPUT, input);
+}
+
+static struct ureg register_output( struct tnl_program *p, GLuint output )
+{
+   p->program->Base.OutputsWritten |= (1<<output);
+   return make_ureg(PROGRAM_OUTPUT, output);
+}
+
+static struct ureg register_const4f( struct tnl_program *p, 
+			      GLfloat s0,
+			      GLfloat s1,
+			      GLfloat s2,
+			      GLfloat s3)
+{
+   GLfloat values[4];
+   GLint idx;
+   GLuint swizzle;
+   values[0] = s0;
+   values[1] = s1;
+   values[2] = s2;
+   values[3] = s3;
+   idx = _mesa_add_unnamed_constant( p->program->Base.Parameters, values, 4,
+                                     &swizzle );
+   ASSERT(swizzle == SWIZZLE_NOOP);
+   return make_ureg(PROGRAM_STATE_VAR, idx);
+}
+
+#define register_const1f(p, s0)         register_const4f(p, s0, 0, 0, 1)
+#define register_scalar_const(p, s0)    register_const4f(p, s0, s0, s0, s0)
+#define register_const2f(p, s0, s1)     register_const4f(p, s0, s1, 0, 1)
+#define register_const3f(p, s0, s1, s2) register_const4f(p, s0, s1, s2, 1)
+
+static GLboolean is_undef( struct ureg reg )
+{
+   return reg.file == PROGRAM_UNDEFINED;
+}
+
+static struct ureg get_identity_param( struct tnl_program *p )
+{
+   if (is_undef(p->identity)) 
+      p->identity = register_const4f(p, 0,0,0,1);
+
+   return p->identity;
+}
+
+static struct ureg register_param5(struct tnl_program *p, 
+				   GLint s0,
+				   GLint s1,
+				   GLint s2,
+				   GLint s3,
+                                   GLint s4)
+{
+   gl_state_index tokens[STATE_LENGTH];
+   GLint idx;
+   tokens[0] = s0;
+   tokens[1] = s1;
+   tokens[2] = s2;
+   tokens[3] = s3;
+   tokens[4] = s4;
+   idx = _mesa_add_state_reference( p->program->Base.Parameters, tokens );
+   return make_ureg(PROGRAM_STATE_VAR, idx);
+}
+
+
+#define register_param1(p,s0)          register_param5(p,s0,0,0,0,0)
+#define register_param2(p,s0,s1)       register_param5(p,s0,s1,0,0,0)
+#define register_param3(p,s0,s1,s2)    register_param5(p,s0,s1,s2,0,0)
+#define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
+
+
+static void register_matrix_param5( struct tnl_program *p,
+				    GLint s0, /* modelview, projection, etc */
+				    GLint s1, /* texture matrix number */
+				    GLint s2, /* first row */
+				    GLint s3, /* last row */
+				    GLint s4, /* inverse, transpose, etc */
+				    struct ureg *matrix )
+{
+   GLint i;
+
+   /* This is a bit sad as the support is there to pull the whole
+    * matrix out in one go:
+    */
+   for (i = 0; i <= s3 - s2; i++) 
+      matrix[i] = register_param5( p, s0, s1, i, i, s4 );
+}
+
+
+/**
+ * Convert a ureg source register to a prog_src_register.
+ */
+static void emit_arg( struct prog_src_register *src,
+		      struct ureg reg )
+{
+   assert(reg.file != PROGRAM_OUTPUT);
+   src->File = reg.file;
+   src->Index = reg.idx;
+   src->Swizzle = reg.swz;
+   src->NegateBase = reg.negate ? NEGATE_XYZW : 0;
+   src->Abs = 0;
+   src->NegateAbs = 0;
+   src->RelAddr = 0;
+}
+
 /**
  * XXX This should go away someday, but still referenced by some drivers...
  */
diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c
index 579bbeb4f31..d70b4bb1a11 100644
--- a/src/mesa/vbo/vbo_exec_api.c
+++ b/src/mesa/vbo/vbo_exec_api.c
@@ -1,6 +1,6 @@
 /**************************************************************************
 
-Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
+Copyright 2002-2008 Tungsten Graphics Inc., Cedar Park, Texas.
 
 All Rights Reserved.
 
@@ -31,6 +31,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 
 #include "main/glheader.h"
+#include "main/bufferobj.h"
 #include "main/context.h"
 #include "main/macros.h"
 #include "main/vtxfmt.h"
@@ -695,7 +696,10 @@ void vbo_exec_vtx_init( struct vbo_exec_context *exec )
     * continuously, unless vbo_use_buffer_objects() is called to enable
     * use of real VBOs.
     */
-   exec->vtx.bufferobj = ctx->Array.NullBufferObj;
+   _mesa_reference_buffer_object(ctx,
+                                 &exec->vtx.bufferobj,
+                                 ctx->Array.NullBufferObj);
+
    exec->vtx.buffer_map = ALIGN_MALLOC(VBO_VERT_BUFFER_SIZE * sizeof(GLfloat), 64);
 
    vbo_exec_vtxfmt_init( exec );
diff --git a/src/mesa/vbo/vbo_exec_draw.c b/src/mesa/vbo/vbo_exec_draw.c
index 3609a7452af..1d095c772f9 100644
--- a/src/mesa/vbo/vbo_exec_draw.c
+++ b/src/mesa/vbo/vbo_exec_draw.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  5.1
+ * Version:  7.2
  *
- * Copyright (C) 1999-2003  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -26,6 +26,7 @@
  */
 
 #include "main/glheader.h"
+#include "main/bufferobj.h"
 #include "main/context.h"
 #include "main/enums.h"
 #include "main/state.h"
@@ -155,8 +156,12 @@ static void vbo_exec_bind_arrays( GLcontext *ctx )
     */
    switch (get_program_mode(exec->ctx)) {
    case VP_NONE:
-      memcpy(arrays,      vbo->legacy_currval, 16 * sizeof(arrays[0]));
-      memcpy(arrays + 16, vbo->mat_currval,    MAT_ATTRIB_MAX * sizeof(arrays[0]));
+      for (attr = 0; attr < 16; attr++) {
+         exec->vtx.inputs[attr] = &vbo->legacy_currval[attr];
+      }
+      for (attr = 0; attr < MAT_ATTRIB_MAX; attr++) {
+         exec->vtx.inputs[attr + 16] = &vbo->mat_currval[attr];
+      }
       map = vbo->map_vp_none;
       break;
    case VP_NV:
@@ -165,8 +170,10 @@ static void vbo_exec_bind_arrays( GLcontext *ctx )
        * occurred.  NV vertex programs cannot access material values,
        * nor attributes greater than VERT_ATTRIB_TEX7.  
        */
-      memcpy(arrays,      vbo->legacy_currval,  16 * sizeof(arrays[0]));
-      memcpy(arrays + 16, vbo->generic_currval, 16 * sizeof(arrays[0]));
+      for (attr = 0; attr < 16; attr++) {
+         exec->vtx.inputs[attr] = &vbo->legacy_currval[attr];
+         exec->vtx.inputs[attr + 16] = &vbo->generic_currval[attr];
+      }
       map = vbo->map_vp_arb;
       break;
    }
@@ -178,24 +185,18 @@ static void vbo_exec_bind_arrays( GLcontext *ctx )
       const GLuint src = map[attr];
 
       if (exec->vtx.attrsz[src]) {
-         if (exec->vtx.bufferobj->Name) {
-            /* a real buffer obj: Ptr is an offset, not a pointer*/
-            int offset;
-            assert(exec->vtx.bufferobj->Pointer);  /* buf should be mapped */
-            offset = (GLbyte *) data - (GLbyte *) exec->vtx.bufferobj->Pointer;
-            assert(offset >= 0);
-            arrays[attr].Ptr = (void *) offset;
-         }
-         else {
-            /* Ptr into ordinary app memory */
-            arrays[attr].Ptr = (void *) data;
-         }
+         /* override the default array set above */
+         exec->vtx.inputs[attr] = &arrays[attr];
+
+	 arrays[attr].Ptr = (void *)data;
 	 arrays[attr].Size = exec->vtx.attrsz[src];
 	 arrays[attr].StrideB = exec->vtx.vertex_size * sizeof(GLfloat);
 	 arrays[attr].Stride = exec->vtx.vertex_size * sizeof(GLfloat);
 	 arrays[attr].Type = GL_FLOAT;
 	 arrays[attr].Enabled = 1;
-	 arrays[attr].BufferObj = exec->vtx.bufferobj;
+         _mesa_reference_buffer_object(ctx,
+                                       &arrays[attr].BufferObj,
+                                       exec->vtx.bufferobj);
 	 arrays[attr]._MaxElement = count; /* ??? */
 
 	 data += exec->vtx.attrsz[src] * sizeof(GLfloat);
diff --git a/src/mesa/vbo/vbo_save.c b/src/mesa/vbo/vbo_save.c
index 8dd87141c0c..9757c3d9f61 100644
--- a/src/mesa/vbo/vbo_save.c
+++ b/src/mesa/vbo/vbo_save.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.3
+ * Version:  7.2
  *
- * Copyright (C) 1999-2005  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -27,6 +27,7 @@
 
 
 #include "main/mtypes.h"
+#include "main/bufferobj.h"
 #include "main/dlist.h"
 #include "main/vtxfmt.h"
 #include "main/imports.h"
@@ -71,19 +72,24 @@ void vbo_save_destroy( GLcontext *ctx )
 {
    struct vbo_context *vbo = vbo_context(ctx);
    struct vbo_save_context *save = &vbo->save;
+   GLuint i;
+
    if (save->prim_store) {
       if ( --save->prim_store->refcount == 0 ) {
          FREE( save->prim_store );
          save->prim_store = NULL;
       }
       if ( --save->vertex_store->refcount == 0 ) {
-         if (save->vertex_store->bufferobj)
-            ctx->Driver.DeleteBuffer( ctx, save->vertex_store->bufferobj );
-
+         _mesa_reference_buffer_object(ctx,
+                                       &save->vertex_store->bufferobj, NULL);
          FREE( save->vertex_store );
          save->vertex_store = NULL;
       }
    }
+
+   for (i = 0; i < VBO_ATTRIB_MAX; i++) {
+      _mesa_reference_buffer_object(ctx, &save->arrays[i].BufferObj, NULL);
+   }
 }
 
 
diff --git a/src/mesa/vbo/vbo_save_api.c b/src/mesa/vbo/vbo_save_api.c
index f62be5c14cf..88d573f1283 100644
--- a/src/mesa/vbo/vbo_save_api.c
+++ b/src/mesa/vbo/vbo_save_api.c
@@ -1,6 +1,6 @@
 /**************************************************************************
 
-Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
+Copyright 2002-2008 Tungsten Graphics Inc., Cedar Park, Texas.
 
 All Rights Reserved.
 
@@ -68,6 +68,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 
 #include "main/glheader.h"
+#include "main/bufferobj.h"
 #include "main/context.h"
 #include "main/dlist.h"
 #include "main/enums.h"
@@ -85,6 +86,10 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #endif
 
 
+/* An interesting VBO number/name to help with debugging */
+#define VBO_BUF_ID  12345
+
+
 /*
  * NOTE: Old 'parity' issue is gone, but copying can still be
  * wrong-footed on replay.
@@ -170,7 +175,9 @@ static struct vbo_save_vertex_store *alloc_vertex_store( GLcontext *ctx )
     * user.  Perhaps there could be a special number for internal
     * buffers:
     */
-   vertex_store->bufferobj = ctx->Driver.NewBufferObject(ctx, 1, GL_ARRAY_BUFFER_ARB);
+   vertex_store->bufferobj = ctx->Driver.NewBufferObject(ctx,
+                                                         VBO_BUF_ID,
+                                                         GL_ARRAY_BUFFER_ARB);
 
    ctx->Driver.BufferData( ctx, 
 			   GL_ARRAY_BUFFER_ARB, 
@@ -190,8 +197,9 @@ static void free_vertex_store( GLcontext *ctx, struct vbo_save_vertex_store *ver
 {
    assert(!vertex_store->buffer);
 
-   if (vertex_store->bufferobj)
-      ctx->Driver.DeleteBuffer( ctx, vertex_store->bufferobj );
+   if (vertex_store->bufferobj) {
+      _mesa_reference_buffer_object(ctx, &vertex_store->bufferobj, NULL);
+   }
 
    FREE( vertex_store );
 }
@@ -1139,6 +1147,7 @@ void vbo_save_api_init( struct vbo_save_context *save )
    _save_vtxfmt_init( ctx );
    _save_current_init( ctx );
 
+   /* These will actually get set again when binding/drawing */
    for (i = 0; i < VBO_ATTRIB_MAX; i++)
       save->inputs[i] = &save->arrays[i];
 
diff --git a/src/mesa/vbo/vbo_save_draw.c b/src/mesa/vbo/vbo_save_draw.c
index bf5c6d4eefe..ed82f09958d 100644
--- a/src/mesa/vbo/vbo_save_draw.c
+++ b/src/mesa/vbo/vbo_save_draw.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.1
+ * Version:  7.2
  *
- * Copyright (C) 1999-2004  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -27,6 +27,7 @@
  */
 
 #include "main/glheader.h"
+#include "main/bufferobj.h"
 #include "main/context.h"
 #include "main/imports.h"
 #include "main/mtypes.h"
@@ -115,8 +116,12 @@ static void vbo_bind_vertex_list( GLcontext *ctx,
     */
    switch (get_program_mode(ctx)) {
    case VP_NONE:
-      memcpy(arrays,      vbo->legacy_currval, 16 * sizeof(arrays[0]));
-      memcpy(arrays + 16, vbo->mat_currval,    MAT_ATTRIB_MAX * sizeof(arrays[0]));
+      for (attr = 0; attr < 16; attr++) {
+         save->inputs[attr] = &vbo->legacy_currval[attr];
+      }
+      for (attr = 0; attr < MAT_ATTRIB_MAX; attr++) {
+         save->inputs[attr + 16] = &vbo->mat_currval[attr];
+      }
       map = vbo->map_vp_none;
       break;
    case VP_NV:
@@ -125,8 +130,10 @@ static void vbo_bind_vertex_list( GLcontext *ctx,
        * occurred.  NV vertex programs cannot access material values,
        * nor attributes greater than VERT_ATTRIB_TEX7.  
        */
-      memcpy(arrays,      vbo->legacy_currval,  16 * sizeof(arrays[0]));
-      memcpy(arrays + 16, vbo->generic_currval, 16 * sizeof(arrays[0]));
+      for (attr = 0; attr < 16; attr++) {
+         save->inputs[attr] = &vbo->legacy_currval[attr];
+         save->inputs[attr + 16] = &vbo->generic_currval[attr];
+      }
       map = vbo->map_vp_arb;
       break;
    }
@@ -135,13 +142,18 @@ static void vbo_bind_vertex_list( GLcontext *ctx,
       GLuint src = map[attr];
 
       if (node->attrsz[src]) {
-	 arrays[attr].Ptr = (const GLubyte *)data;
+         /* override the default array set above */
+         save->inputs[attr] = &arrays[attr];
+
+	 arrays[attr].Ptr = (const GLubyte *) data;
 	 arrays[attr].Size = node->attrsz[src];
 	 arrays[attr].StrideB = node->vertex_size * sizeof(GLfloat);
 	 arrays[attr].Stride = node->vertex_size * sizeof(GLfloat);
 	 arrays[attr].Type = GL_FLOAT;
 	 arrays[attr].Enabled = 1;
-	 arrays[attr].BufferObj = node->vertex_store->bufferobj;
+         _mesa_reference_buffer_object(ctx,
+                                       &arrays[attr].BufferObj,
+                                       node->vertex_store->bufferobj);
 	 arrays[attr]._MaxElement = node->count; /* ??? */
 	 
 	 assert(arrays[attr].BufferObj->Name);