Merge the trunk into the branch.

98 files changed, 15927 insertions, 3550 deletions

diff --git a/xc/lib/GL/mesa/dri/dri_mesa.c b/xc/lib/GL/mesa/dri/dri_mesa.c
index 08f3dfcfd..b679b41a4 100644
--- a/xc/lib/GL/mesa/dri/dri_mesa.c
+++ b/xc/lib/GL/mesa/dri/dri_mesa.c
@@ -29,7 +29,7 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 /*
  * Authors:
  *   Kevin E. Martin <kevin@precisioninsight.com>
- *   Brian Paul <brian@precisioninsight.com>
+ *   Brian E. Paul <brian@precisioninsight.com>
  */
 
 #ifdef GLX_DIRECT_RENDERING
@@ -39,16 +39,12 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include <Xext.h>
 #include <extutil.h>
 #include "glxclient.h"
-#include "GL/xmesa.h"
 #include "xf86dri.h"
 #include "sarea.h"
 #include "dri_mesaint.h"
 #include "dri_xmesaapi.h"
-
-
-#if XMESA_MAJOR_VERSION != 3 || XMESA_MINOR_VERSION != 3
-#error using wrong version of Mesa (need 3.3)
-#endif
+#include "../src/context.h"
+#include "../src/mmath.h"
 
 
 /* Context binding */
@@ -109,6 +105,8 @@ static __DRIdrawable *__driMesaFindDrawable(GLXDrawable draw)
     return pdraw;
 }
 
+#if 0
+/* not used yet */
 static void __driMesaRemoveDrawable(__DRIdrawable *pdraw)
 {
     int retcode;
@@ -122,23 +120,22 @@ static void __driMesaRemoveDrawable(__DRIdrawable *pdraw)
 	drmHashDelete(drawHash, pdp->draw);
     }
 }
+#endif
 
 /*****************************************************************/
 
-static void driMesaInitAPI(__XMESAapi *XMesaAPI)
+static void driMesaInitAPI(__MesaAPI *MesaAPI)
 {
-    XMesaAPI->InitDriver = XMesaInitDriver;
-    XMesaAPI->ResetDriver = XMesaResetDriver;
-    XMesaAPI->CreateVisual = XMesaCreateVisual;
-    XMesaAPI->DestroyVisual = XMesaDestroyVisual;
-    XMesaAPI->CreateContext = XMesaCreateContext;
-    XMesaAPI->DestroyContext = XMesaDestroyContext;
-    XMesaAPI->CreateWindowBuffer = XMesaCreateWindowBuffer;
-    XMesaAPI->CreatePixmapBuffer = XMesaCreatePixmapBuffer;
-    XMesaAPI->DestroyBuffer = XMesaDestroyBuffer;
-    XMesaAPI->SwapBuffers = XMesaSwapBuffers;
-    XMesaAPI->MakeCurrent = XMesaMakeCurrent;
-    XMesaAPI->UnbindContext = XMesaUnbindContext;
+    MesaAPI->InitDriver = XMesaInitDriver;
+    MesaAPI->ResetDriver = XMesaResetDriver;
+    MesaAPI->CreateVisual = XMesaCreateVisual;
+    MesaAPI->CreateContext = XMesaCreateContext;
+    MesaAPI->DestroyContext = XMesaDestroyContext;
+    MesaAPI->CreateWindowBuffer = XMesaCreateWindowBuffer;
+    MesaAPI->CreatePixmapBuffer = XMesaCreatePixmapBuffer;
+    MesaAPI->SwapBuffers = XMesaSwapBuffers;
+    MesaAPI->MakeCurrent = XMesaMakeCurrent;
+    MesaAPI->UnbindContext = XMesaUnbindContext;
 }
 
 /*****************************************************************/
@@ -176,7 +173,7 @@ static Bool driMesaUnbindContext(Display *dpy, int scrn,
     }
 
     /* Unbind Mesa's drawable from Mesa's context */
-    (*psp->XMesaAPI.UnbindContext)(pcp->xm_ctx);
+    (*psp->MesaAPI.UnbindContext)(pcp);
 
     if (pdp->refcount == 0) {
 	/* ERROR!!! */
@@ -279,7 +276,7 @@ static Bool driMesaBindContext(Display *dpy, int scrn,
     }
 
     /* Bind Mesa's drawable to Mesa's context */
-    (*psp->XMesaAPI.MakeCurrent)(pcp->xm_ctx, pdp->xm_buf);
+    (*psp->MesaAPI.MakeCurrent)(pcp, pdp, pdp);
 
     return GL_TRUE;
 }
@@ -307,14 +304,26 @@ void driMesaUpdateDrawableInfo(Display *dpy, int scrn,
 	Xfree(pdp->pClipRects); 
     }
 
+    if (pdp->pBackClipRects) {
+	Xfree(pdp->pBackClipRects); 
+    }
+
+
     DRM_SPINUNLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID);
 
     if (!XF86DRIGetDrawableInfo(dpy, scrn, pdp->draw,
 				&pdp->index, &pdp->lastStamp,
 				&pdp->x, &pdp->y, &pdp->w, &pdp->h,
-				&pdp->numClipRects, &pdp->pClipRects)) {
+				&pdp->numClipRects, &pdp->pClipRects,
+				&pdp->backX,
+				&pdp->backY,
+				&pdp->numBackClipRects,
+				&pdp->pBackClipRects
+                                )) {
 	pdp->numClipRects = 0;
 	pdp->pClipRects = NULL;
+	pdp->numBackClipRects = 0;
+	pdp->pBackClipRects = 0;
 	/* ERROR!!! */
     }
 
@@ -332,7 +341,7 @@ static void *driMesaCreateDrawable(Display *dpy, int scrn, GLXDrawable draw,
     __DRIscreenPrivate *psp;
     __DRIdrawablePrivate *pdp;
     int i;
-    XMesaVisual xm_vis = NULL;
+    GLvisual *mesaVis = NULL;
 
     pdp = (__DRIdrawablePrivate *)Xmalloc(sizeof(__DRIdrawablePrivate));
     if (!pdp) {
@@ -354,7 +363,9 @@ static void *driMesaCreateDrawable(Display *dpy, int scrn, GLXDrawable draw,
     pdp->w = 0;
     pdp->h = 0;
     pdp->numClipRects = 0;
+    pdp->numBackClipRects = 0;
     pdp->pClipRects = NULL;
+    pdp->pBackClipRects = NULL;
 
     pDRIScreen = __glXFindDRIScreen(dpy, scrn);
     pdp->driScreenPriv = psp = (__DRIscreenPrivate *)pDRIScreen->private;
@@ -363,20 +374,19 @@ static void *driMesaCreateDrawable(Display *dpy, int scrn, GLXDrawable draw,
 
     for (i = 0; i < psp->numVisuals; i++) {
 	if (vid == psp->visuals[i].vid) {
-	    xm_vis = psp->visuals[i].xm_vis;
+	    mesaVis = psp->visuals[i].mesaVisual;
 	    break;
 	}
     }
 
-    if (1 /* NOT_DONE: Determine if it is a pixmap or not */) {
-	pdp->xm_buf = (*psp->XMesaAPI.CreateWindowBuffer)(xm_vis, draw, pdp);
-    } else {
-	XMesaVisual xm_vis = NULL;
-	XMesaColormap cmap = 0;
-	pdp->xm_buf = (*psp->XMesaAPI.CreatePixmapBuffer)(xm_vis, draw, cmap,
-							  pdp);
+    /* XXX pixmap rendering not implemented yet */
+    if (1) {
+       pdp->mesaBuffer = (*psp->MesaAPI.CreateWindowBuffer)(dpy, psp, pdp, mesaVis);
+    }
+    else {
+       pdp->mesaBuffer = (*psp->MesaAPI.CreatePixmapBuffer)(dpy, psp, pdp, mesaVis);
     }
-    if (!pdp->xm_buf) {
+    if (!pdp->mesaBuffer) {
 	(void)XF86DRIDestroyDrawable(dpy, scrn, pdp->draw);
 	Xfree(pdp);
 	return NULL;
@@ -402,7 +412,7 @@ static void driMesaSwapBuffers(Display *dpy, void *private)
     __DRIdrawablePrivate *pdp = (__DRIdrawablePrivate *)private;
     __DRIscreenPrivate *psp = pdp->driScreenPriv;
 
-    (*psp->XMesaAPI.SwapBuffers)(pdp->xm_buf);
+    (*psp->MesaAPI.SwapBuffers)(pdp);
 }
 
 static void driMesaDestroyDrawable(Display *dpy, void *private)
@@ -412,7 +422,7 @@ static void driMesaDestroyDrawable(Display *dpy, void *private)
     int scrn = psp->myNum;
 
     if (pdp) {
-	(*psp->XMesaAPI.DestroyBuffer)(pdp->xm_buf);
+        gl_destroy_framebuffer(pdp->mesaBuffer);
 	(void)XF86DRIDestroyDrawable(dpy, scrn, pdp->draw);
 	if (pdp->pClipRects)
 	    Xfree(pdp->pClipRects);
@@ -427,8 +437,6 @@ static void *driMesaCreateContext(Display *dpy, XVisualInfo *vis, void *shared,
 {
     __DRIcontextPrivate *pcp;
     __DRIcontextPrivate *pshare = (__DRIcontextPrivate *)shared;
-    XMesaContext shared_xm_ctx = (pshare ?
-				  pshare->xm_ctx : (XMesaContext)NULL);
     __DRIscreenPrivate *psp;
     __DRIscreen *pDRIScreen;
     int i;
@@ -443,7 +451,7 @@ static void *driMesaCreateContext(Display *dpy, XVisualInfo *vis, void *shared,
 	    return NULL;
 	}
 	psp->dummyContextPriv.driScreenPriv = psp;
-	psp->dummyContextPriv.xm_ctx = NULL;
+	psp->dummyContextPriv.mesaContext = NULL;
 	psp->dummyContextPriv.driDrawablePriv = NULL;
 	/* No other fields should be used! */
     }
@@ -453,8 +461,9 @@ static void *driMesaCreateContext(Display *dpy, XVisualInfo *vis, void *shared,
 	return NULL;
     }
 
+    pcp->display = dpy;
     pcp->driScreenPriv = psp;
-    pcp->xm_ctx = NULL;
+    pcp->mesaContext = NULL;
     pcp->driDrawablePriv = NULL;
 
     if (!XF86DRICreateContext(dpy, vis->screen, vis->visual,
@@ -463,12 +472,28 @@ static void *driMesaCreateContext(Display *dpy, XVisualInfo *vis, void *shared,
 	return NULL;
     }
 
-    for (i = 0; i < psp->numVisuals; i++)
-	if (psp->visuals[i].vid == vis->visualid)
-	    pcp->xm_ctx = (*psp->XMesaAPI.CreateContext)
-		(psp->visuals[i].xm_vis, shared_xm_ctx, pcp);
-
-    if (!pcp->xm_ctx) {
+    for (i = 0; i < psp->numVisuals; i++) {
+        if (psp->visuals[i].vid == vis->visualid) {
+            GLvisual *mesaVis = psp->visuals[i].mesaVisual;
+            GLcontext *sharedMesaCtx = pshare ? pshare->mesaContext : NULL;
+            pcp->mesaContext = gl_create_context(mesaVis,
+                                                 sharedMesaCtx,
+                                                 NULL, /* set below */
+                                                 GL_TRUE);
+            if (pcp->mesaContext) {
+                /* Driver now creates its private context data */
+                if ((*psp->MesaAPI.CreateContext)(dpy, mesaVis, pcp)) {
+                    pcp->mesaContext->DriverCtx = pcp->driverPrivate;
+                }
+                else {
+                    gl_destroy_context(pcp->mesaContext);
+                    pcp->mesaContext = NULL;
+                }
+            }
+        }
+    }
+
+    if (!pcp->mesaContext) {
 	(void)XF86DRIDestroyContext(dpy, vis->screen, pcp->contextID);
 	Xfree(pcp);
 	return NULL;
@@ -487,7 +512,8 @@ static void driMesaDestroyContext(Display *dpy, int scrn, void *private)
 
     if (pcp) {
 	(void)XF86DRIDestroyContext(dpy, scrn, pcp->contextID);
-	(*pcp->driScreenPriv->XMesaAPI.DestroyContext)(pcp->xm_ctx);
+	(*pcp->driScreenPriv->MesaAPI.DestroyContext)(pcp);
+        gl_destroy_context(pcp->mesaContext);
 	Xfree(pcp);
     }
 }
@@ -565,7 +591,7 @@ static void *driMesaCreateScreen(Display *dpy, int scrn, __DRIscreen *psc,
 	return NULL;
     }
 
-    driMesaInitAPI(&psp->XMesaAPI);
+    driMesaInitAPI(&psp->MesaAPI);
 
     if (!XF86DRIGetDeviceInfo(dpy, scrn,
 			      &hFB,
@@ -627,22 +653,13 @@ static void *driMesaCreateScreen(Display *dpy, int scrn, __DRIscreen *psc,
 
     for (i = 0; i < numConfigs; i++, config++) {
 	psp->visuals[i].vid = visinfo[i].visualid;
-	psp->visuals[i].xm_vis =
-	    (*psp->XMesaAPI.CreateVisual)(dpy,
-					  &visinfo[i],
-					  config->rgba,
-					  (config->alphaSize > 0),
-					  config->doubleBuffer,
-					  config->stereo,
-					  GL_TRUE, /* ximage_flag */
-					  config->depthSize,
-					  config->stencilSize,
-					  config->accumRedSize,
-					  config->level);
-	if (!psp->visuals[i].xm_vis) {
+        psp->visuals[i].mesaVisual =
+                 (*psp->MesaAPI.CreateVisual) (dpy, psp, &visinfo[i], config);
+
+	if (!psp->visuals[i].mesaVisual) {
 	    /* Free the visuals so far created */
 	    while (--i >= 0) {
-		(*psp->XMesaAPI.DestroyVisual)(psp->visuals[i].xm_vis);
+               _mesa_destroy_visual(psp->visuals[i].mesaVisual);
 	    }
 	    Xfree(psp->visuals);
 	    XFree(visinfo);
@@ -658,11 +675,10 @@ static void *driMesaCreateScreen(Display *dpy, int scrn, __DRIscreen *psc,
     XFree(visinfo);
 
     /* Initialize the screen specific GLX driver */
-    if (psp->XMesaAPI.InitDriver) {
-	if (!(*psp->XMesaAPI.InitDriver)(psp)) {
+    if (psp->MesaAPI.InitDriver) {
+	if (!(*psp->MesaAPI.InitDriver)(psp)) {
 	    while (--psp->numVisuals >= 0) {
-		(*psp->XMesaAPI.DestroyVisual)
-		    (psp->visuals[psp->numVisuals].xm_vis);
+               _mesa_destroy_visual(psp->visuals[psp->numVisuals].mesaVisual);
 	    }
 	    Xfree(psp->visuals);
 	    (void)drmUnmap((drmAddress)psp->pSAREA, SAREA_MAX);
@@ -699,11 +715,10 @@ static void driMesaDestroyScreen(Display *dpy, int scrn, void *private)
 	    (void)XF86DRIDestroyContext(dpy, scrn,
 					psp->dummyContextPriv.contextID);
 	}
-	if (psp->XMesaAPI.ResetDriver)
-	    (*psp->XMesaAPI.ResetDriver)(psp);
+	if (psp->MesaAPI.ResetDriver)
+	    (*psp->MesaAPI.ResetDriver)(psp);
 	while (--psp->numVisuals >= 0) {
-	    (*psp->XMesaAPI.DestroyVisual)
-		(psp->visuals[psp->numVisuals].xm_vis);
+           _mesa_destroy_visual(psp->visuals[psp->numVisuals].mesaVisual);
 	}
 	Xfree(psp->visuals);
 	(void)drmUnmap((drmAddress)psp->pSAREA, SAREA_MAX);
diff --git a/xc/lib/GL/mesa/dri/dri_mesaint.h b/xc/lib/GL/mesa/dri/dri_mesaint.h
index 6b64114d0..e8918724c 100644
--- a/xc/lib/GL/mesa/dri/dri_mesaint.h
+++ b/xc/lib/GL/mesa/dri/dri_mesaint.h
@@ -29,6 +29,7 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 /*
  * Authors:
  *   Kevin E. Martin <kevin@precisioninsight.com>
+ *   Brian E. Paul <brian@precisioninsight.com>
  *
  */
 
@@ -40,10 +41,10 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include <GL/glx.h>
 #include "xf86dri.h"
 #include "sarea.h"
-#include "GL/xmesa.h"
 #include "dri_mesa.h"
 #include "dri_xmesaapi.h"
 
+
 #define DRI_MESA_VALIDATE_DRAWABLE_INFO(dpy,scrn,pDrawPriv)  \
     do {                                                     \
 	if (*(pDrawPriv->pStamp) != pDrawPriv->lastStamp) {  \
@@ -59,9 +60,9 @@ struct __DRIdrawablePrivateRec {
     drmDrawable hHWDrawable;
 
     /*
-    ** Mesa's private context information.  This structure is opaque.
+    ** Mesa's private frame buffer information.  This structure is opaque.
     */
-    XMesaBuffer xm_buf;
+    GLframebuffer *mesaBuffer;
 
     /*
     ** X's drawable ID associated with this private drawable.
@@ -104,6 +105,16 @@ struct __DRIdrawablePrivateRec {
     XF86DRIClipRectPtr pClipRects;
 
     /*
+    ** Information about the back and depthbuffer where different
+    ** from above.
+    */
+    int backX;
+    int backY;
+    int backClipRectType;
+    int numBackClipRects;
+    XF86DRIClipRectPtr pBackClipRects;
+
+    /*
     ** Pointer to context to which this drawable is currently bound.
     */
     __DRIcontextPrivate *driContextPriv;
@@ -128,7 +139,17 @@ struct __DRIcontextPrivateRec {
     /*
     ** Mesa's private context information.  This structure is opaque.
     */
-    XMesaContext xm_ctx;
+    GLcontext *mesaContext;
+
+    /*
+    ** Device driver's private context data.  This structure is opaque.
+    */
+    void *driverPrivate;
+
+    /*
+    ** This context's display pointer.
+    */
+    Display *display;
 
     /*
     ** Pointer to drawable currently bound to this context.
@@ -145,7 +166,7 @@ struct __DRIvisualPrivateRec {
     /*
     ** Mesa's private visual information.  This structure is opaque.
     */
-    XMesaVisual xm_vis;
+    GLvisual *mesaVisual;
 
     /*
     ** X's visual ID associated with this private visual.
@@ -169,7 +190,7 @@ struct __DRIscreenPrivateRec {
     /*
     ** Function pointers associated with Mesa's GLX functions.
     */
-    __XMESAapi XMesaAPI;
+    __MesaAPI MesaAPI;
 
     /*
     ** Core rendering library's driver version information.
diff --git a/xc/lib/GL/mesa/dri/dri_xmesaapi.h b/xc/lib/GL/mesa/dri/dri_xmesaapi.h
index 18aa3db8d..54f22df70 100644
--- a/xc/lib/GL/mesa/dri/dri_xmesaapi.h
+++ b/xc/lib/GL/mesa/dri/dri_xmesaapi.h
@@ -29,48 +29,110 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 /*
  * Authors:
  *   Kevin E. Martin <kevin@precisioninsight.com>
- *
+ *   Brian E. Paul <brian@precisioninsight.com>
  */
 
+
 #ifndef _DRI_XMESAAPI_H_
 #define _DRI_XMESAAPI_H_
 
-#ifdef GLX_DIRECT_RENDERING
-
-#include "GL/xmesa.h"
+#include "glxclient.h"
 #include "dri_mesa.h"
+#include "../src/types.h"
+
+
+extern GLboolean XMesaInitDriver( __DRIscreenPrivate *driScrnPriv );
+
+extern void XMesaResetDriver( __DRIscreenPrivate *driScrnPriv );
+
+/* Driver creates a GLvisual and returns pointer to it.
+ */
+extern GLvisual *XMesaCreateVisual(Display *dpy,
+                                   __DRIscreenPrivate *driScrnPriv,
+                                   const XVisualInfo *visinfo,
+                                   const __GLXvisualConfig *config);
+
+/* Driver creates its private context struct and hooks it into
+ * the driContextPriv->driverPrivate field.  Return true/false for
+ * success/error.
+ */
+extern GLboolean XMesaCreateContext( Display *dpy, GLvisual *mesaVis,
+                                     __DRIcontextPrivate *driContextPriv );
+
+/* Driver frees data attached to driContextPriv->driverPrivate pointer.
+ */
+extern void XMesaDestroyContext( __DRIcontextPrivate *driContextPriv );
+
+/* Driver creates a GLframebuffer struct indicating which ancillary
+ * buffers are in hardware or software.
+ */
+extern GLframebuffer *XMesaCreateWindowBuffer( Display *dpy,
+                                           __DRIscreenPrivate *driScrnPriv,
+                                           __DRIdrawablePrivate *driDrawPriv,
+                                           GLvisual *mesaVis);
 
-typedef struct __XMESAapiRec __XMESAapi;
+extern GLframebuffer *XMesaCreatePixmapBuffer( Display *dpy,
+                                           __DRIscreenPrivate *driScrnPriv,
+                                           __DRIdrawablePrivate *driDrawPriv,
+                                           GLvisual *mesaVis);
 
-struct __XMESAapiRec {
+extern GLboolean XMesaMakeCurrent( __DRIcontextPrivate *driContextPriv,
+                                   __DRIdrawablePrivate *driDrawPriv,
+                                   __DRIdrawablePrivate *driReadPriv );
+
+extern GLboolean XMesaUnbindContext( __DRIcontextPrivate *driContextPriv );
+
+extern void XMesaSwapBuffers( __DRIdrawablePrivate *driDrawPriv );
+
+
+
+
+
+#define XMESA_VALIDATE_DRAWABLE_INFO(dpy, psp, pdp)                     \
+do {                                                                    \
+    while (*(pdp->pStamp) != pdp->lastStamp) {                          \
+	DRM_UNLOCK(psp->fd, &psp->pSAREA->lock,                         \
+		   pdp->driContextPriv->hHWContext);                    \
+                                                                        \
+	DRM_SPINLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID);     \
+	DRI_MESA_VALIDATE_DRAWABLE_INFO(dpy, psp->myNum, pdp);          \
+	DRM_SPINUNLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID);   \
+                                                                        \
+	DRM_LIGHT_LOCK(psp->fd, &psp->pSAREA->lock,                     \
+		       pdp->driContextPriv->hHWContext);                \
+    }                                                                   \
+} while (0)
+
+
+
+typedef struct __MesaAPIRec __MesaAPI;
+
+struct __MesaAPIRec {
     /* XMESA Functions */
     GLboolean    (*InitDriver)(__DRIscreenPrivate *driScrnPriv);
     void         (*ResetDriver)(__DRIscreenPrivate *driScrnPriv);
-    XMesaVisual  (*CreateVisual)(XMesaDisplay *display,
-				 XMesaVisualInfo visinfo,
-				 GLboolean rgb_flag,
-				 GLboolean alpha_flag,
-				 GLboolean db_flag,
-				 GLboolean stereo_flag,
-				 GLboolean ximage_flag,
-				 GLint depth_size,
-				 GLint stencil_size,
-				 GLint accum_size,
-				 GLint level);
-    void         (*DestroyVisual)(XMesaVisual v);
-    XMesaContext (*CreateContext)(XMesaVisual v, XMesaContext share_list,
-				  __DRIcontextPrivate *driContextPriv);
-    void         (*DestroyContext)(XMesaContext c);
-    XMesaBuffer  (*CreateWindowBuffer)(XMesaVisual v, XMesaWindow w,
-				       __DRIdrawablePrivate *driDrawPriv);
-    XMesaBuffer  (*CreatePixmapBuffer)(XMesaVisual v, XMesaPixmap p,
-				       XMesaColormap c,
-				       __DRIdrawablePrivate *driDrawPriv);
-    void         (*DestroyBuffer)(XMesaBuffer b);
-    void         (*SwapBuffers)(XMesaBuffer b);
-    GLboolean    (*MakeCurrent)(XMesaContext c, XMesaBuffer b);
-    GLboolean    (*UnbindContext)(XMesaContext c);
+    GLvisual *   (*CreateVisual)(Display *display,
+                                 __DRIscreenPrivate *driScrnPriv,
+                                 const XVisualInfo *visinfo,
+                                 const __GLXvisualConfig *config);
+    GLboolean    (*CreateContext)(Display *dpy, GLvisual *mesaVis,
+                                  __DRIcontextPrivate *driContextPriv);
+    void         (*DestroyContext)(__DRIcontextPrivate *driContextPriv);
+    GLframebuffer * (*CreateWindowBuffer)(Display *dpy,
+                                          __DRIscreenPrivate *driScrnPriv,
+                                          __DRIdrawablePrivate *driDrawPriv,
+                                          GLvisual *mesaVis);
+    GLframebuffer * (*CreatePixmapBuffer)(Display *dpy,
+                                          __DRIscreenPrivate *driScrnPriv,
+                                          __DRIdrawablePrivate *driDrawPriv,
+                                          GLvisual *mesaVis);
+    void         (*SwapBuffers)(__DRIdrawablePrivate *driDrawPriv);
+    GLboolean    (*MakeCurrent)(__DRIcontextPrivate *driContextPriv,
+                                __DRIdrawablePrivate *driDrawPriv,
+                                __DRIdrawablePrivate *driReadPriv);
+    GLboolean    (*UnbindContext)(__DRIcontextPrivate *driContextPriv);
 };
 
-#endif
+
+
 #endif /* _DRI_XMESAAPI_H_ */
diff --git a/xc/lib/GL/mesa/src/Imakefile b/xc/lib/GL/mesa/src/Imakefile
index d93aa3f7e..4f5af973c 100644
--- a/xc/lib/GL/mesa/src/Imakefile
+++ b/xc/lib/GL/mesa/src/Imakefile
@@ -16,6 +16,9 @@ ALLOC_DEFINES = -DMALLOC_0_RETURNS_NULL
     TDFX_DEFS = -DFX
 #endif
 
+LinkSourceFile(aatriangle.c, ../../../../extras/Mesa/src)
+LinkSourceFile(aatriangle.h, ../../../../extras/Mesa/src)
+LinkSourceFile(aatritemp.h, ../../../../extras/Mesa/src)
 LinkSourceFile(accum.c, ../../../../extras/Mesa/src)
 LinkSourceFile(accum.h, ../../../../extras/Mesa/src)
 LinkSourceFile(all.h, ../../../../extras/Mesa/src)
@@ -120,6 +123,8 @@ LinkSourceFile(pipeline.c, ../../../../extras/Mesa/src)
 LinkSourceFile(pipeline.h, ../../../../extras/Mesa/src)
 LinkSourceFile(pixel.c, ../../../../extras/Mesa/src)
 LinkSourceFile(pixel.h, ../../../../extras/Mesa/src)
+LinkSourceFile(pixeltex.c, ../../../../extras/Mesa/src)
+LinkSourceFile(pixeltex.h, ../../../../extras/Mesa/src)
 LinkSourceFile(points.c, ../../../../extras/Mesa/src)
 LinkSourceFile(points.h, ../../../../extras/Mesa/src)
 LinkSourceFile(polygon.c, ../../../../extras/Mesa/src)
@@ -156,6 +161,8 @@ LinkSourceFile(texstate.c, ../../../../extras/Mesa/src)
 LinkSourceFile(texstate.h, ../../../../extras/Mesa/src)
 LinkSourceFile(texture.c, ../../../../extras/Mesa/src)
 LinkSourceFile(texture.h, ../../../../extras/Mesa/src)
+LinkSourceFile(texutil.c, ../../../../extras/Mesa/src)
+LinkSourceFile(texutil.h, ../../../../extras/Mesa/src)
 LinkSourceFile(trans_tmp.h, ../../../../extras/Mesa/src)
 LinkSourceFile(translate.c, ../../../../extras/Mesa/src)
 LinkSourceFile(translate.h, ../../../../extras/Mesa/src)
@@ -189,7 +196,8 @@ LinkSourceFile(xform_tmp.h, ../../../../extras/Mesa/src)
 LinkSourceFile(zoom.c, ../../../../extras/Mesa/src)
 LinkSourceFile(zoom.h, ../../../../extras/Mesa/src)
 
-    CORE_SRCS = accum.c \
+    CORE_SRCS = aatriangle.c \
+		accum.c \
 		alpha.c \
 		alphabuf.c \
 		attrib.c \
@@ -220,7 +228,7 @@ LinkSourceFile(zoom.h, ../../../../extras/Mesa/src)
 		hash.c \
 		hint.c \
 		image.c \
-		imaging.o \
+		imaging.c \
 		light.c \
 		lines.c \
 		logic.c \
@@ -231,6 +239,7 @@ LinkSourceFile(zoom.h, ../../../../extras/Mesa/src)
 		pb.c \
 		pipeline.c \
 		pixel.c \
+		pixeltex.c \
 		points.c \
 		polygon.c \
 		quads.c \
@@ -247,6 +256,7 @@ LinkSourceFile(zoom.h, ../../../../extras/Mesa/src)
 		texobj.c \
 		texstate.c \
 		texture.c \
+		texutil.c \
 		translate.c \
 		triangle.c \
 		varray.c \
@@ -262,7 +272,8 @@ LinkSourceFile(zoom.h, ../../../../extras/Mesa/src)
 		xform.c \
 		zoom.c
 
-    CORE_OBJS = accum.o \
+    CORE_OBJS = aatriangle.o \
+		accum.o \
 		alpha.o \
 		alphabuf.o \
 		attrib.o \
@@ -301,6 +312,7 @@ LinkSourceFile(zoom.h, ../../../../extras/Mesa/src)
 		pb.o \
 		pipeline.o \
 		pixel.o \
+		pixeltex.o \
 		points.o \
 		polygon.o \
 		quads.o \
@@ -317,6 +329,7 @@ LinkSourceFile(zoom.h, ../../../../extras/Mesa/src)
 		texobj.o \
 		texstate.o \
 		texture.o \
+		texutil.o \
 		translate.o \
 		triangle.o \
 		varray.o \
@@ -335,8 +348,12 @@ LinkSourceFile(zoom.h, ../../../../extras/Mesa/src)
 #ifdef i386Architecture
      ASM_SRCS = 
      ASM_OBJS = 
+#ifdef MesaUse3DNow
+     ASM_DEFS = -DUSE_MMX_ASM -DUSE_X86_ASM -DUSE_3DNOW_ASM
+#else
      ASM_DEFS = -DUSE_MMX_ASM -DUSE_X86_ASM
 #endif
+#endif
 
       DEFINES = $(ALLOC_DEFINES) GlxDefines $(TDFX_DEFS) $(ASM_DEFS)
      INCLUDES = -I$(XLIBSRC) -I$(EXTINCSRC) -I../include -I../../dri -I. -I../../../../include
@@ -351,9 +368,7 @@ LinkSourceFile(zoom.h, ../../../../extras/Mesa/src)
 #endif
 
 #if GlxBuiltInMesa || GlxDriverUsesMesa || !GlxUseBuiltInDRIDriver
-# We have to go into X to make the xmesaP.h symbolic link
-  MESASUBDIRS = X
-
+  MESASUBDIRS =
   ASMSUBDIRS = X86
 
 #include <Library.tmpl>
@@ -371,13 +386,4 @@ SUBDIRS = $(MESASUBDIRS) $(ASMSUBDIRS) drv
 
 MakeSubdirs($(SUBDIRS))
 DependSubdirs($(SUBDIRS))
-
-#if !GlxUseBuiltInDRIDriver && !GlxDriverUsesMesa
-LIBNAME = mesa_dri.so
-ALL_OBJS = $(CORE_OBJS) X/?*.o X86/?*.o
-ALL_DEPS = $(SUBDIRS) DONE X/DONE X86/DONE
-SharedDepModuleTarget($(LIBNAME),$(ALL_DEPS),$(ALL_OBJS))
-InstallDynamicModule($(LIBNAME),$(MODULEDIR)/dri,.)
-#endif
-
 DependTarget()
diff --git a/xc/lib/GL/mesa/src/X86/Imakefile b/xc/lib/GL/mesa/src/X86/Imakefile
index 858f0cd73..53402ac67 100644
--- a/xc/lib/GL/mesa/src/X86/Imakefile
+++ b/xc/lib/GL/mesa/src/X86/Imakefile
@@ -16,7 +16,6 @@ LinkSourceFile(x86.h, ../../../../../extras/Mesa/src/X86)
 LinkSourceFile(x86a.S, ../../../../../extras/Mesa/src/X86)
 LinkSourceFile(vertex.S, ../../../../../extras/Mesa/src/X86)
 LinkSourceFile(x86flatregs.m4, ../../../../../extras/Mesa/src/X86)
-LinkSourceFile(x86a.S.m4, ../../../../../extras/Mesa/src/X86)
 
 LinkSourceFile(mmx.h, ../../../../../extras/Mesa/src/X86)
 LinkSourceFile(mmx_blend.S, ../../../../../extras/Mesa/src/X86)
@@ -55,9 +54,7 @@ XCOMM We'll learn at runtime whether 3dNow, MMX, etc are really present.
 
       MMX_DEFS = -DUSE_MMX_ASM
 
-
-XCOMM Disabling 3DNow code for the time being
-#if 0
+#ifdef MesaUse3DNow
     3DNOW_SRCS = 3dnow.c 3dnow_norm_raw.S 3dnow_xform_masked1.S \
 		 3dnow_xform_masked2.S 3dnow_xform_masked3.S \
 		 3dnow_xform_masked4.S 3dnow_xform_raw1.S \
@@ -73,6 +70,7 @@ XCOMM Disabling 3DNow code for the time being
     3DNOW_DEFS = -DUSE_3DNOW_ASM
 #endif
 
+
 #endif
 
       DEFINES = $(ALLOC_DEFINES) GlxDefines -DFX $(X86_DEFS) $(MMX_DEFS) $(3DNOW_DEFS)
@@ -96,7 +94,7 @@ STD_CPP_DEFINES = StandardDefines $(PROJECT_DEFINES)
 SubdirLibraryRule($(OBJS))
 NormalLintTarget($(SRCS))
 
-#ifdef HAVE_3DNOW
+#ifdef MesaUse3DNow
 ObjectFromAsmSource(3dnow_norm_raw, NullParameter)
 ObjectFromAsmSource(3dnow_xform_masked1, NullParameter)
 ObjectFromAsmSource(3dnow_xform_masked2, NullParameter)
@@ -109,9 +107,7 @@ ObjectFromAsmSource(3dnow_xform_raw4, NullParameter)
 ObjectFromAsmSource(vertex_3dnow, NullParameter)
 #endif
 
-#ifdef HAVE_MMX
 ObjectFromAsmSource(mmx_blend, NullParameter)
-#endif
 
 ObjectFromAsmSource(common_x86asm, NullParameter)
 ObjectFromAsmSource(vertex, NullParameter)
@@ -119,6 +115,3 @@ ObjectFromAsmSource(x86a, NullParameter)
 
 DependTarget()
 
-x86a.S: x86flatregs.m4
-x86a.S: x86a.S.m4
-	m4 $< >$@
diff --git a/xc/lib/GL/mesa/src/drv/Imakefile b/xc/lib/GL/mesa/src/drv/Imakefile
index fadb29bb2..ed30e7196 100644
--- a/xc/lib/GL/mesa/src/drv/Imakefile
+++ b/xc/lib/GL/mesa/src/drv/Imakefile
@@ -16,6 +16,15 @@ DRIVER += gamma
 #if GlxBuiltInTdfx
 DRIVER += tdfx
 #endif
+#if GlxBuiltInMga
+DRIVER += common mga
+#endif
+#if GlxBuiltInI810
+DRIVER += common i810
+#endif
+#if GlxBuiltInR128
+DRIVER += r128
+#endif
 
 SUBDIRS = $(DRIVER)
 #else
@@ -23,6 +32,10 @@ SUBDIRS += gamma
 #if HasGlide3
 SUBDIRS += tdfx
 #endif
+SUBDIRS += common
+SUBDIRS += mga
+SUBDIRS += i810
+SUBDIRS += r128
 #endif
 
 MakeSubdirs($(SUBDIRS))
diff --git a/xc/lib/GL/mesa/src/drv/common/Imakefile b/xc/lib/GL/mesa/src/drv/common/Imakefile
index 5c3d40dd7..d47b662e8 100644
--- a/xc/lib/GL/mesa/src/drv/common/Imakefile
+++ b/xc/lib/GL/mesa/src/drv/common/Imakefile
@@ -22,13 +22,12 @@ MESA_INCLUDES = -I. -I.. -I../../include
 
 
       DEFINES = $(ALLOC_DEFINES) $(DRI_DEFINES)
-     INCLUDES = -I$(XLIBSRC) -I$(EXTINCSRC) $(MESA_INCLUDES) $(DRI_INCLUDES) \
-		-I/usr/include/glide
+     INCLUDES = -I$(XLIBSRC) -I$(EXTINCSRC) $(MESA_INCLUDES) $(DRI_INCLUDES)
       DRISRCS = hwlog.c mm.c
       DRIOBJS = hwlog.o mm.o
 
-	 SRCS = $(DRISRCS) 
-	 OBJS = $(DRIOBJS) 
+         SRCS = $(DRISRCS) 
+         OBJS = $(DRIOBJS) 
 
 #if !GlxUseBuiltInDRIDriver
 #undef DoNormalLib NormalLibGlx
@@ -44,7 +43,8 @@ LibraryObjectRule()
 SubdirLibraryRule($(OBJS))
 NormalLintTarget($(SRCS))
 
-#if !GlxUseBuiltInDRIDriver
+XCOMM #if !GlxUseBuiltInDRIDriver
+#if 0
 LIBNAME = i810_dri.so
 ALL_OBJS = $(OBJS)
 ALL_DEPS = DONE
diff --git a/xc/lib/GL/mesa/src/drv/gamma/Imakefile b/xc/lib/GL/mesa/src/drv/gamma/Imakefile
index 05e35561e..2a690a3e1 100644
--- a/xc/lib/GL/mesa/src/drv/gamma/Imakefile
+++ b/xc/lib/GL/mesa/src/drv/gamma/Imakefile
@@ -1,4 +1,5 @@
-XCOMM $XFree86: xc/lib/GL/mesa/src/drv/gamma/Imakefile,v 1.8 2000/03/02 16:07:35 martin Exp $
+
+#include <Threads.tmpl>
 
 #define DoNormalLib NormalLibGlx
 #define DoSharedLib SharedLibGlx
@@ -6,28 +7,35 @@ XCOMM $XFree86: xc/lib/GL/mesa/src/drv/gamma/Imakefile,v 1.8 2000/03/02 16:07:35
 #define DoDebugLib DebugLibGlx
 #define DoProfileLib ProfileLibGlx
 
-LinkSourceFile(xmesaP.h, ../../../../../../extras/Mesa/src/X)
-LinkSourceFile(glapi.h, ../../../../../../extras/Mesa/src)
-
-
 #if Malloc0ReturnsNull
 ALLOC_DEFINES = -DMALLOC_0_RETURNS_NULL
 #endif
 
 #if BuildXF86DRI
-  DRI_DEFINES = GlxDefines
+  DRI_DEFINES = GlxDefines -DDRIVERTS
  DRI_INCLUDES = -I../../../../dri -I../../../../glx \
 		-I../../../dri \
 		-I$(TOP)/include -I$(TOP)/include/GL \
 		-I$(XF86OSSRC) -I$(SERVERSRC)/GL/dri \
 		-I$(XF86DRIVERSRC)/glint \
-		-I../../../include -I../.. -I../../X
+		-I../../../include -I../.. -I../../X -I../common \
+	        -I$(XF86OSSRC)/linux/drm/kernel
 #endif
 
 MESA_INCLUDES = -I. -I.. -I../../include
 
+
+
       DEFINES = $(ALLOC_DEFINES) $(DRI_DEFINES)
-     INCLUDES = -I$(XLIBSRC) -I$(EXTINCSRC) $(MESA_INCLUDES) $(DRI_INCLUDES)
+     INCLUDES = -I$(XLIBSRC) -I$(EXTINCSRC) $(MESA_INCLUDES) $(DRI_INCLUDES) 
+
+#if 0
+	LOSRC = ../../../../lowpc.c
+	LOOBJ = ../../../../lowpc.o
+
+	HISRC = ../../../../highpc.c
+	HIOBJ = ../../../../highpc.o
+#endif
 
       DRISRCS = ../../../dri/dri_mesa.c \
 		../../../../dri/dri_tmm.c
@@ -46,16 +54,216 @@ MESA_INCLUDES = -I. -I.. -I../../include
 		../../../../dri/drm/xf86drmSL.o
 
     GAMMASRCS = gamma_gl.c gamma_xmesa.c gamma_init.c gamma_matrix.c \
-                gamma_inithw.c gamma_texture.c
+                gamma_inithw.c gamma_texture.c /* gamma_dlist.c */
 
     GAMMAOBJS = gamma_gl.o gamma_xmesa.o gamma_init.o gamma_matrix.o \
-                gamma_inithw.o gamma_texture.o
+                gamma_inithw.o gamma_texture.o /* gamma_dlist.o */
+
+     MESASRCS = ../../aatriangle.c \
+		../../accum.c \
+		../../alpha.c \
+		../../alphabuf.c \
+		../../attrib.c \
+		../../bbox.c \
+		../../bitmap.c \
+		../../blend.c \
+		../../buffers.c \
+		../../clip.c \
+		../../colortab.c \
+		../../config.c \
+		../../context.c \
+		../../copypix.c \
+		../../cva.c \
+		../../debug_xform.c \
+		../../depth.c \
+		../../dlist.c \
+		../../drawpix.c \
+		../../enable.c \
+		../../enums.c \
+		../../eval.c \
+		../../extensions.c \
+		../../feedback.c \
+		../../fog.c \
+		../../get.c \
+		../../glapi.c \
+		../../glapinoop.c \
+		../../glthread.c \
+		../../hash.c \
+		../../image.c \
+		../../imaging.c \
+		../../light.c \
+		../../lines.c \
+		../../logic.c \
+		../../masking.c \
+		../../matrix.c \
+		../../mem.c \
+		../../mmath.c \
+		../../pb.c \
+		../../pipeline.c \
+		../../pixel.c \
+		../../pixeltex.c \
+		../../points.c \
+		../../polygon.c \
+		../../quads.c \
+		../../rastpos.c \
+		../../readpix.c \
+		../../rect.c \
+		../../scissor.c \
+		../../shade.c \
+		../../span.c \
+		../../stages.c \
+		../../state.c \
+		../../stencil.c \
+		../../teximage.c \
+		../../texobj.c \
+		../../texstate.c \
+		../../texture.c \
+		../../texutil.c \
+		../../translate.c \
+		../../triangle.c \
+		../../varray.c \
+		../../vb.c \
+		../../vbcull.c \
+		../../vbfill.c \
+		../../vbindirect.c \
+		../../vbrender.c \
+		../../vbxform.c \
+		../../vector.c \
+		../../vertices.c \
+		../../winpos.c \
+		../../xform.c \
+		../../zoom.c \
+		../../X86/common_x86.c
+
+     MESAOBJS = ../../aatriangle.o \
+		../../accum.o \
+		../../alpha.o \
+		../../alphabuf.o \
+		../../attrib.o \
+		../../bbox.o \
+		../../bitmap.o \
+		../../blend.o \
+		../../buffers.o \
+		../../clip.o \
+		../../colortab.o \
+		../../config.o \
+		../../context.o \
+		../../copypix.o \
+		../../cva.o \
+		../../debug_xform.o \
+		../../depth.o \
+		../../dlist.o \
+		../../drawpix.o \
+		../../enable.o \
+		../../enums.o \
+		../../eval.o \
+		../../extensions.o \
+		../../feedback.o \
+		../../fog.o \
+		../../get.o \
+		../../hash.o \
+		../../hint.o \
+		../../image.o \
+		../../imaging.o \
+		../../light.o \
+		../../lines.o \
+		../../logic.o \
+		../../masking.o \
+		../../matrix.o \
+		../../mem.o \
+		../../mmath.o \
+		../../pb.o \
+		../../pipeline.o \
+		../../pixel.o \
+		../../pixeltex.o \
+		../../points.o \
+		../../polygon.o \
+		../../quads.o \
+		../../rastpos.o \
+		../../readpix.o \
+		../../rect.o \
+		../../scissor.o \
+		../../shade.o \
+		../../span.o \
+		../../stages.o \
+		../../state.o \
+		../../stencil.o \
+		../../teximage.o \
+		../../texobj.o \
+		../../texstate.o \
+		../../texture.o \
+		../../texutil.o \
+		../../translate.o \
+		../../triangle.o \
+		../../varray.o \
+		../../vb.o \
+		../../vbcull.o \
+		../../vbfill.o \
+		../../vbindirect.o \
+		../../vbrender.o \
+		../../vbxform.o \
+		../../vector.o \
+		../../vertices.o \
+		../../winpos.o \
+		../../xform.o \
+		../../zoom.o
+
+#ifdef i386Architecture
+     X86_SRCS = ../../X86/x86.c \
+		../../X86/x86a.S \
+		../../X86/common_x86.c \
+		../../X86/common_x86asm.S \
+		../../X86/vertex.S
 
-XCOMM	 SRCS = $(DRISRCS) $(DRMSRCS) $(GAMMASRCS)
-XCOMM	 OBJS = $(DRIOBJS) $(DRMOBJS) $(GAMMAOBJS)
+     X86_OBJS = ../../X86/x86.o \
+		../../X86/x86a.o \
+		../../X86/common_x86.o \
+		../../X86/common_x86asm.o \
+		../../X86/vertex.o
 
-	 SRCS = $(GAMMASRCS)
-	 OBJS = $(DRIOBJS) $(DRMOBJS) $(GAMMAOBJS)
+     MMX_SRCS = ../../X86/mmx_blend.S
+
+     MMX_OBJS = ../../X86/mmx_blend.o
+
+XCOMM Disabling 3Dnow code for the time being.
+#if 0
+   3DNOW_SRCS = ../../X86/3dnow.c \
+		../../X86/3dnow_norm_raw.S \
+		../../X86/3dnow_xform_masked1.S \
+		../../X86/3dnow_xform_masked2.S \
+		../../X86/3dnow_xform_masked3.S \
+		../../X86/3dnow_xform_masked4.S \
+		../../X86/3dnow_xform_raw1.S \
+		../../X86/3dnow_xform_raw2.S \
+		../../X86/3dnow_xform_raw3.S \
+		../../X86/3dnow_xform_raw4.S \
+		../../X86/vertex_3dnow.S
+
+   3DNOW_OBJS = ../../X86/3dnow.o \
+		../../X86/3dnow_norm_raw.o \
+		../../X86/3dnow_xform_masked1.o \
+		../../X86/3dnow_xform_masked2.o \
+		../../X86/3dnow_xform_masked3.o \
+		../../X86/3dnow_xform_masked4.o \
+		../../X86/3dnow_xform_raw1.o \
+		../../X86/3dnow_xform_raw2.o \
+		../../X86/3dnow_xform_raw3.o \
+		../../X86/3dnow_xform_raw4.o \
+		../../X86/vertex_3dnow.o
+#endif
+
+#endif
+
+      ASMSRCS = $(X86_SRCS) $(MMX_SRCS) $(3DNOW_SRCS)
+      ASMOBJS = $(X86_OBJS) $(MMX_OBJS) $(3DNOW_OBJS)
+
+	 SRCS = $(LOWSRC) $(DRISRCS) $(DRMSRCS) $(MESASRCS) $(ASMSRCS) $(GAMMASRCS) $(HISRC)
+	 OBJS = $(LOWOBJ) $(DRIOBJS) $(DRMOBJS) $(MESAOBJS) $(ASMOBJS) $(GAMMAOBJS) $(HIOBJ)
+
+REQUIREDLIBS += -lm
+#if !GlxBuiltInGamma
+REQUIREDLIBS += -L../../../.. -lGL
+#endif
 
 
 #if !GlxUseBuiltInDRIDriver
diff --git a/xc/lib/GL/mesa/src/drv/gamma/gamma_gl.c b/xc/lib/GL/mesa/src/drv/gamma/gamma_gl.c
index 3bd2fd0f3..541f4d69c 100644
--- a/xc/lib/GL/mesa/src/drv/gamma/gamma_gl.c
+++ b/xc/lib/GL/mesa/src/drv/gamma/gamma_gl.c
@@ -30,13 +30,16 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  * Authors:
  *   Kevin E. Martin <kevin@precisioninsight.com>
  *   Brian Paul <brian@precisioninsight.com>
+ *   Alan Hourihane <Alan.Hourihane@btinternet.com>
  */
 
 #ifdef GLX_DIRECT_RENDERING
 
+#include <Xarch.h>
 #include <math.h>
 #include "gamma_gl.h"
 #include "gamma_init.h"
+#include "glint_dri.h"
 #ifdef RANDOMIZE_COLORS
 #include <stdlib.h>
 #endif
@@ -114,7 +117,7 @@ void _gamma_Begin(GLenum mode)
     DEBUG_GLCMDS(("Begin: %04x\n", (int)mode));
 
     if ((gCCPriv->Begin & B_PrimType_Mask) != B_PrimType_Null) {
-	/* ERROR!!! */
+        DEBUG_ERROR(("Begin: Error\n"));
 	return;
     }
 
@@ -177,8 +180,9 @@ void _gamma_BindTexture(GLenum target, GLuint texture)
     gCCPriv->curTexObj = gammaTOFind(texture);
 
     /* Make the new texture images resident */
-    if (!driTMMMakeImagesResident(gCCPriv->tmm, MIPMAP_LEVELS,
-				  gCCPriv->curTexObj->image, addrs)) {
+    if (driTMMMakeImagesResident(gCCPriv->tmm, MIPMAP_LEVELS,
+				  gCCPriv->curTexObj->image, addrs) < 0) {
+    	DEBUG_ERROR(("BindTexture: unable\n"));
 	/* NOT_DONE: Handle error */
     }
 
@@ -358,10 +362,12 @@ void _gamma_CallLists(GLsizei n, GLenum type, const GLvoid *lists)
 
 void _gamma_Clear(GLbitfield mask)
 {
+    int temp;
     unsigned int depth = 0;
     int do_clear = 0;
+    GLINTDRIPtr gDRIPriv = (GLINTDRIPtr)gCC->driScreenPriv->pDevPriv;
 #ifdef DO_VALIDATE
-    __DRIscreenPrivate *driScrnPriv = gCC->driContextPriv->driScreenPriv;
+    __DRIscreenPrivate *driScrnPriv = gCC->driScreenPriv;
 #endif
 
     DEBUG_GLCMDS(("Clear: %04x\n", (int)mask));
@@ -437,11 +443,12 @@ void _gamma_Clear(GLbitfield mask)
 	gCCPriv->FrameCount &= 0xff;
 #endif
 
+	temp = (gCCPriv->LBReadMode & LBPartialProdMask) | LBWindowOriginBot;
+	/* UGH - move this later ! */
+	if (gDRIPriv->numMXDevices == 2) temp |= LBScanLineInt2;
+
 	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
-	WRITE(gCCPriv->buf, LBReadMode, 
-	      ((gCCPriv->LBReadMode & LBPartialProdMask) |
-	       LBScanLineInt2 |
-	       LBWindowOriginBot));
+	WRITE(gCCPriv->buf, LBReadMode, temp);
 
 	/* Force FCP to be written */
 	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
@@ -627,34 +634,48 @@ void _gamma_ClipPlane(GLenum plane, const GLdouble *equation)
 
 void _gamma_Color3b(GLbyte red, GLbyte green, GLbyte blue)
 {
+    GLfloat r,g,b;
+
     DEBUG_GLCMDS(("Color3b: %d %d %d\n", red, green, blue));
+  
+    r = BYTE_TO_FLOAT(red);
+    g = BYTE_TO_FLOAT(green);
+    b = BYTE_TO_FLOAT(blue);
+
+    _gamma_Color3f(r,g,b);
 }
 
 void _gamma_Color3bv(const GLbyte *v)
 {
+    GLfloat p[3];
+
     DEBUG_GLCMDS(("Color3bv: %d %d %d\n", v[0], v[1], v[2]));
+
+    p[0] = BYTE_TO_FLOAT(v[0]);
+    p[1] = BYTE_TO_FLOAT(v[1]);
+    p[2] = BYTE_TO_FLOAT(v[2]);
+
+    _gamma_Color3fv(p);
 }
 
 void _gamma_Color3d(GLdouble red, GLdouble green, GLdouble blue)
 {
     DEBUG_GLCMDS(("Color3d: %f %f %f\n", red, green, blue));
+
+    _gamma_Color3f((GLfloat)red,(GLfloat)green,(GLfloat)blue);
 }
 
 void _gamma_Color3dv(const GLdouble *v)
 {
     DEBUG_GLCMDS(("Color3dv: %f %f %f\n", v[0], v[1], v[2]));
+
+    _gamma_Color3fv((GLfloat*)v);
 }
 
 void _gamma_Color3f(GLfloat red, GLfloat green, GLfloat blue)
 {
     DEBUG_GLCMDS(("Color3f: %f %f %f\n", red, green, blue));
 
-#ifdef RANDOMIZE_COLORS
-    red = (random() / (double)RAND_MAX);
-    green = (random() / (double)RAND_MAX);
-    blue = (random() / (double)RAND_MAX);
-#endif
-
     CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
     WRITEF(gCCPriv->buf, Cb,  blue);
     WRITEF(gCCPriv->buf, Cg,  green);
@@ -665,107 +686,188 @@ void _gamma_Color3fv(const GLfloat *v)
 {
     DEBUG_GLCMDS(("Color3fv: %f %f %f\n", v[0], v[1], v[2]));
 
-#ifdef RANDOMIZE_COLORS
-    {
-	float r, g, b;
-	r = (random() / (double)RAND_MAX);
-	g = (random() / (double)RAND_MAX);
-	b = (random() / (double)RAND_MAX);
-	CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
-	WRITEF(gCCPriv->buf, Cb,  b);
-	WRITEF(gCCPriv->buf, Cg,  g);
-	WRITEF(gCCPriv->buf, Cr3, r);
-    }
-#else
     CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
     WRITEF(gCCPriv->buf, Cb,  v[2]);
     WRITEF(gCCPriv->buf, Cg,  v[1]);
     WRITEF(gCCPriv->buf, Cr3, v[0]);
-#endif
 }
 
 void _gamma_Color3i(GLint red, GLint green, GLint blue)
 {
+    GLfloat r,g,b;
+
     DEBUG_GLCMDS(("Color3i: %d %d %d\n", (int)red, (int)green, (int)blue));
+
+    r = INT_TO_FLOAT(red);
+    g = INT_TO_FLOAT(green);
+    b = INT_TO_FLOAT(blue);
+
+    _gamma_Color3f(r,g,b);
 }
 
 void _gamma_Color3iv(const GLint *v)
 {
+    GLfloat p[3];
+
     DEBUG_GLCMDS(("Color3iv: %d %d %d\n", (int)v[0], (int)v[1], (int)v[2]));
+
+    p[0] = INT_TO_FLOAT(v[0]);
+    p[1] = INT_TO_FLOAT(v[1]);
+    p[2] = INT_TO_FLOAT(v[2]);
+
+    _gamma_Color3fv(p);
 }
 
 void _gamma_Color3s(GLshort red, GLshort green, GLshort blue)
 {
+    GLfloat r,g,b;
+
     DEBUG_GLCMDS(("Color3s: %d %d %d\n", red, green, blue));
+
+    r = SHORT_TO_FLOAT(red);
+    g = SHORT_TO_FLOAT(green);
+    b = SHORT_TO_FLOAT(blue);
+
+    _gamma_Color3f(r,g,b);
 }
 
 void _gamma_Color3sv(const GLshort *v)
 {
+    GLfloat p[3];
+
     DEBUG_GLCMDS(("Color3sv: %d %d %d\n", v[0], v[1], v[2]));
+
+    p[0] = SHORT_TO_FLOAT(v[0]);
+    p[1] = SHORT_TO_FLOAT(v[1]);
+    p[2] = SHORT_TO_FLOAT(v[2]);
+
+    _gamma_Color3fv(p);
 }
 
 void _gamma_Color3ub(GLubyte red, GLubyte green, GLubyte blue)
 {
+    GLuint c;
+
     DEBUG_GLCMDS(("Color3ub: %d %d %d\n", red, green, blue));
+
+    c = (blue << 16) | (green << 8) | red;
+
+    CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+    WRITE(gCCPriv->buf, PackedColor3,  c);
 }
 
 void _gamma_Color3ubv(const GLubyte *v)
 {
+    GLuint c;
+
     DEBUG_GLCMDS(("Color3ubv: %d %d %d\n", v[0], v[1], v[2]));
+
+    c = (v[2] << 16) | (v[1] << 8) | v[0];
+
+    CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+    WRITE(gCCPriv->buf, PackedColor3,  c);
 }
 
 void _gamma_Color3ui(GLuint red, GLuint green, GLuint blue)
 {
+    GLfloat r,g,b;
+
     DEBUG_GLCMDS(("Color3ui: %d %d %d\n",
 		  (unsigned int)red, (unsigned int)green, (unsigned int)blue));
+
+    r = UINT_TO_FLOAT(red);
+    g = UINT_TO_FLOAT(green);
+    b = UINT_TO_FLOAT(blue);
+
+    _gamma_Color3f(r,g,b);
 }
 
 void _gamma_Color3uiv(const GLuint *v)
 {
+    GLfloat p[3];
+
     DEBUG_GLCMDS(("Color3uiv: %d %d %d\n",
 		  (unsigned int)v[0], (unsigned int)v[1], (unsigned int)v[2]));
+
+    p[0] = UINT_TO_FLOAT(v[0]);
+    p[1] = UINT_TO_FLOAT(v[1]);
+    p[2] = UINT_TO_FLOAT(v[2]);
+
+    _gamma_Color3fv(p);
 }
 
 void _gamma_Color3us(GLushort red, GLushort green, GLushort blue)
 {
+    GLfloat r,g,b;
+
     DEBUG_GLCMDS(("Color3us: %d %d %d\n", red, green, blue));
+
+    r = USHORT_TO_FLOAT(red);
+    g = USHORT_TO_FLOAT(green);
+    b = USHORT_TO_FLOAT(blue);
+
+    _gamma_Color3f(r,g,b);
 }
 
 void _gamma_Color3usv(const GLushort *v)
 {
+    GLfloat p[3];
+
     DEBUG_GLCMDS(("Color3usv: %d %d %d\n", v[0], v[1], v[2]));
+
+    p[0] = USHORT_TO_FLOAT(v[0]);
+    p[1] = USHORT_TO_FLOAT(v[1]);
+    p[2] = USHORT_TO_FLOAT(v[2]);
+
+    _gamma_Color3fv(p);
 }
 
 void _gamma_Color4b(GLbyte red, GLbyte green, GLbyte blue, GLbyte alpha)
 {
+    GLfloat r,g,b,a;
+
     DEBUG_GLCMDS(("Color4b: %d %d %d %d\n", red, green, blue, alpha));
+
+    r = BYTE_TO_FLOAT(red);
+    g = BYTE_TO_FLOAT(green);
+    b = BYTE_TO_FLOAT(blue);
+    a = BYTE_TO_FLOAT(alpha);
+
+    _gamma_Color4f(r,g,b,a);
 }
 
 void _gamma_Color4bv(const GLbyte *v)
 {
+    GLfloat p[4];
+
     DEBUG_GLCMDS(("Color4bv: %d %d %d %d\n", v[0], v[1], v[2], v[3]));
+
+    p[0] = BYTE_TO_FLOAT(v[0]);
+    p[1] = BYTE_TO_FLOAT(v[1]);
+    p[2] = BYTE_TO_FLOAT(v[2]);
+    p[3] = BYTE_TO_FLOAT(v[3]);
+
+    _gamma_Color4fv(p);
 }
 
 void _gamma_Color4d(GLdouble red, GLdouble green, GLdouble blue, GLdouble alpha)
 {
     DEBUG_GLCMDS(("Color4d: %f %f %f %f\n", red, green, blue, alpha));
+
+    _gamma_Color4f((GLfloat)red,(GLfloat)green,(GLfloat)blue,(GLfloat)alpha);
 }
 
 void _gamma_Color4dv(const GLdouble *v)
 {
     DEBUG_GLCMDS(("Color4dv: %f %f %f %f\n", v[0], v[1], v[2], v[3]));
+
+    _gamma_Color4fv((GLfloat*)v);
 }
 
 void _gamma_Color4f(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha)
 {
     DEBUG_GLCMDS(("Color4f: %f %f %f %f\n", red, green, blue, alpha));
 
-#ifdef RANDOMIZE_COLORS
-    red = (random() / (double)RAND_MAX);
-    green = (random() / (double)RAND_MAX);
-    blue = (random() / (double)RAND_MAX);
-#endif
-
     CHECK_DMA_BUFFER(gCC, gCCPriv, 4);
     WRITEF(gCCPriv->buf, Ca,  alpha);
     WRITEF(gCCPriv->buf, Cb,  blue);
@@ -777,47 +879,69 @@ void _gamma_Color4fv(const GLfloat *v)
 {
     DEBUG_GLCMDS(("Color4fv: %f %f %f %f\n", v[0], v[1], v[2], v[3]));
 
-#ifdef RANDOMIZE_COLORS
-    {
-	float r, g, b;
-	r = (random() / (double)RAND_MAX);
-	g = (random() / (double)RAND_MAX);
-	b = (random() / (double)RAND_MAX);
-	CHECK_DMA_BUFFER(gCC, gCCPriv, 4);
-	WRITEF(gCCPriv->buf, Ca,  v[3]);
-	WRITEF(gCCPriv->buf, Cb,  b);
-	WRITEF(gCCPriv->buf, Cg,  g);
-	WRITEF(gCCPriv->buf, Cr3, r);
-    }
-#else
     CHECK_DMA_BUFFER(gCC, gCCPriv, 4);
     WRITEF(gCCPriv->buf, Ca,  v[3]);
     WRITEF(gCCPriv->buf, Cb,  v[2]);
     WRITEF(gCCPriv->buf, Cg,  v[1]);
     WRITEF(gCCPriv->buf, Cr4, v[0]);
-#endif
 }
 
 void _gamma_Color4i(GLint red, GLint green, GLint blue, GLint alpha)
 {
+    GLfloat r,g,b,a;
+
     DEBUG_GLCMDS(("Color4i: %d %d %d %d\n", (int)red, (int)green, (int)blue,
 		  (int)alpha));
+
+    r = INT_TO_FLOAT(red);
+    g = INT_TO_FLOAT(green);
+    b = INT_TO_FLOAT(blue);
+    a = INT_TO_FLOAT(alpha);
+
+    _gamma_Color4f(r,g,b,a);
 }
 
 void _gamma_Color4iv(const GLint *v)
 {
+    GLfloat p[4];
+
     DEBUG_GLCMDS(("Color4iv: %d %d %d %d\n", (int)v[0], (int)v[1], (int)v[2],
 		  (int)v[3]));
+
+    p[0] = INT_TO_FLOAT(v[0]);
+    p[1] = INT_TO_FLOAT(v[1]);
+    p[2] = INT_TO_FLOAT(v[2]);
+    p[3] = INT_TO_FLOAT(v[3]);
+
+    _gamma_Color4fv(p);
 }
 
 void _gamma_Color4s(GLshort red, GLshort green, GLshort blue, GLshort alpha)
 {
+    GLfloat r,g,b,a;
+
     DEBUG_GLCMDS(("Color4s: %d %d %d %d\n", red, green, blue, alpha));
+
+    r = SHORT_TO_FLOAT(red);
+    g = SHORT_TO_FLOAT(green);
+    b = SHORT_TO_FLOAT(blue);
+    a = SHORT_TO_FLOAT(alpha);
+
+    _gamma_Color4f(r,g,b,a);
 }
 
 void _gamma_Color4sv(const GLshort *v)
 {
+    GLfloat p[4];
+
     DEBUG_GLCMDS(("Color4sv: %d %d %d %d\n", v[0], v[1], v[2], v[3]));
+
+    p[0] = SHORT_TO_FLOAT(v[0]);
+    p[1] = SHORT_TO_FLOAT(v[1]);
+    p[2] = SHORT_TO_FLOAT(v[2]);
+    p[3] = SHORT_TO_FLOAT(v[3]);
+
+    _gamma_Color4fv(p);
 }
 
 void _gamma_Color4ub(GLubyte red, GLubyte green, GLubyte blue, GLubyte alpha)
@@ -826,11 +950,7 @@ void _gamma_Color4ub(GLubyte red, GLubyte green, GLubyte blue, GLubyte alpha)
 
     DEBUG_GLCMDS(("Color4ub: %d %d %d %d\n", red, green, blue, alpha));
 
-#ifdef RANDOMIZE_COLORS
-    c = (random() / (double)RAND_MAX) * 16777216;
-#else
     c = (alpha << 24) | (blue << 16) | (green << 8) | red;
-#endif
 
     CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
     WRITE(gCCPriv->buf, PackedColor4,  c);
@@ -842,17 +962,7 @@ void _gamma_Color4ubv(const GLubyte *v)
 
     DEBUG_GLCMDS(("Color4ubv: %d %d %d %d\n", v[0], v[1], v[2], v[3]));
 
-#ifdef RANDOMIZE_COLORS
-    c = (random() / (double)RAND_MAX) * 16777216;
-#else
-/* NOT_DONE: Is there a standard define for endianness? */
-#define IS_LITTLE_ENDIAN 1
-#if IS_LITTLE_ENDIAN
-    c = *((GLuint *)v);
-#else
     c = (v[3] << 24) | (v[2] << 16) | (v[1] << 8) | v[0];
-#endif
-#endif
 
     CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
     WRITE(gCCPriv->buf, PackedColor4,  c);
@@ -860,26 +970,62 @@ void _gamma_Color4ubv(const GLubyte *v)
 
 void _gamma_Color4ui(GLuint red, GLuint green, GLuint blue, GLuint alpha)
 {
+    GLfloat r,g,b,a;
+
     DEBUG_GLCMDS(("Color4ui: %d %d %d %d\n",
 		  (unsigned int)red, (unsigned int)green,
 		  (unsigned int)blue, (unsigned int)alpha));
+
+    r = UINT_TO_FLOAT(red);
+    g = UINT_TO_FLOAT(green);
+    b = UINT_TO_FLOAT(blue);
+    a = UINT_TO_FLOAT(alpha);
+
+    _gamma_Color4f(r,g,b,a);
 }
 
 void _gamma_Color4uiv(const GLuint *v)
 {
+    GLfloat p[4];
+
     DEBUG_GLCMDS(("Color4uiv: %d %d %d %d\n",
 		  (unsigned int)v[0], (unsigned int)v[1],
 		  (unsigned int)v[2], (unsigned int)v[3]));
+
+    p[0] = UINT_TO_FLOAT(v[0]);
+    p[1] = UINT_TO_FLOAT(v[1]);
+    p[2] = UINT_TO_FLOAT(v[2]);
+    p[3] = UINT_TO_FLOAT(v[3]);
+
+    _gamma_Color4fv(p);
 }
 
 void _gamma_Color4us(GLushort red, GLushort green, GLushort blue, GLushort alpha)
 {
+    GLfloat r,g,b,a;
+
     DEBUG_GLCMDS(("Color4us: %d %d %d %d\n", red, green, blue, alpha));
+
+    r = USHORT_TO_FLOAT(red);
+    g = USHORT_TO_FLOAT(green);
+    b = USHORT_TO_FLOAT(blue);
+    a = USHORT_TO_FLOAT(alpha);
+
+    _gamma_Color4f(r,g,b,a);
 }
 
 void _gamma_Color4usv(const GLushort *v)
 {
+    GLfloat p[4];
+
     DEBUG_GLCMDS(("Color4usv: %d %d %d %d\n", v[0], v[1], v[2], v[3]));
+
+    p[0] = USHORT_TO_FLOAT(v[0]);
+    p[1] = USHORT_TO_FLOAT(v[1]);
+    p[2] = USHORT_TO_FLOAT(v[2]);
+    p[3] = USHORT_TO_FLOAT(v[3]);
+
+    _gamma_Color4fv(p);
 }
 
 void _gamma_ColorMask(GLboolean red, GLboolean green, GLboolean blue, GLboolean alpha)
@@ -890,6 +1036,43 @@ void _gamma_ColorMask(GLboolean red, GLboolean green, GLboolean blue, GLboolean
 void _gamma_ColorMaterial(GLenum face, GLenum mode)
 {
     DEBUG_GLCMDS(("ColorMaterial: %04x %04x\n", (int)face, (int)mode));
+
+    gCCPriv->MaterialMode &= ~MaterialModeEnable;
+    gCCPriv->ColorMaterialMode = ColorMaterialModeDisable;
+
+    switch (face) {
+    case GL_FRONT:
+	gCCPriv->ColorMaterialMode |= ColorMaterialModeFront;
+	break;
+    case GL_BACK:
+	gCCPriv->ColorMaterialMode |= ColorMaterialModeBack;
+	break;
+    case GL_FRONT_AND_BACK:
+	gCCPriv->ColorMaterialMode |= ColorMaterialModeFrontAndBack;
+	break;
+    }
+
+    switch (mode) {
+    case GL_AMBIENT:
+	gCCPriv->ColorMaterialMode |= ColorMaterialModeAmbient;
+	break;
+    case GL_EMISSION:
+	gCCPriv->ColorMaterialMode |= ColorMaterialModeEmission;
+	break;
+    case GL_DIFFUSE:
+	gCCPriv->ColorMaterialMode |= ColorMaterialModeDiffuse;
+	break;
+    case GL_SPECULAR:
+	gCCPriv->ColorMaterialMode |= ColorMaterialModeSpecular;
+	break;
+    case GL_AMBIENT_AND_DIFFUSE:
+	gCCPriv->ColorMaterialMode |= ColorMaterialModeAmbAndDiff;
+	break;
+    }
+
+    CHECK_DMA_BUFFER(gCC, gCCPriv, 2);
+    WRITE(gCCPriv->buf, ColorMaterialMode, gCCPriv->ColorMaterialMode);
+    WRITE(gCCPriv->buf, MaterialMode, gCCPriv->MaterialMode);
 }
 
 void _gamma_ColorPointer(GLint size, GLenum type, GLsizei stride, const GLvoid *pointer)
@@ -938,9 +1121,6 @@ void _gamma_CullFace(GLenum mode)
 
     gCCPriv->GeometryMode &= ~GM_PolyCullMask;
 
-#ifdef CULL_ALL_PRIMS
-    gCCPriv->GeometryMode |= GM_PolyCullBoth;
-#else
     switch (mode) {
     case GL_FRONT:
 	gCCPriv->GeometryMode |= GM_PolyCullFront;
@@ -955,7 +1135,7 @@ void _gamma_CullFace(GLenum mode)
 	/* ERROR!! */
 	break;
     }
-#endif
+
     CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
     WRITE(gCCPriv->buf, GeometryMode, gCCPriv->GeometryMode);
 }
@@ -980,7 +1160,8 @@ void _gamma_DeleteTextures(GLsizei n, const GLuint *textures)
 
     for (i = 0; i < n; i++) {
 	gammaTexObj *t = gammaTOFind(textures[i]);
-	if (!driTMMDeleteImages(gCCPriv->tmm, MIPMAP_LEVELS, t->image)) {
+	if (driTMMDeleteImages(gCCPriv->tmm, MIPMAP_LEVELS, t->image) < 0) {
+    	    DEBUG_ERROR(("DeleteTextures: unable\n"));
 	    /* NOT_DONE: Handle error */
 	}
 	gammaTODelete(textures[i]);
@@ -1068,11 +1249,9 @@ void _gamma_Disable(GLenum cap)
 
     switch (cap) {
     case GL_CULL_FACE:
-#ifdef CULL_ALL_PRIMS
 	gCCPriv->GeometryMode &= ~GM_PolyCullEnable;
 	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
 	WRITE(gCCPriv->buf, GeometryMode, gCCPriv->GeometryMode);
-#endif
 	break;
     case GL_DEPTH_TEST:
 	if (gCCPriv->Flags & GAMMA_DEPTH_BUFFER) {
@@ -1106,6 +1285,35 @@ void _gamma_Disable(GLenum cap)
 	WRITE(gCCPriv->buf, FBReadMode, (gCCPriv->FBReadMode |
 					 gCCPriv->AB_FBReadMode));
 	break;
+    case GL_COLOR_MATERIAL:
+	gCCPriv->ColorMaterialMode &= ~ColorMaterialModeEnable;
+	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	WRITE(gCCPriv->buf, ColorMaterialMode, gCCPriv->ColorMaterialMode);
+	break;
+    case GL_FOG:
+	gCCPriv->Begin &= ~B_FogEnable;
+	gCCPriv->GeometryMode &= ~GM_FogEnable;
+	gCCPriv->DeltaMode &= ~DM_FogEnable;
+	CHECK_DMA_BUFFER(gCC, gCCPriv, 2);
+	WRITE(gCCPriv->buf, GeometryMode, gCCPriv->GeometryMode);
+	WRITE(gCCPriv->buf, DeltaMode, gCCPriv->DeltaMode);
+	WRITE(gCCPriv->buf, FogMode, FogModeDisable);
+	break;
+    case GL_LIGHTING:
+	gCCPriv->LightingMode &= ~LightingModeEnable;
+	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	WRITE(gCCPriv->buf, LightingMode, gCCPriv->LightingMode);
+	break;
+    case GL_LIGHT0:
+	gCCPriv->Light0Mode &= ~LNM_On;
+	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	WRITE(gCCPriv->buf, Light0Mode, gCCPriv->Light0Mode);
+	break;
+    case GL_LIGHT1:
+	gCCPriv->Light1Mode &= ~LNM_On;
+	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	WRITE(gCCPriv->buf, Light1Mode, gCCPriv->Light1Mode);
+	break;
     case GL_TEXTURE_2D:
 	gCCPriv->Texture2DEnabled = GL_FALSE;
 	gCCPriv->Begin                         &= ~B_TextureEnable;
@@ -1227,6 +1435,35 @@ void _gamma_Enable(GLenum cap)
 	WRITE(gCCPriv->buf, FBReadMode, (gCCPriv->FBReadMode |
 					 gCCPriv->AB_FBReadMode));
 	break;
+    case GL_COLOR_MATERIAL:
+	gCCPriv->ColorMaterialMode |= ColorMaterialModeEnable;
+	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	WRITE(gCCPriv->buf, ColorMaterialMode, gCCPriv->ColorMaterialMode);
+	break;
+    case GL_FOG:
+	gCCPriv->Begin |= B_FogEnable;
+	gCCPriv->GeometryMode |= GM_FogEnable;
+	gCCPriv->DeltaMode |= DM_FogEnable;
+	CHECK_DMA_BUFFER(gCC, gCCPriv, 2);
+	WRITE(gCCPriv->buf, GeometryMode, gCCPriv->GeometryMode);
+	WRITE(gCCPriv->buf, DeltaMode, gCCPriv->DeltaMode);
+	WRITE(gCCPriv->buf, FogMode, FogModeEnable);
+	break;
+    case GL_LIGHTING:
+	gCCPriv->LightingMode |= LightingModeEnable | 16<<6;
+	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	WRITE(gCCPriv->buf, LightingMode, gCCPriv->LightingMode);
+	break;
+    case GL_LIGHT0:
+	gCCPriv->Light0Mode |= LNM_On;
+	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	WRITE(gCCPriv->buf, Light0Mode, gCCPriv->Light0Mode);
+	break;
+    case GL_LIGHT1:
+	gCCPriv->Light1Mode |= LNM_On;
+	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	WRITE(gCCPriv->buf, Light1Mode, gCCPriv->Light1Mode);
+	break;
     case GL_TEXTURE_2D:
 	gCCPriv->Texture2DEnabled = GL_TRUE;
 #ifndef TURN_OFF_TEXTURES
@@ -1294,34 +1531,42 @@ void _gamma_EvalCoord1d(GLdouble u)
 
 void _gamma_EvalCoord1dv(const GLdouble *u)
 {
+    DEBUG_GLCMDS(("EvalCoord1dv: %f\n", *u));
 }
 
 void _gamma_EvalCoord1f(GLfloat u)
 {
+    DEBUG_GLCMDS(("EvalCoord1f: %f\n", u));
 }
 
 void _gamma_EvalCoord1fv(const GLfloat *u)
 {
+    DEBUG_GLCMDS(("EvalCoord1fv: %f\n", *u));
 }
 
 void _gamma_EvalCoord2d(GLdouble u, GLdouble v)
 {
+    DEBUG_GLCMDS(("EvalCoord2d: %f %f\n", u, v));
 }
 
 void _gamma_EvalCoord2dv(const GLdouble *u)
 {
+    DEBUG_GLCMDS(("EvalCoord2dv: %f %f\n", u[0], u[1]));
 }
 
 void _gamma_EvalCoord2f(GLfloat u, GLfloat v)
 {
+    DEBUG_GLCMDS(("EvalCoord2f: %f %f\n", u, v));
 }
 
 void _gamma_EvalCoord2fv(const GLfloat *u)
 {
+    DEBUG_GLCMDS(("EvalCoord1fv: %f %f\n", u[0], u[1]));
 }
 
 void _gamma_EvalMesh1(GLenum mode, GLint i1, GLint i2)
 {
+    DEBUG_GLCMDS(("EvalMesh1: %d %d %d\n", mode, i1, i2));
 }
 
 void _gamma_EvalMesh2(GLenum mode, GLint i1, GLint i2, GLint j1, GLint j2)
@@ -1358,16 +1603,55 @@ void _gamma_Flush(void)
 void _gamma_Fogf(GLenum pname, GLfloat param)
 {
     DEBUG_GLCMDS(("Fogf: %04x %f\n", (int)pname, param));
+
+    switch (pname) {
+    case GL_FOG_DENSITY:
+    	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	WRITEF(gCCPriv->buf, FogDensity, param);
+	break;
+    }
 }
 
 void _gamma_Fogfv(GLenum pname, const GLfloat *params)
 {
+    int color;
+
     DEBUG_GLCMDS(("Fogfv: %04x %f\n", (int)pname, *params));
+
+    switch (pname) {
+    case GL_FOG_COLOR:
+	color = ((int)params[3]<<24) | ((int)params[2]<<16) |
+	 	((int)params[1]<<8)  | ((int)params[0]);
+    	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	WRITE(gCCPriv->buf, FogColor, color);
+	break;
+    }
 }
 
 void _gamma_Fogi(GLenum pname, GLint param)
 {
     DEBUG_GLCMDS(("Fogi: %04x %d\n", (int)pname, (int)param));
+
+    gCCPriv->GeometryMode &= ~GM_FogMask;
+
+    switch (pname) {
+    case GL_FOG_MODE:
+	switch (param) {
+	case GL_EXP:
+	    gCCPriv->GeometryMode |= GM_FogExp;
+	    break;
+	case GL_EXP2:
+	    gCCPriv->GeometryMode |= GM_FogExpSquared;
+	    break;
+	case GL_LINEAR:
+	    gCCPriv->GeometryMode |= GM_FogLinear;
+	    break;
+	}
+ 	break;
+    }
+
+    CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+    WRITE(gCCPriv->buf, GeometryMode, gCCPriv->GeometryMode);
 }
 
 void _gamma_Fogiv(GLenum pname, const GLint *params)
@@ -1378,6 +1662,16 @@ void _gamma_Fogiv(GLenum pname, const GLint *params)
 void _gamma_FrontFace(GLenum mode)
 {
     DEBUG_GLCMDS(("FrontFace: %04x\n", (int)mode));
+
+    gCCPriv->GeometryMode &= ~GM_FFMask;
+
+    if (mode == GL_CCW)
+	gCCPriv->GeometryMode |= GM_FrontFaceCCW;
+    else
+	gCCPriv->GeometryMode |= GM_FrontFaceCW;
+
+    CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+    WRITE(gCCPriv->buf, GeometryMode, gCCPriv->GeometryMode);
 }
 
 void _gamma_Frustum(GLdouble left, GLdouble right, GLdouble bottom, GLdouble top, GLdouble zNear, GLdouble zFar)
@@ -1463,14 +1757,24 @@ void _gamma_GetFloatv(GLenum val, GLfloat *f)
 void _gamma_GetIntegerv(GLenum val, GLint *i)
 {
     DEBUG_GLCMDS(("GetIntegerv: %04x\n", (int)val));
+
+    switch (val) {
+    case GL_MAX_TEXTURE_SIZE:
+	*i = 2048;
+	break;
+    }
 }
 
 void _gamma_GetLightfv(GLenum light, GLenum pname, GLfloat *params)
 {
+    DEBUG_GLCMDS(("GetLightfv: %04x %04x %f\n",
+		  (int)light, (int)pname, *params));
 }
 
 void _gamma_GetLightiv(GLenum light, GLenum pname, GLint *params)
 {
+    DEBUG_GLCMDS(("GetLightiv: %04x %04x %d\n",
+		  (int)light, (int)pname, *params));
 }
 
 void _gamma_GetMapdv(GLenum target, GLenum query, GLdouble *v)
@@ -1517,7 +1821,7 @@ void _gamma_GetPolygonStipple(GLubyte *mask)
 const GLubyte *_gamma_GetString(GLenum name)
 {
     static unsigned char vendor[] = "Precision Insight, Inc.";
-    static unsigned char renderer[] = "DRI Glint-Gamma 20000228";
+    static unsigned char renderer[] = "DRI Glint-Gamma 20000605";
     static unsigned char version[] = "1.1";
     static unsigned char ext[] = "";
 
@@ -1654,6 +1958,12 @@ GLboolean _gamma_IsEnabled(GLenum cap)
 {
     DEBUG_GLCMDS(("IsEnabled: %04x\n", (int)cap));
 
+    switch (cap) {
+    case GL_LIGHTING:
+	return ((gCCPriv->LightingMode & LightingModeEnable)?GL_TRUE:GL_FALSE);
+        break;
+    }
+
     return GL_TRUE;
 }
 
@@ -1673,10 +1983,44 @@ GLboolean _gamma_IsTexture(GLuint texture)
 
 void _gamma_LightModelf(GLenum pname, GLfloat param)
 {
+    DEBUG_GLCMDS(("LightModelf: %04x %f\n",
+		  (int)pname, param));
 }
 
 void _gamma_LightModelfv(GLenum pname, const GLfloat *params)
 {
+    DEBUG_GLCMDS(("LightModelfv: %04x %f\n",
+		  (int)pname, *params));
+
+    switch (pname) {
+    case GL_LIGHT_MODEL_AMBIENT:
+	CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
+	/* We don't do alpha */
+	WRITEF(gCCPriv->buf, SceneAmbientColorBlue, params[2]);
+	WRITEF(gCCPriv->buf, SceneAmbientColorGreen, params[1]);
+	WRITEF(gCCPriv->buf, SceneAmbientColorRed, params[0]);
+	break;
+    case GL_LIGHT_MODEL_LOCAL_VIEWER:
+	if ((int)params[0] != 0)
+	    gCCPriv->LightingMode |= LightingModeLocalViewer;
+	else
+	    gCCPriv->LightingMode &= ~LightingModeLocalViewer;
+	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	WRITE(gCCPriv->buf, LightingMode, gCCPriv->LightingMode);
+	break;
+    case GL_LIGHT_MODEL_TWO_SIDE:
+	if ((int)params[0] != 0) {
+	    gCCPriv->LightingMode |= LightingModeTwoSides;
+	    gCCPriv->MaterialMode |= MaterialModeTwoSides;
+	} else {
+	    gCCPriv->LightingMode &= ~LightingModeTwoSides;
+	    gCCPriv->MaterialMode &= ~MaterialModeTwoSides;
+	}
+	CHECK_DMA_BUFFER(gCC, gCCPriv, 2);
+	WRITE(gCCPriv->buf, LightingMode, gCCPriv->LightingMode);
+	WRITE(gCCPriv->buf, MaterialMode, gCCPriv->MaterialMode);
+	break;
+    }
 }
 
 void _gamma_LightModeli(GLenum pname, GLint param)
@@ -1689,10 +2033,173 @@ void _gamma_LightModeliv(GLenum pname, const GLint *params)
 
 void _gamma_Lightf(GLenum light, GLenum pname, GLfloat param)
 {
+    DEBUG_GLCMDS(("Lightf: %04x %04x %f\n",
+		  (int)light, (int)pname, param));
 }
 
 void _gamma_Lightfv(GLenum light, GLenum pname, const GLfloat *params)
 {
+    GLfloat l,x,y,z;
+    DEBUG_GLCMDS(("Lightfv: %04x %04x %f\n",
+		  (int)light, (int)pname, *params));
+
+    switch(light) {
+    case GL_LIGHT0:
+	switch (pname) {
+	case GL_AMBIENT:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
+	    /* We don't do alpha */
+	    WRITEF(gCCPriv->buf, Light0AmbientIntensityBlue, params[2]);
+	    WRITEF(gCCPriv->buf, Light0AmbientIntensityGreen, params[1]);
+	    WRITEF(gCCPriv->buf, Light0AmbientIntensityRed, params[0]);
+	    break;
+	case GL_DIFFUSE:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
+	    /* We don't do alpha */
+	    WRITEF(gCCPriv->buf, Light0DiffuseIntensityBlue, params[2]);
+	    WRITEF(gCCPriv->buf, Light0DiffuseIntensityGreen, params[1]);
+	    WRITEF(gCCPriv->buf, Light0DiffuseIntensityRed, params[0]);
+	    break;
+	case GL_SPECULAR:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
+	    /* We don't do alpha */
+	    WRITEF(gCCPriv->buf, Light0SpecularIntensityBlue, params[2]);
+	    WRITEF(gCCPriv->buf, Light0SpecularIntensityGreen, params[1]);
+	    WRITEF(gCCPriv->buf, Light0SpecularIntensityRed, params[0]);
+	    break;
+	case GL_POSITION:
+    	    /* Normalize <x,y,z> */
+	    x = params[0]; y = params[1]; z = params[2];
+	    l = sqrt(x*x + y*y + z*z);
+	    x /= l;
+	    y /= l;
+	    z /= l;
+	    if ((int)params[3] != 0) {
+		gCCPriv->Light0Mode |= Light0ModeAttenuation;
+		gCCPriv->Light0Mode |= Light0ModeLocal;
+	    } else {
+		gCCPriv->Light0Mode &= ~Light0ModeAttenuation;
+		gCCPriv->Light0Mode &= ~Light0ModeLocal;
+	    }
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 5);
+	    WRITE(gCCPriv->buf, Light0Mode, gCCPriv->Light0Mode);
+	    WRITEF(gCCPriv->buf, Light0PositionW, params[3]);
+	    WRITEF(gCCPriv->buf, Light0PositionZ, z);
+	    WRITEF(gCCPriv->buf, Light0PositionY, y);
+	    WRITEF(gCCPriv->buf, Light0PositionX, x);
+	    break;
+	case GL_SPOT_DIRECTION:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
+	    WRITEF(gCCPriv->buf, Light0SpotlightDirectionZ, params[3]);
+	    WRITEF(gCCPriv->buf, Light0SpotlightDirectionY, params[2]);
+	    WRITEF(gCCPriv->buf, Light0SpotlightDirectionX, params[1]);
+	    /* WRITEF(gCCPriv->buf, Light0SpotlightDirectionW, params[0]); */
+	    break;
+	case GL_SPOT_EXPONENT:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	    WRITEF(gCCPriv->buf, Light0SpotlightExponent, params[0]);
+	    break;
+	case GL_SPOT_CUTOFF:
+	    if ((int)params[0] != -1) 
+		gCCPriv->Light0Mode |= Light0ModeSpotLight;
+	    else
+		gCCPriv->Light0Mode &= ~Light0ModeSpotLight;
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 2);
+	    WRITE(gCCPriv->buf, Light0Mode, gCCPriv->Light0Mode);
+	    WRITEF(gCCPriv->buf, Light0CosSpotlightCutoffAngle, params[0]);
+	    break;
+	case GL_CONSTANT_ATTENUATION:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	    WRITEF(gCCPriv->buf, Light0ConstantAttenuation, params[0]);
+	    break;
+	case GL_LINEAR_ATTENUATION:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	    WRITEF(gCCPriv->buf, Light0LinearAttenuation, params[0]);
+	    break;
+	case GL_QUADRATIC_ATTENUATION:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	    WRITEF(gCCPriv->buf, Light0QuadraticAttenuation, params[0]);
+	    break;
+	}
+	break;
+    case GL_LIGHT1:
+	switch (pname) {
+	case GL_AMBIENT:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
+	    /* We don't do alpha */
+	    WRITEF(gCCPriv->buf, Light1AmbientIntensityBlue, params[2]);
+	    WRITEF(gCCPriv->buf, Light1AmbientIntensityGreen, params[1]);
+	    WRITEF(gCCPriv->buf, Light1AmbientIntensityRed, params[0]);
+	    break;
+	case GL_DIFFUSE:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
+	    /* We don't do alpha */
+	    WRITEF(gCCPriv->buf, Light1DiffuseIntensityBlue, params[2]);
+	    WRITEF(gCCPriv->buf, Light1DiffuseIntensityGreen, params[1]);
+	    WRITEF(gCCPriv->buf, Light1DiffuseIntensityRed, params[0]);
+	    break;
+	case GL_SPECULAR:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
+	    /* We don't do alpha */
+	    WRITEF(gCCPriv->buf, Light1SpecularIntensityBlue, params[2]);
+	    WRITEF(gCCPriv->buf, Light1SpecularIntensityGreen, params[1]);
+	    WRITEF(gCCPriv->buf, Light1SpecularIntensityRed, params[0]);
+	    break;
+	case GL_POSITION:
+    	    /* Normalize <x,y,z> */
+	    x = params[0]; y = params[1]; z = params[2];
+	    l = sqrt(x*x + y*y + z*z);
+	    x /= l;
+	    y /= l;
+	    z /= l;
+	    if ((int)params[3] != 0) {
+		gCCPriv->Light1Mode |= Light1ModeAttenuation;
+		gCCPriv->Light1Mode |= Light1ModeLocal;
+	    } else {
+		gCCPriv->Light0Mode &= ~Light0ModeAttenuation;
+		gCCPriv->Light0Mode &= ~Light0ModeLocal;
+	    }
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 5);
+	    WRITE(gCCPriv->buf, Light1Mode, gCCPriv->Light1Mode);
+	    WRITEF(gCCPriv->buf, Light1PositionW, params[3]);
+	    WRITEF(gCCPriv->buf, Light1PositionZ, z);
+	    WRITEF(gCCPriv->buf, Light1PositionY, y);
+	    WRITEF(gCCPriv->buf, Light1PositionX, x);
+	    break;
+	case GL_SPOT_DIRECTION:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
+	    WRITEF(gCCPriv->buf, Light1SpotlightDirectionZ, params[3]);
+	    WRITEF(gCCPriv->buf, Light1SpotlightDirectionY, params[2]);
+	    WRITEF(gCCPriv->buf, Light1SpotlightDirectionX, params[1]);
+	    /* WRITEF(gCCPriv->buf, Light1SpotlightDirectionW, params[0]); */
+	    break;
+	case GL_SPOT_EXPONENT:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	    WRITEF(gCCPriv->buf, Light1SpotlightExponent, params[0]);
+	    break;
+	case GL_SPOT_CUTOFF:
+	    if ((int)params[0] != -1) 
+		gCCPriv->Light1Mode |= Light1ModeSpotLight;
+	    else
+		gCCPriv->Light1Mode &= ~Light1ModeSpotLight;
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 2);
+	    WRITE(gCCPriv->buf, Light1Mode, gCCPriv->Light1Mode);
+	    WRITEF(gCCPriv->buf, Light1CosSpotlightCutoffAngle, params[0]);
+	    break;
+	case GL_CONSTANT_ATTENUATION:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	    WRITEF(gCCPriv->buf, Light1ConstantAttenuation, params[0]);
+	    break;
+	case GL_LINEAR_ATTENUATION:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	    WRITEF(gCCPriv->buf, Light1LinearAttenuation, params[0]);
+	    break;
+	case GL_QUADRATIC_ATTENUATION:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	    WRITEF(gCCPriv->buf, Light1QuadraticAttenuation, params[0]);
+	    break;
+	}
+    }
 }
 
 void _gamma_Lighti(GLenum light, GLenum pname, GLint param)
@@ -1817,6 +2324,104 @@ void _gamma_Materialfv(GLenum face, GLenum pname, const GLfloat *params)
 {
     DEBUG_GLCMDS(("Materialfv: %04x %04x %f\n",
 		  (int)face, (int)pname, *params));
+
+    gCCPriv->MaterialMode |= MaterialModeEnable;
+    CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+    WRITE(gCCPriv->buf, MaterialMode, gCCPriv->MaterialMode);
+
+    if ((face == GL_FRONT) || (face == GL_FRONT_AND_BACK)) {
+	switch (pname) {
+	case GL_AMBIENT:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
+	    WRITEF(gCCPriv->buf, FrontAmbientColorBlue, params[2]);
+	    WRITEF(gCCPriv->buf, FrontAmbientColorGreen, params[1]);
+	    WRITEF(gCCPriv->buf, FrontAmbientColorRed, params[0]);
+	    break;
+	case GL_DIFFUSE:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 4);
+	    WRITEF(gCCPriv->buf, FrontAlpha, params[3]);
+	    WRITEF(gCCPriv->buf, FrontDiffuseColorBlue, params[2]);
+	    WRITEF(gCCPriv->buf, FrontDiffuseColorGreen, params[1]);
+	    WRITEF(gCCPriv->buf, FrontDiffuseColorRed, params[0]);
+	    break;
+	case GL_SPECULAR:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
+	    WRITEF(gCCPriv->buf, FrontSpecularColorBlue, params[2]);
+	    WRITEF(gCCPriv->buf, FrontSpecularColorGreen, params[1]);
+	    WRITEF(gCCPriv->buf, FrontSpecularColorRed, params[0]);
+	    break;
+	case GL_EMISSION:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
+	    WRITEF(gCCPriv->buf, FrontEmissiveColorBlue, params[2]);
+	    WRITEF(gCCPriv->buf, FrontEmissiveColorGreen, params[1]);
+	    WRITEF(gCCPriv->buf, FrontEmissiveColorRed, params[0]);
+	    break;
+	case GL_SHININESS:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	    WRITEF(gCCPriv->buf, FrontSpecularExponent, params[0]);
+	    break;
+	case GL_AMBIENT_AND_DIFFUSE:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 7);
+	    WRITEF(gCCPriv->buf, FrontAmbientColorBlue, params[2]);
+	    WRITEF(gCCPriv->buf, FrontAmbientColorGreen, params[1]);
+	    WRITEF(gCCPriv->buf, FrontAmbientColorRed, params[0]);
+	    WRITEF(gCCPriv->buf, FrontAlpha, params[3]);
+	    WRITEF(gCCPriv->buf, FrontDiffuseColorBlue, params[2]);
+	    WRITEF(gCCPriv->buf, FrontDiffuseColorGreen, params[1]);
+	    WRITEF(gCCPriv->buf, FrontDiffuseColorRed, params[0]);
+	    break;
+	case GL_COLOR_INDEXES:
+	    /* NOT_DONE */
+	    break;
+	}
+    }
+
+    if ((face == GL_BACK) || (face == GL_FRONT_AND_BACK)) {
+	switch (pname) {
+	case GL_AMBIENT:
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
+	WRITEF(gCCPriv->buf, BackAmbientColorBlue, params[2]);
+	WRITEF(gCCPriv->buf, BackAmbientColorGreen, params[1]);
+	WRITEF(gCCPriv->buf, BackAmbientColorRed, params[0]);
+	break;
+    case GL_DIFFUSE:
+	CHECK_DMA_BUFFER(gCC, gCCPriv, 4);
+	WRITEF(gCCPriv->buf, BackAlpha, params[3]);
+	WRITEF(gCCPriv->buf, BackDiffuseColorBlue, params[2]);
+	WRITEF(gCCPriv->buf, BackDiffuseColorGreen, params[1]);
+	WRITEF(gCCPriv->buf, BackDiffuseColorRed, params[0]);
+	break;
+    case GL_SPECULAR:
+	CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
+	WRITEF(gCCPriv->buf, BackSpecularColorBlue, params[2]);
+	WRITEF(gCCPriv->buf, BackSpecularColorGreen, params[1]);
+	WRITEF(gCCPriv->buf, BackSpecularColorRed, params[0]);
+	break;
+    case GL_EMISSION:
+	CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
+	WRITEF(gCCPriv->buf, BackEmissiveColorBlue, params[2]);
+	WRITEF(gCCPriv->buf, BackEmissiveColorGreen, params[1]);
+	WRITEF(gCCPriv->buf, BackEmissiveColorRed, params[0]);
+	break;
+    case GL_SHININESS:
+	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	WRITEF(gCCPriv->buf, BackSpecularExponent, params[0]);
+	break;
+    case GL_AMBIENT_AND_DIFFUSE:
+	CHECK_DMA_BUFFER(gCC, gCCPriv, 7);
+	WRITEF(gCCPriv->buf, BackAmbientColorBlue, params[2]);
+	WRITEF(gCCPriv->buf, BackAmbientColorGreen, params[1]);
+	WRITEF(gCCPriv->buf, BackAmbientColorRed, params[0]);
+	WRITEF(gCCPriv->buf, BackAlpha, params[3]);
+	WRITEF(gCCPriv->buf, BackDiffuseColorBlue, params[2]);
+	WRITEF(gCCPriv->buf, BackDiffuseColorGreen, params[1]);
+	WRITEF(gCCPriv->buf, BackDiffuseColorRed, params[0]);
+	break;
+    case GL_COLOR_INDEXES:
+	/* NOT_DONE */
+	break;
+    }
+    }
 }
 
 void _gamma_Materiali(GLenum face, GLenum pname, GLint param)
@@ -1837,9 +2442,9 @@ void _gamma_MatrixMode(GLenum mode)
 
     switch (mode) {
     case GL_TEXTURE:
-	/* Eanble the Texture transform */
+	/* Enable the Texture transform */
 	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
-	WRITE(gCCPriv->buf, TransformModeOr, 0x00000010);
+	WRITE(gCCPriv->buf, TransformModeOr, XM_XformTexture);
     case GL_MODELVIEW:
     case GL_PROJECTION:
         gCCPriv->MatrixMode = mode;
@@ -1878,52 +2483,114 @@ void _gamma_NewList(GLuint list, GLenum mode)
 
 void _gamma_Normal3b(GLbyte nx, GLbyte ny, GLbyte nz)
 {
+    GLfloat x,y,z;
+
     DEBUG_GLCMDS(("Normal3b: %d %d %d\n", nx, ny, nz));
+
+    x = BYTE_TO_FLOAT(nx);
+    y = BYTE_TO_FLOAT(ny);
+    z = BYTE_TO_FLOAT(nz);
+   
+    _gamma_Normal3f(x,y,z);
 }
 
 void _gamma_Normal3bv(const GLbyte *v)
 {
+    GLfloat p[3];
+
     DEBUG_GLCMDS(("Normal3bv: %d %d %d\n", v[0], v[1], v[2]));
+
+    p[0] = BYTE_TO_FLOAT(v[0]);
+    p[1] = BYTE_TO_FLOAT(v[1]);
+    p[2] = BYTE_TO_FLOAT(v[2]);
+   
+    _gamma_Normal3fv(p);
 }
 
 void _gamma_Normal3d(GLdouble nx, GLdouble ny, GLdouble nz)
 {
     DEBUG_GLCMDS(("Normal3d: %f %f %f\n", nx, ny, nz));
+
+    _gamma_Normal3f((GLfloat)nx,(GLfloat)ny,(GLfloat)nz);
 }
 
 void _gamma_Normal3dv(const GLdouble *v)
 {
     DEBUG_GLCMDS(("Normal3dv: %f %f %f\n", v[0], v[1], v[2]));
+
+    _gamma_Normal3fv((GLfloat*)v);
 }
 
 void _gamma_Normal3f(GLfloat nx, GLfloat ny, GLfloat nz)
 {
     DEBUG_GLCMDS(("Normal3f: %f %f %f\n", nx, ny, nz));
+
+    CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
+    WRITEF(gCCPriv->buf, Nz, nz);
+    WRITEF(gCCPriv->buf, Ny, ny);
+    WRITEF(gCCPriv->buf, Nx, nx);
 }
 
 void _gamma_Normal3fv(const GLfloat *v)
 {
     DEBUG_GLCMDS(("Normal3fv: %f %f %f\n", v[0], v[1], v[2]));
+
+    CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
+    WRITEF(gCCPriv->buf, Nz, v[2]);
+    WRITEF(gCCPriv->buf, Ny, v[1]);
+    WRITEF(gCCPriv->buf, Nx, v[0]);
 }
 
 void _gamma_Normal3i(GLint nx, GLint ny, GLint nz)
 {
+    GLfloat x,y,z;
+
     DEBUG_GLCMDS(("Normal3i: %d %d %d\n", (int)nx, (int)ny, (int)nz));
+
+    x = INT_TO_FLOAT(nx);
+    y = INT_TO_FLOAT(ny);
+    z = INT_TO_FLOAT(nz);
+
+    _gamma_Normal3f(x,y,z);
 }
 
 void _gamma_Normal3iv(const GLint *v)
 {
+    GLfloat p[3];
+
     DEBUG_GLCMDS(("Normal3iv: %d %d %d\n", (int)v[0], (int)v[1], (int)v[2]));
+
+    p[0] = INT_TO_FLOAT(v[0]);
+    p[1] = INT_TO_FLOAT(v[1]);
+    p[2] = INT_TO_FLOAT(v[2]);
+
+    _gamma_Normal3fv(p);
 }
 
 void _gamma_Normal3s(GLshort nx, GLshort ny, GLshort nz)
 {
+    GLfloat x,y,z;
+
     DEBUG_GLCMDS(("Normal3s: %d %d %d\n", nx, ny, nz));
+
+    x = SHORT_TO_FLOAT(nx);
+    y = SHORT_TO_FLOAT(ny);
+    z = SHORT_TO_FLOAT(nz);
+
+    _gamma_Normal3f(x,y,z);
 }
 
 void _gamma_Normal3sv(const GLshort *v)
 {
+    GLfloat p[3];
+
     DEBUG_GLCMDS(("Normal3sv: %d %d %d\n", v[0], v[1], v[2]));
+
+    p[0] = SHORT_TO_FLOAT(v[0]);
+    p[1] = SHORT_TO_FLOAT(v[1]);
+    p[2] = SHORT_TO_FLOAT(v[2]);
+
+    _gamma_Normal3fv(p);
 }
 
 void _gamma_NormalPointer(GLenum type, GLsizei stride, const GLvoid *pointer)
@@ -2020,6 +2687,56 @@ void _gamma_PointSize(GLfloat size)
 void _gamma_PolygonMode(GLenum face, GLenum mode)
 {
     DEBUG_GLCMDS(("PolygonMode: %04x %04x\n", (int)face, (int)mode));
+
+    gCCPriv->GeometryMode &= ~GM_FB_PolyMask;
+
+    switch (mode) {
+    case GL_FILL:
+	switch (face) {
+	case GL_FRONT:
+	    gCCPriv->GeometryMode |= GM_FrontPolyFill;
+	    break;
+	case GL_BACK:
+	    gCCPriv->GeometryMode |= GM_BackPolyFill;
+	    break;
+	case GL_FRONT_AND_BACK:
+	    gCCPriv->GeometryMode |= GM_FrontPolyFill;
+	    gCCPriv->GeometryMode |= GM_BackPolyFill;
+	    break;
+	}
+	break;
+    case GL_LINE:
+	switch (face) {
+	case GL_FRONT:
+	    gCCPriv->GeometryMode |= GM_FrontPolyLine;
+	    break;
+	case GL_BACK:
+	    gCCPriv->GeometryMode |= GM_BackPolyLine;
+	    break;
+	case GL_FRONT_AND_BACK:
+	    gCCPriv->GeometryMode |= GM_FrontPolyLine;
+	    gCCPriv->GeometryMode |= GM_BackPolyLine;
+	    break;
+	}
+	break;
+    case GL_POINT:
+	switch (face) {
+	case GL_FRONT:
+	    gCCPriv->GeometryMode |= GM_FrontPolyPoint;
+	    break;
+	case GL_BACK:
+	    gCCPriv->GeometryMode |= GM_BackPolyPoint;
+	    break;
+	case GL_FRONT_AND_BACK:
+	    gCCPriv->GeometryMode |= GM_FrontPolyPoint;
+	    gCCPriv->GeometryMode |= GM_BackPolyPoint;
+	    break;
+	}
+	break;
+    }
+
+    CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+    WRITE(gCCPriv->buf, GeometryMode, gCCPriv->GeometryMode);
 }
 
 void _gamma_PolygonOffset(GLfloat factor, GLfloat units)
@@ -2272,6 +2989,15 @@ void _gamma_Rectdv(const GLdouble *v1, const GLdouble *v2)
 void _gamma_Rectf(GLfloat x1, GLfloat y1, GLfloat x2, GLfloat y2)
 {
     DEBUG_GLCMDS(("Rectf: %f %f %f %f\n", x1, y1, x2, y2));
+ 
+    /* This should be done with Gamma's Rectangle engine - later */
+
+    _gamma_Begin(GL_POLYGON);
+    _gamma_Vertex2f(x1,y1);
+    _gamma_Vertex2f(x2,y1);
+    _gamma_Vertex2f(x2,y2);
+    _gamma_Vertex2f(x1,y2);
+    _gamma_End();
 }
 
 void _gamma_Rectfv(const GLfloat *v1, const GLfloat *v2)
@@ -2474,51 +3200,87 @@ void _gamma_StencilOp(GLenum fail, GLenum zfail, GLenum zpass)
 void _gamma_TexCoord1d(GLdouble s)
 {
     DEBUG_GLCMDS(("TexCoord1d: %f\n", s));
+
+    _gamma_TexCoord1f((GLfloat)s);
 }
 
 void _gamma_TexCoord1dv(const GLdouble *v)
 {
     DEBUG_GLCMDS(("TexCoord1dv: %f\n", *v));
+
+    _gamma_TexCoord1fv((GLfloat*)v);
 }
 
 void _gamma_TexCoord1f(GLfloat s)
 {
     DEBUG_GLCMDS(("TexCoord1f: %f\n", s));
+
+    WRITEF(gCCPriv->buf, Ts1, s);
 }
 
 void _gamma_TexCoord1fv(const GLfloat *v)
 {
     DEBUG_GLCMDS(("TexCoord1fv: %f\n", *v));
+
+    WRITEF(gCCPriv->buf, Ts1, v[0]);
 }
 
 void _gamma_TexCoord1i(GLint s)
 {
+    GLfloat x;
+
     DEBUG_GLCMDS(("TexCoord1i: %d\n", (int)s));
+
+    x = INT_TO_FLOAT(s);
+
+    _gamma_TexCoord1f(x);
 }
 
 void _gamma_TexCoord1iv(const GLint *v)
 {
+    GLfloat p[1];
+
     DEBUG_GLCMDS(("TexCoord1iv: %d\n", (int)*v));
+
+    p[0] = INT_TO_FLOAT(v[0]);
+
+    _gamma_TexCoord1fv(p);
 }
 
 void _gamma_TexCoord1s(GLshort s)
 {
+    GLfloat x;
+
     DEBUG_GLCMDS(("TexCoord1s: %d\n", s));
+
+    x = SHORT_TO_FLOAT(s);
+
+    _gamma_TexCoord1f(x);
 }
 
 void _gamma_TexCoord1sv(const GLshort *v)
 {
+    GLfloat p[1];
+
     DEBUG_GLCMDS(("TexCoord1sv: %d\n", *v));
+
+    p[0] = SHORT_TO_FLOAT(v[0]);
+
+    _gamma_TexCoord1fv(p);
 }
 
 void _gamma_TexCoord2d(GLdouble s, GLdouble t)
 {
     DEBUG_GLCMDS(("TexCoord2d: %f %f\n", s, t));
+
+    _gamma_TexCoord2f((GLfloat)s,(GLfloat)t);
 }
 
 void _gamma_TexCoord2dv(const GLdouble *v)
 {
     DEBUG_GLCMDS(("TexCoord2dv: %f %f\n", v[0], v[1]));
+
+    _gamma_TexCoord2fv((GLfloat*)v);
 }
 
 void _gamma_TexCoord2f(GLfloat s, GLfloat t)
@@ -2533,107 +3295,240 @@ void _gamma_TexCoord2f(GLfloat s, GLfloat t)
 void _gamma_TexCoord2fv(const GLfloat *v)
 {
     DEBUG_GLCMDS(("TexCoord2fv: %f %f\n", v[0], v[1]));
+
+    CHECK_DMA_BUFFER(gCC, gCCPriv, 2);
+    WRITEF(gCCPriv->buf, Tt2, v[1]);
+    WRITEF(gCCPriv->buf, Ts2, v[0]);
 }
 
 void _gamma_TexCoord2i(GLint s, GLint t)
 {
+    GLfloat x,y;
+
     DEBUG_GLCMDS(("TexCoord2i: %d %d\n", (int)s, (int)t));
+
+    x = INT_TO_FLOAT(s);
+    y = INT_TO_FLOAT(t);
+
+    _gamma_TexCoord2f(x,y);
 }
 
 void _gamma_TexCoord2iv(const GLint *v)
 {
+    GLfloat p[2];
+
     DEBUG_GLCMDS(("TexCoord2iv: %d %d\n", (int)v[0], (int)v[1]));
+
+    p[0] = INT_TO_FLOAT(v[0]);
+    p[1] = INT_TO_FLOAT(v[1]);
+
+    _gamma_TexCoord2fv(p);
 }
 
 void _gamma_TexCoord2s(GLshort s, GLshort t)
 {
+    GLfloat x,y;
+
     DEBUG_GLCMDS(("TexCoord2s: %d %d\n", s, t));
+
+    x = SHORT_TO_FLOAT(s);
+    y = SHORT_TO_FLOAT(t);
+
+    _gamma_TexCoord2f(x,y);
 }
 
 void _gamma_TexCoord2sv(const GLshort *v)
 {
+    GLfloat p[2];
+
     DEBUG_GLCMDS(("TexCoord2sv: %d %d\n", v[0], v[1]));
+
+    p[0] = SHORT_TO_FLOAT(v[0]);
+    p[1] = SHORT_TO_FLOAT(v[1]);
+
+    _gamma_TexCoord2fv(p);
 }
 
 void _gamma_TexCoord3d(GLdouble s, GLdouble t, GLdouble r)
 {
     DEBUG_GLCMDS(("TexCoord3d: %f %f %f\n", s, t, r));
+
+    _gamma_TexCoord3f((GLfloat)s,(GLfloat)t,(GLfloat)r);
 }
 
 void _gamma_TexCoord3dv(const GLdouble *v)
 {
     DEBUG_GLCMDS(("TexCoord3dv: %f %f %f\n", v[0], v[1], v[2]));
+
+    _gamma_TexCoord3fv((GLfloat*)v);
 }
 
 void _gamma_TexCoord3f(GLfloat s, GLfloat t, GLfloat r)
 {
     DEBUG_GLCMDS(("TexCoord3f: %f %f %f\n", s, t, r));
+
+    _gamma_TexCoord4f(s,t,r,1.0f);
 }
 
 void _gamma_TexCoord3fv(const GLfloat *v)
 {
+    GLfloat p[4];
+
     DEBUG_GLCMDS(("TexCoord3fv: %f %f %f\n", v[0], v[1], v[2]));
+
+    p[0] = v[0];
+    p[1] = v[1];
+    p[2] = v[2];
+    p[3] = 1.0f;
+
+    _gamma_TexCoord4fv(p);
 }
 
 void _gamma_TexCoord3i(GLint s, GLint t, GLint r)
 {
+    GLfloat x,y,z;
+
     DEBUG_GLCMDS(("TexCoord3i: %d %d %d\n", (int)s, (int)t, (int)r));
+
+    x = INT_TO_FLOAT(s);
+    y = INT_TO_FLOAT(t);
+    z = INT_TO_FLOAT(r);
+
+    _gamma_TexCoord4f(x,y,z,1.0f);
 }
 
 void _gamma_TexCoord3iv(const GLint *v)
 {
+    GLfloat p[4];
+
     DEBUG_GLCMDS(("TexCoord3iv: %d %d %d\n", (int)v[0], (int)v[1], (int)v[2]));
+
+    p[0] = INT_TO_FLOAT(v[0]);
+    p[1] = INT_TO_FLOAT(v[1]);
+    p[2] = INT_TO_FLOAT(v[2]);
+    p[3] = 1.0f;
+
+    _gamma_TexCoord4fv(p);
 }
 
 void _gamma_TexCoord3s(GLshort s, GLshort t, GLshort r)
 {
+    GLfloat x,y,z;
+
     DEBUG_GLCMDS(("TexCoord3s: %d %d %d\n", s, t, r));
+  
+    x = SHORT_TO_FLOAT(s);
+    y = SHORT_TO_FLOAT(t);
+    z = SHORT_TO_FLOAT(r);
+
+    _gamma_TexCoord4f(x,y,z,1.0f);
 }
 
 void _gamma_TexCoord3sv(const GLshort *v)
 {
+    GLfloat p[4];
+
     DEBUG_GLCMDS(("TexCoord3sv: %d %d %d\n", v[0], v[1], v[2]));
+
+    p[0] = SHORT_TO_FLOAT(v[0]);
+    p[1] = SHORT_TO_FLOAT(v[1]);
+    p[2] = SHORT_TO_FLOAT(v[2]);
+    p[3] = 1.0f;
+
+    _gamma_TexCoord4fv(p);
 }
 
 void _gamma_TexCoord4d(GLdouble s, GLdouble t, GLdouble r, GLdouble q)
 {
     DEBUG_GLCMDS(("TexCoord4d: %f %f %f %f\n", s, t, r, q));
+
+    _gamma_TexCoord4f((GLfloat)s,(GLfloat)t,(GLfloat)r,(GLfloat)q);
 }
 
 void _gamma_TexCoord4dv(const GLdouble *v)
 {
     DEBUG_GLCMDS(("TexCoord4dv: %f %f %f %f\n", v[0], v[1], v[2], v[3]));
+
+    _gamma_TexCoord4fv((GLfloat*)v);
 }
 
 void _gamma_TexCoord4f(GLfloat s, GLfloat t, GLfloat r, GLfloat q)
 {
     DEBUG_GLCMDS(("TexCoord4f: %f %f %f %f\n", s, t, r, q));
+
+    CHECK_DMA_BUFFER(gCC, gCCPriv, 4);
+    WRITEF(gCCPriv->buf, Tq4, q);
+    WRITEF(gCCPriv->buf, Tr4, r);
+    WRITEF(gCCPriv->buf, Tt4, t);
+    WRITEF(gCCPriv->buf, Ts4, s);
 }
 
 void _gamma_TexCoord4fv(const GLfloat *v)
 {
     DEBUG_GLCMDS(("TexCoord4fv: %f %f %f %f\n", v[0], v[1], v[2], v[3]));
+
+    CHECK_DMA_BUFFER(gCC, gCCPriv, 4);
+    WRITEF(gCCPriv->buf, Tq4, v[3]);
+    WRITEF(gCCPriv->buf, Tr4, v[2]);
+    WRITEF(gCCPriv->buf, Tt4, v[1]);
+    WRITEF(gCCPriv->buf, Ts4, v[0]);
 }
 
 void _gamma_TexCoord4i(GLint s, GLint t, GLint r, GLint q)
 {
+    GLfloat x,y,z,a;
+
     DEBUG_GLCMDS(("TexCoord4i: %d %d %d %d\n", (int)s, (int)t, (int)r, (int)q));
+
+    x = INT_TO_FLOAT(s);
+    y = INT_TO_FLOAT(t);
+    z = INT_TO_FLOAT(r);
+    a = INT_TO_FLOAT(q);
+
+    _gamma_TexCoord4f(x,y,z,a);
 }
 
 void _gamma_TexCoord4iv(const GLint *v)
 {
+    GLfloat p[4];
+
     DEBUG_GLCMDS(("TexCoord4iv: %d %d %d %d\n",
 		  (int)v[0], (int)v[1], (int)v[2], (int)v[3]));
+
+    p[0] = INT_TO_FLOAT(v[0]);
+    p[1] = INT_TO_FLOAT(v[1]);
+    p[2] = INT_TO_FLOAT(v[2]);
+    p[3] = INT_TO_FLOAT(v[3]);
+
+    _gamma_TexCoord4fv(p);
 }
 
 void _gamma_TexCoord4s(GLshort s, GLshort t, GLshort r, GLshort q)
 {
+    GLfloat x,y,z,a;
+
     DEBUG_GLCMDS(("TexCoord4s: %d %d %d %d\n", s, t, r, q));
+
+    x = SHORT_TO_FLOAT(s);
+    y = SHORT_TO_FLOAT(t);
+    z = SHORT_TO_FLOAT(r);
+    a = SHORT_TO_FLOAT(q);
+
+    _gamma_TexCoord4f(s,t,r,q);
 }
 
 void _gamma_TexCoord4sv(const GLshort *v)
 {
+    GLfloat p[4];
+
     DEBUG_GLCMDS(("TexCoord4sv: %d %d %d %d\n", v[0], v[1], v[2], v[3]));
+
+    p[0] = SHORT_TO_FLOAT(v[0]);
+    p[1] = SHORT_TO_FLOAT(v[1]);
+    p[2] = SHORT_TO_FLOAT(v[2]);
+    p[3] = SHORT_TO_FLOAT(v[3]);
+
+    _gamma_TexCoord4fv(p);
 }
 
 void _gamma_TexCoordPointer(GLint size, GLenum type, GLsizei stride, const GLvoid *pointer)
@@ -2678,6 +3573,33 @@ void _gamma_TexEnvfv(GLenum target, GLenum pname, const GLfloat *params)
 {
     DEBUG_GLCMDS(("TexEnvfv: %04x %04x %f\n",
 		  (int)target, (int)pname, *params));
+
+    if (target != GL_TEXTURE_ENV) {
+	/* ERROR !! */
+    }
+
+    switch (pname) {
+    case GL_TEXTURE_ENV_MODE:
+	gCCPriv->curTexObj->TextureColorMode &= ~TCM_ApplicationMask;
+	switch ((int)params[0]) {
+	case GL_MODULATE:
+	    gCCPriv->curTexObj->TextureColorMode |= TCM_Modulate;
+	    break;
+	case GL_DECAL:
+	    gCCPriv->curTexObj->TextureColorMode |= TCM_Decal;
+	    break;
+	case GL_BLEND:
+	    gCCPriv->curTexObj->TextureColorMode |= TCM_Blend;
+	    break;
+	case GL_REPLACE:
+	    gCCPriv->curTexObj->TextureColorMode |= TCM_Replace;
+	    break;
+	}
+    }
+
+    CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+    WRITE(gCCPriv->buf, TextureColorMode,
+	  gCCPriv->curTexObj->TextureColorMode);
 }
 
 void _gamma_TexEnvi(GLenum target, GLenum pname, GLint param)
@@ -2745,10 +3667,12 @@ void _gamma_TexImage2D(GLenum target, GLint level, GLint components, GLsizei wid
 
     if (target == GL_TEXTURE_1D) {
 	/* NOT_DONE */
+	DEBUG_ERROR(("TexImage2D: 1D requested - ERROR\n"));
     } else if (target == GL_TEXTURE_2D) {
 	/* NOT_DONE: The follow are not currently supported... */
-	if (border != 0 || format != GL_RGBA || type != GL_UNSIGNED_BYTE ||
+	if (border != 0 /* || format != GL_RGBA */ || type != GL_UNSIGNED_BYTE ||
 	    (components != 3 && components != 4)) {
+	    DEBUG_ERROR(("TexImage2D: 2D op not supported - ERROR\n"));
 	    return;
 	}
 
@@ -2790,7 +3714,11 @@ void _gamma_TexImage2D(GLenum target, GLint level, GLint components, GLsizei wid
 		break;
 	    }
 
+#if X_BYTE_ORDER == X_LITTLE_ENDIAN
 	    t->TextureFormat = (TF_LittleEndian |
+#else
+	    t->TextureFormat = (TF_BigEndian |
+#endif
 				TF_ColorOrder_BGR |
 				TF_Compnents_4 |
 				TF_OutputFmt_Texel);
@@ -2804,14 +3732,17 @@ void _gamma_TexImage2D(GLenum target, GLint level, GLint components, GLsizei wid
 	t->image[level] = driTMMInsertImage(gCCPriv->tmm,
 					    width, height, 1<<l2d,
 					    image, NULL);
+
 	if (!t->image[level]) {
 	    /* NOT_DONE: Handle error */
+	    DEBUG_ERROR(("TexImage2D: unable1\n"));
 	}
 
 	/* Make the new image resident (and all of the other mipmaps) */
-	if (!driTMMMakeImagesResident(gCCPriv->tmm, MIPMAP_LEVELS,
-				      t->image, addrs)) {
+	if (driTMMMakeImagesResident(gCCPriv->tmm, MIPMAP_LEVELS,
+				      t->image, addrs) < 0) {
 	    /* NOT_DONE: Handle error */
+	    DEBUG_ERROR(("TexImage2D: unable2\n"));
 	}
 
 	for (i = 0; i < MIPMAP_LEVELS; i++)
@@ -3010,6 +3941,123 @@ void _gamma_TexParameterfv(GLenum target, GLenum pname, const GLfloat *params)
 {
     DEBUG_GLCMDS(("TexParameterfv: %04x %04x %f\n",
 		  (int)target, (int)pname, *params));
+
+    if (target == GL_TEXTURE_1D) {
+	/* NOT_DONE */
+    } else if (target == GL_TEXTURE_2D) {
+	switch (pname) {
+	case GL_TEXTURE_MAG_FILTER:
+	    gCCPriv->curTexObj2D->TextureReadMode &= ~TRM_Mag_Mask;
+	    switch ((int)params[0]) {
+	    case GL_NEAREST:
+		gCCPriv->curTexObj2D->TextureReadMode |=
+		    TRM_Mag_Nearest;
+		break;
+	    case GL_LINEAR:
+		gCCPriv->curTexObj2D->TextureReadMode |=
+		    TRM_Mag_Linear;
+		break;
+	    default:
+		break;
+	    }
+
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
+	    WRITE(gCCPriv->buf, TextureReadMode,
+		  gCCPriv->curTexObj2D->TextureReadMode);
+	    break;
+	case GL_TEXTURE_MIN_FILTER:
+	    gCCPriv->curTexObj2D->TextureReadMode    &= ~TRM_Min_Mask;
+	    gCCPriv->curTexObj2D->TextureReadMode    |= TRM_MipMapEnable;
+	    gCCPriv->curTexObj2D->TextureAddressMode |= TAM_LODEnable;
+	    switch ((int)params[0]) {
+	    case GL_NEAREST:
+		gCCPriv->curTexObj2D->TextureReadMode |=
+		    TRM_Min_Nearest;
+		gCCPriv->curTexObj2D->TextureReadMode    &= ~TRM_MipMapEnable;
+		gCCPriv->curTexObj2D->TextureAddressMode &= ~TAM_LODEnable;
+		break;
+	    case GL_LINEAR:
+		gCCPriv->curTexObj2D->TextureReadMode |=
+		    TRM_Min_Linear;
+		gCCPriv->curTexObj2D->TextureReadMode    &= ~TRM_MipMapEnable;
+		gCCPriv->curTexObj2D->TextureAddressMode &= ~TAM_LODEnable;
+		break;
+	    case GL_NEAREST_MIPMAP_NEAREST:
+		gCCPriv->curTexObj2D->TextureReadMode |=
+		    TRM_Min_NearestMMNearest;
+		break;
+	    case GL_LINEAR_MIPMAP_NEAREST:
+		gCCPriv->curTexObj2D->TextureReadMode |=
+		    TRM_Min_NearestMMLinear;
+		break;
+	    case GL_NEAREST_MIPMAP_LINEAR:
+		gCCPriv->curTexObj2D->TextureReadMode |=
+		    TRM_Min_LinearMMNearest;
+		break;
+	    case GL_LINEAR_MIPMAP_LINEAR:
+		gCCPriv->curTexObj2D->TextureReadMode |=
+		    TRM_Min_LinearMMLinear;
+		break;
+	    default:
+		break;
+	    }
+
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 2);
+	    WRITE(gCCPriv->buf, TextureReadMode,
+		  gCCPriv->curTexObj2D->TextureReadMode);
+	    WRITE(gCCPriv->buf, TextureAddressMode,
+		  gCCPriv->curTexObj2D->TextureAddressMode);
+	    break;
+	case GL_TEXTURE_WRAP_S:
+	    gCCPriv->curTexObj2D->TextureAddressMode &= ~TAM_SWrap_Mask;
+	    gCCPriv->curTexObj2D->TextureReadMode    &= ~TRM_UWrap_Mask;
+	    switch ((int)params[0]) {
+	    case GL_CLAMP:
+		gCCPriv->curTexObj2D->TextureAddressMode |= TAM_SWrap_Clamp;
+		gCCPriv->curTexObj2D->TextureReadMode    |= TRM_UWrap_Clamp;
+		break;
+	    case GL_REPEAT:
+		gCCPriv->curTexObj2D->TextureAddressMode |= TAM_SWrap_Repeat;
+		gCCPriv->curTexObj2D->TextureReadMode    |= TRM_UWrap_Repeat;
+		break;
+	    default:
+		break;
+	    }
+
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 2);
+	    WRITE(gCCPriv->buf, TextureReadMode,
+		  gCCPriv->curTexObj2D->TextureReadMode);
+	    WRITE(gCCPriv->buf, TextureAddressMode,
+		  gCCPriv->curTexObj2D->TextureAddressMode);
+	    break;
+	case GL_TEXTURE_WRAP_T:
+	    gCCPriv->curTexObj2D->TextureAddressMode &= ~TAM_TWrap_Mask;
+	    gCCPriv->curTexObj2D->TextureReadMode    &= ~TRM_VWrap_Mask;
+	    switch ((int)params[0]) {
+	    case GL_CLAMP:
+		gCCPriv->curTexObj2D->TextureAddressMode |= TAM_TWrap_Clamp;
+		gCCPriv->curTexObj2D->TextureReadMode    |= TRM_VWrap_Clamp;
+		break;
+	    case GL_REPEAT:
+		gCCPriv->curTexObj2D->TextureAddressMode |= TAM_TWrap_Repeat;
+		gCCPriv->curTexObj2D->TextureReadMode    |= TRM_VWrap_Repeat;
+		break;
+	    default:
+		break;
+	    }
+
+	    CHECK_DMA_BUFFER(gCC, gCCPriv, 2);
+	    WRITE(gCCPriv->buf, TextureReadMode,
+		  gCCPriv->curTexObj2D->TextureReadMode);
+	    WRITE(gCCPriv->buf, TextureAddressMode,
+		  gCCPriv->curTexObj2D->TextureAddressMode);
+	    break;
+	default:
+	    break;
+	}
+    } else {
+	/* ERROR !! */
+    }
 }
 
 void _gamma_TexParameteri(GLenum target, GLenum pname, GLint param)
@@ -3054,8 +4102,8 @@ void _gamma_TexSubImage2D(GLenum target, GLint level, GLint xoffset, GLint yoffs
 	** format for the image that it is replacing.
 	*/
 
-	if (!driTMMSubImage(gCCPriv->tmm, gCCPriv->curTexObj2D->image[level],
-			    xoffset, yoffset, width, height, image)) {
+	if (driTMMSubImage(gCCPriv->tmm, gCCPriv->curTexObj2D->image[level],
+			    xoffset, yoffset, width, height, image) < 0) {
 	    /* NOT_DONE: Handle error */
 	}
     } else {
@@ -3089,7 +4137,7 @@ void _gamma_Translatef(GLfloat x, GLfloat y, GLfloat z)
     GLfloat m[16];
     int i;
 
-    DEBUG_GLCMDS(("Tranlatef: %f %f %f\n", x, y, z));
+    DEBUG_GLCMDS(("Translatef: %f %f %f\n", x, y, z));
 
     for (i = 0; i < 16; i++)
 	if (i % 5 == 0)
@@ -3108,26 +4156,21 @@ void _gamma_Translatef(GLfloat x, GLfloat y, GLfloat z)
 void _gamma_Vertex2d(GLdouble x, GLdouble y)
 {
     DEBUG_GLCMDS(("Vertex2d: %f %f\n", x, y));
+
+    _gamma_Vertex2f((GLfloat)x,(GLfloat)y);
 }
 
 void _gamma_Vertex2dv(const GLdouble *v)
 {
     DEBUG_GLCMDS(("Vertex2dv: %f %f\n", v[0], v[1]));
+
+    _gamma_Vertex2fv((GLfloat*)v);
 }
 
 void _gamma_Vertex2f(GLfloat x, GLfloat y)
 {
     DEBUG_GLCMDS(("Vertex2f: %f %f\n", x, y));
 
-#ifdef RANDOMIZE_COLORS
-    {
-	GLuint c;
-	c = (random() / (double)RAND_MAX) * 16777216;
-	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
-	WRITE(gCCPriv->buf, PackedColor4,  c);
-    }
-#endif
-
     CHECK_DMA_BUFFER(gCC, gCCPriv, 2);
     WRITEF(gCCPriv->buf, Vy,  y);
     WRITEF(gCCPriv->buf, Vx2, x);
@@ -3136,51 +4179,78 @@ void _gamma_Vertex2f(GLfloat x, GLfloat y)
 void _gamma_Vertex2fv(const GLfloat *v)
 {
     DEBUG_GLCMDS(("Vertex2fv: %f %f\n", v[0], v[1]));
+
+    CHECK_DMA_BUFFER(gCC, gCCPriv, 2);
+    WRITEF(gCCPriv->buf, Vy,  v[1]);
+    WRITEF(gCCPriv->buf, Vx2, v[0]);
 }
 
 void _gamma_Vertex2i(GLint x, GLint y)
 {
+    GLfloat a,b;
+
     DEBUG_GLCMDS(("Vertex2i: %d %d\n", (int)x, (int)y));
+
+    a = INT_TO_FLOAT(x);
+    b = INT_TO_FLOAT(y);
+
+    _gamma_Vertex2f(a,b);
 }
 
 void _gamma_Vertex2iv(const GLint *v)
 {
+    GLfloat p[2];
+
     DEBUG_GLCMDS(("Vertex2iv: %d %d\n", (int)v[0], (int)v[1]));
+
+    p[0] = INT_TO_FLOAT(v[0]);
+    p[1] = INT_TO_FLOAT(v[1]);
+    
+    _gamma_Vertex2fv(p);
 }
 
 void _gamma_Vertex2s(GLshort x, GLshort y)
 {
+    GLfloat a,b;
+
     DEBUG_GLCMDS(("Vertex2s: %d %d\n", x, y));
+
+    a = SHORT_TO_FLOAT(x);
+    b = SHORT_TO_FLOAT(y);
+
+    _gamma_Vertex2f(a,b);
 }
 
 void _gamma_Vertex2sv(const GLshort *v)
 {
+    GLfloat p[2];
+
     DEBUG_GLCMDS(("Vertex2sv: %d %d\n", v[0], v[1]));
+
+    p[0] = SHORT_TO_FLOAT(v[0]);
+    p[1] = SHORT_TO_FLOAT(v[1]);
+
+    _gamma_Vertex2fv(p);
 }
 
 void _gamma_Vertex3d(GLdouble x, GLdouble y, GLdouble z)
 {
     DEBUG_GLCMDS(("Vertex3d: %f %f %f\n", x, y, z));
+
+    _gamma_Vertex3f((GLfloat)x,(GLfloat)y,(GLfloat)z);
 }
 
 void _gamma_Vertex3dv(const GLdouble *v)
 {
     DEBUG_GLCMDS(("Vertex2fv: %f %f %f\n", v[0], v[1], v[2]));
+
+    _gamma_Vertex3fv((GLfloat*)v);
 }
 
 void _gamma_Vertex3f(GLfloat x, GLfloat y, GLfloat z)
 {
     DEBUG_GLCMDS(("Vertex3f: %f %f %f\n", x, y, z));
 
-#ifdef RANDOMIZE_COLORS
-    {
-	GLuint c;
-	c = (random() / (double)RAND_MAX) * 16777216;
-	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
-	WRITE(gCCPriv->buf, PackedColor4,  c);
-    }
-#endif
-
     CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
     WRITEF(gCCPriv->buf, Vz,  z);
     WRITEF(gCCPriv->buf, Vy,  y);
@@ -3191,15 +4261,6 @@ void _gamma_Vertex3fv(const GLfloat *v)
 {
     DEBUG_GLCMDS(("Vertex3fv: %f %f %f\n", v[0], v[1], v[2]));
 
-#ifdef RANDOMIZE_COLORS
-    {
-	GLuint c;
-	c = (random() / (double)RAND_MAX) * 16777216;
-	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
-	WRITE(gCCPriv->buf, PackedColor4,  c);
-    }
-#endif
-
     CHECK_DMA_BUFFER(gCC, gCCPriv, 3);
     WRITEF(gCCPriv->buf, Vz,  v[2]);
     WRITEF(gCCPriv->buf, Vy,  v[1]);
@@ -3208,63 +4269,147 @@ void _gamma_Vertex3fv(const GLfloat *v)
 
 void _gamma_Vertex3i(GLint x, GLint y, GLint z)
 {
+    GLfloat a,b,c;
+
     DEBUG_GLCMDS(("Vertex3i: %d %d %d\n", (int)x, (int)y, (int)z));
+
+    a = INT_TO_FLOAT(x);
+    b = INT_TO_FLOAT(y);
+    c = INT_TO_FLOAT(z);
+
+    _gamma_Vertex3f(a,b,c);
 }
 
 void _gamma_Vertex3iv(const GLint *v)
 {
+    GLfloat p[3];
+
     DEBUG_GLCMDS(("Vertex3iv: %d %d %d\n", (int)v[0], (int)v[1], (int)v[2]));
+
+    p[0] = INT_TO_FLOAT(v[0]);
+    p[1] = INT_TO_FLOAT(v[1]);
+    p[2] = INT_TO_FLOAT(v[2]);
+
+    _gamma_Vertex3fv(p);
 }
 
 void _gamma_Vertex3s(GLshort x, GLshort y, GLshort z)
 {
+    GLfloat a,b,c;
+
     DEBUG_GLCMDS(("Vertex3s: %d %d %d\n", x, y, z));
+
+    a = SHORT_TO_FLOAT(x);
+    b = SHORT_TO_FLOAT(y);
+    c = SHORT_TO_FLOAT(z);
+
+    _gamma_Vertex3f(a,b,c);
 }
 
 void _gamma_Vertex3sv(const GLshort *v)
 {
+    GLfloat p[3];
+
     DEBUG_GLCMDS(("Vertex3sv: %d %d %d\n", v[0], v[1], v[2]));
+
+    p[0] = SHORT_TO_FLOAT(v[0]);
+    p[1] = SHORT_TO_FLOAT(v[1]);
+    p[2] = SHORT_TO_FLOAT(v[2]);
+
+    _gamma_Vertex3fv(p);
 }
 
 void _gamma_Vertex4d(GLdouble x, GLdouble y, GLdouble z, GLdouble w)
 {
     DEBUG_GLCMDS(("Vertex4d: %f %f %f %f\n", x, y, z, w));
+
+    _gamma_Vertex4f((GLfloat)x,(GLfloat)y,(GLfloat)z,(GLfloat)w);
 }
 
 void _gamma_Vertex4dv(const GLdouble *v)
 {
     DEBUG_GLCMDS(("Vertex4dv: %f %f %f %f\n", v[0], v[1], v[2], v[3]));
+
+    _gamma_Vertex4fv((GLfloat*)v);
 }
 
 void _gamma_Vertex4f(GLfloat x, GLfloat y, GLfloat z, GLfloat w)
 {
     DEBUG_GLCMDS(("Vertex4f: %f %f %f %f\n", x, y, z, w));
+
+    CHECK_DMA_BUFFER(gCC, gCCPriv, 4);
+    WRITEF(gCCPriv->buf, Vw,  w);
+    WRITEF(gCCPriv->buf, Vz,  z);
+    WRITEF(gCCPriv->buf, Vy,  y);
+    WRITEF(gCCPriv->buf, Vx4, x);
 }
 
 void _gamma_Vertex4fv(const GLfloat *v)
 {
     DEBUG_GLCMDS(("Vertex4fv: %f %f %f %f\n", v[0], v[1], v[2], v[3]));
+
+    CHECK_DMA_BUFFER(gCC, gCCPriv, 4);
+    WRITEF(gCCPriv->buf, Vw,  v[3]);
+    WRITEF(gCCPriv->buf, Vz,  v[2]);
+    WRITEF(gCCPriv->buf, Vy,  v[1]);
+    WRITEF(gCCPriv->buf, Vx4, v[0]);
 }
 
 void _gamma_Vertex4i(GLint x, GLint y, GLint z, GLint w)
 {
+    GLfloat a,b,c,d;
+
     DEBUG_GLCMDS(("Vertex4i: %d %d %d %d\n", (int)x, (int)y, (int)z, (int)w));
+
+    a = INT_TO_FLOAT(x);
+    b = INT_TO_FLOAT(y);
+    c = INT_TO_FLOAT(z);
+    d = INT_TO_FLOAT(w);
+
+    _gamma_Vertex4f(a,b,c,d);
 }
 
 void _gamma_Vertex4iv(const GLint *v)
 {
+    GLfloat p[4];
+
     DEBUG_GLCMDS(("Vertex4iv: %d %d %d %d\n",
 		  (int)v[0], (int)v[1], (int)v[2], (int)v[3]));
+
+    p[0] = INT_TO_FLOAT(v[0]);
+    p[1] = INT_TO_FLOAT(v[1]);
+    p[2] = INT_TO_FLOAT(v[2]);
+    p[3] = INT_TO_FLOAT(v[3]);
+
+    _gamma_Vertex4fv(p);
 }
 
 void _gamma_Vertex4s(GLshort x, GLshort y, GLshort z, GLshort w)
 {
+    GLfloat a,b,c,d;
+
     DEBUG_GLCMDS(("Vertex4s: %d %d %d %d\n", x, y, z, w));
+
+    a = SHORT_TO_FLOAT(x);
+    b = SHORT_TO_FLOAT(y);
+    c = SHORT_TO_FLOAT(z);
+    d = SHORT_TO_FLOAT(w);
+
+    _gamma_Vertex4f(a,b,c,d);
 }
 
 void _gamma_Vertex4sv(const GLshort *v)
 {
+    GLfloat p[4];
+
     DEBUG_GLCMDS(("Vertex4sv: %d %d %d %d\n", v[0], v[1], v[2], v[3]));
+
+    p[0] = SHORT_TO_FLOAT(v[0]);
+    p[1] = SHORT_TO_FLOAT(v[1]);
+    p[2] = SHORT_TO_FLOAT(v[2]);
+    p[3] = SHORT_TO_FLOAT(v[3]);
+ 
+    _gamma_Vertex4fv(p);
 }
 
 void _gamma_VertexPointer(GLint size, GLenum type, GLsizei stride, const GLvoid *pointer)
@@ -3280,10 +4425,10 @@ void _gamma_Viewport(GLint x, GLint y, GLsizei width, GLsizei height)
     DEBUG_GLCMDS(("Viewport: %d %d %d %d\n",
 		  (int)x, (int)y, (int)width, (int)height));
 
-    gCCPriv->x = gCC->driContextPriv->driDrawablePriv->x + x;
-    gCCPriv->y = gCC->driContextPriv->driScreenPriv->fbHeight -
-	(gCC->driContextPriv->driDrawablePriv->y +
-	 gCC->driContextPriv->driDrawablePriv->h) + y;
+    gCCPriv->x = gCC->driDrawablePriv->x + x;
+    gCCPriv->y = gCC->driScreenPriv->fbHeight -
+	(gCC->driDrawablePriv->y +
+	 gCC->driDrawablePriv->h) + y;
     gCCPriv->w = width;
     gCCPriv->h = height;
 
@@ -3296,16 +4441,21 @@ void _gamma_Viewport(GLint x, GLint y, GLsizei width, GLsizei height)
     oy = y + sy;
 
     CHECK_DMA_BUFFER(gCC, gCCPriv, 4);
-    WRITEF(gCCPriv->buf, ViewPortOffsetX, ox);
-    WRITEF(gCCPriv->buf, ViewPortOffsetY, oy);
     WRITEF(gCCPriv->buf, ViewPortScaleX,  sx);
     WRITEF(gCCPriv->buf, ViewPortScaleY,  sy);
+    WRITEF(gCCPriv->buf, ViewPortOffsetX, ox);
+    WRITEF(gCCPriv->buf, ViewPortOffsetY, oy);
+#if 1 /* Err - this shouldn't be needed, but something isn't flushing */
+    FLUSH_DMA_BUFFER(gCC,gCCPriv);
+#endif
 }
 
 
 static GLint
 generic_noop(void)
 {
+   DEBUG_GLCMDS(("OOPS GENERIC NOOP CALLED\n"));
+
    return 0;
 }
 
diff --git a/xc/lib/GL/mesa/src/drv/gamma/gamma_init.h b/xc/lib/GL/mesa/src/drv/gamma/gamma_init.h
index ce804d018..e8fd5a0df 100644
--- a/xc/lib/GL/mesa/src/drv/gamma/gamma_init.h
+++ b/xc/lib/GL/mesa/src/drv/gamma/gamma_init.h
@@ -43,7 +43,6 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "gamma_regs.h"
 #include "gamma_macros.h"
 #include "gamma_texture.h"
-#include "xmesaP.h"
 
 typedef struct {
     int           regionCount;       /* Count of register regions */
@@ -109,6 +108,12 @@ typedef struct {
     int                 AB_FBReadMode;
     int                 AB_FBReadMode_Save;
     int                 DeltaMode;
+    int			ColorMaterialMode;
+    int			MaterialMode;
+    int			LightingMode;
+    int			Light0Mode;
+    int			Light1Mode;
+    int			ScissorMode;
     int                 Window; /* GID part probably should be in draw priv */
 
     gammaTexObj        *curTexObj;
@@ -140,8 +145,8 @@ extern void       gammaLoadHWMatrix(void);
 extern void       gammaInitHW(gammaContextPrivate *gcp);
 
 extern float                IdentityMatrix[16];
-extern XMesaContext         nullCC;
-extern XMesaContext         gCC;
+extern __DRIcontextPrivate *nullCC;
+extern __DRIcontextPrivate *gCC;
 extern gammaContextPrivate *gCCPriv;
 
 #endif
diff --git a/xc/lib/GL/mesa/src/drv/gamma/gamma_inithw.c b/xc/lib/GL/mesa/src/drv/gamma/gamma_inithw.c
index edb1d3da0..87d68278b 100644
--- a/xc/lib/GL/mesa/src/drv/gamma/gamma_inithw.c
+++ b/xc/lib/GL/mesa/src/drv/gamma/gamma_inithw.c
@@ -29,27 +29,32 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 /*
  * Authors:
  *   Kevin E. Martin <kevin@precisioninsight.com>
+ *   Alan Hourihane <Alan.Hourihane@btinternet.com>
  *
  */
 
 #ifdef GLX_DIRECT_RENDERING
 
 #include "gamma_init.h"
+#include "glint_dri.h"
 
 void gammaInitHW(gammaContextPrivate *gcp)
 {
     __DRIscreenPrivate *driScrnPriv = gcp->gammaScrnPriv->driScrnPriv;
+    GLINTDRIPtr         gDRIPriv = (GLINTDRIPtr)driScrnPriv->pDevPriv;
 
-    /* Set up each MX's ScanLineOwnership for OpenGL */
-    CHECK_DMA_BUFFER(nullCC, gcp, 4);
-    WRITE(gcp->buf, BroadcastMask, 1);
-    WRITE(gcp->buf, ScanLineOwnership, 5); /* Use bottom left as [0,0] */
-    WRITE(gcp->buf, BroadcastMask, 2);
-    WRITE(gcp->buf, ScanLineOwnership, 1); /* Use bottom left as [0,0] */
+    if (gDRIPriv->numMXDevices == 2) {
+	/* Set up each MX's ScanLineOwnership for OpenGL */
+	CHECK_DMA_BUFFER(nullCC, gcp, 4);
+	WRITE(gcp->buf, BroadcastMask, 1);
+	WRITE(gcp->buf, ScanLineOwnership, 5); /* Use bottom left as [0,0] */
+	WRITE(gcp->buf, BroadcastMask, 2);
+	WRITE(gcp->buf, ScanLineOwnership, 1); /* Use bottom left as [0,0] */
 
-    /* Broadcast to both MX's */
-    CHECK_DMA_BUFFER(nullCC, gcp, 1);
-    WRITE(gcp->buf, BroadcastMask, 3);
+	/* Broadcast to both MX's */
+	CHECK_DMA_BUFFER(nullCC, gcp, 1);
+	WRITE(gcp->buf, BroadcastMask, 3);
+    }
 
     /* Set MXs to known state */
     CHECK_DMA_BUFFER(nullCC, gcp, 27);
@@ -82,6 +87,7 @@ void gammaInitHW(gammaContextPrivate *gcp)
     WRITE(gcp->buf, PixelSize, 0);
 
     /* Set Gamma to known state */
+    CHECK_DMA_BUFFER(nullCC, gcp, 10);
     WRITE(gcp->buf, TriangleMode, 0);
     WRITE(gcp->buf, GeometryMode, 0);
     WRITE(gcp->buf, NormalizeMode, 0);
@@ -136,7 +142,10 @@ void gammaInitHW(gammaContextPrivate *gcp)
     WRITE(gcp->buf, AlphaTestMode, gcp->AlphaTestMode);
     WRITE(gcp->buf, AlphaBlendMode, gcp->AlphaBlendMode);
     WRITE(gcp->buf, DitherMode, DitherModeEnable | DM_ColorOrder_RGB);
-    WRITE(gcp->buf, RasterizerMode, RM_MultiGLINT | RM_BiasCoordNearHalf);
+    if (gDRIPriv->numMXDevices == 2)
+    	WRITE(gcp->buf, RasterizerMode, RM_MultiGLINT | RM_BiasCoordNearHalf);
+    else
+    	WRITE(gcp->buf, RasterizerMode, RM_BiasCoordNearHalf);
     WRITE(gcp->buf, GLINTWindow, gcp->Window);
     WRITE(gcp->buf, FastClearDepth, 0xffffffff);
     WRITE(gcp->buf, GLINTDepth, 0xffffffff);
@@ -242,7 +251,7 @@ void gammaInitHW(gammaContextPrivate *gcp)
     CHECK_DMA_BUFFER(nullCC, gcp, 5);
     WRITE(gcp->buf, GeometryMode, gcp->GeometryMode);
     WRITE(gcp->buf, NormalizeMode, NormalizeModeDisable);
-    WRITE(gcp->buf, LightingMode, LightingModeDisable);
+    WRITE(gcp->buf, LightingMode, gcp->LightingMode);
     WRITE(gcp->buf, ColorMaterialMode, ColorMaterialModeDisable);
     WRITE(gcp->buf, MaterialMode, MaterialModeDisable);
 
@@ -251,18 +260,18 @@ void gammaInitHW(gammaContextPrivate *gcp)
     WRITE(gcp->buf, BackSpecularExponent,  0); /* fixed point */
 
     CHECK_DMA_BUFFER(nullCC, gcp, 29);
-    WRITEF(gcp->buf, FrontAmbientColorRed,    0.0);
-    WRITEF(gcp->buf, FrontAmbientColorGreen,  0.0);
-    WRITEF(gcp->buf, FrontAmbientColorBlue,   0.0);
-    WRITEF(gcp->buf, BackAmbientColorRed,     0.0);
-    WRITEF(gcp->buf, BackAmbientColorGreen,   0.0);
-    WRITEF(gcp->buf, BackAmbientColorBlue,    0.0);
-    WRITEF(gcp->buf, FrontDiffuseColorRed,    0.0);
-    WRITEF(gcp->buf, FrontDiffuseColorGreen,  0.0);
-    WRITEF(gcp->buf, FrontDiffuseColorBlue,   0.0);
-    WRITEF(gcp->buf, BackDiffuseColorRed,     0.0);
-    WRITEF(gcp->buf, BackDiffuseColorGreen,   0.0);
-    WRITEF(gcp->buf, BackDiffuseColorBlue,    0.0);
+    WRITEF(gcp->buf, FrontAmbientColorRed,    0.2);
+    WRITEF(gcp->buf, FrontAmbientColorGreen,  0.2);
+    WRITEF(gcp->buf, FrontAmbientColorBlue,   0.2);
+    WRITEF(gcp->buf, BackAmbientColorRed,     0.2);
+    WRITEF(gcp->buf, BackAmbientColorGreen,   0.2);
+    WRITEF(gcp->buf, BackAmbientColorBlue,    0.2);
+    WRITEF(gcp->buf, FrontDiffuseColorRed,    0.8);
+    WRITEF(gcp->buf, FrontDiffuseColorGreen,  0.8);
+    WRITEF(gcp->buf, FrontDiffuseColorBlue,   0.8);
+    WRITEF(gcp->buf, BackDiffuseColorRed,     0.8);
+    WRITEF(gcp->buf, BackDiffuseColorGreen,   0.8);
+    WRITEF(gcp->buf, BackDiffuseColorBlue,    0.8);
     WRITEF(gcp->buf, FrontSpecularColorRed,   0.0);
     WRITEF(gcp->buf, FrontSpecularColorGreen, 0.0);
     WRITEF(gcp->buf, FrontSpecularColorBlue,  0.0);
@@ -275,11 +284,11 @@ void gammaInitHW(gammaContextPrivate *gcp)
     WRITEF(gcp->buf, BackEmissiveColorRed,    0.0);
     WRITEF(gcp->buf, BackEmissiveColorGreen,  0.0);
     WRITEF(gcp->buf, BackEmissiveColorBlue,   0.0);
-    WRITEF(gcp->buf, SceneAmbientColorRed,    0.0);
-    WRITEF(gcp->buf, SceneAmbientColorGreen,  0.0);
-    WRITEF(gcp->buf, SceneAmbientColorBlue,   0.0);
-    WRITEF(gcp->buf, FrontAlpha,              0.0);
-    WRITEF(gcp->buf, BackAlpha,               0.0);
+    WRITEF(gcp->buf, SceneAmbientColorRed,    0.2);
+    WRITEF(gcp->buf, SceneAmbientColorGreen,  0.2);
+    WRITEF(gcp->buf, SceneAmbientColorBlue,   0.2);
+    WRITEF(gcp->buf, FrontAlpha,              1.0);
+    WRITEF(gcp->buf, BackAlpha,               1.0);
 
     CHECK_DMA_BUFFER(nullCC, gcp, 8);
     WRITE(gcp->buf, PointMode, (PM_AntialiasDisable |
@@ -319,28 +328,28 @@ void gammaInitHW(gammaContextPrivate *gcp)
     WRITE(gcp->buf, Light14Mode, LNM_Off);
     WRITE(gcp->buf, Light15Mode, LNM_Off);
 
-    CHECK_DMA_BUFFER(nullCC, gcp, 21);
-    WRITEF(gcp->buf, FrontAmbientColorRed,    1.0);
-    WRITEF(gcp->buf, FrontAmbientColorGreen,  1.0);
-    WRITEF(gcp->buf, FrontAmbientColorBlue,   1.0);
-    WRITEF(gcp->buf, BackAmbientColorRed,     1.0);
-    WRITEF(gcp->buf, BackAmbientColorGreen,   1.0);
-    WRITEF(gcp->buf, BackAmbientColorBlue,    1.0);
-    WRITEF(gcp->buf, FrontDiffuseColorRed,    1.0);
-    WRITEF(gcp->buf, FrontDiffuseColorGreen,  1.0);
-    WRITEF(gcp->buf, FrontDiffuseColorBlue,   1.0);
-    WRITEF(gcp->buf, BackDiffuseColorRed,     1.0);
-    WRITEF(gcp->buf, BackDiffuseColorGreen,   1.0);
-    WRITEF(gcp->buf, BackDiffuseColorBlue,    1.0);
-    WRITEF(gcp->buf, FrontSpecularColorRed,   0.0);
-    WRITEF(gcp->buf, FrontSpecularColorGreen, 0.0);
-    WRITEF(gcp->buf, FrontSpecularColorBlue,  0.0);
-    WRITEF(gcp->buf, BackSpecularColorRed,    0.0);
-    WRITEF(gcp->buf, BackSpecularColorGreen,  0.0);
-    WRITEF(gcp->buf, BackSpecularColorBlue,   0.0);
-    WRITEF(gcp->buf, SceneAmbientColorRed,    0.0);
-    WRITEF(gcp->buf, SceneAmbientColorGreen,  0.0);
-    WRITEF(gcp->buf, SceneAmbientColorBlue,   0.0);
+    CHECK_DMA_BUFFER(nullCC, gcp, 22);
+    WRITEF(gcp->buf, Light0AmbientIntensityBlue, 0.0);
+    WRITEF(gcp->buf, Light0AmbientIntensityGreen, 0.0);
+    WRITEF(gcp->buf, Light0AmbientIntensityRed, 0.0);
+    WRITEF(gcp->buf, Light0DiffuseIntensityBlue, 1.0);
+    WRITEF(gcp->buf, Light0DiffuseIntensityGreen, 1.0);
+    WRITEF(gcp->buf, Light0DiffuseIntensityRed, 1.0);
+    WRITEF(gcp->buf, Light0SpecularIntensityBlue, 1.0);
+    WRITEF(gcp->buf, Light0SpecularIntensityGreen, 1.0);
+    WRITEF(gcp->buf, Light0SpecularIntensityRed, 1.0);
+    WRITEF(gcp->buf, Light0SpotlightDirectionZ, 0.0);
+    WRITEF(gcp->buf, Light0SpotlightDirectionY, 0.0);
+    WRITEF(gcp->buf, Light0SpotlightDirectionX, -1.0);
+    WRITEF(gcp->buf, Light0SpotlightExponent, 0.0);
+    WRITEF(gcp->buf, Light0PositionZ, 0.0);
+    WRITEF(gcp->buf, Light0PositionY, 0.0);
+    WRITEF(gcp->buf, Light0PositionX, 1.0);
+    WRITEF(gcp->buf, Light0PositionW, 0.0);
+    WRITEF(gcp->buf, Light0CosSpotlightCutoffAngle, -1.0);
+    WRITEF(gcp->buf, Light0ConstantAttenuation, 1.0);
+    WRITEF(gcp->buf, Light0LinearAttenuation,   0.0);
+    WRITEF(gcp->buf, Light0QuadraticAttenuation,0.0);
 
     CHECK_DMA_BUFFER(nullCC, gcp, 6);
     WRITEF(gcp->buf, ViewPortScaleX,  (gcp->w)/2.0);
diff --git a/xc/lib/GL/mesa/src/drv/gamma/gamma_macros.h b/xc/lib/GL/mesa/src/drv/gamma/gamma_macros.h
index ae923c7bf..c458bf642 100644
--- a/xc/lib/GL/mesa/src/drv/gamma/gamma_macros.h
+++ b/xc/lib/GL/mesa/src/drv/gamma/gamma_macros.h
@@ -228,14 +228,14 @@ do {                                                                       \
 #ifdef DO_VALIDATE
 #define VALIDATE_DRAWABLE_INFO_NO_LOCK(gcc,gcp)                            \
 do {                                                                       \
-    __DRIcontextPrivate *pcp = gcc->driContextPriv;                        \
+    __DRIcontextPrivate *pcp = gcc;                                        \
     __DRIscreenPrivate *psp = pcp->driScreenPriv;                          \
     __DRIdrawablePrivate *pdp = pcp->driDrawablePriv;                      \
                                                                            \
     if (*(pdp->pStamp) != pdp->lastStamp) {                                \
 	int old_index = pdp->index;                                        \
 	while (*(pdp->pStamp) != pdp->lastStamp) {                         \
-	    DRI_MESA_VALIDATE_DRAWABLE_INFO(gcc->display, psp->myNum, pdp);\
+	    DRI_MESA_VALIDATE_DRAWABLE_INFO(pcp->display, psp->myNum, pdp);\
         }                                                                  \
                                                                            \
 	if (pdp->index != old_index) {                                     \
@@ -303,7 +303,7 @@ do {                                                                       \
 
 #define VALIDATE_DRAWABLE_INFO(gcc,gcp)                                    \
 do {                                                                       \
-    __DRIcontextPrivate *pcp = gcc->driContextPriv;                        \
+    __DRIcontextPrivate *pcp = gcc;                                        \
     __DRIscreenPrivate *psp = pcp->driScreenPriv;                          \
                                                                            \
     DRM_SPINLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID);            \
diff --git a/xc/lib/GL/mesa/src/drv/gamma/gamma_matrix.c b/xc/lib/GL/mesa/src/drv/gamma/gamma_matrix.c
index f6c166609..17220fbb4 100644
--- a/xc/lib/GL/mesa/src/drv/gamma/gamma_matrix.c
+++ b/xc/lib/GL/mesa/src/drv/gamma/gamma_matrix.c
@@ -163,9 +163,7 @@ void gammaLoadHWMatrix(void)
 	WRITEF(gCCPriv->buf, ModelViewMatrix13, gCCPriv->ModelView[13]);
 	WRITEF(gCCPriv->buf, ModelViewMatrix14, gCCPriv->ModelView[14]);
 	WRITEF(gCCPriv->buf, ModelViewMatrix15, gCCPriv->ModelView[15]);
-	/* Fall through to load ModelViewProjectionMatrix */
 #endif
-
     case GL_PROJECTION:
 	CHECK_DMA_BUFFER(gCC, gCCPriv, 16);
 	WRITEF(gCCPriv->buf,
diff --git a/xc/lib/GL/mesa/src/drv/gamma/gamma_regs.h b/xc/lib/GL/mesa/src/drv/gamma/gamma_regs.h
index 3dd4b8783..b13dbdba5 100644
--- a/xc/lib/GL/mesa/src/drv/gamma/gamma_regs.h
+++ b/xc/lib/GL/mesa/src/drv/gamma/gamma_regs.h
@@ -163,6 +163,10 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define FBWriteModeEnable             0x00000001
 #define FBW_UploadColorData           0x00000008
 
+/* FogMode */
+#define FogModeDisable                0x00000000
+#define FogModeEnable                 0x00000001
+
 /* LBWriteMode */
 #define LBWriteModeDisable            0x00000000
 #define LBWriteModeEnable             0x00000001
@@ -391,14 +395,17 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define GM_FogLinear                  0x00000000
 #define GM_FogExp                     0x00000004
 #define GM_FogExpSquared              0x00000008
+#define GM_FogMask                    0x0000000C
 #define GM_FrontPolyPoint             0x00000000
 #define GM_FrontPolyLine              0x00000010
 #define GM_FrontPolyFill              0x00000020
 #define GM_BackPolyPoint              0x00000000
 #define GM_BackPolyLine               0x00000040
 #define GM_BackPolyFill               0x00000080
+#define GM_FB_PolyMask                0x000000F0
 #define GM_FrontFaceCW                0x00000000
 #define GM_FrontFaceCCW               0x00000100
+#define GM_FFMask                     0x00000100
 #define GM_PolyCullDisable            0x00000000
 #define GM_PolyCullEnable             0x00000200
 #define GM_PolyCullFront              0x00000000
@@ -480,18 +487,44 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 /* LightingMode */
 #define LightingModeDisable           0x00000000
 #define LightingModeEnable            0x00000001
+#define LightingModeTwoSides          0x00000002
+#define LightingModeLocalViewer       0x00000008
+
+/* Light0Mode */
+#define Light0ModeDisable             0x00000000
+#define Light0ModeEnable              0x00000001
+#define Light0ModeSpotLight           0x00000002
+#define Light0ModeAttenuation         0x00000004
+#define Light0ModeLocal               0x00000008
+
+/* Light0Mode */
+#define Light1ModeDisable             0x00000000
+#define Light1ModeEnable              0x00000001
+#define Light1ModeSpotLight           0x00000002
+#define Light1ModeAttenuation         0x00000004
+#define Light1ModeLocal               0x00000008
 
 /* ColorMaterialMode */
 #define ColorMaterialModeDisable      0x00000000
 #define ColorMaterialModeEnable       0x00000001
+#define ColorMaterialModeFront        0x00000000
+#define ColorMaterialModeBack         0x00000002
+#define ColorMaterialModeFrontAndBack 0x00000004
+#define ColorMaterialModeEmission     0x00000000
+#define ColorMaterialModeAmbient      0x00000008
+#define ColorMaterialModeDiffuse      0x00000010
+#define ColorMaterialModeSpecular     0x00000018
+#define ColorMaterialModeAmbAndDiff   0x00000020
+#define ColorMaterialModeMask         0x0000003e
 
 /* MaterialMode */
 #define MaterialModeDisable           0x00000000
 #define MaterialModeEnable            0x00000001
+#define MaterialModeTwoSides          0x00000080
 
 /* DeltaMode */
 #define DM_Target300SX                0x00000000
-#define DM_Target300TXMX              0x00000001
+#define DM_Target500TXMX              0x00000001
 #define DM_Depth16                    0x00000004
 #define DM_Depth24                    0x00000008
 #define DM_Depth32                    0x0000000c
diff --git a/xc/lib/GL/mesa/src/drv/gamma/gamma_texture.c b/xc/lib/GL/mesa/src/drv/gamma/gamma_texture.c
index 14a993969..ffa437ed2 100644
--- a/xc/lib/GL/mesa/src/drv/gamma/gamma_texture.c
+++ b/xc/lib/GL/mesa/src/drv/gamma/gamma_texture.c
@@ -34,6 +34,7 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #ifdef GLX_DIRECT_RENDERING
 
+#include <Xarch.h>
 #include <X11/Xlibint.h>
 #include "gamma_init.h"
 #include <string.h>
@@ -74,7 +75,11 @@ void gammaTOInit(gammaTexObj *t)
 
     t->TextureFilterMode = (TextureFilterModeDisable);
 
+#if X_BYTE_ORDER == X_LITTLE_ENDIAN
     t->TextureFormat = (TF_LittleEndian |
+#else
+    t->TextureFormat = (TF_BigEndian |
+#endif
 			TF_16Bit_565 |
 			TF_ColorOrder_RGB |
 			TF_Compnents_4 |
diff --git a/xc/lib/GL/mesa/src/drv/gamma/gamma_xmesa.c b/xc/lib/GL/mesa/src/drv/gamma/gamma_xmesa.c
index 1f968e39b..3c1fe64fc 100644
--- a/xc/lib/GL/mesa/src/drv/gamma/gamma_xmesa.c
+++ b/xc/lib/GL/mesa/src/drv/gamma/gamma_xmesa.c
@@ -30,31 +30,26 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  * Authors:
  *   Kevin E. Martin <kevin@precisioninsight.com>
  *   Brian Paul <brian@precisioninsight.com>
+ *   Alan Hourihane <Alan.Hourihane@btinternet.com>
  */
 
 #ifdef GLX_DIRECT_RENDERING
 
 #include <X11/Xlibint.h>
 #include "gamma_init.h"
+#include "gamma_gl.h"
 #include "glapi.h"
+#include "glint_dri.h"
+#include "context.h"
+#include "mmath.h"
 
 
-XMesaContext         nullCC  = NULL;
-XMesaContext         gCC     = NULL;
+__DRIcontextPrivate *nullCC  = NULL;
+__DRIcontextPrivate *gCC = NULL;
 gammaContextPrivate *gCCPriv = NULL;
 
 static struct _glapi_table *Dispatch = NULL;
 
-static int count_bits(unsigned int n)
-{
-   int bits = 0;
-
-   while (n > 0) {
-      if (n & 1) bits++;
-      n >>= 1;
-   }
-   return bits;
-}
 
 GLboolean XMesaInitDriver(__DRIscreenPrivate *driScrnPriv)
 {
@@ -83,78 +78,41 @@ void XMesaResetDriver(__DRIscreenPrivate *driScrnPriv)
     Xfree(driScrnPriv->private);
 }
 
-XMesaVisual XMesaCreateVisual(XMesaDisplay *display,
-				      XMesaVisualInfo visinfo,
-				      GLboolean rgb_flag,
-				      GLboolean alpha_flag,
-				      GLboolean db_flag,
-				      GLboolean stereo_flag,
-				      GLboolean ximage_flag,
-				      GLint depth_size,
-				      GLint stencil_size,
-				      GLint accum_size,
-				      GLint level)
+GLvisual *XMesaCreateVisual(Display *dpy,
+                            __DRIscreenPrivate *driScrnPriv,
+                            const XVisualInfo *visinfo,
+                            const __GLXvisualConfig *config)
 {
-    XMesaVisual v;
-
-    /* Only RGB visuals are supported on the GMX2000 */
-    if (!rgb_flag) {
-	return NULL;
-    }
-
-    v = (XMesaVisual)Xmalloc(sizeof(struct xmesa_visual));
-    if (!v) {
-	return NULL;
-    }
-
-    v->visinfo = (XVisualInfo *)Xmalloc(sizeof(*visinfo));
-    if(!v->visinfo) {
-	Xfree(v);
-	return NULL;
-    }
-    memcpy(v->visinfo, visinfo, sizeof(*visinfo));
-
-    v->display = display;
-    v->level = level;
-
-    v->gl_visual = (GLvisual *)Xmalloc(sizeof(GLvisual));
-    if (!v->gl_visual) {
-	Xfree(v->visinfo);
-	XFree(v);
-	return NULL;
-    }
-
-    v->gl_visual->RGBAflag   = rgb_flag;
-    v->gl_visual->DBflag     = db_flag;
-    v->gl_visual->StereoFlag = stereo_flag;
-
-    v->gl_visual->RedBits   = count_bits(visinfo->red_mask);
-    v->gl_visual->GreenBits = count_bits(visinfo->green_mask);
-    v->gl_visual->BlueBits  = count_bits(visinfo->blue_mask);
-    v->gl_visual->AlphaBits = 0; /* Not currently supported */
-
-    v->gl_visual->AccumBits   = accum_size;
-    v->gl_visual->DepthBits   = depth_size;
-    v->gl_visual->StencilBits = stencil_size;
-
-    return v;
+   /* Drivers may change the args to _mesa_create_visual() in order to
+    * setup special visuals.
+    */
+   return _mesa_create_visual( config->rgba,
+                               config->doubleBuffer,
+                               config->stereo,
+                               _mesa_bitcount(visinfo->red_mask),
+                               _mesa_bitcount(visinfo->green_mask),
+                               _mesa_bitcount(visinfo->blue_mask),
+                               config->alphaSize,
+                               0, /* index bits */
+                               config->depthSize,
+                               config->stencilSize,
+                               config->accumRedSize,
+                               config->accumGreenSize,
+                               config->accumBlueSize,
+                               config->accumAlphaSize,
+                               0 /* num samples */ );
 }
 
-void XMesaDestroyVisual(XMesaVisual v)
-{
-    Xfree(v->gl_visual);
-    Xfree(v->visinfo);
-    Xfree(v);
-}
 
-XMesaContext XMesaCreateContext(XMesaVisual v, XMesaContext share_list,
-					__DRIcontextPrivate *driContextPriv)
+GLboolean XMesaCreateContext( Display *dpy,
+                              GLvisual *mesaVis,
+                              __DRIcontextPrivate *driContextPriv )
 {
     int i;
-    XMesaContext c;
     gammaContextPrivate *cPriv;
     __DRIscreenPrivate *driScrnPriv = driContextPriv->driScreenPriv;
     gammaScreenPrivate *gPriv = (gammaScreenPrivate *)driScrnPriv->private;
+    GLINTDRIPtr         gDRIPriv = (GLINTDRIPtr)driScrnPriv->pDevPriv;
 
     if (!Dispatch) {
        GLuint size = _glapi_get_dispatch_table_size() * sizeof(GLvoid *);
@@ -162,20 +120,9 @@ XMesaContext XMesaCreateContext(XMesaVisual v, XMesaContext share_list,
        _gamma_init_dispatch(Dispatch);
     }
 
-    c = (XMesaContext)Xmalloc(sizeof(struct xmesa_context));
-    if (!c) {
-	return NULL;
-    }
-
-    c->driContextPriv = driContextPriv;
-    c->xm_visual = v;
-    c->xm_buffer = NULL; /* Set by MakeCurrent */
-    c->display = v->display;
-
     cPriv = (gammaContextPrivate *)Xmalloc(sizeof(gammaContextPrivate));
     if (!cPriv) {
-	Xfree(c);
-	return NULL;
+	return GL_FALSE;
     }
 
     cPriv->hHWContext = driContextPriv->hHWContext;
@@ -216,27 +163,26 @@ XMesaContext XMesaCreateContext(XMesaVisual v, XMesaContext share_list,
 		cPriv->ModelViewProj[i] =
 		cPriv->Texture[i] = 0.0;
 
-    /*
-    ** NOT_DONE: The 0xe4 in LBReadMode and FBReadMode refers to the
-    ** partial products for the 640x480 mode.  We will need to look up
-    ** the partial products in a table (c.f., glint_driver.c) to support
-    ** other FB sizes.
-    */
     cPriv->LBReadMode = (LBReadSrcDisable |
 			 LBReadDstDisable |
 			 LBDataTypeDefault |
 			 LBWindowOriginBot |
-			 LBScanLineInt2 |
-			 0xe4); /* NOT_DONE: 640x480 partial products */
+			 gDRIPriv->pprod);
     cPriv->FBReadMode = (FBReadSrcDisable |
 			 FBReadDstDisable |
 			 FBDataTypeDefault |
 			 FBWindowOriginBot |
-			 FBScanLineInt2 |
-			 0xe4); /* NOT_DONE: 640x480 partial products */
-
-    cPriv->FBWindowBase = driScrnPriv->fbWidth * (driScrnPriv->fbHeight/2 - 1);
-    cPriv->LBWindowBase = driScrnPriv->fbWidth * (driScrnPriv->fbHeight/2 - 1);
+			 gDRIPriv->pprod);
+ 
+    if (gDRIPriv->numMXDevices == 2) {
+	cPriv->LBReadMode |= LBScanLineInt2;
+	cPriv->FBReadMode |= FBScanLineInt2;
+    	cPriv->FBWindowBase =driScrnPriv->fbWidth*(driScrnPriv->fbHeight/2 - 1);
+    	cPriv->LBWindowBase =driScrnPriv->fbWidth*(driScrnPriv->fbHeight/2 - 1);
+    } else {
+    	cPriv->FBWindowBase = driScrnPriv->fbWidth * driScrnPriv->fbHeight;
+    	cPriv->LBWindowBase = driScrnPriv->fbWidth * driScrnPriv->fbHeight;
+    }
 
     cPriv->Begin = (B_AreaStippleDisable |
 		    B_LineStippleDisable |
@@ -328,7 +274,7 @@ XMesaContext XMesaCreateContext(XMesaVisual v, XMesaContext share_list,
 			      4, 1,
 			      gammaTOLoad,
 			      gammaTOLoadSub);
-	
+
     cPriv->curTexObj = gammaTOFind(0);
     cPriv->curTexObj1D = cPriv->curTexObj;
     cPriv->curTexObj2D = cPriv->curTexObj;
@@ -338,17 +284,17 @@ XMesaContext XMesaCreateContext(XMesaVisual v, XMesaContext share_list,
 #ifdef FORCE_DEPTH32
     cPriv->DepthSize = 32;
 #else
-    cPriv->DepthSize = v->gl_visual->DepthBits;
+    cPriv->DepthSize = mesaVis->DepthBits;
 #endif
     cPriv->zNear = 0.0;
     cPriv->zFar  = 1.0;
 
     cPriv->Flags  = GAMMA_FRONT_BUFFER;
-    cPriv->Flags |= (v->gl_visual->DBflag ? GAMMA_BACK_BUFFER  : 0);
+    cPriv->Flags |= (mesaVis->DBflag ? GAMMA_BACK_BUFFER  : 0);
     cPriv->Flags |= (cPriv->DepthSize > 0 ? GAMMA_DEPTH_BUFFER : 0);
 
     cPriv->EnabledFlags = GAMMA_FRONT_BUFFER;
-    cPriv->EnabledFlags |= (v->gl_visual->DBflag ? GAMMA_BACK_BUFFER  : 0);
+    cPriv->EnabledFlags |= (mesaVis->DBflag ? GAMMA_BACK_BUFFER  : 0);
 
     cPriv->DepthMode = (DepthModeDisable |
 			DM_WriteMask |
@@ -356,7 +302,7 @@ XMesaContext XMesaCreateContext(XMesaVisual v, XMesaContext share_list,
 
     cPriv->DeltaMode = (DM_SubPixlCorrectionEnable |
 			DM_SmoothShadingEnable |
-			DM_Target300TXMX);
+			DM_Target500TXMX);
 
     switch (cPriv->DepthSize) {
     case 16:
@@ -374,36 +320,68 @@ XMesaContext XMesaCreateContext(XMesaVisual v, XMesaContext share_list,
 
     cPriv->gammaScrnPriv = gPriv;
 
-    c->private = (void *)cPriv;
+    cPriv->LightingMode = LightingModeDisable;
+    cPriv->Light0Mode = LNM_Off;
+    cPriv->Light1Mode = LNM_Off;
+
+    cPriv->MaterialMode = MaterialModeDisable;
+
+    cPriv->ScissorMode = UserScissorDisable | ScreenScissorDisable;
+
+    driContextPriv->driverPrivate = cPriv;
 
     /* Initialize the HW to a known state */
     gammaInitHW(cPriv);
 
-    return c;
+    return GL_TRUE;
 }
 
-void XMesaDestroyContext(XMesaContext c)
+void XMesaDestroyContext(__DRIcontextPrivate *driContextPriv)
 {
+    gammaContextPrivate *cPriv;
+    cPriv = (gammaContextPrivate *) driContextPriv->driverPrivate;
+    if (cPriv) {
+       /* XXX free driver context data? */
+    }
 }
 
-XMesaBuffer XMesaCreateWindowBuffer(XMesaVisual v, XMesaWindow w,
-					    __DRIdrawablePrivate *driDrawPriv)
-{
-    return (XMesaBuffer)1;
-}
 
-XMesaBuffer XMesaCreatePixmapBuffer(XMesaVisual v, XMesaPixmap p,
-					    XMesaColormap c,
-					    __DRIdrawablePrivate *driDrawPriv)
+GLframebuffer *XMesaCreateWindowBuffer( Display *dpy,
+                                        __DRIscreenPrivate *driScrnPriv,
+                                        __DRIdrawablePrivate *driDrawPriv,
+                                        GLvisual *mesaVis)
 {
-    return (XMesaBuffer)1;
+   return gl_create_framebuffer(mesaVis,
+                                GL_FALSE,  /* software depth buffer? */
+                                mesaVis->StencilBits > 0,
+                                mesaVis->AccumRedBits > 0,
+                                mesaVis->AlphaBits > 0
+                                );
 }
 
-void XMesaDestroyBuffer(XMesaBuffer b)
+
+GLframebuffer *XMesaCreatePixmapBuffer( Display *dpy,
+                                        __DRIscreenPrivate *driScrnPriv,
+                                        __DRIdrawablePrivate *driDrawPriv,
+                                        GLvisual *mesaVis)
 {
+#if 0
+   /* Different drivers may have different combinations of hardware and
+    * software ancillary buffers.
+    */
+   return gl_create_framebuffer(mesaVis,
+                                GL_FALSE,  /* software depth buffer? */
+                                mesaVis->StencilBits > 0,
+                                mesaVis->AccumRedBits > 0,
+                                mesaVis->AlphaBits > 0
+                                );
+#else
+   return NULL;  /* not implemented yet */
+#endif
 }
 
-void XMesaSwapBuffers(XMesaBuffer b)
+
+void XMesaSwapBuffers(__DRIdrawablePrivate *driDrawPriv)
 {
     /*
     ** NOT_DONE: This assumes buffer is currently bound to a context.
@@ -420,11 +398,9 @@ void XMesaSwapBuffers(XMesaBuffer b)
     if (gCCPriv->EnabledFlags & GAMMA_BACK_BUFFER) {
 	int src, dst, x0, y0, x1, h;
 	int i;
-	__DRIdrawablePrivate *driDrawPriv =
-	    gCC->driContextPriv->driDrawablePriv;
 	int nRect = driDrawPriv->numClipRects;
 	XF86DRIClipRectPtr pRect = driDrawPriv->pClipRects;
-	__DRIscreenPrivate *driScrnPriv = gCC->driContextPriv->driScreenPriv;
+	__DRIscreenPrivate *driScrnPriv = gCC->driScreenPriv;
 
 #ifdef DO_VALIDATE
 	DRM_SPINLOCK(&driScrnPriv->pSAREA->drawable_lock,
@@ -487,19 +463,23 @@ void XMesaSwapBuffers(XMesaBuffer b)
     }
 }
 
-GLboolean XMesaMakeCurrent(XMesaContext c, XMesaBuffer b)
+GLboolean XMesaMakeCurrent(__DRIcontextPrivate *driContextPriv,
+                           __DRIdrawablePrivate *driDrawPriv,
+                           __DRIdrawablePrivate *driReadPriv)
 {
-    if (c) {
-	gCC     = c;
-	gCCPriv = (gammaContextPrivate *)c->private;
+    if (driContextPriv) {
+	gCC     = driContextPriv;
+	gCCPriv = (gammaContextPrivate *) driContextPriv->driverPrivate;
 
 	gCCPriv->Window &= ~W_GIDMask;
-	gCCPriv->Window |= (gCC->driContextPriv->driDrawablePriv->index << 5);
+	gCCPriv->Window |= (driDrawPriv->index << 5);
 
 	CHECK_DMA_BUFFER(gCC, gCCPriv, 1);
 	WRITE(gCCPriv->buf, GLINTWindow, gCCPriv->Window);
 
         _glapi_set_dispatch(Dispatch);
+
+	_gamma_Viewport(0, 0, driDrawPriv->w, driDrawPriv->h);
     } else {
 	gCC     = NULL;
 	gCCPriv = NULL;
@@ -508,7 +488,7 @@ GLboolean XMesaMakeCurrent(XMesaContext c, XMesaBuffer b)
 }
 
 
-GLboolean XMesaUnbindContext( XMesaContext c )
+GLboolean XMesaUnbindContext( __DRIcontextPrivate *driContextPriv )
 {
    /* XXX not 100% sure what's supposed to be done here */
    return GL_TRUE;
diff --git a/xc/lib/GL/mesa/src/drv/i810/Imakefile b/xc/lib/GL/mesa/src/drv/i810/Imakefile
index 7260846fd..6b72c80e4 100644
--- a/xc/lib/GL/mesa/src/drv/i810/Imakefile
+++ b/xc/lib/GL/mesa/src/drv/i810/Imakefile
@@ -1,4 +1,6 @@
 
+#include <Threads.tmpl>
+
 #define DoNormalLib NormalLibGlx
 #define DoSharedLib SharedLibGlx
 #define DoExtraLib SharedLibGlx
@@ -9,13 +11,22 @@
 ALLOC_DEFINES = -DMALLOC_0_RETURNS_NULL
 #endif
 
+#if defined(LinuxArchitecture)
+OS_SUBDIR = linux
+#endif
+#if defined(FreeBSDArchitecture)
+OS_SUBDIR = bsd
+#endif
+
 #if BuildXF86DRI
   DRI_DEFINES = GlxDefines -DDRIVERTS
  DRI_INCLUDES = -I../../../../dri -I../../../../glx \
+		-I../../../dri \
 		-I$(TOP)/include -I$(TOP)/include/GL \
 		-I$(XF86OSSRC) -I$(SERVERSRC)/GL/dri \
 		-I$(XF86DRIVERSRC)/i810  \
-		-I../../../include -I../.. -I../common -I../../X
+		-I../../../include -I../.. -I../common -I../../X \
+	        -I$(XF86OSSRC)/$(OS_SUBDIR)/drm/kernel
 #endif
 
 MESA_INCLUDES = -I. -I.. -I../../include
@@ -23,20 +34,244 @@ MESA_INCLUDES = -I. -I.. -I../../include
 
 
       DEFINES = $(ALLOC_DEFINES) $(DRI_DEFINES)
-     INCLUDES = -I$(XLIBSRC) -I$(EXTINCSRC) $(MESA_INCLUDES) $(DRI_INCLUDES) \
-		-I/usr/include/glide
-      DRISRCS = i810_xmesa.c i810clear.c	\
-		i810dd.c i810depth.c i810dma.c i810ring.c		\
-		i810pipeline.c i810span.c i810state.c i810swap.c	\
-		i810tex.c i810tris.c i810vb.c i810fastpath.c
+     INCLUDES = -I$(XLIBSRC) -I$(EXTINCSRC) $(MESA_INCLUDES) $(DRI_INCLUDES) 
+
+      DRISRCS = ../../../dri/dri_mesa.c \
+		../../../../dri/dri_tmm.c
+
+      DRIOBJS = ../../../dri/dri_mesa.o \
+		../../../../dri/dri_tmm.o
+
+      DRMSRCS = ../../../../dri/drm/xf86drm.c \
+		../../../../dri/drm/xf86drmHash.c \
+		../../../../dri/drm/xf86drmRandom.c \
+		../../../../dri/drm/xf86drmSL.c
+
+      DRMOBJS = ../../../../dri/drm/xf86drm.o \
+		../../../../dri/drm/xf86drmHash.o \
+		../../../../dri/drm/xf86drmRandom.o \
+		../../../../dri/drm/xf86drmSL.o
+
+
+     I810SRCS = i810_xmesa.c	\
+		i810dd.c  	\
+		i810pipeline.c i810span.c i810state.c	\
+		i810tex.c i810tris.c i810vb.c i810fastpath.c i810ioctl.c
+
+     I810OBJS = i810_xmesa.o \
+		i810dd.o  		\
+		i810pipeline.o i810span.o i810state.o	\
+		i810tex.o i810tris.o i810vb.o i810fastpath.o i810ioctl.o
+
+     MESASRCS = ../../aatriangle.c \
+		../../accum.c \
+		../../alpha.c \
+		../../alphabuf.c \
+		../../attrib.c \
+		../../bbox.c \
+		../../bitmap.c \
+		../../blend.c \
+		../../buffers.c \
+		../../clip.c \
+		../../colortab.c \
+		../../config.c \
+		../../context.c \
+		../../copypix.c \
+		../../cva.c \
+		../../debug_xform.c \
+		../../depth.c \
+		../../dlist.c \
+		../../drawpix.c \
+		../../enable.c \
+		../../enums.c \
+		../../eval.c \
+		../../extensions.c \
+		../../feedback.c \
+		../../fog.c \
+		../../get.c \
+		../../glapi.c \
+		../../glapinoop.c \
+		../../glthread.c \
+		../../hash.c \
+		../../image.c \
+		../../imaging.c \
+		../../light.c \
+		../../lines.c \
+		../../logic.c \
+		../../masking.c \
+		../../matrix.c \
+		../../mem.c \
+		../../mmath.c \
+		../../pb.c \
+		../../pipeline.c \
+		../../pixel.c \
+		../../pixeltex.c \
+		../../points.c \
+		../../polygon.c \
+		../../quads.c \
+		../../rastpos.c \
+		../../readpix.c \
+		../../rect.c \
+		../../scissor.c \
+		../../shade.c \
+		../../span.c \
+		../../stages.c \
+		../../state.c \
+		../../stencil.c \
+		../../teximage.c \
+		../../texobj.c \
+		../../texstate.c \
+		../../texture.c \
+		../../texutil.c \
+		../../translate.c \
+		../../triangle.c \
+		../../varray.c \
+		../../vb.c \
+		../../vbcull.c \
+		../../vbfill.c \
+		../../vbindirect.c \
+		../../vbrender.c \
+		../../vbxform.c \
+		../../vector.c \
+		../../vertices.c \
+		../../winpos.c \
+		../../xform.c \
+		../../zoom.c \
+		../../X86/common_x86.c
+
+     MESAOBJS = ../../aatriangle.o \
+		../../accum.o \
+		../../alpha.o \
+		../../alphabuf.o \
+		../../attrib.o \
+		../../bbox.o \
+		../../bitmap.o \
+		../../blend.o \
+		../../buffers.o \
+		../../clip.o \
+		../../colortab.o \
+		../../config.o \
+		../../context.o \
+		../../copypix.o \
+		../../cva.o \
+		../../debug_xform.o \
+		../../depth.o \
+		../../dlist.o \
+		../../drawpix.o \
+		../../enable.o \
+		../../enums.o \
+		../../eval.o \
+		../../extensions.o \
+		../../feedback.o \
+		../../fog.o \
+		../../get.o \
+		../../hash.o \
+		../../hint.o \
+		../../image.o \
+		../../imaging.o \
+		../../light.o \
+		../../lines.o \
+		../../logic.o \
+		../../masking.o \
+		../../matrix.o \
+		../../mem.o \
+		../../mmath.o \
+		../../pb.o \
+		../../pipeline.o \
+		../../pixel.o \
+		../../pixeltex.o \
+		../../points.o \
+		../../polygon.o \
+		../../quads.o \
+		../../rastpos.o \
+		../../readpix.o \
+		../../rect.o \
+		../../scissor.o \
+		../../shade.o \
+		../../span.o \
+		../../stages.o \
+		../../state.o \
+		../../stencil.o \
+		../../teximage.o \
+		../../texobj.o \
+		../../texstate.o \
+		../../texture.o \
+		../../texutil.o \
+		../../translate.o \
+		../../triangle.o \
+		../../varray.o \
+		../../vb.o \
+		../../vbcull.o \
+		../../vbfill.o \
+		../../vbindirect.o \
+		../../vbrender.o \
+		../../vbxform.o \
+		../../vector.o \
+		../../vertices.o \
+		../../winpos.o \
+		../../xform.o \
+		../../zoom.o
 
-      DRIOBJS = i810_xmesa.o i810clear.o	\
-		i810dd.o i810depth.o i810dma.o i810ring.o		\
-		i810pipeline.o i810span.o i810state.o i810swap.o	\
-		i810tex.o i810tris.o i810vb.o i810fastpath.o
+#ifdef i386Architecture
+     X86_SRCS = ../../X86/x86.c \
+		../../X86/x86a.S \
+		../../X86/common_x86.c \
+		../../X86/common_x86asm.S \
+		../../X86/vertex.S
+
+     X86_OBJS = ../../X86/x86.o \
+		../../X86/x86a.o \
+		../../X86/common_x86.o \
+		../../X86/common_x86asm.o \
+		../../X86/vertex.o
+
+     MMX_SRCS = ../../X86/mmx_blend.S
+
+     MMX_OBJS = ../../X86/mmx_blend.o
+
+XCOMM Disabling 3Dnow code for the time being.
+#if 0
+   3DNOW_SRCS = ../../X86/3dnow.c \
+		../../X86/3dnow_norm_raw.S \
+		../../X86/3dnow_xform_masked1.S \
+		../../X86/3dnow_xform_masked2.S \
+		../../X86/3dnow_xform_masked3.S \
+		../../X86/3dnow_xform_masked4.S \
+		../../X86/3dnow_xform_raw1.S \
+		../../X86/3dnow_xform_raw2.S \
+		../../X86/3dnow_xform_raw3.S \
+		../../X86/3dnow_xform_raw4.S \
+		../../X86/vertex_3dnow.S
+
+   3DNOW_OBJS = ../../X86/3dnow.o \
+		../../X86/3dnow_norm_raw.o \
+		../../X86/3dnow_xform_masked1.o \
+		../../X86/3dnow_xform_masked2.o \
+		../../X86/3dnow_xform_masked3.o \
+		../../X86/3dnow_xform_masked4.o \
+		../../X86/3dnow_xform_raw1.o \
+		../../X86/3dnow_xform_raw2.o \
+		../../X86/3dnow_xform_raw3.o \
+		../../X86/3dnow_xform_raw4.o \
+		../../X86/vertex_3dnow.o
+#endif
+
+#endif
+
+      ASMSRCS = $(X86_SRCS) $(MMX_SRCS) $(3DNOW_SRCS)
+      ASMOBJS = $(X86_OBJS) $(MMX_OBJS) $(3DNOW_OBJS)
+
+      COMMONSRCS = ../common/mm.c ../common/hwlog.c
+      COMMONOBJS = ../common/mm.o ../common/hwlog.o
+
+         SRCS = $(DRISRCS) $(DRMSRCS) $(MESASRCS) $(ASMSRCS) $(COMMONSRCS) $(I810SRCS) 
+         OBJS = $(DRIOBJS) $(DRMOBJS) $(MESAOBJS) $(ASMOBJS) $(COMMONOBJS) $(I810OBJS)
+
+REQUIREDLIBS += -lm
+#if !GlxBuiltInI810
+REQUIREDLIBS += -L../../../.. -lGL
+#endif
 
-	 SRCS = $(DRISRCS) 
-	 OBJS = $(DRIOBJS) 
 
 #if !GlxUseBuiltInDRIDriver
 #undef DoNormalLib NormalLibGlx
@@ -57,7 +292,7 @@ LIBNAME = i810_dri.so
 ALL_OBJS = $(OBJS)
 ALL_DEPS = DONE
 SharedDepModuleTarget($(LIBNAME),$(ALL_DEPS),$(ALL_OBJS))
-InstallDynamicModule($(LIBNAME),$(MODULEDIR),.)
+InstallDynamicModule($(LIBNAME),$(MODULEDIR)/dri,.)
 #endif
 
 DependTarget()
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810_3d_reg.h b/xc/lib/GL/mesa/src/drv/i810/i810_3d_reg.h
index 8e6942497..a4b947b30 100644
--- a/xc/lib/GL/mesa/src/drv/i810/i810_3d_reg.h
+++ b/xc/lib/GL/mesa/src/drv/i810/i810_3d_reg.h
@@ -108,6 +108,8 @@
 #define LCS_UPDATE_LINEWIDTH    (0x1<<15)
 #define LCS_LINEWIDTH_MASK      (0x7<<12)
 #define LCS_LINEWIDTH_SHIFT           12
+#define LCS_LINEWIDTH_0_5       (0x1<<12)
+#define LCS_LINEWIDTH_1_0       (0x2<<12)
 #define LCS_UPDATE_ALPHA_INTERP (0x1<<11)
 #define LCS_ALPHA_FLAT          (0x0<<10)
 #define LCS_ALPHA_INTERP        (0x1<<10)
@@ -616,9 +618,6 @@ typedef struct {
 #define DV_PF_555           (0x1<<8)
 #define DV_PF_565           (0x2<<8)
 
-
-
-
 #define GFX_OP_ANTIALIAS         ((0x3<<29)|(0x6<<24))
 #define AA_UPDATE_EDGEFLAG       (1<<13)
 #define AA_ENABLE_EDGEFLAG       (1<<12)
@@ -641,84 +640,16 @@ typedef struct {
 #define ST1_ENABLE               (1<<16)
 #define ST1_MASK                 (0xffff)
 
+#define I810_SET_FIELD( var, mask, value ) (var &= ~(mask), var |= value)
 
-/* Indices into buf.Setup where various bits of state are mirrored per
- * context and per buffer.  These can be fired at the card as a unit,
- * or in a piecewise fashion as required.
- */
-
-/* Destbuffer state 
- *    - backbuffer linear offset and pitch -- invarient in the current dri
- *    - zbuffer linear offset and pitch -- also invarient
- *    - drawing origin in back and depth buffers.
- *
- * Keep the depth/back buffer state here to acommodate private buffers
- * in the future.
- */
-#define I810_DESTREG_DI0  0	/* CMD_OP_DESTBUFFER_INFO (2 dwords) */
-#define I810_DESTREG_DI1  1
-#define I810_DESTREG_ZB0  2	/* CMD_OP_Z_BUFFER_INFO */
-#define I810_DESTREG_ZB1  3	/* CMD_OP_Z_BUFFER_INFO */
-#define I810_DESTREG_DV0  4	/* GFX_OP_DESTBUFFER_VARS (2 dwords) */
-#define I810_DESTREG_DV1  5
-#define I810_DESTREG_DR0  6	/* GFX_OP_DRAWRECT_INFO (4 dwords) */
-#define I810_DESTREG_DR1  7
-#define I810_DESTREG_DR2  8
-#define I810_DESTREG_DR3  9
-#define I810_DESTREG_DR4  10
-
-#define I810_DEST_SETUP_SIZE (11+1)
-
-/* Context state
- */
-#define I810_CTXREG_VF    0	/* GFX_OP_VERTEX_FMT */
-#define I810_CTXREG_MT    1	/* GFX_OP_MAP_TEXELS */
-#define I810_CTXREG_MC0   2	/* GFX_OP_MAP_COLOR_STAGES - stage 0 */
-#define I810_CTXREG_MC1   3	/* GFX_OP_MAP_COLOR_STAGES - stage 1 */
-#define I810_CTXREG_MC2   4	/* GFX_OP_MAP_COLOR_STAGES - stage 2 */
-#define I810_CTXREG_MA0   5	/* GFX_OP_MAP_ALPHA_STAGES - stage 0 */
-#define I810_CTXREG_MA1   6	/* GFX_OP_MAP_ALPHA_STAGES - stage 1 */
-#define I810_CTXREG_MA2   7	/* GFX_OP_MAP_ALPHA_STAGES - stage 2 */
-#define I810_CTXREG_SDM   8	/* GFX_OP_SRC_DEST_MONO */
-#define I810_CTXREG_CF0   9	/* GFX_OP_COLOR_FACTOR */
-#define I810_CTXREG_CF1   10	
-#define I810_CTXREG_FOG   11	/* GFX_OP_FOG_COLOR */
-#define I810_CTXREG_B1    12	/* GFX_OP_BOOL_1 */
-#define I810_CTXREG_B2    13	/* GFX_OP_BOOL_2 */
-#define I810_CTXREG_LCS   14	/* GFX_OP_LINEWIDTH_CULL_SHADE_MODE */
-#define I810_CTXREG_PV    15	/* GFX_OP_PV_RULE -- Invarient! */
-#define I810_CTXREG_ZA    16	/* GFX_OP_ZBIAS_ALPHAFUNC */
-#define I810_CTXREG_ST0   17    /* GFX_OP_STIPPLE */
-#define I810_CTXREG_ST1   18
-#define I810_CTXREG_AA    19	/* GFX_OP_ANTIALIAS */
-#define I810_CTX_SETUP_SIZE (20) /* pad to qword */
-
-
-/* Cliprect state - keep seperate from context as it is used for
- *                  drawing triangles to the shared backbuffer, and
- *                  changes more frequently than the 'normal' context.
- */
-#define I810_CLIPREG_SCI0  0 	/* GFX_OP_SCISSOR_INFO (3 dwords) */
-#define I810_CLIPREG_SCI1  1	
-#define I810_CLIPREG_SCI2  2	
-#define I810_CLIPREG_SC    3	/* GFX_OP_SCISSOR */
-
-#define I810_CLIP_SETUP_SIZE 4
-
-/* Texture state (per tex_buffer)
- */
-#define I810_TEXREG_MI0  0	/* GFX_OP_MAP_INFO (4 dwords) */
-#define I810_TEXREG_MI1  1	
-#define I810_TEXREG_MI2  2	
-#define I810_TEXREG_MI3  3	
-#define I810_TEXREG_MF   4	/* GFX_OP_MAP_FILTER */
-#define I810_TEXREG_MLC  5	/* GFX_OP_MAP_LOD_CTL */
-#define I810_TEXREG_MLL  6	/* GFX_OP_MAP_LOD_LIMITS */
-#define I810_TEXREG_MCS  7	/* GFX_OP_MAP_COORD_SETS ??? */
-
-#define I810_TEX_SETUP_SIZE 8
+#define I810PACKCOLOR4444(r,g,b,a) \
+  ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4))
 
+#define I810PACKCOLOR1555(r,g,b,a) \
+  ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \
+    ((a) ? 0x8000 : 0))
 
-#define I810_SET_FIELD( var, mask, value ) (var &= ~(mask), var |= value)
+#define I810PACKCOLOR565(r,g,b) \
+  ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3))
 
 #endif
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810_init.h b/xc/lib/GL/mesa/src/drv/i810/i810_init.h
index f6aa8bd30..df5362787 100644
--- a/xc/lib/GL/mesa/src/drv/i810/i810_init.h
+++ b/xc/lib/GL/mesa/src/drv/i810/i810_init.h
@@ -28,7 +28,6 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 /*
  * Authors:
  *   Keith Whitwell <keithw@precisioninsight.com>
- *   Daryll Strauss <daryll@precisioninsight.com> (Origninal tdfx driver).
  *
  */
 
@@ -42,16 +41,18 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "dri_mesaint.h"
 #include "dri_mesa.h"
 #include "types.h"
-#include "xmesaP.h"
 
 typedef struct {
-  drmHandle handle;
-  drmSize size;
-  drmAddress map;
+   drmHandle handle;
+   drmSize size;
+   char *map;
 } i810Region, *i810RegionPtr;
 
 typedef struct {
-   i810Region regs;
+   i810Region front;
+   i810Region back;
+   i810Region depth;
+   i810Region tex;
 
    int deviceID;
    int width;
@@ -68,21 +69,22 @@ typedef struct {
    int backOffset;
    int depthOffset;
 
-   int auxPitch;
-   int auxPitchBits;
+   int backPitch;
+   int backPitchBits;
 
    int textureOffset;
    int textureSize;
    int logTextureGranularity;
 
    __DRIscreenPrivate *driScrnPriv;
+   drmBufMapPtr  bufs;
 
 } i810ScreenPrivate;
 
 
 #include "i810context.h"
 
-extern void i810XMesaUpdateState( i810ContextPtr imesa );
+extern void i810GetLock( i810ContextPtr imesa, GLuint flags );
 extern void i810EmitHwStateLocked( i810ContextPtr imesa );
 extern void i810EmitScissorValues( i810ContextPtr imesa, int box_nr, int emit );
 extern void i810EmitDrawingRectangle( i810ContextPtr imesa );
@@ -90,6 +92,10 @@ extern void i810XMesaSetBackClipRects( i810ContextPtr imesa );
 extern void i810XMesaSetFrontClipRects( i810ContextPtr imesa );
 
 
+#define GET_DISPATCH_AGE( imesa ) imesa->sarea->last_dispatch
+#define GET_ENQUEUE_AGE( imesa ) imesa->sarea->last_enqueue
+
+
 /* Lock the hardware and validate our state.  
  */
 #define LOCK_HARDWARE( imesa )				\
@@ -97,38 +103,28 @@ extern void i810XMesaSetFrontClipRects( i810ContextPtr imesa );
     char __ret=0;					\
     DRM_CAS(imesa->driHwLock, imesa->hHWContext,	\
 	    (DRM_LOCK_HELD|imesa->hHWContext), __ret);	\
-    if (__ret) {					\
-        drmGetLock(imesa->driFd, imesa->hHWContext, 0);	\
-        i810XMesaUpdateState( imesa );			\
-    }							\
+    if (__ret)						\
+        i810GetLock( imesa, 0 );			\
   } while (0)
 
 
+
 /* Unlock the hardware using the global current context 
  */
 #define UNLOCK_HARDWARE(imesa)					\
     DRM_UNLOCK(imesa->driFd, imesa->driHwLock, imesa->hHWContext);	
 
 
-/*
-  This pair of macros makes a loop over the drawing operations
-  so it is not self contained and doesn't have the nice single 
-  statement semantics of most macros
-*/
-#define BEGIN_CLIP_LOOP(imesa)						\
-  do {									\
-    int _nc;								\
-    LOCK_HARDWARE( imesa );						\
-    for (_nc = imesa->numClipRects; _nc ; _nc--) {			\
-      if (imesa->needClip)						\
-        i810EmitScissorValues(imesa, _nc-1, 1);
-
-
-
-#define END_CLIP_LOOP(imesa)			\
-    }						\
-    UNLOCK_HARDWARE(imesa);			\
-  } while (0)
+/* This is the wrong way to do it, I'm sure.  Otherwise the drm
+ * bitches that I've already got the heavyweight lock.  At worst,
+ * this is 3 ioctls.  The best solution probably only gets me down 
+ * to 2 ioctls in the worst case.
+ */
+#define LOCK_HARDWARE_QUIESCENT( imesa ) do {	\
+   LOCK_HARDWARE( imesa );			\
+   i810RegetLockQuiescent( imesa );		\
+} while(0)
+     
 
 #endif
 #endif
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810_xmesa.c b/xc/lib/GL/mesa/src/drv/i810/i810_xmesa.c
index 50b7e0689..5edcf2620 100644
--- a/xc/lib/GL/mesa/src/drv/i810/i810_xmesa.c
+++ b/xc/lib/GL/mesa/src/drv/i810/i810_xmesa.c
@@ -46,10 +46,9 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "i810state.h"
 #include "i810tex.h"
 #include "i810span.h"
-#include "i810depth.h"
 #include "i810tris.h"
-#include "i810swap.h"
 #include "i810pipeline.h"
+#include "i810ioctl.h"
 
 #include "i810_dri.h"
 
@@ -57,36 +56,25 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #ifndef I810_DEBUG
 int I810_DEBUG = (0
-   		  | DEBUG_ALWAYS_SYNC 
-		  | DEBUG_VERBOSE_RING   
-		  | DEBUG_VERBOSE_OUTREG 
+/*     		  | DEBUG_ALWAYS_SYNC  */
+/*  		  | DEBUG_VERBOSE_RING    */
+/*  		  | DEBUG_VERBOSE_OUTREG  */
 /*  		  | DEBUG_VERBOSE_MSG */
 /*  		  | DEBUG_NO_OUTRING */
 /*  		  | DEBUG_NO_OUTREG */
 /*  		  | DEBUG_VERBOSE_API */
 /*  		  | DEBUG_VERBOSE_2D */
 /*  		  | DEBUG_VERBOSE_DRI */
-		  | DEBUG_VALIDATE_RING
+/*  		  | DEBUG_VALIDATE_RING */
+/*  		  | DEBUG_VERBOSE_IOCTL */
 		  );
 #endif
 
 
 static i810ContextPtr      i810Ctx = 0;
 
-i810Glx_t	i810glx;
 
-static int count_bits(unsigned int n)
-{
-   int bits = 0;
-
-   while (n > 0) {
-      if (n & 1) bits++;
-      n >>= 1;
-   }
-   return bits;
-}
-
-/* These functions are accessed by dlsym from dri_mesa_init.c:
+/* These functions are accessed externally to the driver:
  *
  * XMesaInitDriver
  * XMesaResetDriver
@@ -100,14 +88,26 @@ static int count_bits(unsigned int n)
  * XMesaSwapBuffers
  * XMesaMakeCurrent
  *
- * So this is kind of the public interface to the driver.  The driver
- * uses the X11 mesa driver context as a kind of wrapper around its
- * own driver context - but there isn't much justificiation for doing
- * it that way - the DRI might as well use a (void *) to refer to the
- * driver contexts.  Nothing in the X context really gets used.
  */
 
 
+static drmBufMapPtr i810_create_empty_buffers(void)
+{
+   drmBufMapPtr retval;
+   
+   retval = (drmBufMapPtr)Xmalloc(sizeof(drmBufMap));
+   if(retval == NULL) return NULL;
+   memset(retval, 0, sizeof(drmBufMap));
+   retval->list = (drmBufPtr)Xmalloc(sizeof(drmBuf) * I810_DMA_BUF_NR);
+   if(retval->list == NULL) {
+      Xfree(retval);
+      return NULL;
+   }
+   memset(retval->list, 0, sizeof(drmBuf) * I810_DMA_BUF_NR);
+   fprintf(stderr, "retval : %p, retval->list : %p\n", retval, retval->list);
+   return retval;
+}
+
 GLboolean XMesaInitDriver(__DRIscreenPrivate *sPriv)
 {
    i810ScreenPrivate *i810Screen;
@@ -120,8 +120,6 @@ GLboolean XMesaInitDriver(__DRIscreenPrivate *sPriv)
    i810Screen->driScrnPriv = sPriv;
    sPriv->private = (void *)i810Screen;
 
-   i810Screen->regs.handle=gDRIPriv->regs;
-   i810Screen->regs.size=gDRIPriv->regsSize;
    i810Screen->deviceID=gDRIPriv->deviceID;
    i810Screen->width=gDRIPriv->width;
    i810Screen->height=gDRIPriv->height;
@@ -137,39 +135,60 @@ GLboolean XMesaInitDriver(__DRIscreenPrivate *sPriv)
 
    i810Screen->backOffset=gDRIPriv->backOffset; 
    i810Screen->depthOffset=gDRIPriv->depthOffset;
-   i810Screen->auxPitch = gDRIPriv->auxPitch;
-   i810Screen->auxPitchBits = gDRIPriv->auxPitchBits;
+   i810Screen->backPitch = gDRIPriv->auxPitch;
+   i810Screen->backPitchBits = gDRIPriv->auxPitchBits;
    i810Screen->textureOffset=gDRIPriv->textureOffset;
    i810Screen->textureSize=gDRIPriv->textureSize;
    i810Screen->logTextureGranularity = gDRIPriv->logTextureGranularity;
 
-
-   if (0)
+   if (1)
       fprintf(stderr, "Tex heap size %x, granularity %x bytes\n",
 	      i810Screen->textureSize, 1<<(i810Screen->logTextureGranularity));
+
+   i810Screen->bufs = i810_create_empty_buffers();
+   if(i810Screen->bufs == NULL)
+   {
+      Xfree(i810Screen);
+      return GL_FALSE;
+   }
     
+   i810Screen->back.handle = gDRIPriv->backbuffer;
+   i810Screen->back.size = gDRIPriv->backbufferSize;
+
    if (drmMap(sPriv->fd, 
-	      i810Screen->regs.handle, 
-	      i810Screen->regs.size, 
-	      &i810Screen->regs.map) != 0) 
+	      i810Screen->back.handle, 
+	      i810Screen->back.size, 
+	      (drmAddress *)&i810Screen->back.map) != 0) 
    {
       Xfree(i810Screen);
       return GL_FALSE;
    }
 
-   /* Ditch i810glx in favor of i810Screen?
-    */
-   memset(&i810glx, 0, sizeof(i810glx));
+   i810Screen->depth.handle = gDRIPriv->depthbuffer;
+   i810Screen->depth.size = gDRIPriv->depthbufferSize;
+
+   if (drmMap(sPriv->fd, 
+	      i810Screen->depth.handle, 
+	      i810Screen->depth.size, 
+	      (drmAddress *)&i810Screen->depth.map) != 0) 
+   {
+      Xfree(i810Screen);
+      return GL_FALSE;
+   }
+
+   i810Screen->tex.handle = gDRIPriv->textures;
+   i810Screen->tex.size = gDRIPriv->textureSize;
+
+   if (drmMap(sPriv->fd, 
+	      i810Screen->tex.handle, 
+	      i810Screen->tex.size, 
+	      (drmAddress *)&i810Screen->tex.map) != 0) 
+   {
+      Xfree(i810Screen);
+      return GL_FALSE;
+   }
 
-   i810glx.LpRing.mem.Start = gDRIPriv->ringOffset;
-   i810glx.LpRing.mem.Size = gDRIPriv->ringSize;
-   i810glx.LpRing.mem.End = gDRIPriv->ringOffset + gDRIPriv->ringSize;
-   i810glx.LpRing.virtual_start = sPriv->pFB + i810glx.LpRing.mem.Start;
-   i810glx.LpRing.tail_mask = i810glx.LpRing.mem.Size - 1;
-   i810glx.MMIOBase = i810Screen->regs.map;
-   i810glx.texVirtual = sPriv->pFB + i810Screen->textureOffset;
 
-   
    i810DDFastPathInit();
    i810DDTrifuncInit();
    i810DDSetupInit();
@@ -183,111 +202,75 @@ void XMesaResetDriver(__DRIscreenPrivate *sPriv)
 {
    i810ScreenPrivate *i810Screen = (i810ScreenPrivate *)sPriv->private;
 
-   drmUnmap(i810Screen->regs.map, i810Screen->regs.size);
-   Xfree(i810Screen);
-}
-
-/* Accessed by dlsym from dri_mesa_init.c
- */
-XMesaVisual XMesaCreateVisual(XMesaDisplay *display,
-			      XMesaVisualInfo visinfo,
-			      GLboolean rgb_flag,
-			      GLboolean alpha_flag,
-			      GLboolean db_flag,
-			      GLboolean stereo_flag,
-			      GLboolean ximage_flag,
-			      GLint depth_size,
-			      GLint stencil_size,
-			      GLint accum_size,
-			      GLint level)
-{
-   XMesaVisual v;
-
-   /* Only RGB visuals are supported on the I810 boards */
-   if (!rgb_flag) return 0;
-
-   v = (XMesaVisual)Xmalloc(sizeof(struct xmesa_visual));
-   if (!v) return 0;
-
-   v->visinfo = (XVisualInfo *)Xmalloc(sizeof(*visinfo));
-   if(!v->visinfo) {
-      Xfree(v);
-      return 0;
-   }
-   memcpy(v->visinfo, visinfo, sizeof(*visinfo));
-
-   v->display = display;
-   v->level = level;
-  
-   v->gl_visual = (GLvisual *)Xmalloc(sizeof(GLvisual));
-   if (!v->gl_visual) {
-      Xfree(v->visinfo);
-      XFree(v);
-      return 0;
-   }
-
-   v->gl_visual->RGBAflag   = rgb_flag;
-   v->gl_visual->DBflag     = db_flag;
-   v->gl_visual->StereoFlag = stereo_flag;
-
-   v->gl_visual->RedBits   = count_bits(visinfo->red_mask);
-   v->gl_visual->GreenBits = count_bits(visinfo->green_mask);
-   v->gl_visual->BlueBits  = count_bits(visinfo->blue_mask);
-   v->gl_visual->AlphaBits = 0; /* Not currently supported */
+   /* Need to unmap all the bufs and maps here:
+    */
 
-   v->gl_visual->AccumBits   = accum_size;
-   v->gl_visual->DepthBits   = depth_size;
-   v->gl_visual->StencilBits = stencil_size;
 
-   return v;
+   Xfree(i810Screen);
 }
 
-void XMesaDestroyVisual(XMesaVisual v)
+
+GLvisual *XMesaCreateVisual(Display *dpy,
+                            __DRIscreenPrivate *driScrnPriv,
+                            const XVisualInfo *visinfo,
+                            const __GLXvisualConfig *config)
 {
-   Xfree(v->gl_visual);
-   Xfree(v->visinfo);
-   Xfree(v);
+   /* Drivers may change the args to _mesa_create_visual() in order to
+    * setup special visuals.
+    */
+   return _mesa_create_visual( config->rgba,
+                               config->doubleBuffer,
+                               config->stereo,
+                               _mesa_bitcount(visinfo->red_mask),
+                               _mesa_bitcount(visinfo->green_mask),
+                               _mesa_bitcount(visinfo->blue_mask),
+                               config->alphaSize,
+                               0, /* index bits */
+                               config->depthSize,
+                               config->stencilSize,
+                               config->accumRedSize,
+                               config->accumGreenSize,
+                               config->accumBlueSize,
+                               config->accumAlphaSize,
+                               0 /* num samples */ );
 }
 
-XMesaContext XMesaCreateContext(XMesaVisual v, XMesaContext share_list,
-				__DRIcontextPrivate *driContextPriv)
+
+GLboolean XMesaCreateContext( Display *dpy, GLvisual *mesaVis,
+                              __DRIcontextPrivate *driContextPriv )
 {
-   GLcontext *ctx;
-   XMesaContext c;
+   GLcontext *ctx = driContextPriv->mesaContext;
    i810ContextPtr imesa;
    __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
    i810ScreenPrivate *i810Screen = (i810ScreenPrivate *)sPriv->private;
-   I810SAREAPriv *saPriv=(I810SAREAPriv*)(((char*)sPriv->pSAREA)+
-					  sizeof(XF86DRISAREARec));
+   drm_i810_sarea_t *saPriv=(drm_i810_sarea_t *)(((char*)sPriv->pSAREA)+
+						 sizeof(XF86DRISAREARec));
 
-   GLcontext *shareCtx = 0;
-
-   c = (XMesaContext)Xmalloc(sizeof(struct xmesa_context));
-   if (!c) {
-      return 0;
-   }
-
-   imesa = (i810ContextPtr)Xmalloc(sizeof(i810Context));
+   imesa = (i810ContextPtr)Xcalloc(sizeof(i810Context), 1);
    if (!imesa) {
-      Xfree(c);
-      return 0;
+      return GL_FALSE;
    }
 
-   c->driContextPriv = driContextPriv;
-   c->xm_visual = v;
-   c->xm_buffer = 0; /* Set by MakeCurrent */
-   c->display = v->display;
-   c->private = (void *)imesa;
 
-   if (share_list)
-      shareCtx=((i810ContextPtr)(share_list->private))->glCtx;
+   /* Set the maximum texture size small enough that we can guarentee
+    * that both texture units can bind a maximal texture and have them
+    * in memory at once.
+    */
+   if (i810Screen->textureSize < 2*1024*1024) {
+      ctx->Const.MaxTextureLevels = 9;
+      ctx->Const.MaxTextureSize = 1<<8;
+   } else if (i810Screen->textureSize < 8*1024*1024) {
+      ctx->Const.MaxTextureLevels = 10;
+      ctx->Const.MaxTextureSize = 1<<9;     
+   } else {
+      ctx->Const.MaxTextureLevels = 11;
+      ctx->Const.MaxTextureSize = 1<<10;
+   }      
 
-   ctx = imesa->glCtx = gl_create_context(v->gl_visual, shareCtx, 
-					  (void*)imesa, GL_TRUE);
 
    /* Dri stuff
     */
-   imesa->display = v->display;
+   imesa->display = dpy;
    imesa->hHWContext = driContextPriv->hHWContext;
    imesa->driFd = sPriv->fd;
    imesa->driHwLock = &sPriv->pSAREA->lock;
@@ -295,10 +278,7 @@ XMesaContext XMesaCreateContext(XMesaVisual v, XMesaContext share_list,
    imesa->i810Screen = i810Screen;
    imesa->driScreen = sPriv;
    imesa->sarea = saPriv;
-
-   imesa->glBuffer = gl_create_framebuffer(v->gl_visual);
-
-   imesa->needClip=1;
+   imesa->glBuffer = NULL;
 
    imesa->texHeap = mmInit( 0, i810Screen->textureSize );
 
@@ -315,17 +295,21 @@ XMesaContext XMesaCreateContext(XMesaVisual v, XMesaContext share_list,
    imesa->TextureMode = ctx->Texture.Unit[0].EnvMode;
    imesa->CurrentTexObj[0] = 0;
    imesa->CurrentTexObj[1] = 0;
+
+   ctx->DriverCtx = (void *) imesa;
+   imesa->glCtx = ctx;
    
    i810DDExtensionsInit( ctx );
 
    i810DDInitStateFuncs( ctx );
    i810DDInitTextureFuncs( ctx );
    i810DDInitSpanFuncs( ctx );
-   i810DDInitDepthFuncs( ctx );
    i810DDInitDriverFuncs( ctx );
+   i810DDInitIoctlFuncs( ctx );
 
    ctx->Driver.TriangleCaps = (DD_TRI_CULL|
 			       DD_TRI_LIGHT_TWOSIDE|
+			       DD_TRI_STIPPLE|
 			       DD_TRI_OFFSET);
 
    /* Ask mesa to clip fog coordinates for us.
@@ -346,18 +330,19 @@ XMesaContext XMesaCreateContext(XMesaVisual v, XMesaContext share_list,
 
    i810DDInitState( imesa );
 
-   return c;
+   driContextPriv->driverPrivate = (void *) imesa;
+
+   return GL_TRUE;
 }
 
-void XMesaDestroyContext(XMesaContext c)
+void XMesaDestroyContext(__DRIcontextPrivate *driContextPriv)
 {
-   i810ContextPtr imesa = (i810ContextPtr) c->private;
+   i810ContextPtr imesa = (i810ContextPtr) driContextPriv->driverPrivate;
 
    if (imesa) {
       i810TextureObjectPtr next_t, t;
 
       gl_destroy_context(imesa->glCtx);
-      gl_destroy_framebuffer(imesa->glBuffer);
 
       foreach_s (t, next_t, &(imesa->TexObjList))
 	 i810DestroyTexObj(imesa, t);
@@ -366,58 +351,53 @@ void XMesaDestroyContext(XMesaContext c)
 	 i810DestroyTexObj(imesa, t);
 
       Xfree(imesa);
-
-      c->private = 0;
    }
 }
 
-XMesaBuffer XMesaCreateWindowBuffer(XMesaVisual v, XMesaWindow w,
-				    __DRIdrawablePrivate *driDrawPriv)
+GLframebuffer *XMesaCreateWindowBuffer( Display *dpy,
+                                        __DRIscreenPrivate *driScrnPriv,
+                                        __DRIdrawablePrivate *driDrawPriv,
+                                        GLvisual *mesaVis)
 {
-  return (XMesaBuffer)1;
+   return gl_create_framebuffer(mesaVis,
+                                GL_FALSE,  /* software depth buffer? */
+                                mesaVis->StencilBits > 0,
+                                mesaVis->AccumRedBits > 0,
+                                mesaVis->AlphaBits > 0
+                                );
 }
 
-XMesaBuffer XMesaCreatePixmapBuffer(XMesaVisual v, XMesaPixmap p,
-				    XMesaColormap c, 
-				    __DRIdrawablePrivate *driDrawPriv)
-{
-  return (XMesaBuffer)1;
-}
 
-void XMesaDestroyBuffer(XMesaBuffer b)
+GLframebuffer *XMesaCreatePixmapBuffer( Display *dpy,
+                                        __DRIscreenPrivate *driScrnPriv,
+                                        __DRIdrawablePrivate *driDrawPriv,
+                                        GLvisual *mesaVis)
 {
+#if 0
+   /* Different drivers may have different combinations of hardware and
+    * software ancillary buffers.
+    */
+   return gl_create_framebuffer(mesaVis,
+                                GL_FALSE,  /* software depth buffer? */
+                                mesaVis->StencilBits > 0,
+                                mesaVis->AccumRedBits > 0,
+                                mesaVis->AlphaBits > 0
+                                );
+#else
+   return NULL;  /* not implemented yet */
+#endif
 }
 
-void XMesaSwapBuffers(XMesaBuffer bogus)
+
+void XMesaSwapBuffers(__DRIdrawablePrivate *driDrawPriv)
 {
+   /* XXX should do swap according to the buffer, not the context! */
    i810ContextPtr imesa = i810Ctx; 
 
    FLUSH_VB( imesa->glCtx, "swap buffers" );
    i810SwapBuffers(imesa);
 }
 
-static void i810InitClipRects( i810ContextPtr imesa )
-{
-   switch (imesa->numClipRects) {
-   case 0:
-      imesa->ClipSetup[I810_CLIPREG_SC] = ( GFX_OP_SCISSOR |
-					    SC_UPDATE_SCISSOR );
-      imesa->ClipSetup[I810_CLIPREG_SCI1] = 0;
-      imesa->ClipSetup[I810_CLIPREG_SCI2] = 0;
-      imesa->needClip = 0;
-      imesa->dirty |= I810_EMIT_CLIPRECT;
-      break;
-   case 1:
-      imesa->needClip = 0;
-      i810EmitScissorValues( imesa, 0, 0 );
-      imesa->dirty |= I810_EMIT_CLIPRECT;
-      break;
-   default:
-      imesa->needClip=1;
-      break;
-   }
-}
-
 
 
 void i810XMesaSetFrontClipRects( i810ContextPtr imesa )
@@ -426,12 +406,12 @@ void i810XMesaSetFrontClipRects( i810ContextPtr imesa )
 
    imesa->numClipRects = dPriv->numClipRects;
    imesa->pClipRects = dPriv->pClipRects;
+   imesa->dirty |= I810_UPLOAD_CLIPRECTS;
    imesa->drawX = dPriv->x;
    imesa->drawY = dPriv->y;
 
-   imesa->drawOffset = imesa->i810Screen->fbOffset;
+   imesa->drawMap = imesa->driScreen->pFB;
    i810EmitDrawingRectangle( imesa );
-   i810InitClipRects( imesa );
 }
 
 
@@ -440,7 +420,7 @@ void i810XMesaSetBackClipRects( i810ContextPtr imesa )
    __DRIdrawablePrivate *dPriv = imesa->driDrawable;
    int i;
 
-   if (dPriv->numAuxClipRects == 0) 
+   if (dPriv->numBackClipRects == 0) 
    {
       if (I810_DEBUG & DEBUG_VERBOSE_DRI)
 	 fprintf(stderr, "FRONT_CLIPRECTS, %d rects\n", 
@@ -452,18 +432,18 @@ void i810XMesaSetBackClipRects( i810ContextPtr imesa )
       imesa->drawY = dPriv->y;
    } else {
       if (I810_DEBUG & DEBUG_VERBOSE_DRI)
-	 fprintf(stderr, "AUX_RECTS, %d rects\n", 
-		 dPriv->numAuxClipRects);
+	 fprintf(stderr, "BACK_RECTS, %d rects\n", 
+		 dPriv->numBackClipRects);
 
-      imesa->numClipRects = dPriv->numAuxClipRects;
-      imesa->pClipRects = dPriv->pAuxClipRects;
-      imesa->drawX = dPriv->auxX;
-      imesa->drawY = dPriv->auxY;
+      imesa->numClipRects = dPriv->numBackClipRects;
+      imesa->pClipRects = dPriv->pBackClipRects;
+      imesa->drawX = dPriv->backX;
+      imesa->drawY = dPriv->backY;
    }
 
-   imesa->drawOffset = imesa->i810Screen->backOffset;
+   imesa->drawMap = imesa->i810Screen->back.map;
    i810EmitDrawingRectangle( imesa );
-   i810InitClipRects( imesa );
+   imesa->dirty |= I810_UPLOAD_CLIPRECTS;
 
    if (I810_DEBUG & DEBUG_VERBOSE_DRI)
       for (i = 0 ; i < imesa->numClipRects ; i++) 
@@ -478,6 +458,9 @@ void i810XMesaSetBackClipRects( i810ContextPtr imesa )
 
 static void i810XMesaWindowMoved( i810ContextPtr imesa ) 
 {
+   if (0)
+   fprintf(stderr, "i810XMesaWindowMoved\n\n");
+
    switch (imesa->glCtx->Color.DriverDrawBuffer) {
    case GL_FRONT_LEFT:
       i810XMesaSetFrontClipRects( imesa );
@@ -489,35 +472,40 @@ static void i810XMesaWindowMoved( i810ContextPtr imesa )
       fprintf(stderr, "fallback buffer\n");
       break;
    }
-
 }
 
 
-/* This looks buggy to me - the 'b' variable isn't used anywhere...
- * Hmm - It seems that the drawable is already hooked in to
- * driDrawablePriv.  
- */
-GLboolean XMesaMakeCurrent(XMesaContext c, XMesaBuffer b)
+GLboolean XMesaUnbindContext(__DRIcontextPrivate *driContextPriv)
 {
+   i810ContextPtr i810 = (i810ContextPtr) driContextPriv->driverPrivate;
+   if (i810)
+      i810->dirty = ~0;
 
-   if (c->private==(void *)i810Ctx) return GL_TRUE;
-
-   if (c) {
-      __DRIdrawablePrivate *dPriv = c->driContextPriv->driDrawablePriv;
-
-      i810Ctx = (i810ContextPtr)c->private;
+   return GL_TRUE;
+}
 
-      gl_make_current(i810Ctx->glCtx, i810Ctx->glBuffer);
 
-      i810Ctx->driDrawable = dPriv;
+GLboolean XMesaMakeCurrent(__DRIcontextPrivate *driContextPriv,
+			   __DRIdrawablePrivate *driDrawPriv,
+			   __DRIdrawablePrivate *driReadPriv)
+{
+   if (driContextPriv) {
+      i810Ctx = (i810ContextPtr) driContextPriv->driverPrivate;
+      
+      gl_make_current2(i810Ctx->glCtx, driDrawPriv->mesaBuffer, 
+		       driReadPriv->mesaBuffer);
+      
+      
+      i810Ctx->driDrawable = driDrawPriv;
       i810Ctx->dirty = ~0;
-   
+      
       i810XMesaWindowMoved( i810Ctx );
-
+      
       if (!i810Ctx->glCtx->Viewport.Width)
-	 gl_Viewport(i810Ctx->glCtx, 0, 0, dPriv->w, dPriv->h);
-
-   } else {
+	 gl_Viewport(i810Ctx->glCtx, 0, 0, driDrawPriv->w, driDrawPriv->h);
+   }
+   else 
+   {
       gl_make_current(0,0);
       i810Ctx = NULL;
    }
@@ -525,15 +513,26 @@ GLboolean XMesaMakeCurrent(XMesaContext c, XMesaBuffer b)
 }
 
 
-void i810XMesaUpdateState( i810ContextPtr imesa ) 
+void i810GetLock( i810ContextPtr imesa, GLuint flags ) 
 {
    __DRIdrawablePrivate *dPriv = imesa->driDrawable;
    __DRIscreenPrivate *sPriv = imesa->driScreen;
-   I810SAREAPriv *sarea = imesa->sarea;
+   drm_i810_sarea_t *sarea = imesa->sarea;
    int me = imesa->hHWContext;
    int stamp = dPriv->lastStamp; 
 
 
+   if (0) fprintf(stderr, ".\n");
+
+   /* We know there has been contention.
+    */
+   drmGetLock(imesa->driFd, imesa->hHWContext, flags);	
+
+
+   /* Note contention for throttling hint
+    */
+   imesa->any_contend = 1;
+
    /* If the window moved, may need to set a new cliprect now.
     *
     * NOTE: This releases and regains the hw lock, so all state
@@ -541,26 +540,24 @@ void i810XMesaUpdateState( i810ContextPtr imesa )
     */
    XMESA_VALIDATE_DRAWABLE_INFO(imesa->display, sPriv, dPriv);		
 
-   i810glx.LpRing.synced = 0;
 
-   /* If another client has touched the ringbuffer, need to update 
-    * where we think the pointers are:
-    */
-   if (sarea->ringOwner != me) {
-      i810glx.c_ringlost++;
-      imesa->dirty |= I810_REFRESH_RING;
-   }
-      
+   if (0)
+   fprintf(stderr, "i810GetLock, last enque: %d last dispatch: %d\n",
+	   sarea->last_enqueue, 
+	   sarea->last_dispatch);
+
    /* If we lost context, need to dump all registers to hardware.
     * Note that we don't care about 2d contexts, even if they perform
     * accelerated commands, so the DRI locking in the X server is even
     * more broken than usual.
     */
    if (sarea->ctxOwner != me) {
-      i810glx.c_ctxlost++;
-      imesa->dirty |= I810_UPLOAD_CTX;
-      imesa->dirty |= I810_EMIT_CLIPRECT;
-      imesa->dirty |= I810_UPLOAD_BUFFERS;
+      imesa->dirty |= (I810_UPLOAD_CTX |
+		       I810_UPLOAD_CLIPRECTS |
+		       I810_UPLOAD_BUFFERS |
+		       I810_UPLOAD_TEX0 |
+		       I810_UPLOAD_TEX1);
+      sarea->ctxOwner = me;
    }
 
    /* Shared texture managment - if another client has played with
@@ -587,23 +584,18 @@ void i810XMesaUpdateState( i810ContextPtr imesa )
 	 i810ResetGlobalLRU( imesa );
       }
 
+      if (0) fprintf(stderr, "imesa %d sarea %d\n", imesa->texAge, sarea->texAge);
+      imesa->dirty |= I810_UPLOAD_TEX0IMAGE;
+      imesa->dirty |= I810_UPLOAD_TEX1IMAGE;
       imesa->texAge = sarea->texAge;
-      imesa->dirty |= I810_UPLOAD_TEX0IMAGE | I810_UPLOAD_TEX1IMAGE;
    }
 
    
    if (dPriv->lastStamp != stamp)
       i810XMesaWindowMoved( imesa );
 
-   sarea->ctxOwner=me;
-   sarea->ringOwner=me;
-
-
-   _I810RefreshLpRing(imesa, 0);  
-
-   if (imesa->dirty) 
-       i810EmitHwStateLocked( imesa );  
-
+   
+   sarea->last_quiescent = -1;	/* just kill it for now */
 }
 
 
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810clear.c b/xc/lib/GL/mesa/src/drv/i810/i810clear.c
deleted file mode 100644
index 5ba264059..000000000
--- a/xc/lib/GL/mesa/src/drv/i810/i810clear.c
+++ /dev/null
@@ -1,193 +0,0 @@
-
-
-#include "types.h"
-#include "vbrender.h"
-#include "i810log.h"
-
-#include <stdio.h>
-#include <stdlib.h>
-#include <unistd.h>
-
-#include "mm.h"
-#include "i810lib.h"
-#include "i810dd.h"
-#include "i810clear.h"
-#include "i810state.h"
-#include "i810tris.h"
-
-
-
-/* Clear the depthbuffer.  Always uses the auxillary cliprects and
- * origin from 'dPriv'.
- */
-static void i810_clear_depthbuffer( GLcontext *ctx, 
-				    GLboolean all,
-				    GLint cx, GLint cy, 
-				    GLint cwidth, GLint cheight ) 
-{
-   i810ContextPtr imesa = I810_CONTEXT( ctx );
-   GLuint zval = (GLuint) (ctx->Depth.Clear * DEPTH_SCALE);
-   i810ScreenPrivate *i810Screen = imesa->i810Screen;
-   __DRIdrawablePrivate *dPriv = imesa->driDrawable;
-
-/*     int _nc = dPriv->numAuxClipRects; */
-/*     XF86DRIClipRectPtr box = dPriv->pAuxClipRects; */
-
-   int _nc = imesa->numClipRects;
-   XF86DRIClipRectPtr box = imesa->pClipRects;
-
-   cy = dPriv->h-cy-cheight;
-   cx += dPriv->auxX;
-   cy += dPriv->auxY;
-         
-   while (_nc--) {	 
-      GLint x = box[_nc].x1;
-      GLint y = box[_nc].y1;
-      GLint width = box[_nc].x2 - x;
-      GLint height = box[_nc].y2 - y;
-
-      if (!all) {
-	 if (x < cx) width -= cx - x, x = cx; 
-	 if (y < cy) height -= cy - y, y = cy;
-	 if (x + width > cx + cwidth) width = cx + cwidth - x;
-	 if (y + height > cy + cheight) height = cy + cheight - y;
-	 if (width <= 0) continue;
-	 if (height <= 0) continue;
-      }
-
-      if (I810_DEBUG&DEBUG_VERBOSE_2D)
-	 fprintf(stderr, "clear depth rect %d,%d %dx%d\n", 
-		 (int) x, (int) y, (int) width, (int) height);
-
-      {
-	 int start = (i810Screen->depthOffset + 
-		      y * i810Screen->auxPitch + 
-		      x * 2);
-	    
-	 BEGIN_BATCH(imesa, 6);
-   
-	 OUT_BATCH( BR00_BITBLT_CLIENT | BR00_OP_COLOR_BLT | 0x3 );
-	 OUT_BATCH( BR13_SOLID_PATTERN | 
-		    (0xF0 << 16) |
-		    i810Screen->auxPitch );
-	 OUT_BATCH( (height << 16) | (width * 2));
-	 OUT_BATCH( start );
-	 OUT_BATCH( zval );
-	 OUT_BATCH( 0 );
-
-	 ADVANCE_BATCH();   
-      }
-   }
-}
-
-
-/* Clear the current drawbuffer.  Checks 'imesa' to determine which set of
- * cliprects and origin to use.
- */
-static void i810_clear_colorbuffer( GLcontext *ctx, 
-				    GLboolean all,
-				    GLint cx, GLint cy, 
-				    GLint cwidth, GLint cheight ) 
-{
-   i810ContextPtr imesa = I810_CONTEXT( ctx );
-   GLushort c = imesa->ClearColor;
-   i810ScreenPrivate *i810Screen = imesa->i810Screen;
-   __DRIdrawablePrivate *dPriv = imesa->driDrawable;
-
-   int _nc = imesa->numClipRects;
-   XF86DRIClipRectPtr box = imesa->pClipRects;
-
-   cy = dPriv->h-cy-cheight;
-   cx += imesa->drawX;
-   cy += imesa->drawY;
-         
-   while (_nc--) {	 
-      GLint x = box[_nc].x1;
-      GLint y = box[_nc].y1;
-      GLint width = box[_nc].x2 - x;
-      GLint height = box[_nc].y2 - y;
-
-      if (!all) {
-	 if (x < cx) width -= cx - x, x = cx; 
-	 if (y < cy) height -= cy - y, y = cy;
-	 if (x + width > cx + cwidth) width = cx + cwidth - x;
-	 if (y + height > cy + cheight) height = cy + cheight - y;
-	 if (width <= 0) continue;
-	 if (height <= 0) continue;
-      }
-
-      if (I810_DEBUG&DEBUG_VERBOSE_2D)
-	 fprintf(stderr, "clear color rect %d,%d %dx%d\n", 
-		 (int) x, (int) y, (int) width, (int) height);
-
-
-      {
-	 int start = (imesa->drawOffset +  
-		      y * i810Screen->auxPitch + 
-		      x * i810Screen->cpp);
-	    
-	 BEGIN_BATCH( imesa, 6 );
-	    
-	 OUT_BATCH( BR00_BITBLT_CLIENT | BR00_OP_COLOR_BLT | 0x3 );
-	 OUT_BATCH( BR13_SOLID_PATTERN | 
-		    (0xF0 << 16) |
-		    i810Screen->auxPitch );
-	 OUT_BATCH( (height << 16) | (width * i810Screen->cpp));
-	 OUT_BATCH( start );
-	 OUT_BATCH( c );
-	 OUT_BATCH( 0 );
-
-	 ADVANCE_BATCH();
-      }
-   }
-}
-
-
-
-
-/*
- * i810Clear
- *
- * Clear the color and/or depth buffers.  If 'all' is GL_TRUE, clear
- * whole buffer, otherwise clear the region defined by the remaining
- * parameters.  
- */
-GLbitfield i810Clear( GLcontext *ctx, GLbitfield mask, GLboolean all,
-		      GLint cx, GLint cy, GLint cwidth, GLint cheight ) 
-{
-   i810ContextPtr imesa = I810_CONTEXT( ctx );
-
-   if (I810_DEBUG&DEBUG_VERBOSE_API)
-      fprintf(stderr, "i810Clear( %d, %d, %d, %d, %d )\n",
-	      (int)mask, (int)cx, (int)cy, (int)cwidth, (int)cheight );
-	
-   LOCK_HARDWARE(imesa);
-
-   {
-      BEGIN_BATCH( imesa, 2 );   
-      OUT_BATCH( INST_PARSER_CLIENT | INST_OP_FLUSH );
-      OUT_BATCH( 0 );
-      ADVANCE_BATCH();      
-   }
-
-   if (mask & GL_COLOR_BUFFER_BIT) {
-      i810_clear_colorbuffer( ctx, all, cx, cy, cwidth, cheight );
-      mask &= ~GL_COLOR_BUFFER_BIT;
-   }
-
-   if ( (mask & GL_DEPTH_BUFFER_BIT) && ctx->Depth.Mask ) {
-      i810_clear_depthbuffer( ctx, all, cx, cy, cwidth, cheight );
-      mask &= ~GL_DEPTH_BUFFER_BIT;
-   }      
-
-   {
-      BEGIN_BATCH( imesa, 2 );   
-      OUT_BATCH( INST_PARSER_CLIENT | INST_OP_FLUSH );
-      OUT_BATCH( 0 );
-      ADVANCE_BATCH();      
-   }
-
-   UNLOCK_HARDWARE( imesa );
-   return mask;
-}
-
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810clear.h b/xc/lib/GL/mesa/src/drv/i810/i810clear.h
deleted file mode 100644
index aedbf84d9..000000000
--- a/xc/lib/GL/mesa/src/drv/i810/i810clear.h
+++ /dev/null
@@ -1,7 +0,0 @@
-#ifndef _I810_CLEAR_H
-#define _I810_CLEAR_H
-
-extern GLbitfield i810Clear( GLcontext *ctx, GLbitfield mask, GLboolean all,
-			    GLint x, GLint y, GLint width, GLint height );
-
-#endif
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810context.h b/xc/lib/GL/mesa/src/drv/i810/i810context.h
index 9f316f6fd..629abf0b8 100644
--- a/xc/lib/GL/mesa/src/drv/i810/i810context.h
+++ b/xc/lib/GL/mesa/src/drv/i810/i810context.h
@@ -28,24 +28,32 @@
 
 typedef struct i810_context_t i810Context;
 typedef struct i810_context_t *i810ContextPtr;
+typedef struct i810_texture_object_t *i810TextureObjectPtr;
 
 #include <X11/Xlibint.h>
 #include "dri_tmm.h"
 #include "dri_mesaint.h"
 #include "dri_mesa.h"
-#include "xmesaP.h"
 
 #include "types.h"
 #include "i810_init.h"
-#include "i810_sarea.h"
+#include "drm.h"
+#include "mm.h"
+#include "i810log.h"
 
 #include "i810tex.h"
 #include "i810vb.h"
 
 
-
-#define I810_FALLBACK_TEXTURE   0x1
-#define I810_FALLBACK_BUFFER    0x2
+/* Reasons to fallback on all primitives.  (see also
+ * imesa->IndirectTriangles).  
+ */
+#define I810_FALLBACK_TEXTURE        0x1
+#define I810_FALLBACK_DRAW_BUFFER    0x2
+#define I810_FALLBACK_READ_BUFFER    0x4
+#define I810_FALLBACK_COLORMASK      0x8  
+#define I810_FALLBACK_STIPPLE        0x10  
+#define I810_FALLBACK_SPECULAR       0x20 
 
 
 
@@ -53,17 +61,6 @@ typedef struct i810_context_t *i810ContextPtr;
  */
 #define I810_NEW_TEXTURE 0x1
 
-/* for i810ctx.dirty - manage driver->hw state changes, including
- * lost contexts.
- */
-#define I810_REQUIRE_QUIESCENT 0x1
-#define I810_UPLOAD_TEX0IMAGE  0x2
-#define I810_UPLOAD_TEX1IMAGE  0x4
-#define I810_UPLOAD_CTX        0x8
-#define I810_UPLOAD_BUFFERS    0x10
-#define I810_REFRESH_RING      0x20
-#define I810_EMIT_CLIPRECT     0x40
-
 
 typedef void (*i810_interp_func)( GLfloat t, 
 				  GLfloat *result,
@@ -88,7 +85,6 @@ struct i810_context_t {
     */
    GLuint Setup[I810_CTX_SETUP_SIZE];
    GLuint BufferSetup[I810_DEST_SETUP_SIZE];
-   GLuint ClipSetup[I810_CLIP_SETUP_SIZE];
    
 
    /* Support for CVA and the fast paths.
@@ -125,24 +121,38 @@ struct i810_context_t {
 
    /* DRI stuff
     */
+   drmBufPtr  vertex_dma_buffer;
+   GLuint vertex_prim;
+
    GLframebuffer *glBuffer;
    
+   /* Two flags to keep track of fallbacks.
+    */
+   GLuint IndirectTriangles;
    GLuint Fallback;
+
+
    GLuint needClip;
 
    /* These refer to the current draw (front vs. back) buffer:
     */
-   int drawOffset;		/* draw buffer address in agp space */
+   char *drawMap;		/* draw buffer address in virtual mem */
+   char *readMap;	
    int drawX;			/* origin of drawable in draw buffer */
    int drawY;
    GLuint numClipRects;		/* cliprects for that buffer */
    XF86DRIClipRectPtr pClipRects;
 
-
    int lastSwap;
+   int secondLastSwap;
    int texAge;
+   int ctxAge;
+   int dirtyAge;
+   int any_contend;		/* throttle me harder */
 
-   XF86DRIClipRectRec draw_rect;
+   int scissor;
+   drm_clip_rect_t draw_rect;
+   drm_clip_rect_t scissor_rect;
 
    drmContext hHWContext;
    drmLock *driHwLock;
@@ -152,7 +162,7 @@ struct i810_context_t {
    __DRIdrawablePrivate *driDrawable;
    __DRIscreenPrivate *driScreen;
    i810ScreenPrivate *i810Screen; 
-   I810SAREAPriv *sarea;
+   drm_i810_sarea_t *sarea;
 };
 
 
@@ -161,7 +171,7 @@ struct i810_context_t {
  */
 #define I810_DEBUG 0   
 #ifndef I810_DEBUG
-/*  #warning "Debugging enabled - expect reduced performance" */
+#warning "Debugging enabled - expect reduced performance"
 extern int I810_DEBUG;
 #endif
 
@@ -176,6 +186,7 @@ extern int I810_DEBUG;
 #define DEBUG_VALIDATE_RING  0x800
 #define DEBUG_VERBOSE_LRU    0x1000
 #define DEBUG_VERBOSE_DRI    0x2000
+#define DEBUG_VERBOSE_IOCTL  0x4000
 
 
 
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810dd.c b/xc/lib/GL/mesa/src/drv/i810/i810dd.c
index 17436bf0d..7c04f6d56 100644
--- a/xc/lib/GL/mesa/src/drv/i810/i810dd.c
+++ b/xc/lib/GL/mesa/src/drv/i810/i810dd.c
@@ -29,10 +29,7 @@
 #include <stdio.h>
 
 #include "mm.h"
-#include "i810lib.h"
-#include "i810clear.h"
 #include "i810dd.h"
-#include "i810depth.h"
 #include "i810log.h"
 #include "i810state.h"
 #include "i810span.h"
@@ -59,7 +56,7 @@ static const GLubyte *i810DDGetString( GLcontext *ctx, GLenum name )
    case GL_VENDOR:
       return "Keith Whitwell, Precision Insight Inc.";
    case GL_RENDERER:
-      return "DRI-I810";
+      return "Mesa DRI I810 20000415";
    default:
       return 0;
    }
@@ -100,10 +97,13 @@ void i810DDExtensionsInit( GLcontext *ctx )
    /* The imaging subset of 1.2 isn't supported by any mesa driver.
     */
    gl_extensions_disable( ctx, "ARB_imaging" );
+   gl_extensions_disable( ctx, "GL_EXT_blend_color" );
    gl_extensions_disable( ctx, "GL_EXT_blend_minmax" );
    gl_extensions_disable( ctx, "GL_EXT_blend_logic_op" );
    gl_extensions_disable( ctx, "GL_EXT_blend_subtract" );
    gl_extensions_disable( ctx, "GL_INGR_blend_func_separate" );
+   gl_extensions_disable( ctx, "GL_EXT_texture_lod_bias" );   
+   gl_extensions_disable( ctx, "GL_MESA_resize_buffers" );   
 
 
    if (0) gl_extensions_disable( ctx, "GL_ARB_multitexture" );
@@ -115,18 +115,6 @@ void i810DDExtensionsInit( GLcontext *ctx )
 }
 
 
-static void i810DDFlush( GLcontext *ctx )
-{
-   i810DmaFlush( I810_CONTEXT(ctx) );
-}
-
-static void i810DDFinish( GLcontext *ctx )
-{
-   i810ContextPtr imesa = I810_CONTEXT(ctx);
-   LOCK_HARDWARE( imesa );
-   i810DmaFinish( imesa );
-   UNLOCK_HARDWARE( imesa );
-}
 
 
 void i810DDInitDriverFuncs( GLcontext *ctx )
@@ -139,8 +127,5 @@ void i810DDInitDriverFuncs( GLcontext *ctx )
    ctx->Driver.Clear = i810Clear;
 
 
-   ctx->Driver.Flush          = i810DDFlush;
-   ctx->Driver.Finish         = i810DDFinish;
-
    ctx->Driver.BuildPrecalcPipeline = i810DDBuildPrecalcPipeline;
 }
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810dma.h b/xc/lib/GL/mesa/src/drv/i810/i810dma.h
deleted file mode 100644
index 9d472719a..000000000
--- a/xc/lib/GL/mesa/src/drv/i810/i810dma.h
+++ /dev/null
@@ -1,186 +0,0 @@
-/*
-  GLX Hardware Device Driver for Intel i810
-  Copyright (C) 1999 Keith Whitwell <keithw@precisioninsight.com>
-  
-  This program is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published by
-  the Free Software Foundation; either version 2 of the License, or
-  any later version.
-  
-  This program is distributed in the hope that it will be useful,
-  but WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-  GNU General Public License for more details.
-  
-  You should have received a copy of the GNU General Public License
-  along with this program; if not, write to the Free Software
-  Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
-  
- original by Jeff Hartmann <slicer@ionet.net>
-  6/16/99: rewrite by John Carmack <johnc@idsoftware.com>
-  Oct 99: port to i180 by Keith Whitwell <keithw@precisioninsight.com>
-*/
-
-#ifndef I810DMA_H
-#define I810DMA_H
-
-#include "i810lib.h"
-#include "mm.h"
-
-/* a flush command will guarantee that all data added to the dma buffer
-is on its way to the card, and will eventually complete with no more
-intervention.
-*/
-void i810DmaFlush( i810ContextPtr imesa );
-
-/* the overflow function is called when a block can't be allocated
-in the current dma buffer.  It flushes the current buffer and
-records some information */
-void i810DmaOverflow(int newDwords);
-
-/* a finish command will guarantee that all dma commands have actually
-been consumed by the card.  Note that there may still be a couple primitives
-that have not yet been executed out of the internal FIFO, so this does not
-guarantee that the drawing engine is idle. */
-void i810DmaFinish( i810ContextPtr imesa );
-
-
-
-typedef struct {
-   unsigned long Start;
-   unsigned long End;
-   unsigned long Size;
-} I810MemRange;
-
-
-
-typedef struct i810_batch_buffer {
-   I810MemRange mem;
-   char *virtual_start;
-   int head;
-   int space;
-   int additional_space;
-   int texture_age;
-} i810BatchBuffer;
-
-
-
-#define I810_USE_BATCH 0
-#if I810_USE_BATCH
-
-#define BEGIN_BATCH( imesa, n )						\
-   unsigned int outbatch;						\
-   volatile char *virt;							\
-   if (I810_DEBUG & DEBUG_VERBOSE_RING)					\
-      fprintf(stderr,							\
-      "BEGIN_BATCH(%d) in %s\n"						\
-      "(spc left %d/%ld, head %x vstart %p start %lx)\n",		\
-      n, __FUNCTION__, i810glx.dma_buffer->space,			\
-      i810glx.dma_buffer->mem.Size - i810glx.dma_buffer->head,		\
-      i810glx.dma_buffer->head,						\
-      i810glx.dma_buffer->virtual_start,				\
-      i810glx.dma_buffer->mem.Start );					\
-   if (i810glx.dma_buffer->space < n*4)					\
-       i810DmaOverflow(n);						\
-   outbatch = i810glx.dma_buffer->head;					\
-   virt = i810glx.dma_buffer->virtual_start;			
-
-
-#define OUT_BATCH(val) {					\
-   *(volatile unsigned int *)(virt + outbatch) = val;		\
-   if (I810_DEBUG & DEBUG_VERBOSE_RING)				\
-      fprintf(stderr, "OUT_BATCH %x: %x\n", (int)(outbatch/4), (int)(val));	\
-   outbatch += 4;						\
-}
-   
-
-#define ADVANCE_BATCH() {						\
-   if (I810_DEBUG & DEBUG_VERBOSE_RING)					\
-      fprintf(stderr, "ADVANCE_BATCH(%f) in %s\n",			\
-                  (outbatch - i810glx.dma_buffer->head) / 4.0,		\
-                  __FUNCTION__);					\
-   i810glx.dma_buffer->space -= outbatch - i810glx.dma_buffer->head;	\
-   i810glx.dma_buffer->head = outbatch;					\
-}
-
-#define FINISH_PRIM()
-
-#else
-
-
-/* As an alternate path to dma, we can export the registers (and hence
- * ringbuffer) to the client.  For security we should alloc these
- * things in agp memory which isn't exported to the client.  But then
- * again, the client could blit over such things if it wanted to hang
- * the system.
- *
- * Malicious clients aren't really delt with by the DRI.
- */
-
-
-extern void _I810RefreshLpRing( i810ContextPtr imesa, int update );
-extern int _I810WaitLpRing( i810ContextPtr imesa, int n, int timeout_usec );
-extern int _I810Sync( i810ContextPtr imesa );
-
-
-#define OUT_RING(n) {					\
-   if (I810_DEBUG & DEBUG_VERBOSE_RING)			\
-      fprintf(stderr, "OUT_RING %x: %x\n", outring, (int)(n));	\
-   if (!(I810_DEBUG & DEBUG_NO_OUTRING)) {		\
-      *(volatile unsigned int *)(virt + outring) = n;	\
-      outring += 4;					\
-      outring &= ringmask;				\
-   }							\
-}
-
-#define ADVANCE_LP_RING() {			\
-    i810glx.LpRing.tail = outring;		\
-    OUTREG(LP_RING + RING_TAIL, outring);	\
-}
-
-#define BEGIN_LP_RING(imesa, n)						\
-   unsigned int outring, ringmask;					\
-   volatile char *virt;							\
-   if ((I810_DEBUG&DEBUG_ALWAYS_SYNC) && n>2) _I810Sync( imesa );	\
-   if (i810glx.LpRing.space < n*4) _I810WaitLpRing( imesa, n*4, 0);	\
-   i810glx.LpRing.space -= n*4;						\
-   if (I810_DEBUG & DEBUG_VERBOSE_RING) 				\
-      fprintf(stderr, "BEGIN_LP_RING %d in %s\n", n, __FUNCTION__);	\
-   if (I810_DEBUG & DEBUG_VALIDATE_RING) { 				\
-      CARD32 tail = INREG(LP_RING+RING_TAIL);				\
-      if (tail != i810glx.LpRing.tail) {				\
-	 fprintf(stderr, "tail %x pI810->LpRing.tail %x\n",		\
-		    (int) tail, (int) i810glx.LpRing.tail);				\
-         exit(1);							\
-      }									\
-   }									\
-   i810glx.LpRing.synced = 0;						\
-   outring = i810glx.LpRing.tail;					\
-   ringmask = i810glx.LpRing.tail_mask;					\
-   virt = i810glx.LpRing.virtual_start;			
-
-
-#define EXTRAAA \
-
-#define INREG(addr)         *(volatile CARD32 *)(i810glx.MMIOBase + (addr))
-#define INREG16(addr)       *(volatile CARD16 *)(i810glx.MMIOBase + (addr))
-#define INREG8(addr)        *(volatile CARD8  *)(i810glx.MMIOBase + (addr))
-
-#define OUTREG(addr, val) do {					\
-   if (I810_DEBUG&DEBUG_VERBOSE_OUTREG)				\
-      fprintf(stderr, "OUTREG(%x, %x)\n", addr, val);		\
-   if (!(I810_DEBUG&DEBUG_NO_OUTREG))				\
-      *(volatile CARD32 *)(i810glx.MMIOBase + (addr)) = (val);	\
-} while (0)
-
-
-#define BEGIN_BATCH(imesa, n) BEGIN_LP_RING(imesa, n)
-#define ADVANCE_BATCH() ADVANCE_LP_RING()
-#define OUT_BATCH(val) OUT_RING(val)
-#define FINISH_PRIM() OUTREG(LP_RING + RING_TAIL, i810glx.LpRing.tail)
-
-#endif
-
-
-
-#endif
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810ioctl.c b/xc/lib/GL/mesa/src/drv/i810/i810ioctl.c
new file mode 100644
index 000000000..b4e86cff2
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/i810/i810ioctl.c
@@ -0,0 +1,477 @@
+#include <stdio.h>
+#include <unistd.h>
+
+
+#include "types.h"
+#include "pb.h"
+#include "dd.h"
+
+#include "mm.h"
+#include "i810context.h"
+#include "i810log.h"
+#include "i810ioctl.h"
+
+#include "drm.h"
+#include <sys/ioctl.h>
+
+static drmBufPtr i810_get_buffer_ioctl( i810ContextPtr imesa )
+{
+   drm_i810_dma_t dma;
+   drmBufPtr buf;
+   int retcode;
+   
+   if (I810_DEBUG&DEBUG_VERBOSE_IOCTL)
+      fprintf(stderr,  "Getting dma buffer\n");
+
+   while (1) {
+      retcode = ioctl(imesa->driFd, DRM_IOCTL_I810_GETBUF, &dma);
+
+      if (dma.granted == 1 && retcode == 0) 
+	 break;
+
+      if (I810_DEBUG&DEBUG_VERBOSE_IOCTL)
+	 fprintf(stderr, "Retcode : %d, granted : %d\n", retcode, dma.granted);
+
+      ioctl(imesa->driFd, DRM_IOCTL_I810_FLUSH);
+   }
+
+   if (I810_DEBUG&DEBUG_VERBOSE_IOCTL)
+      fprintf(stderr, 
+	      "imesa->i810Screen->bufs->list : %p, "
+	      "dma.request_idx : %d\n", 
+	      imesa->i810Screen->bufs->list, dma.request_idx);
+
+   buf = &(imesa->i810Screen->bufs->list[dma.request_idx]);
+   buf->idx = dma.request_idx;
+   buf->used = 0;
+   buf->total = dma.request_size;
+   buf->address = (drmAddress)dma.virtual;
+   return buf;
+}
+
+
+
+#define DEPTH_SCALE ((1<<16)-1)
+
+GLbitfield i810Clear( GLcontext *ctx, GLbitfield mask, GLboolean all,
+		      GLint cx, GLint cy, GLint cw, GLint ch ) 
+{
+   i810ContextPtr imesa = I810_CONTEXT( ctx );
+   __DRIdrawablePrivate *dPriv = imesa->driDrawable;
+   const GLuint colorMask = *((GLuint *) &ctx->Color.ColorMask);
+   drm_i810_clear_t clear;
+   int i;
+
+   clear.flags = 0;
+   clear.clear_color = imesa->ClearColor;
+   clear.clear_depth = (GLuint) (ctx->Depth.Clear * DEPTH_SCALE);
+
+   FLUSH_BATCH( imesa );
+	
+   if ((mask & DD_FRONT_LEFT_BIT) && colorMask == ~0) {
+      clear.flags |= I810_FRONT;
+      mask &= ~DD_FRONT_LEFT_BIT;
+   }
+
+   if ((mask & DD_BACK_LEFT_BIT) && colorMask == ~0) {
+      clear.flags |= I810_BACK;
+      mask &= ~DD_BACK_LEFT_BIT;
+   }
+
+   if ((mask & DD_DEPTH_BIT) && ctx->Depth.Mask) {
+      clear.flags |= I810_DEPTH;
+      mask &= ~DD_DEPTH_BIT;
+   }
+
+   if (!clear.flags)
+      return mask;
+
+   LOCK_HARDWARE( imesa );
+
+   /* flip top to bottom */
+   cy = dPriv->h-cy-ch;
+   cx += imesa->drawX;
+   cy += imesa->drawY;
+
+   if (I810_DEBUG&DEBUG_VERBOSE_IOCTL)
+      fprintf(stderr, "Clear, bufs %x nbox %d\n", 
+	      (int)clear.flags, (int)imesa->numClipRects);
+
+   for (i = 0 ; i < imesa->numClipRects ; ) 
+   { 	 
+      int nr = MIN2(i + I810_NR_SAREA_CLIPRECTS, imesa->numClipRects);
+      XF86DRIClipRectRec *box = imesa->pClipRects;	 
+      drm_clip_rect_t *b = imesa->sarea->boxes;
+      int n = 0;
+
+      if (!all) {
+	 for ( ; i < nr ; i++) {
+	    GLint x = box[i].x1;
+	    GLint y = box[i].y1;
+	    GLint w = box[i].x2 - x;
+	    GLint h = box[i].y2 - y;
+
+	    if (x < cx) w -= cx - x, x = cx; 
+	    if (y < cy) h -= cy - y, y = cy;
+	    if (x + w > cx + cw) w = cx + cw - x;
+	    if (y + h > cy + ch) h = cy + ch - y;
+	    if (w <= 0) continue;
+	    if (h <= 0) continue;
+
+	    b->x1 = x;
+	    b->y1 = y;
+	    b->x2 = x + w;
+	    b->y2 = y + h;
+	    b++;
+	    n++;
+	 }
+      } else {
+	 for ( ; i < nr ; i++) {
+	    *b++ = *(drm_clip_rect_t *)&box[i];
+	    n++;
+	 }
+      }
+
+      imesa->sarea->nbox = n;
+      ioctl(imesa->driFd, DRM_IOCTL_I810_CLEAR, &clear);
+   }
+
+   UNLOCK_HARDWARE( imesa );
+   imesa->dirty |= I810_UPLOAD_CLIPRECTS;
+
+   return mask;
+}
+
+
+
+
+/*
+ * Copy the back buffer to the front buffer. 
+ */
+void i810SwapBuffers( i810ContextPtr imesa ) 
+{
+   __DRIdrawablePrivate *dPriv = imesa->driDrawable;
+   XF86DRIClipRectPtr pbox;
+   int nbox;
+   int i;
+   int tmp;
+
+   FLUSH_BATCH( imesa );
+   LOCK_HARDWARE( imesa );
+   
+   pbox = dPriv->pClipRects;
+   nbox = dPriv->numClipRects;
+
+   for (i = 0 ; i < nbox ; )
+   {
+      int nr = MIN2(i + I810_NR_SAREA_CLIPRECTS, dPriv->numClipRects);
+      XF86DRIClipRectRec *b = (XF86DRIClipRectRec *)imesa->sarea->boxes;
+
+      imesa->sarea->nbox = nr - i;
+
+      for ( ; i < nr ; i++) 
+	 *b++ = pbox[i];
+
+      ioctl(imesa->driFd, DRM_IOCTL_I810_SWAP);
+   }
+
+   tmp = GET_ENQUEUE_AGE(imesa);
+   UNLOCK_HARDWARE( imesa );
+
+   if (GET_DISPATCH_AGE(imesa) < imesa->lastSwap)
+      i810WaitAge(imesa, imesa->lastSwap);
+
+   imesa->lastSwap = tmp;
+   imesa->dirty |= I810_UPLOAD_CLIPRECTS;
+}
+
+
+
+
+
+
+/* This waits for *everybody* to finish rendering -- overkill.
+ */
+void i810DmaFinish( i810ContextPtr imesa  ) 
+{
+   FLUSH_BATCH( imesa );
+
+   if (imesa->sarea->last_quiescent != imesa->sarea->last_enqueue) {
+     
+      if (I810_DEBUG&DEBUG_VERBOSE_IOCTL) 
+	 fprintf(stderr, "i810DmaFinish\n");
+
+      LOCK_HARDWARE( imesa );
+      i810RegetLockQuiescent( imesa );
+      UNLOCK_HARDWARE( imesa );
+      imesa->sarea->last_quiescent = imesa->sarea->last_enqueue;
+   }
+}
+
+
+void i810RegetLockQuiescent( i810ContextPtr imesa  ) 
+{
+   if (imesa->sarea->last_quiescent != imesa->sarea->last_enqueue) {
+      if (I810_DEBUG&DEBUG_VERBOSE_IOCTL)
+	 fprintf(stderr, "i810RegetLockQuiescent\n");
+
+      drmUnlock(imesa->driFd, imesa->hHWContext);
+      i810GetLock( imesa, DRM_LOCK_QUIESCENT ); 
+      imesa->sarea->last_quiescent = imesa->sarea->last_enqueue;
+   }
+}
+
+void i810WaitAgeLocked( i810ContextPtr imesa, int age  ) 
+{
+   int i = 0;
+
+
+   while (++i < 500000 && GET_DISPATCH_AGE(imesa) < age) {
+      ioctl(imesa->driFd, DRM_IOCTL_I810_GETAGE);
+   }
+
+   if (GET_DISPATCH_AGE(imesa) < age) {
+      if (0)
+	 fprintf(stderr, "wait locked %d %d\n", age, GET_DISPATCH_AGE(imesa));
+      ioctl(imesa->driFd, DRM_IOCTL_I810_FLUSH);
+   }
+}
+
+
+void i810WaitAge( i810ContextPtr imesa, int age  ) 
+{
+   int i = 0;
+
+   while (++i < 500000 && GET_DISPATCH_AGE(imesa) < age) {
+      ioctl(imesa->driFd, DRM_IOCTL_I810_GETAGE);
+   }
+
+   if (GET_DISPATCH_AGE(imesa) >= age)
+      return;
+
+   i = 0;
+   while (++i < 1000 && GET_DISPATCH_AGE(imesa) < age) {
+      ioctl(imesa->driFd, DRM_IOCTL_I810_GETAGE);
+      usleep(1000);
+   }
+   
+   /* To be effective at letting other clients at the hardware,
+    * particularly the X server which regularly needs quiescence to
+    * touch the framebuffer, we really need to sleep *beyond* the
+    * point where our last buffer clears the hardware.  
+    */
+   if (imesa->any_contend) {
+      usleep(3000); 
+   }
+
+   imesa->any_contend = 0;
+
+   if (GET_DISPATCH_AGE(imesa) < age) {
+      LOCK_HARDWARE(imesa);
+      if (GET_DISPATCH_AGE(imesa) < age) 
+	 ioctl(imesa->driFd, DRM_IOCTL_I810_FLUSH);
+      UNLOCK_HARDWARE(imesa);
+   }
+}
+
+
+
+void i810FlushVertices( i810ContextPtr imesa ) 
+{
+   if (!imesa->vertex_dma_buffer) return;
+
+   LOCK_HARDWARE( imesa );
+   i810FlushVerticesLocked( imesa );
+   UNLOCK_HARDWARE( imesa );
+}
+
+
+static int intersect_rect( drm_clip_rect_t *out,
+			    drm_clip_rect_t *a,
+			    drm_clip_rect_t *b )
+{
+   *out = *a;
+   if (b->x1 > out->x1) out->x1 = b->x1;
+   if (b->y1 > out->y1) out->y1 = b->y1;
+   if (b->x2 < out->x2) out->x2 = b->x2;
+   if (b->y2 < out->y2) out->y2 = b->y2;
+   if (out->x1 >= out->x2) return 0;
+   if (out->y1 >= out->y2) return 0;
+   return 1;
+}
+
+
+static void age_imesa( i810ContextPtr imesa, int age )
+{
+   if (imesa->CurrentTexObj[0]) imesa->CurrentTexObj[0]->age = age;
+   if (imesa->CurrentTexObj[1]) imesa->CurrentTexObj[1]->age = age;
+}
+
+void i810FlushVerticesLocked( i810ContextPtr imesa )
+{
+   drm_clip_rect_t *pbox = (drm_clip_rect_t *)imesa->pClipRects;
+   int nbox = imesa->numClipRects;
+   drmBufPtr buffer = imesa->vertex_dma_buffer;
+   drm_i810_vertex_t vertex;
+   int i;
+
+   if (I810_DEBUG&DEBUG_VERBOSE_IOCTL)
+      fprintf(stderr, "i810FlushVerticesLocked, buf->used %d\n", 
+	      buffer->used);
+
+   if (!buffer)
+      return;
+
+   if (imesa->dirty & ~I810_UPLOAD_CLIPRECTS)
+      i810EmitHwStateLocked( imesa );
+
+   if (I810_DEBUG&DEBUG_VERBOSE_IOCTL)
+      fprintf(stderr, "i810FlushVerticesLocked, used %d\n",
+	      buffer->used);
+   
+   imesa->vertex_dma_buffer = 0;
+
+   vertex.idx = buffer->idx;
+   vertex.used = buffer->used;
+   vertex.discard = 0;
+
+   if (!nbox)
+      vertex.used = 0;
+
+   if (nbox > I810_NR_SAREA_CLIPRECTS)
+      imesa->dirty |= I810_UPLOAD_CLIPRECTS;
+
+   imesa->sarea->vertex_prim = imesa->vertex_prim;
+
+   if (!nbox || !(imesa->dirty & I810_UPLOAD_CLIPRECTS)) 
+   {
+      if (nbox == 1) 
+	 imesa->sarea->nbox = 0;
+      else
+	 imesa->sarea->nbox = nbox;
+
+      if (I810_DEBUG&DEBUG_VERBOSE_IOCTL)
+	 fprintf(stderr, "DRM_IOCTL_I810_VERTEX CASE1 nbox %d used %d\n", 
+		 nbox, vertex.used);
+
+      vertex.discard = 1;
+      ioctl(imesa->driFd, DRM_IOCTL_I810_VERTEX, &vertex);
+      age_imesa(imesa, imesa->sarea->last_enqueue);
+   }  
+   else 
+   {
+      for (i = 0 ; i < nbox ; )
+      {
+	 int nr = MIN2(i + I810_NR_SAREA_CLIPRECTS, nbox);
+	 drm_clip_rect_t *b = imesa->sarea->boxes;
+
+	 if (imesa->scissor) {
+	    imesa->sarea->nbox = 0;
+	 
+	    for ( ; i < nr ; i++) {
+	       b->x1 = pbox[i].x1 - imesa->drawX;
+	       b->y1 = pbox[i].y1 - imesa->drawY;
+	       b->x2 = pbox[i].x2 - imesa->drawX;
+	       b->y2 = pbox[i].y2 - imesa->drawY;
+
+	       if (intersect_rect(b, b, &imesa->scissor_rect)) {
+		  imesa->sarea->nbox++;
+		  b++;
+	       }
+	    }
+
+	    /* Culled?
+	     */
+	    if (!imesa->sarea->nbox) {
+	       if (nr < nbox) continue;
+	       vertex.used = 0;
+	    }
+	 } else {
+	    imesa->sarea->nbox = nr - i;
+	    for ( ; i < nr ; i++, b++) {
+	       b->x1 = pbox[i].x1 - imesa->drawX;
+	       b->y1 = pbox[i].y1 - imesa->drawY;
+	       b->x2 = pbox[i].x2 - imesa->drawX;
+	       b->y2 = pbox[i].y2 - imesa->drawY;
+	    }
+	 }
+	 
+	 /* Finished with the buffer?
+	  */
+	 if (nr == nbox) 
+	    vertex.discard = 1;
+
+     	 if (I810_DEBUG&DEBUG_VERBOSE_IOCTL)
+	    fprintf(stderr, "DRM_IOCTL_I810_VERTEX nbox %d used %d\n", 
+		    nbox, vertex.used);
+
+	 ioctl(imesa->driFd, DRM_IOCTL_I810_VERTEX, &vertex);
+	 age_imesa(imesa, imesa->sarea->last_enqueue);
+      }
+   }
+
+   imesa->dirty = 0;
+   if (I810_DEBUG&DEBUG_VERBOSE_IOCTL)
+      fprintf(stderr, "finished i810FlushVerticesLocked\n");
+}
+
+
+GLuint *i810AllocDwords( i810ContextPtr imesa, int dwords, GLuint prim )
+{
+   GLuint *start;
+
+   if (!imesa->vertex_dma_buffer) 
+   {
+      if (I810_DEBUG&DEBUG_VERBOSE_IOCTL)
+	 fprintf(stderr, "i810AllocPrimitiveVerts -- get buf\n");
+      LOCK_HARDWARE(imesa);
+      imesa->vertex_dma_buffer = i810_get_buffer_ioctl( imesa );
+      imesa->vertex_dma_buffer->used = 4;
+      imesa->vertex_prim = prim;
+      UNLOCK_HARDWARE(imesa);
+   } 
+   else if (imesa->vertex_prim != prim ||
+	    imesa->vertex_dma_buffer->used + dwords * 4 > 
+	    imesa->vertex_dma_buffer->total) 
+   {
+      if (I810_DEBUG&DEBUG_VERBOSE_IOCTL)
+	 fprintf(stderr, "i810AllocPrimitiveVerts -- flush\n");
+      i810FlushVertices( imesa );
+      LOCK_HARDWARE(imesa);
+      imesa->vertex_dma_buffer = i810_get_buffer_ioctl( imesa );
+      imesa->vertex_dma_buffer->used = 4;
+      imesa->vertex_prim = prim;
+      UNLOCK_HARDWARE(imesa);
+   }
+
+
+   if (0)
+      fprintf(stderr, "i810AllocPrimitiveVerts %d, buf %d, used %d\n",
+	      dwords, imesa->vertex_dma_buffer->idx,
+	      imesa->vertex_dma_buffer->used);
+      
+   start = (GLuint *)((char *)imesa->vertex_dma_buffer->address + 
+		      imesa->vertex_dma_buffer->used);
+
+   imesa->vertex_dma_buffer->used += dwords * 4;
+
+   return start;
+}
+
+static void i810DDFlush( GLcontext *ctx )
+{
+   i810ContextPtr imesa = I810_CONTEXT( ctx );
+   FLUSH_BATCH( imesa );
+}
+
+static void i810DDFinish( GLcontext *ctx  ) 
+{
+   i810ContextPtr imesa = I810_CONTEXT( ctx );
+   i810DmaFinish( imesa );
+}
+
+void i810DDInitIoctlFuncs( GLcontext *ctx )
+{
+   ctx->Driver.Flush = i810DDFlush;
+   ctx->Driver.Finish = i810DDFinish;
+}
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810ioctl.h b/xc/lib/GL/mesa/src/drv/i810/i810ioctl.h
new file mode 100644
index 000000000..004b723b4
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/i810/i810ioctl.h
@@ -0,0 +1,36 @@
+#ifndef MGA_IOCTL_H
+#define MGA_IOCTL_H
+
+#include "i810context.h"
+
+
+GLuint *i810AllocDwords( i810ContextPtr imesa, int dwords, GLuint prim );
+
+void i810GetGeneralDmaBufferLocked( i810ContextPtr mmesa ); 
+
+void i810FlushVertices( i810ContextPtr mmesa ); 
+void i810FlushVerticesLocked( i810ContextPtr mmesa );
+
+void i810FlushGeneralLocked( i810ContextPtr imesa );
+void i810WaitAgeLocked( i810ContextPtr imesa, int age );
+void i810WaitAge( i810ContextPtr imesa, int age );
+
+void i810DmaFinish( i810ContextPtr imesa );
+
+void i810RegetLockQuiescent( i810ContextPtr imesa );
+
+void i810DDInitIoctlFuncs( GLcontext *ctx );
+
+void i810SwapBuffers( i810ContextPtr imesa );
+
+GLbitfield i810Clear( GLcontext *ctx, GLbitfield mask, GLboolean all,
+		      GLint cx, GLint cy, GLint cw, GLint ch );
+
+#define FLUSH_BATCH(imesa) do {						\
+        if (I810_DEBUG&DEBUG_VERBOSE_IOCTL)  				\
+              fprintf(stderr, "FLUSH_BATCH in %s\n", __FUNCTION__);	\
+	if (imesa->vertex_dma_buffer) i810FlushVertices(imesa);		\
+} while (0)
+
+
+#endif
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810lib.h b/xc/lib/GL/mesa/src/drv/i810/i810lib.h
deleted file mode 100644
index ad80674f0..000000000
--- a/xc/lib/GL/mesa/src/drv/i810/i810lib.h
+++ /dev/null
@@ -1,130 +0,0 @@
-/*
- * GLX Hardware Device Driver for Intel i810
- * Copyright (C) 1999 Keith Whitwell
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included
- * in all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * KEITH WHITWELL, OR ANY OTHER CONTRIBUTORS 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 I810LIB_INC
-#define I810LIB_INC
-
-#include <stdio.h>
-
-#include "i810context.h"
-#include "mm.h"
-#include "i810log.h"
-
-
-struct i810_mem_range {
-   unsigned long Start;
-   unsigned long End;
-   unsigned long Size;
-};
-
-struct i810_batch_buffer;
-
-struct i810_ring_buffer {
-   int tail_mask;
-   struct i810_mem_range mem;
-   char *virtual_start;
-   int head;
-   int tail;
-   int space;
-   int synced;
-};
-
-typedef struct {
-   /* logging stuff */
-   GLuint logLevel;
-   FILE *logFile;
-
-   /* bookkeeping for texture swaps */
-   GLuint dma_buffer_age;
-   GLuint current_texture_age;
-        
-   /* options */
-   GLuint nullprims;		/* skip all primitive generation */
-   GLuint boxes;		/* draw performance boxes */
-   GLuint noFallback;		/* don't fall back to software */
-   GLuint skipDma;		/* don't send anything to hardware */
-
-   /* performance counters */	
-   GLuint c_setupPointers;
-   GLuint c_triangles;
-   GLuint c_points;
-   GLuint c_lines;
-   GLuint c_drawWaits;
-   GLuint c_textureSwaps;
-   GLuint c_dmaFlush;
-   GLuint c_overflows;
-
-   GLuint c_ringlost;
-   GLuint c_texlost;
-   GLuint c_ctxlost;
-	
-   GLuint hardwareWentIdle;	/* cleared each swapbuffers, set if a
-				   waitfordmacompletion ever exited
-				   without having to wait */
-
-   unsigned char *texVirtual;
-
-   
-   struct i810_batch_buffer *dma_buffer;
-   struct i810_ring_buffer LpRing;
-   unsigned char *MMIOBase;
-
-} i810Glx_t;
-
-extern i810Glx_t	i810glx;
-
-
-#define I810PACKCOLOR1555(r,g,b,a) \
-  ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \
-    ((a) ? 0x8000 : 0))
-
-#define I810PACKCOLOR565(r,g,b) \
-  ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3))
-
-#define I810PACKCOLOR4444(r,g,b,a) \
-  ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4))
-
-
-#include "i810_3d_reg.h"
-
-static __inline__ GLuint i810PackColor(GLuint format, 
-				       GLubyte r, GLubyte g, 
-				       GLubyte b, GLubyte a)
-{
-  switch (format) {
-  case DV_PF_555:
-    return I810PACKCOLOR1555(r,g,b,a);
-  case DV_PF_565:
-    return I810PACKCOLOR565(r,g,b);
-  default:
-     fprintf(stderr, "unknown format %d\n", (int)format);
-    return 0;
-  }
-}
-
-
-
-
-#endif
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810pipeline.c b/xc/lib/GL/mesa/src/drv/i810/i810pipeline.c
index fb2d368fe..34171d943 100644
--- a/xc/lib/GL/mesa/src/drv/i810/i810pipeline.c
+++ b/xc/lib/GL/mesa/src/drv/i810/i810pipeline.c
@@ -1,13 +1,15 @@
 /*  #include "i810pipeline.h" */
 
 #include <stdio.h>
+
+#include "types.h"
+#include "fog.h"
+
 #include "i810vb.h"
 #include "i810dd.h"
-#include "i810lib.h"
 #include "i810tris.h"
 #include "i810pipeline.h"
 
-#include "fog.h"
 
 static struct gl_pipeline_stage i810_fast_stage = {
    "I810 fast path",
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810span.c b/xc/lib/GL/mesa/src/drv/i810/i810span.c
index 5ea27fd84..aa4f22d0c 100644
--- a/xc/lib/GL/mesa/src/drv/i810/i810span.c
+++ b/xc/lib/GL/mesa/src/drv/i810/i810span.c
@@ -1,31 +1,39 @@
 #include "types.h"
 #include "i810dd.h"
-#include "i810lib.h"
-#include "i810dma.h"
 #include "i810log.h"
 #include "i810span.h"
+#include "i810ioctl.h"
 
 
 #define DBG 0
 
-
 #define LOCAL_VARS					\
-   i810ContextPtr imesa = I810_CONTEXT(ctx);		\
    __DRIdrawablePrivate *dPriv = imesa->driDrawable;	\
-   __DRIscreenPrivate *sPriv = imesa->driScreen;	\
    i810ScreenPrivate *i810Screen = imesa->i810Screen;	\
-   GLuint pitch = i810Screen->auxPitch;			\
+   GLuint pitch = i810Screen->backPitch;		\
    GLuint height = dPriv->h;				\
-   char *buf = (char *)(sPriv->pFB + 			\
-			imesa->drawOffset +	\
-			dPriv->x * 2 + 			\
+   char *buf = (char *)(imesa->drawMap +		\
+			dPriv->x * 2 +			\
+			dPriv->y * pitch);		\
+   char *read_buf = (char *)(imesa->readMap +		\
+			     dPriv->x * 2 +		\
+			     dPriv->y * pitch); 	\
+   GLushort p = I810_CONTEXT( ctx )->MonoColor;         \
+   (void) read_buf; (void) buf; (void) p
+
+#define LOCAL_DEPTH_VARS				\
+   __DRIdrawablePrivate *dPriv = imesa->driDrawable;	\
+   i810ScreenPrivate *i810Screen = imesa->i810Screen;	\
+   GLuint pitch = i810Screen->backPitch;		\
+   GLuint height = dPriv->h;				\
+   char *buf = (char *)(i810Screen->depth.map +	\
+			dPriv->x * 2 +			\
 			dPriv->y * pitch)
 
-#define INIT_MONO_PIXEL(p) \
-   GLushort p = I810_CONTEXT( ctx )->MonoColor;
+#define INIT_MONO_PIXEL(p) 
 
-#define CLIPPIXEL(_x,_y) (_x >= minx && _x <= maxx && \
-			  _y >= miny && _y <= maxy)
+#define CLIPPIXEL(_x,_y) (_x >= minx && _x < maxx && \
+			  _y >= miny && _y < maxy)
 
 
 #define CLIPSPAN(_x,_y,_n,_x1,_n1,_i)				\
@@ -34,15 +42,20 @@
             _n1 = _n;						\
 	    _x1 = _x;						\
 	    if (_x1 < minx) _i += (minx - _x1), _x1 = minx;	\
-	    if (_x1 + _n1 > maxx) n1 -= (_x1 + n1 - maxx);	\
+	    if (_x1 + _n1 >= maxx) n1 -= (_x1 + n1 - maxx) + 1;	\
          }
 
+#define Y_FLIP(_y) (height - _y - 1)
+
+
+#define HW_LOCK()				\
+   i810ContextPtr imesa = I810_CONTEXT(ctx);	\
+   LOCK_HARDWARE_QUIESCENT(imesa);
+
 #define HW_CLIPLOOP()						\
   do {								\
     __DRIdrawablePrivate *dPriv = imesa->driDrawable;		\
     int _nc = dPriv->numClipRects;				\
-    LOCK_HARDWARE(imesa);						\
-    if (!i810glx.LpRing.synced) _I810Sync( imesa );			\
     while (_nc--) {						\
        int minx = dPriv->pClipRects[_nc].x1 - dPriv->x;		\
        int miny = dPriv->pClipRects[_nc].y1 - dPriv->y; 	\
@@ -52,13 +65,16 @@
 
 #define HW_ENDCLIPLOOP()			\
     }						\
-    UNLOCK_HARDWARE(imesa);			\
   } while (0)
 
+#define HW_UNLOCK()				\
+    UNLOCK_HARDWARE(imesa);
+
 
 
 
-#define Y_FLIP(_y) (height - _y)
+/* 16 bit, 565 rgb color spanline and pixel functions
+ */
 #define WRITE_RGBA( _x, _y, r, g, b, a )				\
    *(GLushort *)(buf + _x*2 + _y*pitch)  = ( (((int)r & 0xf8) << 8) |	\
 		                             (((int)g & 0xfc) << 3) |	\
@@ -66,13 +82,13 @@
 #define WRITE_PIXEL( _x, _y, p )  \
    *(GLushort *)(buf + _x*2 + _y*pitch) = p
 
-#define READ_RGBA( rgba, _x, _y )			\
-do {							\
-   GLushort p = *(GLushort *)(buf + _x*2 + _y*pitch);	\
-   rgba[0] = (p >> 8) & 0xf8;				\
-   rgba[1] = (p >> 3) & 0xfc;				\
-   rgba[2] = (p << 3) & 0xf8;				\
-   rgba[3] = 0;			/* or 255? */		\
+#define READ_RGBA( rgba, _x, _y )				\
+do {								\
+   GLushort p = *(GLushort *)(read_buf + _x*2 + _y*pitch);	\
+   rgba[0] = (p >> 8) & 0xf8;					\
+   rgba[1] = (p >> 3) & 0xfc;					\
+   rgba[2] = (p << 3) & 0xf8;					\
+   rgba[3] = 255;						\
 } while(0)
 
 #define TAG(x) i810##x##_565
@@ -81,28 +97,46 @@ do {							\
 
 
 
-
-#define WRITE_RGBA( _x, _y, r, g, b, a )				\
+/* 15 bit, 555 rgb color spanline and pixel functions
+ */
+#define WRITE_RGBA( _x, _y, r, g, b, a )			\
    *(GLushort *)(buf + _x*2 + _y*pitch)  = (((r & 0xf8) << 7) |	\
-		                            ((g & 0xf8) << 3) |		\
+		                            ((g & 0xf8) << 3) |	\
                          		    ((b & 0xf8) >> 3))
 
 #define WRITE_PIXEL( _x, _y, p )  \
    *(GLushort *)(buf + _x*2 + _y*pitch)  = p
 
-#define READ_RGBA( rgba, _x, _y )			\
-do {							\
-   GLushort p = *(GLushort *)(buf + _x*2 + _y*pitch);	\
-   rgba[0] = (p >> 7) & 0xf8;				\
-   rgba[1] = (p >> 3) & 0xf8;				\
-   rgba[2] = (p << 3) & 0xf8;				\
-   rgba[3] = 0;			/* or 255? */		\
+#define READ_RGBA( rgba, _x, _y )				\
+do {								\
+   GLushort p = *(GLushort *)(read_buf + _x*2 + _y*pitch);	\
+   rgba[0] = (p >> 7) & 0xf8;					\
+   rgba[1] = (p >> 3) & 0xf8;					\
+   rgba[2] = (p << 3) & 0xf8;					\
+   rgba[3] = 255;						\
 } while(0)
 
 #define TAG(x) i810##x##_555
 #include "spantmp.h"
 
 
+
+
+/* 16 bit depthbuffer functions.
+ */
+#define WRITE_DEPTH( _x, _y, d ) \
+   *(GLushort *)(buf + _x*2 + _y*pitch)  = d;
+
+#define READ_DEPTH( d, _x, _y )	\
+   d = *(GLushort *)(buf + _x*2 + _y*pitch);	 
+
+/*     d = 0xffff; */
+
+#define TAG(x) i810##x##_16
+#include "depthtmp.h"
+
+
+
 void i810DDInitSpanFuncs( GLcontext *ctx )
 {
    if (1) {
@@ -123,6 +157,11 @@ void i810DDInitSpanFuncs( GLcontext *ctx )
       ctx->Driver.ReadRGBAPixels = i810ReadRGBAPixels_555;
    }
 
+   ctx->Driver.ReadDepthSpan = i810ReadDepthSpan_16;
+   ctx->Driver.WriteDepthSpan = i810WriteDepthSpan_16;
+   ctx->Driver.ReadDepthPixels = i810ReadDepthPixels_16;
+   ctx->Driver.WriteDepthPixels = i810WriteDepthPixels_16;
+
    ctx->Driver.WriteCI8Span        =NULL;
    ctx->Driver.WriteCI32Span       =NULL;
    ctx->Driver.WriteMonoCISpan     =NULL;
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810state.c b/xc/lib/GL/mesa/src/drv/i810/i810state.c
index bc3391e42..80332e3b3 100644
--- a/xc/lib/GL/mesa/src/drv/i810/i810state.c
+++ b/xc/lib/GL/mesa/src/drv/i810/i810state.c
@@ -7,24 +7,40 @@
 #include "dd.h"
 
 #include "mm.h"
-#include "i810lib.h"
 #include "i810dd.h"
 #include "i810context.h"
 #include "i810state.h"
-#include "i810depth.h"
 #include "i810tex.h"
 #include "i810log.h"
 #include "i810vb.h"
 #include "i810tris.h"
+#include "i810ioctl.h"
 
+  
+
+static __inline__ GLuint i810PackColor(GLuint format, 
+				       GLubyte r, GLubyte g, 
+				       GLubyte b, GLubyte a)
+{
+   switch (format) {
+   case DV_PF_555:
+      return I810PACKCOLOR1555(r,g,b,a);
+   case DV_PF_565:
+      return I810PACKCOLOR565(r,g,b);
+   default:
+      fprintf(stderr, "unknown format %d\n", (int)format);
+      return 0;
+   }
+}
 
 
 static void i810DDAlphaFunc(GLcontext *ctx, GLenum func, GLclampf ref)
 {
    i810ContextPtr imesa = I810_CONTEXT(ctx);
-
    CARD32 a = (ZA_UPDATE_ALPHAFUNC|ZA_UPDATE_ALPHAREF);
 
+   FLUSH_BATCH(imesa);
+
    switch (ctx->Color.AlphaFunc) {
    case GL_NEVER:    a |= ZA_ALPHA_NEVER;    break;
    case GL_LESS:     a |= ZA_ALPHA_LESS;     break;
@@ -44,10 +60,6 @@ static void i810DDAlphaFunc(GLcontext *ctx, GLenum func, GLclampf ref)
    imesa->Setup[I810_CTXREG_ZA] |= a;
 }
 
-/* This shouldn't get called, as the extension is disabled.  However,
- * there are internal Mesa calls, and rogue use of the api which must be
- * caught.
- */
 static void i810DDBlendEquation(GLcontext *ctx, GLenum mode) 
 {
    if (mode != GL_FUNC_ADD_EXT) {
@@ -60,9 +72,8 @@ static void i810DDBlendEquation(GLcontext *ctx, GLenum mode)
 static void i810DDBlendFunc(GLcontext *ctx, GLenum sfactor, GLenum dfactor)
 {
    i810ContextPtr imesa = I810_CONTEXT(ctx);
-   GLuint a;
-
-   a = SDM_UPDATE_SRC_BLEND | SDM_UPDATE_DST_BLEND;
+   GLuint a = SDM_UPDATE_SRC_BLEND | SDM_UPDATE_DST_BLEND;
+   FLUSH_BATCH(imesa);
 
    switch (ctx->Color.BlendSrcRGB) {
    case GL_ZERO:                a |= SDM_SRC_ZERO; break;
@@ -73,13 +84,8 @@ static void i810DDBlendFunc(GLcontext *ctx, GLenum sfactor, GLenum dfactor)
    case GL_ONE_MINUS_SRC_ALPHA: a |= SDM_SRC_INV_SRC_ALPHA; break;
    case GL_DST_ALPHA:           a |= SDM_SRC_ONE; break;
    case GL_ONE_MINUS_DST_ALPHA: a |= SDM_SRC_ZERO; break;
-   case GL_SRC_ALPHA_SATURATE:      
-      a |= SDM_SRC_SRC_ALPHA;	/* use GFXRENDERSTATE_COLOR_FACTOR ??? */
-      break;
-   default: 
-      i810Error("unknown blend source func");
-      exit(1);
-      return;
+   case GL_SRC_ALPHA_SATURATE:  a |= SDM_SRC_SRC_ALPHA; break;
+   default: return;
    }
 
    switch (ctx->Color.BlendDstRGB) {
@@ -91,10 +97,7 @@ static void i810DDBlendFunc(GLcontext *ctx, GLenum sfactor, GLenum dfactor)
    case GL_ONE_MINUS_SRC_COLOR: a |= SDM_DST_INV_SRC_COLOR; break;
    case GL_DST_ALPHA:           a |= SDM_DST_ONE; break;
    case GL_ONE_MINUS_DST_ALPHA: a |= SDM_DST_ZERO; break;
-   default: 
-      i810Error( "unknown blend dst func");
-      exit(1);
-      return;      
+   default: return;      
    }  
 
    imesa->dirty |= I810_UPLOAD_CTX;
@@ -123,6 +126,8 @@ static void i810DDDepthFunc(GLcontext *ctx, GLenum func)
    i810ContextPtr imesa = I810_CONTEXT(ctx);
    int zmode;
 
+   FLUSH_BATCH(imesa);
+
    switch(func)  { 
    case GL_NEVER: zmode = LCS_Z_NEVER; break;
    case GL_ALWAYS: zmode = LCS_Z_ALWAYS; break;
@@ -134,9 +139,9 @@ static void i810DDDepthFunc(GLcontext *ctx, GLenum func)
    case GL_NOTEQUAL: zmode = LCS_Z_NOTEQUAL; break;
    default: return;
    }
-
+   
    imesa->Setup[I810_CTXREG_LCS] &= ~LCS_Z_MASK;
-   imesa->Setup[I810_CTXREG_LCS] |= LCS_UPDATE_ZMODE | zmode;
+   imesa->Setup[I810_CTXREG_LCS] |= LCS_UPDATE_ZMODE | zmode;   
    imesa->dirty |= I810_UPLOAD_CTX;
 }
 
@@ -144,20 +149,62 @@ static void i810DDDepthMask(GLcontext *ctx, GLboolean flag)
 {
    i810ContextPtr imesa = I810_CONTEXT(ctx);
 
+   FLUSH_BATCH(imesa);
+
    imesa->dirty |= I810_UPLOAD_CTX;
    imesa->Setup[I810_CTXREG_B2] &= ~B2_ZB_WRITE_ENABLE;
-   imesa->Setup[I810_CTXREG_B2] |= B2_UPDATE_ZB_WRITE_ENABLE;
 
    if (flag)
      imesa->Setup[I810_CTXREG_B2] |= B2_ZB_WRITE_ENABLE;
 }
 
 
+/* =============================================================
+ * Polygon stipple 
+ * 
+ * The i810 supports a 4x4 stipple natively, GL wants 32x32. 
+ * Fortunately stipple is usually a repeating pattern.  Could
+ * also consider using a multitexturing mechanism for this, but
+ * that has real issues, too.
+ */
+static void i810DDPolygonStipple( GLcontext *ctx, const GLubyte *mask )
+{
+   i810ContextPtr imesa = I810_CONTEXT(ctx);
+   const GLubyte *m = mask;
+   GLubyte p[4];
+   int i,j,k;
+   int active = (ctx->Polygon.StippleFlag && ctx->PB->primitive == GL_POLYGON);
 
+   FLUSH_BATCH(imesa);
+   ctx->Driver.TriangleCaps |= DD_TRI_STIPPLE;
 
+   if (active) {
+      imesa->dirty |= I810_UPLOAD_CTX;
+      imesa->Setup[I810_CTXREG_ST1] &= ~ST1_ENABLE;
+   }
 
-
-
+   p[0] = mask[0] & 0xf; p[0] |= p[0] << 4;
+   p[1] = mask[4] & 0xf; p[1] |= p[1] << 4;
+   p[2] = mask[8] & 0xf; p[2] |= p[2] << 4;
+   p[3] = mask[12] & 0xf; p[3] |= p[3] << 4;
+
+   for (k = 0 ; k < 8 ; k++)
+      for (j = 0 ; j < 4; j++) 
+	 for (i = 0 ; i < 4 ; i++) 
+	    if (*m++ != p[j]) {
+	       ctx->Driver.TriangleCaps &= ~DD_TRI_STIPPLE;
+	       return;
+	    }
+   
+   imesa->Setup[I810_CTXREG_ST1] &= ~0xffff;
+   imesa->Setup[I810_CTXREG_ST1] |= ( ((p[0] & 0xf) << 0) |
+				      ((p[1] & 0xf) << 4) |
+				      ((p[2] & 0xf) << 8) |
+				      ((p[3] & 0xf) << 12) );   
+
+   if (active)
+      imesa->Setup[I810_CTXREG_ST1] |= ST1_ENABLE;
+}
 
 
 
@@ -167,40 +214,26 @@ static void i810DDDepthMask(GLcontext *ctx, GLboolean flag)
 
 
 static void i810DDScissor( GLcontext *ctx, GLint x, GLint y, 
-			  GLsizei w, GLsizei h )
+			   GLsizei w, GLsizei h )
 { 
    i810ContextPtr imesa = I810_CONTEXT(ctx);
-   __DRIdrawablePrivate *dPriv = imesa->driDrawable;
-   int x1,x2,y1,y2;
 
-   x1 = ctx->Scissor.X;
-   x2 = ctx->Scissor.X + ctx->Scissor.Width - 1;
-   y1 = dPriv->h - ctx->Scissor.Y - ctx->Scissor.Height;
-   y2 = dPriv->h - ctx->Scissor.Y - 1;
-  
-   if (x1 < 0) x1 = 0;
-   if (y1 < 0) y1 = 0;
-   if (x2 >= dPriv->w) x2 = dPriv->w-1;
-   if (y2 >= dPriv->h) y2 = dPriv->h-1;
-
-   if (x1 > x2 || y1 > y2) {
-      x1 = 0; x2 = 0;
-      y2 = 0; y1 = 1;
-   }
+   FLUSH_BATCH(imesa);
+   imesa->scissor_rect.x1 = x;
+   imesa->scissor_rect.y1 = imesa->driDrawable->h - (y+h);
+   imesa->scissor_rect.x2 = x+w;
+   imesa->scissor_rect.y2 = imesa->driDrawable->h - y;
 
-   /* Need to push this into drawing rectangle.
-    */
-#if 0
-   imesa->Setup[I810_CTXREG_SCI0] = GFX_OP_SCISSOR_INFO;
-   imesa->Setup[I810_CTXREG_SCI1] = (y1<<16)|x1;
-   imesa->Setup[I810_CTXREG_SCI2] = (y2<<16)|x2;
 
+   if (I810_DEBUG&DEBUG_VERBOSE_2D) 
+      fprintf(stderr, "SET SCISSOR %d,%d-%d,%d\n", 
+	      imesa->scissor_rect.x1,
+	      imesa->scissor_rect.y1,
+	      imesa->scissor_rect.x2,
+	      imesa->scissor_rect.y2);
 
-   /* Need to intersect with cliprects???
-    */
-   imesa->dirty |= I810_UPLOAD_CTX;
-#endif
 
+   imesa->dirty |= I810_UPLOAD_CLIPRECTS;
 }
 
 
@@ -209,38 +242,56 @@ static void i810DDDither(GLcontext *ctx, GLboolean enable)
 }
 
 
-static GLboolean i810DDSetBuffer(GLcontext *ctx, GLenum mode )
+static GLboolean i810DDSetDrawBuffer(GLcontext *ctx, GLenum mode )
 {
    i810ContextPtr imesa = I810_CONTEXT(ctx);
 
+   FLUSH_BATCH(imesa);
 
-   fprintf(stderr, "i810DDSetBuffer %s\n", gl_lookup_enum_by_nr( mode ));
-
-   imesa->Fallback &= ~I810_FALLBACK_BUFFER;
+   imesa->Fallback &= ~I810_FALLBACK_DRAW_BUFFER;
 
    if (mode == GL_FRONT_LEFT) 
    {
-      imesa->drawOffset = imesa->i810Screen->fbOffset;
+      imesa->drawMap = imesa->driScreen->pFB;
       imesa->BufferSetup[I810_DESTREG_DI1] = (imesa->i810Screen->fbOffset | 
-					      imesa->i810Screen->auxPitchBits);
+					      imesa->i810Screen->backPitchBits);
       imesa->dirty |= I810_UPLOAD_BUFFERS;
       i810XMesaSetFrontClipRects( imesa );
       return GL_TRUE;
    } 
    else if (mode == GL_BACK_LEFT) 
    {
-      imesa->drawOffset = imesa->i810Screen->backOffset;
+      imesa->drawMap = imesa->i810Screen->back.map;
       imesa->BufferSetup[I810_DESTREG_DI1] = (imesa->i810Screen->backOffset | 
-					      imesa->i810Screen->auxPitchBits);
+					      imesa->i810Screen->backPitchBits);
       imesa->dirty |= I810_UPLOAD_BUFFERS;
       i810XMesaSetBackClipRects( imesa );
       return GL_TRUE;
    }
 
-   imesa->Fallback |= I810_FALLBACK_BUFFER;
+   imesa->Fallback |= I810_FALLBACK_DRAW_BUFFER;
    return GL_FALSE;
 }
 
+static void i810DDSetReadBuffer(GLcontext *ctx, GLframebuffer *colorBuffer,
+				GLenum mode )
+{
+   i810ContextPtr imesa = I810_CONTEXT(ctx);
+
+   if (mode == GL_FRONT_LEFT) 
+   {
+      imesa->readMap = imesa->driScreen->pFB;
+      imesa->Fallback &= ~I810_FALLBACK_READ_BUFFER;
+   } 
+   else if (mode == GL_BACK_LEFT) 
+   {
+      imesa->readMap = imesa->i810Screen->back.map;
+      imesa->Fallback &= ~I810_FALLBACK_READ_BUFFER;
+   }
+   else
+      imesa->Fallback |= I810_FALLBACK_READ_BUFFER;
+}
+
 
 
 static void i810DDSetColor(GLcontext *ctx, 
@@ -248,9 +299,7 @@ static void i810DDSetColor(GLcontext *ctx,
 			   GLubyte b, GLubyte a )
 {
    i810ContextPtr imesa = I810_CONTEXT(ctx);
-
-   imesa->MonoColor = i810PackColor( imesa->i810Screen->fbFormat,
-				      r, g, b, a );
+   imesa->MonoColor = i810PackColor( imesa->i810Screen->fbFormat, r, g, b, a );
 }
 
 
@@ -259,9 +308,7 @@ static void i810DDClearColor(GLcontext *ctx,
 			     GLubyte b, GLubyte a )
 {
    i810ContextPtr imesa = I810_CONTEXT(ctx);
-
-   imesa->ClearColor = i810PackColor( imesa->i810Screen->fbFormat,
-				      r, g, b, a );
+   imesa->ClearColor = i810PackColor( imesa->i810Screen->fbFormat, r, g, b, a );
 }
 
 
@@ -274,6 +321,8 @@ static void i810DDCullFaceFrontFace(GLcontext *ctx, GLenum unused)
    i810ContextPtr imesa = I810_CONTEXT(ctx);
    GLuint mode = LCS_CULL_BOTH;
    
+   FLUSH_BATCH(imesa);
+
    if (ctx->Polygon.CullFaceMode != GL_FRONT_AND_BACK) {
       mode = LCS_CULL_CW;
       if (ctx->Polygon.CullFaceMode == GL_FRONT)
@@ -288,27 +337,49 @@ static void i810DDCullFaceFrontFace(GLcontext *ctx, GLenum unused)
    {
       imesa->dirty |= I810_UPLOAD_CTX;
       imesa->Setup[I810_CTXREG_LCS] &= ~LCS_CULL_MASK;
-      imesa->Setup[I810_CTXREG_LCS] |= (LCS_UPDATE_CULL_MODE | mode);
+      imesa->Setup[I810_CTXREG_LCS] |= mode;
    }
 }
 
 
 static void i810DDReducedPrimitiveChange( GLcontext *ctx, GLenum prim )
 {
-   if (ctx->Polygon.CullFlag) {
-      i810ContextPtr imesa = I810_CONTEXT(ctx);
-      GLuint mode = imesa->LcsCullMode;
-
-      if (ctx->PB->primitive != GL_POLYGON)
-	 mode = LCS_CULL_DISABLE;
-
-      imesa->dirty |= I810_UPLOAD_CTX;
-      imesa->Setup[I810_CTXREG_LCS] &= ~LCS_CULL_MASK;
-      imesa->Setup[I810_CTXREG_LCS] |= (LCS_UPDATE_CULL_MODE | mode);
+   i810ContextPtr imesa = I810_CONTEXT(ctx);
+   FLUSH_BATCH(imesa);
 
-      LOCK_HARDWARE(imesa);
-      i810EmitHwStateLocked( imesa );	
-      UNLOCK_HARDWARE(imesa);
+   imesa->dirty |= I810_UPLOAD_CTX;
+   imesa->Setup[I810_CTXREG_LCS] &= ~LCS_CULL_MASK;
+   imesa->Setup[I810_CTXREG_ST1] &= ~ST1_ENABLE;
+   imesa->Setup[I810_CTXREG_AA] &= ~AA_ENABLE;
+
+   switch (ctx->PB->primitive) {
+   case GL_POLYGON:
+      if (ctx->Polygon.StippleFlag && 
+	  (ctx->Driver.TriangleCaps & DD_TRI_STIPPLE))
+	 imesa->Setup[I810_CTXREG_ST1] |= ST1_ENABLE;
+      if (ctx->Polygon.CullFlag) 
+	 imesa->Setup[I810_CTXREG_LCS] |= imesa->LcsCullMode;
+      else 
+	 imesa->Setup[I810_CTXREG_LCS] |= LCS_CULL_DISABLE;
+      if (ctx->Polygon.SmoothFlag)
+	 imesa->Setup[I810_CTXREG_AA] |= AA_ENABLE;
+      break;
+   case GL_LINES:
+      imesa->Setup[I810_CTXREG_LCS] &= ~LCS_LINEWIDTH_0_5;
+      if (ctx->Line.SmoothFlag) {
+	 imesa->Setup[I810_CTXREG_AA] |= AA_ENABLE;  
+	 imesa->Setup[I810_CTXREG_LCS] |= LCS_LINEWIDTH_0_5;
+      } 
+      imesa->Setup[I810_CTXREG_LCS] |= LCS_CULL_DISABLE;
+      break;
+   case GL_POINTS:
+      if (ctx->Point.SmoothFlag)
+	 imesa->Setup[I810_CTXREG_AA] |= AA_ENABLE;      
+      imesa->Setup[I810_CTXREG_LCS] |= LCS_CULL_DISABLE;
+      break;
+   default:
+      imesa->Setup[I810_CTXREG_LCS] |= LCS_CULL_DISABLE;
+      break;
    }
 }
 
@@ -318,15 +389,36 @@ static void i810DDReducedPrimitiveChange( GLcontext *ctx, GLenum prim )
  * Color masks
  */
 
-/* Mesa calls this from the wrong place.
+/* Mesa calls this from the wrong place - it is called a very large
+ * number of redundant times.
  *
- * Its a fallback...
+ * Colormask can be simulated by multipass or multitexture techniques.
  */
 static GLboolean i810DDColorMask(GLcontext *ctx, 
 				 GLboolean r, GLboolean g, 
 				 GLboolean b, GLboolean a )
 {
-   return 1;
+   i810ContextPtr imesa = I810_CONTEXT( ctx );
+   GLuint tmp = 0;
+   GLuint rv = 1;
+
+   imesa->Fallback &= ~I810_FALLBACK_COLORMASK;
+
+   if (r && g && b) {
+      tmp = imesa->Setup[I810_CTXREG_B2] | B2_FB_WRITE_ENABLE;
+   } else if (!r && !g && !b) {
+      tmp = imesa->Setup[I810_CTXREG_B2] & ~B2_FB_WRITE_ENABLE;
+   } else {
+      rv = 0;
+      imesa->Fallback |= I810_FALLBACK_COLORMASK;
+   }
+      
+   if (tmp != imesa->Setup[I810_CTXREG_B2]) {
+      FLUSH_BATCH(imesa);
+      imesa->Setup[I810_CTXREG_B2] = tmp;
+   }
+
+   return rv;
 }
 
 /* Seperate specular not fully implemented in hardware...  Needs
@@ -336,8 +428,14 @@ static GLboolean i810DDColorMask(GLcontext *ctx,
 static void i810DDLightModelfv(GLcontext *ctx, GLenum pname, 
 			      const GLfloat *param)
 {
-   if (pname == GL_LIGHT_MODEL_COLOR_CONTROL) {
-      I810_CONTEXT(ctx)->new_state |= I810_NEW_TEXTURE;
+   if (pname == GL_LIGHT_MODEL_COLOR_CONTROL) 
+   {            
+      i810ContextPtr imesa = I810_CONTEXT( ctx );
+      FLUSH_BATCH(imesa);
+
+      imesa->Fallback &= ~I810_FALLBACK_SPECULAR;
+      if (ctx->Light.Model.ColorControl == GL_SEPARATE_SPECULAR_COLOR)
+	 imesa->Fallback |= I810_FALLBACK_SPECULAR;
    }
 }
 
@@ -372,64 +470,105 @@ static void i810DDEnable(GLcontext *ctx, GLenum cap, GLboolean state)
 
    switch(cap) {
    case GL_ALPHA_TEST:
+      FLUSH_BATCH(imesa);
       imesa->dirty |= I810_UPLOAD_CTX;
       imesa->Setup[I810_CTXREG_B1] &= ~B1_ALPHA_TEST_ENABLE;
-      imesa->Setup[I810_CTXREG_B1] |= B1_UPDATE_ALPHA_TEST_ENABLE;
       if (state)
 	 imesa->Setup[I810_CTXREG_B1] |= B1_ALPHA_TEST_ENABLE;
       break;
    case GL_BLEND:
+      FLUSH_BATCH(imesa);
       imesa->dirty |= I810_UPLOAD_CTX;
       imesa->Setup[I810_CTXREG_B1] &= ~B1_BLEND_ENABLE;
-      imesa->Setup[I810_CTXREG_B1] |= B1_UPDATE_BLEND_ENABLE;
       if (state)
 	 imesa->Setup[I810_CTXREG_B1] |= B1_BLEND_ENABLE;
       break;
    case GL_DEPTH_TEST:
+      FLUSH_BATCH(imesa);
       imesa->dirty |= I810_UPLOAD_CTX;
       imesa->Setup[I810_CTXREG_B1] &= ~B1_Z_TEST_ENABLE;
-      imesa->Setup[I810_CTXREG_B1] |= B1_UPDATE_Z_TEST_ENABLE;
       if (state)
 	 imesa->Setup[I810_CTXREG_B1] |= B1_Z_TEST_ENABLE;
       break;
    case GL_SCISSOR_TEST:
-#if 0
-      imesa->dirty |= I810_UPLOAD_CTX;
-      imesa->Setup[I810_CTXREG_SC] &= ~SC_ENABLE_MASK;
-      imesa->Setup[I810_CTXREG_SC] |= SC_UPDATE_SCISSOR;
-      if (state)
-	 imesa->Setup[I810_CTXREG_SC] |= SC_ENABLE;
-#endif
+      FLUSH_BATCH(imesa);
+      imesa->scissor = state;
+      imesa->dirty |= I810_UPLOAD_CLIPRECTS;
+      break;
+   case GL_POLYGON_STIPPLE:      
+      if ((ctx->Driver.TriangleCaps & DD_TRI_STIPPLE) &&
+	  ctx->PB->primitive == GL_POLYGON) 
+      {
+	 FLUSH_BATCH(imesa);
+	 imesa->dirty |= I810_UPLOAD_CTX;
+	 imesa->Setup[I810_CTXREG_ST1] &= ~ST1_ENABLE;
+	 if (state)
+	    imesa->Setup[I810_CTXREG_ST1] |= ST1_ENABLE;
+      }
+      break;
+   case GL_LINE_SMOOTH:
+      if (ctx->PB->primitive == GL_LINE) {
+	 FLUSH_BATCH(imesa);
+	 imesa->dirty |= I810_UPLOAD_CTX;
+	 imesa->Setup[I810_CTXREG_AA] &= ~AA_ENABLE;
+	 imesa->Setup[I810_CTXREG_LCS] &= ~LCS_LINEWIDTH_0_5;
+	 if (state) {
+	    imesa->Setup[I810_CTXREG_AA] |= AA_ENABLE;
+	    imesa->Setup[I810_CTXREG_LCS] |= LCS_LINEWIDTH_0_5;
+	 }
+      }
+      break;
+   case GL_POINT_SMOOTH:
+      if (ctx->PB->primitive == GL_POINT) {
+	 FLUSH_BATCH(imesa);
+	 imesa->dirty |= I810_UPLOAD_CTX;
+	 imesa->Setup[I810_CTXREG_AA] &= ~AA_ENABLE;
+	 if (state) 
+	    imesa->Setup[I810_CTXREG_AA] |= AA_ENABLE;
+      }
+      break;
+   case GL_POLYGON_SMOOTH:
+      if (ctx->PB->primitive == GL_POLYGON) {
+	 FLUSH_BATCH(imesa);
+	 imesa->dirty |= I810_UPLOAD_CTX;
+	 imesa->Setup[I810_CTXREG_AA] &= ~AA_ENABLE;
+	 if (state) 
+	    imesa->Setup[I810_CTXREG_AA] |= AA_ENABLE;
+      }
       break;
    case GL_FOG:
+      FLUSH_BATCH(imesa);
       imesa->dirty |= I810_UPLOAD_CTX;
       imesa->Setup[I810_CTXREG_B1] &= ~B1_FOG_ENABLE;
-      imesa->Setup[I810_CTXREG_B1] |= B1_UPDATE_FOG_ENABLE;
       if (state)
 	 imesa->Setup[I810_CTXREG_B1] |= B1_FOG_ENABLE;
       break;
    case GL_CULL_FACE:
       if (ctx->PB->primitive == GL_POLYGON) {
+	 FLUSH_BATCH(imesa);
 	 imesa->dirty |= I810_UPLOAD_CTX;
 	 imesa->Setup[I810_CTXREG_LCS] &= ~LCS_CULL_MASK;
-	 imesa->Setup[I810_CTXREG_LCS] |= LCS_UPDATE_CULL_MODE;
 	 if (state)
 	    imesa->Setup[I810_CTXREG_LCS] |= imesa->LcsCullMode;
 	 else
 	    imesa->Setup[I810_CTXREG_LCS] |= LCS_CULL_DISABLE;
       }
       break;
+   case GL_TEXTURE_1D:      
+   case GL_TEXTURE_3D:      
+      FLUSH_BATCH(imesa);
+      imesa->new_state |= I810_NEW_TEXTURE;
+      break;
    case GL_TEXTURE_2D:      
+      FLUSH_BATCH(imesa);
       imesa->new_state |= I810_NEW_TEXTURE;
       imesa->dirty |= I810_UPLOAD_CTX;
       if (ctx->Texture.CurrentUnit == 0) {
 	 imesa->Setup[I810_CTXREG_MT] &= ~MT_TEXEL0_ENABLE;
-	 imesa->Setup[I810_CTXREG_MT] |= MT_UPDATE_TEXEL0_STATE;
 	 if (state)
 	    imesa->Setup[I810_CTXREG_MT] |= MT_TEXEL0_ENABLE;
       } else {
 	 imesa->Setup[I810_CTXREG_MT] &= ~MT_TEXEL1_ENABLE;
-	 imesa->Setup[I810_CTXREG_MT] |= MT_UPDATE_TEXEL1_STATE;
 	 if (state)
 	    imesa->Setup[I810_CTXREG_MT] |= MT_TEXEL1_ENABLE;
       }
@@ -444,92 +583,20 @@ static void i810DDEnable(GLcontext *ctx, GLenum cap, GLboolean state)
 /* =============================================================
  */
 
-/* Delightfully few possibilities:
- */
-void i810DDPrintState( const char *msg, GLuint state )
-{
-   fprintf(stderr, "%s (0x%x): %s\n",
-	   msg,
-	   (unsigned int) state,
-	   (state & I810_NEW_TEXTURE) ? "texture, " : "");
-}
 
 void i810DDUpdateHwState( GLcontext *ctx )
 {
    i810ContextPtr imesa = I810_CONTEXT(ctx);
 
-   if (imesa->new_state & I810_NEW_TEXTURE)
+   if (imesa->new_state & I810_NEW_TEXTURE) {
+      FLUSH_BATCH(imesa);
       i810UpdateTextureState( ctx );
-
-   imesa->new_state = 0;
-
-   if (imesa->dirty) {
-      LOCK_HARDWARE(imesa);
-      i810EmitHwStateLocked( imesa );	
-      UNLOCK_HARDWARE(imesa);
-   }
-}
-
-
-
-/*
- * i810DmaExecute
- * Add a block of data to the dma buffer
- *
- * -- In ring buffer mode, must be called with lock held, and translates to 
- *    OUT_RING rather than outbatch
- *
- * -- In dma mode, probably won't be used because state updates will have
- *    to be done via the kernel for security...
- */
-static void i810DmaExecute( i810ContextPtr imesa,
-			    GLuint *code, int dwords, const char *where ) 
-{	
-   if (I810_DEBUG & DEBUG_VERBOSE_RING)
-      fprintf(stderr, "i810DmaExecute (%s)\n", where);
-
-   if (dwords & 1) {
-      i810Error( "Misaligned buffer in i810DmaExecute\n" );
-      exit(1);
    }
 
-   {
-      int i;
-      BEGIN_BATCH( imesa, dwords);
-
-      for ( i = 0 ; i < dwords ; i++ ) {
-	 if (0) fprintf(stderr, "%d: %x\n", i, code[i]);
-	 OUT_BATCH( code[i] );
-      }
-
-      ADVANCE_BATCH();
-   }
+   imesa->new_state = 0;
 }
 
 
-
-void i810EmitScissorValues( i810ContextPtr imesa, int nr, int emit )
-{
-   int x1 = imesa->pClipRects[nr].x1 - imesa->drawX;
-   int y1 = imesa->pClipRects[nr].y1 - imesa->drawY;
-   int x2 = imesa->pClipRects[nr].x2 - imesa->drawX;
-   int y2 = imesa->pClipRects[nr].y2 - imesa->drawY;
-
-   if (I810_DEBUG&DEBUG_VERBOSE_DRI)
-      fprintf(stderr, "i810EmitScissorValues %d,%d - %d,%d\n",
-	      x1,y1,x2,y2);
-
-   imesa->ClipSetup[I810_CLIPREG_SCI1] = (x1 | (y1 << 16));
-   imesa->ClipSetup[I810_CLIPREG_SCI2] = (x2 | (y2 << 16));
-   imesa->ClipSetup[I810_CLIPREG_SC] = ( GFX_OP_SCISSOR |
-					 SC_UPDATE_SCISSOR |
-					 SC_ENABLE );
-
-   if (emit)
-      i810DmaExecute( imesa, 
-		      imesa->ClipSetup, I810_CLIP_SETUP_SIZE, "cliprect" );
-}
-
 void i810EmitDrawingRectangle( i810ContextPtr imesa )
 {
    __DRIdrawablePrivate *dPriv = imesa->driDrawable;
@@ -545,7 +612,6 @@ void i810EmitDrawingRectangle( i810ContextPtr imesa )
     */
    imesa->BufferSetup[I810_DESTREG_DR4] = ((y0<<16) | 
 					   (((unsigned)x0)&0xFFFF));
-
   
    /* Clip to screen.
     */
@@ -556,74 +622,74 @@ void i810EmitDrawingRectangle( i810ContextPtr imesa )
 
 
    /* Onscreen drawing rectangle.
-    *
-    * TODO: clip again to GL scissor values.
     */
    imesa->BufferSetup[I810_DESTREG_DR2] = ((y0<<16) | x0);
-   imesa->BufferSetup[I810_DESTREG_DR3] = ((y1<<16) | x1);
+   imesa->BufferSetup[I810_DESTREG_DR3] = (((y1+1)<<16) | (x1+1));
    imesa->dirty |= I810_UPLOAD_BUFFERS;
 }
 
 
 static void i810DDPrintDirty( const char *msg, GLuint state )
 {
-   fprintf(stderr, "%s (0x%x): %s%s%s%s%s%s%s\n",	   
+   fprintf(stderr, "%s (0x%x): %s%s%s%s%s\n",	   
 	   msg,
 	   (unsigned int) state,
-	   (state & I810_REFRESH_RING)      ? "read-lp-ring, " : "",
-	   (state & I810_REQUIRE_QUIESCENT) ? "req-quiescent, " : "",
 	   (state & I810_UPLOAD_TEX0IMAGE)  ? "upload-tex0, " : "",
 	   (state & I810_UPLOAD_TEX1IMAGE)  ? "upload-tex1, " : "",
 	   (state & I810_UPLOAD_CTX)        ? "upload-ctx, " : "",
 	   (state & I810_UPLOAD_BUFFERS)    ? "upload-bufs, " : "",
-	   (state & I810_EMIT_CLIPRECT)     ? "emit-single-cliprect, " : ""
+	   (state & I810_UPLOAD_CLIPRECTS)  ? "upload-cliprects, " : ""
 	   );
 }
 
 
-/* Spew the state onto the ringbuffer and/or texture memory.
+/* Push the state into the sarea and/or texture memory.
  */
 void i810EmitHwStateLocked( i810ContextPtr imesa )
 {
    if (I810_DEBUG & DEBUG_VERBOSE_API)
-      i810DDPrintDirty( "i810EmitHwStateLocked", imesa->dirty );
-
-   if (imesa->dirty & I810_REQUIRE_QUIESCENT) {
-      i810glx.c_drawWaits += _I810Sync( imesa );
-   }
+      i810DDPrintDirty( "\n\n\ni810EmitHwStateLocked", imesa->dirty );
 
-   /* TODO: Refine mechanism to be equivalent to UploadSubImage
-    */
-   if ((imesa->dirty & I810_UPLOAD_TEX0IMAGE) && imesa->CurrentTexObj[0])
-      i810UploadTexImages(imesa, imesa->CurrentTexObj[0]);
+   if (imesa->dirty & ~I810_UPLOAD_CLIPRECTS)
+   {
+      if ((imesa->dirty & I810_UPLOAD_TEX0IMAGE) && imesa->CurrentTexObj[0])
+	 i810UploadTexImages(imesa, imesa->CurrentTexObj[0]);
    
-   if ((imesa->dirty & I810_UPLOAD_TEX1IMAGE) && imesa->CurrentTexObj[1])
-      i810UploadTexImages(imesa, imesa->CurrentTexObj[1]);
+      if ((imesa->dirty & I810_UPLOAD_TEX1IMAGE) && imesa->CurrentTexObj[1])
+	 i810UploadTexImages(imesa, imesa->CurrentTexObj[1]);
   
-   if ((imesa->dirty & I810_UPLOAD_CTX)) {
-      i810DmaExecute( imesa, imesa->Setup, I810_CTX_SETUP_SIZE, "context" );
-
-      if (imesa->CurrentTexObj[0])
-	 i810DmaExecute( imesa, imesa->CurrentTexObj[0]->Setup,
-			 I810_TEX_SETUP_SIZE, "tex-0");
+      if (imesa->dirty & I810_UPLOAD_CTX)
+	 memcpy( imesa->sarea->ContextState, 
+		 imesa->Setup, 
+		 sizeof(imesa->Setup) );
+
+      if ((imesa->dirty & I810_UPLOAD_TEX0) && imesa->CurrentTexObj[0]) {
+	 imesa->sarea->dirty |= I810_UPLOAD_TEX0;
+	 memcpy(imesa->sarea->TexState[0],
+		imesa->CurrentTexObj[0]->Setup,
+		sizeof(imesa->sarea->TexState[0]));
+      }
 
-      if (imesa->CurrentTexObj[1]) 
-	 i810DmaExecute( imesa, imesa->CurrentTexObj[1]->Setup,
-			 I810_TEX_SETUP_SIZE, "tex-1");
+      if ((imesa->dirty & I810_UPLOAD_TEX1) && imesa->CurrentTexObj[1]) {
+	 imesa->sarea->dirty |= I810_UPLOAD_TEX1;
+	 memcpy(imesa->sarea->TexState[1],
+		imesa->CurrentTexObj[1]->Setup,
+		sizeof(imesa->sarea->TexState[1]));
+      }
+      
+      if (imesa->dirty & I810_UPLOAD_BUFFERS) 
+	 memcpy( imesa->sarea->BufferState, 
+		 imesa->BufferSetup, 
+		 sizeof(imesa->BufferSetup) );
+
+      imesa->sarea->dirty |= (imesa->dirty & 
+			      ~(I810_UPLOAD_TEX1|I810_UPLOAD_TEX0));
+      imesa->dirty &= I810_UPLOAD_CLIPRECTS;
    }
-
-   if (imesa->dirty & I810_UPLOAD_BUFFERS) 
-      i810DmaExecute( imesa, imesa->BufferSetup, 
-		      I810_DEST_SETUP_SIZE, "buffers" );
-   
-   if (imesa->dirty & I810_EMIT_CLIPRECT)
-      i810DmaExecute( imesa, imesa->ClipSetup, 
-		      I810_CLIP_SETUP_SIZE, "cliprect" );
-
-   imesa->dirty = 0;
 }
 
 
+
 void i810DDInitState( i810ContextPtr imesa )
 {
    i810ScreenPrivate *i810Screen = imesa->i810Screen;
@@ -797,7 +863,7 @@ void i810DDInitState( i810ContextPtr imesa )
 				     LCS_UPDATE_ZMODE |
 				     LCS_Z_LESS |
 				     LCS_UPDATE_LINEWIDTH |
-				     (0x2<<LCS_LINEWIDTH_SHIFT) |
+				     LCS_LINEWIDTH_1_0 |
 				     LCS_UPDATE_ALPHA_INTERP |
 				     LCS_ALPHA_INTERP |
 				     LCS_UPDATE_FOG_INTERP |
@@ -825,11 +891,9 @@ void i810DDInitState( i810ContextPtr imesa )
    imesa->Setup[I810_CTXREG_ST0] = GFX_OP_STIPPLE;
    imesa->Setup[I810_CTXREG_ST1] = 0;
 
-
-
    imesa->Setup[I810_CTXREG_AA] = ( GFX_OP_ANTIALIAS |
 				    AA_UPDATE_EDGEFLAG |
-				    0 |
+				    AA_ENABLE_EDGEFLAG | /* ? */
 				    AA_UPDATE_POLYWIDTH |
 				    AA_POLYWIDTH_05 |
 				    AA_UPDATE_LINEWIDTH |
@@ -839,38 +903,29 @@ void i810DDInitState( i810ContextPtr imesa )
 				    AA_UPDATE_AA_ENABLE |
 				    0 );
 
-
-
-   memset(imesa->ClipSetup, 0, sizeof(imesa->ClipSetup));
-   imesa->ClipSetup[I810_CLIPREG_SCI0] = GFX_OP_SCISSOR_INFO; 
-   imesa->ClipSetup[I810_CLIPREG_SCI1] = 0;
-   imesa->ClipSetup[I810_CLIPREG_SCI2] = 0;
-   imesa->ClipSetup[I810_CLIPREG_SC] = ( GFX_OP_SCISSOR | 
-					 SC_UPDATE_SCISSOR |
-					 0 );
-
-   /* This stuff is all invarient as long as we are using
-    * shared back and depth buffers.  
-    */
    memset(imesa->BufferSetup, 0, sizeof(imesa->BufferSetup));
-   imesa->drawOffset = i810Screen->backOffset;
    imesa->BufferSetup[I810_DESTREG_DI0] = CMD_OP_DESTBUFFER_INFO;
 
-   if (imesa->glCtx->Color.DriverDrawBuffer == GL_BACK_LEFT)
+   if (imesa->glCtx->Color.DriverDrawBuffer == GL_BACK_LEFT) {
+      imesa->drawMap = i810Screen->back.map;
       imesa->BufferSetup[I810_DESTREG_DI1] = (i810Screen->backOffset | 
-					      i810Screen->auxPitchBits);
-   else
+					      i810Screen->backPitchBits);
+   } else {
+      imesa->drawMap = imesa->driScreen->pFB;
       imesa->BufferSetup[I810_DESTREG_DI1] = (i810Screen->fbOffset | 
-					      i810Screen->auxPitchBits);
+					      i810Screen->backPitchBits);
+   }
 
+   if (imesa->glCtx->Color.DriverDrawBuffer == GL_BACK_LEFT) {
+      imesa->readMap = i810Screen->back.map;
+   } else {
+      imesa->readMap = imesa->driScreen->pFB;
+   }
 
    imesa->BufferSetup[I810_DESTREG_DV0] = GFX_OP_DESTBUFFER_VARS;
    imesa->BufferSetup[I810_DESTREG_DV1] = (DV_HORG_BIAS_OGL |
 					   DV_VORG_BIAS_OGL |
 					   i810Screen->fbFormat);
-   imesa->BufferSetup[I810_DESTREG_ZB0] = CMD_OP_Z_BUFFER_INFO;
-   imesa->BufferSetup[I810_DESTREG_ZB1] = (i810Screen->depthOffset | 
-					   i810Screen->auxPitchBits);
 
    imesa->BufferSetup[I810_DESTREG_DR0] = GFX_OP_DRAWRECT_INFO;
    imesa->BufferSetup[I810_DESTREG_DR1] = DR1_RECT_CLIP_ENABLE;
@@ -886,17 +941,31 @@ void i810DDUpdateState( GLcontext *ctx )
 {
    i810ContextPtr imesa = I810_CONTEXT( ctx );
 
+   /* Have to do this here to detect texture fallbacks in time:
+    */
+   if (I810_CONTEXT(ctx)->new_state & I810_NEW_TEXTURE)
+      i810DDUpdateHwState( ctx );
+
+
    if (ctx->NewState & INTERESTED) {
       i810DDChooseRenderState(ctx);  
       i810ChooseRasterSetupFunc(ctx);
    }
+
+   if (0) 
+      fprintf(stderr, "IndirectTriangles %x Fallback %x\n", 
+	      imesa->IndirectTriangles, imesa->Fallback);
    
-   /* TODO: stop mesa from clobbering these.
-    */
-   ctx->Driver.PointsFunc=imesa->PointsFunc;
-   ctx->Driver.LineFunc=imesa->LineFunc;
-   ctx->Driver.TriangleFunc=imesa->TriangleFunc;
-   ctx->Driver.QuadFunc=imesa->QuadFunc;
+   if (!imesa->Fallback)
+   {
+      ctx->IndirectTriangles &= ~DD_SW_RASTERIZE;
+      ctx->IndirectTriangles |= imesa->IndirectTriangles;
+
+      ctx->Driver.PointsFunc=imesa->PointsFunc;
+      ctx->Driver.LineFunc=imesa->LineFunc;
+      ctx->Driver.TriangleFunc=imesa->TriangleFunc;
+      ctx->Driver.QuadFunc=imesa->QuadFunc;
+   }
 }
 
 
@@ -919,8 +988,13 @@ void i810DDInitStateFuncs(GLcontext *ctx)
    ctx->Driver.ReducedPrimitiveChange = i810DDReducedPrimitiveChange;
    ctx->Driver.RenderStart = i810DDUpdateHwState; 
    ctx->Driver.RenderFinish = 0; 
+   
+   ctx->Driver.PolygonStipple = i810DDPolygonStipple;
+   ctx->Driver.LineStipple = 0;
+
+   ctx->Driver.SetReadBuffer = i810DDSetReadBuffer;
+   ctx->Driver.SetDrawBuffer = i810DDSetDrawBuffer;
 
-   ctx->Driver.SetBuffer = i810DDSetBuffer;
    ctx->Driver.Color = i810DDSetColor;
    ctx->Driver.ClearColor = i810DDClearColor;
    ctx->Driver.Dither = i810DDDither;
@@ -928,6 +1002,5 @@ void i810DDInitStateFuncs(GLcontext *ctx)
    ctx->Driver.Index = 0;
    ctx->Driver.ClearIndex = 0;
    ctx->Driver.IndexMask = 0;
-
 }
 
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810state.h b/xc/lib/GL/mesa/src/drv/i810/i810state.h
index d0551631e..b43a1a4ff 100644
--- a/xc/lib/GL/mesa/src/drv/i810/i810state.h
+++ b/xc/lib/GL/mesa/src/drv/i810/i810state.h
@@ -1,13 +1,12 @@
 #ifndef _I810_STATE_H
 #define _I810_STATE_H
 
+#include "i810context.h"
 
 extern void i810DDUpdateHwState( GLcontext *ctx );
 extern void i810DDUpdateState( GLcontext *ctx );
 extern void i810DDInitState( i810ContextPtr imesa );
 extern void i810DDInitStateFuncs( GLcontext *ctx );
 
-extern void i810DDPrintState( const char *msg, GLuint state );
-
 
 #endif
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810swap.c b/xc/lib/GL/mesa/src/drv/i810/i810swap.c
deleted file mode 100644
index 9220ef9f4..000000000
--- a/xc/lib/GL/mesa/src/drv/i810/i810swap.c
+++ /dev/null
@@ -1,259 +0,0 @@
-
-
-#include "types.h"
-#include "vbrender.h"
-#include "i810log.h"
-
-#include <stdio.h>
-#include <stdlib.h>
-
-#include "mm.h"
-#include "i810lib.h"
-#include "i810dd.h"
-#include "i810dma.h"
-#include "i810state.h"
-#include "i810swap.h"
-#include "i810dma.h"
-
-/*
- * i810BackToFront
- *
- * Blit the visible rectangles from the back buffer to the screen.
- * Respects the frontbuffer cliprects, and applies the offset
- * necessary if the backbuffer is positioned elsewhere in the screen.
- */
-static int i810BackToFront( i810ContextPtr imesa ) 
-{
-   __DRIdrawablePrivate *dPriv = imesa->driDrawable;
-   i810ScreenPrivate *i810Screen = imesa->i810Screen;
-
-
-   /* Use the frontbuffer cliprects:
-    */
-   XF86DRIClipRectPtr pbox = dPriv->pClipRects;
-   int nbox = dPriv->numClipRects;
-
-   if( nbox ) 
-   {
-      int i;
-      int pitch = i810Screen->auxPitch;
-
-      int dx = dPriv->auxX - dPriv->x;
-      int dy = dPriv->auxY - dPriv->y;
-      int ofs = ( i810Screen->backOffset + 
-		  dy * i810Screen->auxPitch +
-		  dx * i810Screen->cpp);
-
-      unsigned int BR13 = ((i810Screen->fbStride) |
-			   (0xCC << 16));
-
-
-      for (i=0; i < nbox; i++, pbox++) {
-	 int w = pbox->x2 - pbox->x1;
-	 int h = pbox->y2 - pbox->y1;
-
-	 int start = ofs + pbox->x1*i810Screen->cpp + pbox->y1*pitch;
-	 int dst = pbox->x1*i810Screen->cpp + pbox->y1*i810Screen->fbStride;
-	 
-	 if (I810_DEBUG&DEBUG_VERBOSE_2D) 
-	    fprintf(stderr, "i810BackToFront %d,%d-%dx%d\n",
-		    pbox->x1,pbox->y1,w,h);
-
-	 {
-	    BEGIN_BATCH( imesa, 6 );
-	    OUT_BATCH( BR00_BITBLT_CLIENT | BR00_OP_SRC_COPY_BLT | 0x4 );
-	    OUT_BATCH( BR13 );	/* dest pitch, rop */
-
-	    OUT_BATCH( (h << 16) | (w * i810Screen->cpp));
-	    OUT_BATCH( dst );
-
-	    OUT_BATCH( pitch );	/* src pitch */
-	    OUT_BATCH( start );
-	    ADVANCE_BATCH();
-	 }
-      }
-   }
-  
-   return Success;   
-}
-
-
-/*
- * ClearBox
- *
- * Add hardware commands to draw a filled box for the debugging
- * display.  These are drawn into the current drawbuffer, so should
- * work with both front and backbuffer rendering.  (However, it
- * is only ever called on swapbuffer - ie backbuffer rendering).
- */
-static void ClearBox( i810ContextPtr imesa,
-		      int cx, int cy, int cw, int ch, 
-		      int r, int g, int b ) 
-{
-   __DRIdrawablePrivate *dPriv = imesa->driDrawable;
-   i810ScreenPrivate *i810Screen = imesa->i810Screen;
-
-   int _nc = imesa->numClipRects;
-      
-   while (_nc--) {
-      int x1 = cx + dPriv->x;
-      int y1 = cy + dPriv->y;
-      int x2 = x1 + cw;
-      int y2 = y1 + ch;
-
-      if (imesa->needClip) {
-	 int rx1 = imesa->pClipRects[_nc].x1;
-	 int ry1 = imesa->pClipRects[_nc].y1;
-	 int rx2 = imesa->pClipRects[_nc].x2;
-	 int ry2 = imesa->pClipRects[_nc].y2;
-
-
-	 if (x2 < rx1) continue;
-	 if (y2 < ry1) continue;
-	 if (x1 > rx2) continue;
-	 if (y1 > ry2) continue;
-	 if (x1 < rx1) x1 = rx1;
-	 if (y1 < ry1) y1 = ry1;
-	 if (x2 > rx2) x2 = rx2;
-	 if (y2 > ry2) y2 = ry2;
-      }
-
-
-      {
-	 int start = (imesa->drawOffset + 
-		      y1 * i810Screen->auxPitch + 
-		      x1 * i810Screen->cpp);
-
-	 BEGIN_BATCH( imesa, 6 );
-   
-	 OUT_BATCH( BR00_BITBLT_CLIENT | BR00_OP_COLOR_BLT | 0x3 );
-	 OUT_BATCH( BR13_SOLID_PATTERN | 
-		    (0xF0 << 16) |
-		    i810Screen->auxPitch );
-	 OUT_BATCH( ((y2-y1) << 16) | ((x2-x1) * i810Screen->cpp));
-	 OUT_BATCH( start );
-	 OUT_BATCH( i810PackColor( i810Screen->fbFormat, r, g, b, 0 ) );
-	 OUT_BATCH( 0 );
-
-	 if (I810_DEBUG&DEBUG_VERBOSE_2D) 
-	    fprintf(stderr, "box %d,%x %d,%d col %x (%d,%d,%d)\n", 
-		    (int)x1,(int)y1, (int)x2,(int)y2,
-		    (int)i810PackColor( i810Screen->fbFormat, r, g, b, 0 ), 
-		    (int)r, (int)g, (int)b);
-	 
-	 ADVANCE_BATCH();
-      }
-   }
-}
-
-
-/*
- * performanceBoxes
- * Draw some small boxesin the corner of the buffer
- * based on some performance information
- */
-static void i810PerformanceBoxes( i810ContextPtr imesa ) {
-	
-
-   /* draw a box to show we are active, so if it is't seen it means
-      that it is completely software based.  Purple is traditional for
-      direct rendering */
-   ClearBox( imesa, 4, 4, 8, 8, 255, 0, 255 );
-	
-   /* draw a red box if we had to wait for drawing to complete 
-      (software render or texture swap) */
-   if ( i810glx.c_drawWaits ) {
-      ClearBox( imesa, 16, 4, 8, 8, 255, 0, 0 );
-      i810glx.c_drawWaits = 0;
-   }
-	
-   /* draw a blue box if the 3d context was lost
-    */
-   if ( i810glx.c_ctxlost ) {
-      ClearBox( imesa, 28, 4, 8, 8, 0, 0, 255 );
-      i810glx.c_ctxlost = 0;
-   }
-	
-   /* draw an orange box if texture state was stomped */
-   if ( i810glx.c_texlost ) {
-      ClearBox( imesa, 40, 4, 8, 8, 255, 128, 0 );
-      i810glx.c_texlost = 0;
-   }
-
-   /* draw a yellow box if textures were swapped */
-   if ( i810glx.c_textureSwaps ) {
-      ClearBox( imesa, 56, 4, 8, 8, 255, 255, 0 );
-      i810glx.c_textureSwaps = 0;
-   }
-		
-   /* draw a green box if we had to wait for dma to complete (full utilization) 
-      on the previous frame */
-   if ( !i810glx.hardwareWentIdle ) {
-      ClearBox( imesa, 72, 4, 8, 8, 0, 255, 0 );
-   }
-   i810glx.hardwareWentIdle = 0;
-	
-#if 0
-   /* show buffer utilization */
-   if ( i810glx.c_dmaFlush > 1 ) {
-      /* draw a solid bar if we flushed more than one buffer */
-      ClearBox( imesa, 4, 16, 252, 4, 255, 32, 32 );
-   } else {
-      /* draw bars to represent the utilization of primary and
-         secondary buffers */
-      ClearBox( imesa, 4, 16, 252, 4, 32, 32, 32 );
-      t = i810glx.dma_buffer->mem.Size;
-      w = 252 * i810glx.dma_buffer->head / t;
-      if ( w < 1 ) {
-	 w = 1;
-      }
-      ClearBox( imesa, 4, 16, w, 4, 196, 128, 128 );
-   }
-#endif
-
-   i810glx.c_dmaFlush = 0;
-}
-
-
-static void i810SendDmaFlush( i810ContextPtr imesa )
-{
-   BEGIN_BATCH( imesa, 2 );
-   OUT_BATCH( INST_PARSER_CLIENT | INST_OP_FLUSH );
-   OUT_BATCH( 0 );
-   ADVANCE_BATCH();
-}
-
-void i810SwapBuffers( i810ContextPtr imesa )
-{
-   if (I810_DEBUG & DEBUG_VERBOSE_API)
-      fprintf(stderr, "i810SwapBuffers()\n");
-
-   LOCK_HARDWARE( imesa );
-   i810SendDmaFlush( imesa );
-   if ( i810glx.boxes )  
-      i810PerformanceBoxes( imesa );
-   i810BackToFront( imesa );
-
-   /* Check if we are emitting thousands of tiny frames, and if so
-    * throttle them.  Would be better to simply go to sleep until
-    * our stuff had cleared, rather than polling.  A dma solution
-    * should provide for this.
-    */   
-   if (imesa->lastSwap >= imesa->sarea->lastWrap &&
-       imesa->lastSwap >= imesa->sarea->lastSync) {
-
-      if (I810_DEBUG & DEBUG_VERBOSE_API)
-	 fprintf(stderr, ".");
-
-      _I810Sync( imesa );
-   }
-
-   imesa->lastSwap = imesa->sarea->lastSync;
-   if (imesa->sarea->lastWrap > imesa->lastSwap) 
-      imesa->lastSwap = imesa->sarea->lastWrap;
-      
-   
-   UNLOCK_HARDWARE( imesa );
-
-/*     XSync(imesa->display, 0); */
-}
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810swap.h b/xc/lib/GL/mesa/src/drv/i810/i810swap.h
deleted file mode 100644
index 1657a717d..000000000
--- a/xc/lib/GL/mesa/src/drv/i810/i810swap.h
+++ /dev/null
@@ -1,6 +0,0 @@
-#ifndef I810_SWAP_H
-#define I810_SWAP_H
-
-extern void i810SwapBuffers( i810ContextPtr imesa );
-
-#endif
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810tex.c b/xc/lib/GL/mesa/src/drv/i810/i810tex.c
index 039c0af59..8c88f2b2b 100644
--- a/xc/lib/GL/mesa/src/drv/i810/i810tex.c
+++ b/xc/lib/GL/mesa/src/drv/i810/i810tex.c
@@ -28,9 +28,10 @@
 #include <GL/gl.h>
 
 #include "mm.h"
-#include "i810lib.h"
+#include "i810context.h"
 #include "i810tex.h"
 #include "i810log.h"
+#include "i810ioctl.h"
 #include "simple_list.h"
 #include "enums.h"
 
@@ -204,6 +205,7 @@ static i810TextureObjectPtr i810CreateTexObj(i810ContextPtr imesa,
       log_pitch++;
    
    t->dirty_images = 0;
+   t->bound = 0;
    
    for ( height = i = 0 ; i < I810_TEX_MAXLEVELS && tObj->Image[i] ; i++ ) {
       t->image[i].image = tObj->Image[i];
@@ -278,24 +280,24 @@ static i810TextureObjectPtr i810CreateTexObj(i810ContextPtr imesa,
 
 void i810DestroyTexObj(i810ContextPtr imesa, i810TextureObjectPtr t)
 {
-   int i;
    if (!t) return;
 
    /* This is sad - need to sync *in case* we upload a texture
     * to this newly free memory...
     */
    if (t->MemBlock) {
-      imesa->dirty |= I810_REQUIRE_QUIESCENT;
       mmFreeMem(t->MemBlock);
-      t->MemBlock = 0;      
+      t->MemBlock = 0;
+
+      if (t->age > imesa->dirtyAge)
+	 imesa->dirtyAge = t->age;
    }
 
    if (t->globj)
       t->globj->DriverData = 0;
 
-   for ( i = 0 ; i < 2 ; i++ ) 
-      if ( imesa->CurrentTexObj[i] == t ) 
-	 imesa->CurrentTexObj[i] = 0;
+   if (t->bound)
+      imesa->CurrentTexObj[t->bound - 1] = 0; 
 
    remove_from_list(t);
    free(t);
@@ -305,9 +307,11 @@ void i810DestroyTexObj(i810ContextPtr imesa, i810TextureObjectPtr t)
 static void i810SwapOutTexObj(i810ContextPtr imesa, i810TextureObjectPtr t)
 {
    if (t->MemBlock) {
-      imesa->dirty |= I810_REQUIRE_QUIESCENT;
       mmFreeMem(t->MemBlock);
       t->MemBlock = 0;      
+
+      if (t->age > imesa->dirtyAge)
+	 imesa->dirtyAge = t->age;
    }
 
    t->dirty_images = ~0;
@@ -413,13 +417,28 @@ static void i810UploadTexLevel( i810TextureObjectPtr t, int level )
    }
    break;
 
+   case GL_ALPHA:
+   {
+      GLushort *dst = (GLushort *)(t->BufAddr + t->image[level].offset);
+      GLubyte  *src = (GLubyte *)image->Data;
+      i810Msg(10,"  CopyA: %p %p\n", dst, src);      
+
+      for (j = 0 ; j < image->Height ; j++, dst += (t->Pitch/2)) {
+	 for (i = 0 ; i < image->Width ; i++) {
+	    dst[i] = I810PACKCOLOR4444(255,255,255,src[0]);
+	    src += 1;
+	 }
+      }
+   }
+   break;
+
    /* TODO: Translate color indices *now*:
     */
    case GL_COLOR_INDEX:
       {
 	 GLubyte *dst = (GLubyte *)(t->BufAddr + t->image[level].offset);
 	 GLubyte *src = (GLubyte *)image->Data;
-	 i810Msg(10,"  CopyIndex  %p %p\n", dst, src);      
+	 i810Msg(10,"  CopyIndex: %p %p\n", dst, src);      
 
 	 for (j = 0 ; j < image->Height ; j++, dst += t->Pitch) {
 	    for (i = 0 ; i < image->Width ; i++) {
@@ -431,8 +450,8 @@ static void i810UploadTexLevel( i810TextureObjectPtr t, int level )
    break;
       
    default:
-      i810Error("Not supported texture format %d\n",(int)image->Format);
-      exit(1);
+      i810Error("Not supported texture format %s\n",
+		gl_lookup_enum_by_nr(image->Format));
    }
 }
 
@@ -449,13 +468,19 @@ void i810PrintLocalLRU( i810ContextPtr imesa )
 		 t->MemBlock->ofs / sz,
 		 t->MemBlock->ofs,
 		 t->MemBlock->size);      
+      else
+	 fprintf(stderr, "Texture (bound %d) at %x sz %x\n", 
+		 t->bound,
+		 t->MemBlock->ofs,
+		 t->MemBlock->size);      
+
    }
 }
 
 void i810PrintGlobalLRU( i810ContextPtr imesa )
 {
    int i, j;
-   i810TexRegion *list = imesa->sarea->texList;
+   drm_i810_tex_region_t *list = imesa->sarea->texList;
 
    for (i = 0, j = I810_NR_TEX_REGIONS ; i < I810_NR_TEX_REGIONS ; i++) {
       fprintf(stderr, "list[%d] age %d next %d prev %d\n",
@@ -471,7 +496,7 @@ void i810PrintGlobalLRU( i810ContextPtr imesa )
 
 void i810ResetGlobalLRU( i810ContextPtr imesa )
 {
-   i810TexRegion *list = imesa->sarea->texList;
+   drm_i810_tex_region_t *list = imesa->sarea->texList;
    int sz = 1 << imesa->i810Screen->logTextureGranularity;
    int i;
 
@@ -481,7 +506,7 @@ void i810ResetGlobalLRU( i810ContextPtr imesa )
     * when looking up objects at a particular location in texture
     * memory.  
     */
-   for (i = 0 ; (i+1) * sz < imesa->i810Screen->textureSize ; i++) {
+   for (i = 0 ; (i+1) * sz <= imesa->i810Screen->textureSize ; i++) {
       list[i].prev = i-1;
       list[i].next = i+1;
       list[i].age = 0;
@@ -503,7 +528,7 @@ static void i810UpdateTexLRU( i810ContextPtr imesa, i810TextureObjectPtr t )
    int logsz = imesa->i810Screen->logTextureGranularity;
    int start = t->MemBlock->ofs >> logsz;
    int end = (t->MemBlock->ofs + t->MemBlock->size - 1) >> logsz;
-   i810TexRegion *list = imesa->sarea->texList;
+   drm_i810_tex_region_t *list = imesa->sarea->texList;
    
    imesa->texAge = ++imesa->sarea->texAge;
 
@@ -556,7 +581,7 @@ void i810TexturesGone( i810ContextPtr imesa,
       /* It overlaps - kick it off.  Need to hold onto the currently bound
        * objects, however.
        */
-      if (t == imesa->CurrentTexObj[0] || t == imesa->CurrentTexObj[1]) 
+      if (t->bound)
 	 i810SwapOutTexObj( imesa, t );
       else
 	 i810DestroyTexObj( imesa, t );
@@ -583,7 +608,6 @@ int i810UploadTexImages( i810ContextPtr imesa, i810TextureObjectPtr t )
 {
    int i;
    int ofs;
-   i810glx.c_textureSwaps++;
 
    /* Do we need to eject LRU texture objects?
     */
@@ -594,18 +618,24 @@ int i810UploadTexImages( i810ContextPtr imesa, i810TextureObjectPtr t )
 	 if (t->MemBlock)
 	    break;
 
+	 if (imesa->TexObjList.prev->bound) {
+  	    fprintf(stderr, "Hit bound texture in upload\n"); 
+	    i810PrintLocalLRU( imesa );
+	    return -1;
+	 }
+
 	 if (imesa->TexObjList.prev == &(imesa->TexObjList)) {
-	    fprintf(stderr, "Failed to upload texture, sz %d\n", t->totalSize);
+ 	    fprintf(stderr, "Failed to upload texture, sz %d\n", t->totalSize);
 	    mmDumpMemInfo( imesa->texHeap );
 	    return -1;
 	 }
-
+	 
 	 i810DestroyTexObj( imesa, imesa->TexObjList.prev );
       }
  
       ofs = t->MemBlock->ofs;
       t->Setup[I810_TEXREG_MI3] = imesa->i810Screen->textureOffset + ofs;
-      t->BufAddr = i810glx.texVirtual + ofs;
+      t->BufAddr = imesa->i810Screen->tex.map + ofs;
       imesa->dirty |= I810_UPLOAD_CTX;
    }
 
@@ -613,6 +643,14 @@ int i810UploadTexImages( i810ContextPtr imesa, i810TextureObjectPtr t )
     */
    i810UpdateTexLRU( imesa, t );
 
+   if (I810_DEBUG & DEBUG_VERBOSE_LRU)
+      fprintf(stderr, "dispatch age: %d age freed memory: %d\n",
+	      GET_DISPATCH_AGE(imesa), imesa->dirtyAge);
+
+   if (imesa->dirtyAge >= GET_DISPATCH_AGE(imesa)) 
+      i810WaitAgeLocked( imesa, imesa->dirtyAge );
+   
+
    if (t->dirty_images) {
       if (I810_DEBUG & DEBUG_VERBOSE_LRU)
 	 fprintf(stderr, "*");
@@ -694,21 +732,37 @@ static void i810UpdateTex0State( GLcontext *ctx )
     
    if (t->dirty_images) 
       imesa->dirty |= I810_UPLOAD_TEX0IMAGE;
-
+   
    imesa->CurrentTexObj[0] = t;
+   t->bound = 1;
+
+   if (t->MemBlock)
+      i810UpdateTexLRU( imesa, t );
   
    switch (ctx->Texture.Unit[0].EnvMode) {
-      case GL_REPLACE:
-      imesa->Setup[I810_CTXREG_MC0] = ( GFX_OP_MAP_COLOR_STAGES |
-					MC_STAGE_0 |
-					MC_UPDATE_DEST |
-					MC_DEST_CURRENT |
-					MC_UPDATE_ARG1 |
-					MC_ARG1_TEX0_COLOR | 
-					MC_UPDATE_ARG2 |
-					MC_ARG2_ONE |
-					MC_UPDATE_OP |
-					MC_OP_ARG1 );
+   case GL_REPLACE:
+      if (t->image[0].internalFormat == GL_ALPHA) 
+	 imesa->Setup[I810_CTXREG_MC0] = ( GFX_OP_MAP_COLOR_STAGES |
+					   MC_STAGE_0 |
+					   MC_UPDATE_DEST |
+					   MC_DEST_CURRENT |
+					   MC_UPDATE_ARG1 |
+					   MC_ARG1_TEX0_COLOR | 
+					   MC_UPDATE_ARG2 |
+					   MC_ARG2_ITERATED_COLOR |
+					   MC_UPDATE_OP |
+					   MC_OP_ARG2 );
+      else 
+	 imesa->Setup[I810_CTXREG_MC0] = ( GFX_OP_MAP_COLOR_STAGES |
+					   MC_STAGE_0 |
+					   MC_UPDATE_DEST |
+					   MC_DEST_CURRENT |
+					   MC_UPDATE_ARG1 |
+					   MC_ARG1_TEX0_COLOR | 
+					   MC_UPDATE_ARG2 |
+					   MC_ARG2_ONE |
+					   MC_UPDATE_OP |
+					   MC_OP_ARG1 );
 
       if (t->image[0].internalFormat == GL_RGB) {
 	 ma_modulate_op = MA_OP_ARG1;
@@ -812,16 +866,28 @@ static void i810UpdateTex0State( GLcontext *ctx )
 					MA_OP_ARG1 );
       break;
    case GL_BLEND:
-      imesa->Setup[I810_CTXREG_MC0] = ( GFX_OP_MAP_COLOR_STAGES |
-					MC_STAGE_0 |
-					MC_UPDATE_DEST |
-					MC_DEST_CURRENT |
-					MC_UPDATE_ARG1 |
-					MC_ARG1_COLOR_FACTOR | 
-					MC_UPDATE_ARG2 |
-					MC_ARG2_ITERATED_COLOR |
-					MC_UPDATE_OP |
-					MC_OP_LIN_BLEND_TEX0_COLOR );
+      if (t->image[0].internalFormat == GL_ALPHA) 
+	 imesa->Setup[I810_CTXREG_MC0] = ( GFX_OP_MAP_COLOR_STAGES |
+					   MC_STAGE_0 |
+					   MC_UPDATE_DEST |
+					   MC_DEST_CURRENT |
+					   MC_UPDATE_ARG1 |
+					   MC_ARG1_TEX0_COLOR | 
+					   MC_UPDATE_ARG2 |
+					   MC_ARG2_ITERATED_COLOR |
+					   MC_UPDATE_OP |
+					   MC_OP_ARG2 );
+      else 
+	 imesa->Setup[I810_CTXREG_MC0] = ( GFX_OP_MAP_COLOR_STAGES |
+					   MC_STAGE_0 |
+					   MC_UPDATE_DEST |
+					   MC_DEST_CURRENT |
+					   MC_UPDATE_ARG1 |
+					   MC_ARG1_COLOR_FACTOR | 
+					   MC_UPDATE_ARG2 |
+					   MC_ARG2_ITERATED_COLOR |
+					   MC_UPDATE_OP |
+					   MC_OP_LIN_BLEND_TEX0_COLOR );
 
       if (t->image[0].internalFormat == GL_RGB) {
 	 imesa->Setup[I810_CTXREG_MA0] = ( GFX_OP_MAP_ALPHA_STAGES |
@@ -911,7 +977,10 @@ static void i810UpdateTex1State( GLcontext *ctx )
       imesa->dirty |= I810_UPLOAD_TEX1IMAGE;
 
    imesa->CurrentTexObj[1] = t;
+   t->bound = 2;
 
+   if (t->MemBlock)
+      i810UpdateTexLRU( imesa, t );
 
    switch (ctx->Texture.Unit[1].EnvMode) {
    case GL_REPLACE:
@@ -1078,11 +1147,15 @@ static void i810UpdateTex1State( GLcontext *ctx )
 void i810UpdateTextureState( GLcontext *ctx )
 {
    i810ContextPtr imesa = I810_CONTEXT(ctx);
+   if (imesa->CurrentTexObj[0]) imesa->CurrentTexObj[0]->bound = 0;
+   if (imesa->CurrentTexObj[1]) imesa->CurrentTexObj[1]->bound = 0;
    imesa->CurrentTexObj[0] = 0;
-   imesa->CurrentTexObj[1] = 0;
+   imesa->CurrentTexObj[1] = 0;   
    i810UpdateTex0State( ctx );
    i810UpdateTex1State( ctx );
-   I810_CONTEXT( ctx )->dirty |= I810_UPLOAD_CTX;
+   I810_CONTEXT( ctx )->dirty |= (I810_UPLOAD_CTX |
+				  I810_UPLOAD_TEX0 | 
+				  I810_UPLOAD_TEX1);
 }
 
 
@@ -1091,25 +1164,30 @@ void i810UpdateTextureState( GLcontext *ctx )
  * DRIVER functions
  *****************************************/
 
-static void i810TexEnv( GLcontext *ctx, GLenum pname, const GLfloat *param )
+static void i810TexEnv( GLcontext *ctx, GLenum target, 
+			GLenum pname, const GLfloat *param )
 {
    i810ContextPtr imesa = I810_CONTEXT( ctx );
 
    if (pname == GL_TEXTURE_ENV_MODE) {
 
+      FLUSH_BATCH(imesa);	
       imesa->new_state |= I810_NEW_TEXTURE;
 
    } else if (pname == GL_TEXTURE_ENV_COLOR) {
 
       struct gl_texture_unit *texUnit = 
 	 &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
-      GLubyte c[4];
+      GLfloat *fc = texUnit->EnvColor;
       GLuint col;
 
-      FLOAT_RGBA_TO_UBYTE_RGBA(texUnit->EnvColor, c);
-      col = (c[3]<<24) | (c[0]<<16) | (c[1]<<8) | (c[2]);
+      col = ((((GLubyte)fc[3])<<24) | 
+	     (((GLubyte)fc[0])<<16) | 
+	     (((GLubyte)fc[1])<<8) | 
+	     (((GLubyte)fc[2])<<0));
     
       if (imesa->Setup[I810_CTXREG_CF1] != col) {
+	 FLUSH_BATCH(imesa);	
 	 imesa->Setup[I810_CTXREG_CF1] = col;      
 	 imesa->dirty |= I810_UPLOAD_CTX;
       }
@@ -1137,6 +1215,7 @@ static void i810TexImage( GLcontext *ctx,
 
    t = (i810TextureObjectPtr) tObj->DriverData;
    if (t) {
+      if (t->bound) FLUSH_BATCH(imesa);
       /* if this is the current object, it will force an update */
       i810DestroyTexObj( imesa, t );
       tObj->DriverData = 0;
@@ -1163,10 +1242,10 @@ static void i810TexSubImage( GLcontext *ctx, GLenum target,
    
    t = (i810TextureObjectPtr) tObj->DriverData;
    if (t) {
+      if (t->bound) FLUSH_BATCH( imesa );
       i810DestroyTexObj( imesa, t );
       tObj->DriverData = 0;
       imesa->new_state |= I810_NEW_TEXTURE;
-      i810glx.c_textureSwaps++;
    }
 }
 
@@ -1183,15 +1262,18 @@ static void i810TexParameter( GLcontext *ctx, GLenum target,
    switch (pname) {
    case GL_TEXTURE_MIN_FILTER:
    case GL_TEXTURE_MAG_FILTER:
+      if (t->bound) FLUSH_BATCH( imesa );
       i810SetTexFilter(t,tObj->MinFilter,tObj->MagFilter);
       break;
 
    case GL_TEXTURE_WRAP_S:
    case GL_TEXTURE_WRAP_T:
+      if (t->bound) FLUSH_BATCH( imesa );
       i810SetTexWrapping(t,tObj->WrapS,tObj->WrapT);
       break;
   
    case GL_TEXTURE_BORDER_COLOR:
+      if (t->bound) FLUSH_BATCH( imesa );
       i810SetTexBorderColor(t,tObj->BorderColor);
       break;
 
@@ -1206,7 +1288,14 @@ static void i810BindTexture( GLcontext *ctx, GLenum target,
 			     struct gl_texture_object *tObj )
 {
    i810ContextPtr imesa = I810_CONTEXT( ctx );
-   imesa->CurrentTexObj[ctx->Texture.CurrentUnit] = 0;  
+   
+   FLUSH_BATCH(imesa);
+
+   if (imesa->CurrentTexObj[ctx->Texture.CurrentUnit]) {
+      imesa->CurrentTexObj[ctx->Texture.CurrentUnit]->bound = 0;
+      imesa->CurrentTexObj[ctx->Texture.CurrentUnit] = 0;  
+   }
+
    imesa->new_state |= I810_NEW_TEXTURE;
 }
 
@@ -1216,6 +1305,13 @@ static void i810DeleteTexture( GLcontext *ctx, struct gl_texture_object *tObj )
    i810ContextPtr imesa = I810_CONTEXT( ctx );
 
    if (t) {
+
+      if (t->bound) {
+	 FLUSH_BATCH(imesa);
+	 imesa->CurrentTexObj[t->bound-1] = 0;
+	 imesa->new_state |= I810_NEW_TEXTURE;
+      }
+
       i810DestroyTexObj(imesa,t);
       tObj->DriverData=0;
    }
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810tex.h b/xc/lib/GL/mesa/src/drv/i810/i810tex.h
index d2b153ad8..a16733ec1 100644
--- a/xc/lib/GL/mesa/src/drv/i810/i810tex.h
+++ b/xc/lib/GL/mesa/src/drv/i810/i810tex.h
@@ -61,10 +61,11 @@ struct i810_texture_object_t {
    int Height;
    int texelBytes;
    int totalSize;
+   int bound;
 
    PMemBlock MemBlock;   
    char *BufAddr;
-
+   
    GLuint min_level;
    GLuint max_level;
    GLuint dirty_images;
@@ -87,8 +88,6 @@ struct i810_texture_object_t {
 #define I810_UPDATE_PALETTE    0x2
 #define I810_FALLBACK_PALETTE  0x4
 
-typedef struct i810_texture_object_t *i810TextureObjectPtr;
-
 void i810UpdateTextureState( GLcontext *ctx );
 void i810DDInitTextureFuncs( GLcontext *ctx );
 
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810tris.c b/xc/lib/GL/mesa/src/drv/i810/i810tris.c
index 08e012c66..df28991f1 100644
--- a/xc/lib/GL/mesa/src/drv/i810/i810tris.c
+++ b/xc/lib/GL/mesa/src/drv/i810/i810tris.c
@@ -26,35 +26,17 @@
 #include <stdio.h>
 #include <math.h>
 
+#include "types.h"
 #include "vb.h"
 #include "pipeline.h"
 
 #include "mm.h"
-#include "i810lib.h"
 #include "i810tris.h"
 #include "i810vb.h"
 #include "i810log.h"
 
-
-static void i810_null_quad( GLcontext *ctx, GLuint v0,
-			    GLuint v1, GLuint v2, GLuint v3, GLuint pv )
-{
-}     
-
-static void i810_null_triangle( GLcontext *ctx, GLuint v0,
-				GLuint v1, GLuint v2, GLuint pv )
-{
-}     
-
-static void i810_null_line( GLcontext *ctx, GLuint v1, GLuint v2, GLuint pv ) 
-{
-}
-
-static void i810_null_points( GLcontext *ctx, GLuint first, GLuint last ) 
-{
-}
-
-
+/* Used in i810tritmp.h
+ */
 #define I810_COLOR(to, from) {			\
   (to)[0] = (from)[2];				\
   (to)[1] = (from)[1];				\
@@ -69,17 +51,6 @@ static quad_func     quad_tab[0x20];
 static line_func     line_tab[0x20];  
 static points_func   points_tab[0x20];
 
-void i810PrintRenderState( const char *msg, GLuint state )
-{
-   fprintf(stderr, "%s: (%x) %s%s%s%s%s%s\n",
-	   msg, (int) state,
-	   (state & I810_FLAT_BIT)       ? "flat, "       : "",
-	   (state & I810_OFFSET_BIT)     ? "offset, "     : "",
-	   (state & I810_TWOSIDE_BIT)    ? "twoside, "    : "",
-	   (state & I810_ANTIALIAS_BIT)  ? "antialias, "  : "",
-	   (state & I810_NODRAW_BIT)     ? "no-draw, "    : "",
-	   (state & I810_FALLBACK_BIT)   ? "fallback"     : "");
-}
 
 #define IND (0)
 #define TAG(x) x
@@ -89,33 +60,32 @@ void i810PrintRenderState( const char *msg, GLuint state )
 #define TAG(x) x##_flat
 #include "i810tritmp.h"
 
-#define IND (I810_OFFSET_BIT)
+#define IND (I810_OFFSET_BIT)	/* wide */
 #define TAG(x) x##_offset
 #include "i810tritmp.h"
 
-#define IND (I810_OFFSET_BIT|I810_FLAT_BIT)
+#define IND (I810_OFFSET_BIT|I810_FLAT_BIT) /* wide|flat */
 #define TAG(x) x##_offset_flat
 #include "i810tritmp.h"
 
-#define IND (I810_TWOSIDE_BIT)
+#define IND (I810_TWOSIDE_BIT)	/* stipple */
 #define TAG(x) x##_twoside
 #include "i810tritmp.h"
 
-#define IND (I810_TWOSIDE_BIT|I810_FLAT_BIT)
+#define IND (I810_TWOSIDE_BIT|I810_FLAT_BIT) /* stipple|flat */
 #define TAG(x) x##_twoside_flat
 #include "i810tritmp.h"
 
-#define IND (I810_TWOSIDE_BIT|I810_OFFSET_BIT)
+#define IND (I810_TWOSIDE_BIT|I810_OFFSET_BIT) /* stipple|wide */
 #define TAG(x) x##_twoside_offset
 #include "i810tritmp.h"
 
-#define IND (I810_TWOSIDE_BIT|I810_OFFSET_BIT|I810_FLAT_BIT)
+#define IND (I810_TWOSIDE_BIT|I810_OFFSET_BIT|I810_FLAT_BIT) /* stip|wide|flat*/
 #define TAG(x) x##_twoside_offset_flat
 #include "i810tritmp.h"
 
 void i810DDTrifuncInit()
 {
-   int i;
    init();
    init_flat();
    init_offset();
@@ -124,128 +94,9 @@ void i810DDTrifuncInit()
    init_twoside_flat();
    init_twoside_offset();
    init_twoside_offset_flat();
-
-   /* Hmmm...
-    */
-   for (i = 0 ; i < 0x20 ; i++) {
-      if (i & ~I810_FLAT_BIT) {
-	 points_tab[i] = points_tab[i&I810_FLAT_BIT];
-	 line_tab[i] = line_tab[i&I810_FLAT_BIT];
-      }
-   }
-
-   for (i = 0 ; i < 0x20 ; i++) 
-      if ((i & (I810_NODRAW_BIT|I810_FALLBACK_BIT)) == I810_NODRAW_BIT ||
-	  i810glx.nullprims) 
-      {
-	 quad_tab[i] = i810_null_quad; 
-	 tri_tab[i] = i810_null_triangle; 
-	 line_tab[i] = i810_null_line;
-	 points_tab[i] = i810_null_points;
-      }
-
-   if (i810glx.noFallback) {
-      for (i = 0 ; i < 0x10 ; i++) {
-	 points_tab[i|I810_FALLBACK_BIT] = points_tab[i];
-	 line_tab[i|I810_FALLBACK_BIT] = line_tab[i];
-	 tri_tab[i|I810_FALLBACK_BIT] = tri_tab[i];
-	 quad_tab[i|I810_FALLBACK_BIT] = quad_tab[i];
-      }
-   }
-
 }
 
 
-/* Everything is done via single triangle instructiopns at the moment;
- * this can change fairly easily.
- */
-#if I810_USE_BATCH
-
-GLuint *i810AllocPrimitiveVerts( i810ContextPtr imesa, int dwords ) 
-{
-   GLuint orig_dwords = dwords;
-
-   dwords+=2;
-   dwords&=~1;
-
-   while (1) {
-      if (i810glx.dma_buffer->space < dwords * 4) 
-      {
-	 if (I810_DEBUG & DEBUG_VERBOSE_RING)
-	    fprintf(stderr, "i810AllocPrimitiveVerts: dma buffer overflow\n");
-	 i810DmaOverflow( dwords );
-      }
-      else
-      {
-	 GLuint start = i810glx.dma_buffer->head;
-	 i810glx.dma_buffer->head += dwords * 4;
-	 i810glx.dma_buffer->space -= dwords * 4;
-
-	 if ((orig_dwords & 1) == 0) {
-	    *(GLuint *)(i810glx.dma_buffer->virtual_start + start ) = 0;
-	    start += 4;
-	 }
-
-	 *(GLuint *)(i810glx.dma_buffer->virtual_start + start ) = 
-	    GFX_OP_PRIMITIVE | PR_TRIANGLES | (orig_dwords-1);
-
-	 return (GLuint *)(i810glx.dma_buffer->virtual_start + start + 4);
-      }
-   }
-}
-
-#else
-
-/* Hardware lock is held.
- */
-GLuint *i810AllocPrimitiveVerts( i810ContextPtr imesa, int dwords ) 
-{
-   GLuint orig_dwords = dwords;
-   
-   dwords+=2;
-   dwords&=~1;
-
-   while (1)
-   {
-      BEGIN_LP_RING( imesa, dwords );
-
-      if (outring + dwords * 4 != ((outring + dwords * 4) & ringmask)) 
-      {
-	 int i;
-
-	 if (I810_DEBUG & DEBUG_VERBOSE_RING)
-	    fprintf(stderr, "\n\nwrap case in i810AllocPrimitiveVerts\n\n");
-	 
-	 for (i = 0 ; i < dwords ; i++)
-	    OUT_RING( 0 );
-	 ADVANCE_LP_RING();
-      }
-      else 
-      {
-	 i810glx.LpRing.tail = outring + dwords * 4;
-	 
-	 if ((orig_dwords & 1) == 0)
-	    OUT_RING(0);
-	 
-	 OUT_RING( GFX_OP_PRIMITIVE | PR_TRIANGLES | (orig_dwords-1) );
-	 return (GLuint *)(virt + outring);
-      }
-   }
-}
-
-#if 0
-void i810FinishPrim( void )
-{
-   char *virt = i810glx.LpRing.virtual_start;
-
-   for (i = i810glx.LpRing.head ; i != i810glx.LpRing.tail ; i += 4) 
-      fprintf(stderr, "prim %x (%f)", *(virt + i), *(float *)(virt + i));
-
-   OUTREG(LP_RING + RING_TAIL, i810glx.LpRing.tail);
-}
-#endif 
-#endif
-
 
 
 void i810DDChooseRenderState( GLcontext *ctx )
@@ -253,50 +104,41 @@ void i810DDChooseRenderState( GLcontext *ctx )
    i810ContextPtr imesa = I810_CONTEXT( ctx );
    GLuint flags = ctx->TriangleCaps;
 
-   ctx->IndirectTriangles &= ~DD_SW_RASTERIZE;
+   imesa->IndirectTriangles = 0;
 
    if (flags) {
       GLuint ind = 0;
       GLuint shared = 0;
-      GLuint setup = imesa->setupindex;
-      GLuint fallback = I810_FALLBACK_BIT; 
-
-      if (i810glx.noFallback) fallback = 0; 
 
-      if ((flags & DD_FLATSHADE) && (setup & I810_RGBA_BIT)) shared |= I810_FLAT_BIT;
-      if (flags & DD_MULTIDRAW)                    shared |= fallback;
+      if (flags & DD_FLATSHADE)                    shared |= I810_FLAT_BIT;
+      if (flags & DD_MULTIDRAW)                    shared |= I810_FALLBACK_BIT;
       if (flags & DD_SELECT)                       shared |= I810_FALLBACK_BIT;
       if (flags & DD_FEEDBACK)                     shared |= I810_FALLBACK_BIT;
 
-      ind = shared;
-      if (flags & DD_POINT_SMOOTH)                 ind |= I810_ANTIALIAS_BIT;
-      if (flags & DD_POINT_ATTEN)                  ind |= fallback;
-
-      imesa->renderindex = ind;
-      imesa->PointsFunc = points_tab[ind]; 
-      if (ind & I810_FALLBACK_BIT)  
-	 ctx->IndirectTriangles |= DD_POINT_SW_RASTERIZE;
+      imesa->renderindex = shared;
+      imesa->PointsFunc = points_tab[shared]; 
 
       ind = shared;
-      if (flags & DD_LINE_SMOOTH)                   ind |= I810_ANTIALIAS_BIT;
-      if (flags & DD_LINE_STIPPLE)                  ind |= fallback;
+      if (flags & DD_LINE_WIDTH)                    ind |= I810_WIDE_LINE_BIT;
+      if (flags & DD_LINE_STIPPLE)                  ind |= I810_FALLBACK_BIT;
 
       imesa->renderindex |= ind;
       imesa->LineFunc = line_tab[ind];
       if (ind & I810_FALLBACK_BIT) 
-	 ctx->IndirectTriangles |= DD_LINE_SW_RASTERIZE;
+	 imesa->IndirectTriangles |= DD_LINE_SW_RASTERIZE;
 
       ind = shared;
-      if (flags & DD_TRI_SMOOTH)                    ind |= I810_ANTIALIAS_BIT;
       if (flags & DD_TRI_OFFSET)                    ind |= I810_OFFSET_BIT;
       if (flags & DD_TRI_LIGHT_TWOSIDE)             ind |= I810_TWOSIDE_BIT;
-      if (flags & (DD_TRI_UNFILLED|DD_TRI_STIPPLE)) ind |= fallback;	
+      if (flags & DD_TRI_UNFILLED)                  ind |= I810_FALLBACK_BIT;
+      if ((flags & DD_TRI_STIPPLE) &&
+	  (ctx->IndirectTriangles & DD_TRI_STIPPLE)) ind |= I810_FALLBACK_BIT;	
 
       imesa->renderindex |= ind;
       imesa->TriangleFunc = tri_tab[ind];
       imesa->QuadFunc = quad_tab[ind];
       if (ind & I810_FALLBACK_BIT)
-	 ctx->IndirectTriangles |= (DD_TRI_SW_RASTERIZE|DD_QUAD_SW_RASTERIZE);
+	 imesa->IndirectTriangles |= (DD_TRI_SW_RASTERIZE|DD_QUAD_SW_RASTERIZE);
    } 
    else if (imesa->renderindex)
    {
@@ -307,9 +149,9 @@ void i810DDChooseRenderState( GLcontext *ctx )
       imesa->QuadFunc     = quad_tab[0];
    }
 
+
    if (I810_DEBUG&DEBUG_VERBOSE_API) {
       gl_print_tri_caps("tricaps", ctx->TriangleCaps);
-      i810PrintRenderState("i810: Render state", imesa->renderindex);
    }
 }
 
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810tris.h b/xc/lib/GL/mesa/src/drv/i810/i810tris.h
index 81cdb5159..3b3843683 100644
--- a/xc/lib/GL/mesa/src/drv/i810/i810tris.h
+++ b/xc/lib/GL/mesa/src/drv/i810/i810tris.h
@@ -27,37 +27,48 @@
 #define I810TRIS_INC
 
 #include "types.h"
-#include "i810dma.h"
+#include "i810ioctl.h"
 
 extern void i810PrintRenderState( const char *msg, GLuint state );
 extern void i810DDChooseRenderState(GLcontext *ctx);
 extern void i810DDTrifuncInit( void );
 
-extern GLuint *i810AllocPrimitiveVerts( i810ContextPtr imesa, int dwords );
 
-
-
-/* Todo: 
- *    - multidraw, ...
- *    - Antialiasing (?)
- *    - line and polygon stipple
- *    - select and feedback 
- *    - stencil 
- *    - point parameters
- *    - 
- */
-#define I810_ANTIALIAS_BIT   0       /* ignored for now, no fallback */
+/* shared */
 #define I810_FLAT_BIT 	     0x1
+
+/* triangle */
 #define I810_OFFSET_BIT	     0x2	
 #define I810_TWOSIDE_BIT     0x4
-#define I810_NODRAW_BIT	     0x8
-#define I810_FALLBACK_BIT    0x10
 
-/* Not in use:
- */
-#define I810_FEEDBACK_BIT    0x20
-#define I810_SELECT_BIT      0x40
-#define I810_POINT_PARAM_BIT 0x80	/* not needed? */
+/* line */
+#define I810_WIDE_LINE_BIT    0x2 
+#define I810_STIPPLE_LINE_BIT 0x4 
+
+/* shared */
+#define I810_FALLBACK_BIT    0x8 
+
+
+
+
+
+static i810_vertex __inline__ *i810AllocTriangles( i810ContextPtr imesa, int nr)
+{
+   GLuint *start = i810AllocDwords( imesa, 30*nr, PR_TRIANGLES );
+   return (i810_vertex *)start;
+}
+
+static i810_vertex __inline__ *i810AllocLine( i810ContextPtr imesa ) 
+{
+   GLuint *start = i810AllocDwords( imesa, 20, PR_LINES );
+   return (i810_vertex *)start;
+}
+
+static i810_vertex __inline__ *i810AllocRect( i810ContextPtr imesa ) 
+{
+   GLuint *start = i810AllocDwords( imesa, 30, PR_RECTS );
+   return (i810_vertex *)start;
+}
 
 
 
@@ -66,23 +77,17 @@ static void __inline__ i810_draw_triangle( i810ContextPtr imesa,
 					   i810_vertex *v1, 
 					   i810_vertex *v2 )
 {
-   i810_vertex *wv = (i810_vertex *)i810AllocPrimitiveVerts( imesa, 30 );
+   i810_vertex *wv = i810AllocTriangles( imesa, 1 );
    wv[0] = *v0;
    wv[1] = *v1;
    wv[2] = *v2;
-   FINISH_PRIM();
 }
 
 
-
-
-/* These can go soon, but for the meantime we're using triangles for 
- * everything.
- */
 static __inline__ void i810_draw_point( i810ContextPtr imesa,
 					i810_vertex *tmp, float sz )
 {
-   i810_vertex *wv = (i810_vertex *)i810AllocPrimitiveVerts( imesa, 6*10 );
+   i810_vertex *wv = i810AllocTriangles( imesa, 2 );
 
    wv[0] = *tmp;
    wv[0].x = tmp->x - sz;
@@ -108,16 +113,25 @@ static __inline__ void i810_draw_point( i810ContextPtr imesa,
    wv[5].x = tmp->x - sz;
    wv[5].y = tmp->y - sz;
 
-   FINISH_PRIM();
 }
 
 
-static __inline__ void i810_draw_line( i810ContextPtr imesa, 
-				       i810_vertex *tmp0, 
-				       i810_vertex *tmp1,
-				       float width )
+static __inline__ void i810_draw_line_line( i810ContextPtr imesa, 
+					    i810_vertex *tmp0, 
+					    i810_vertex *tmp1 )
 {
-   i810_vertex *wv = (i810_vertex *)i810AllocPrimitiveVerts( imesa, 6 * 10 );
+   i810_vertex *wv = i810AllocLine( imesa );
+   wv[0] = *tmp0;
+   wv[1] = *tmp1;
+}
+
+static __inline__ void i810_draw_tri_line( i810ContextPtr imesa, 
+					   i810_vertex *tmp0, 
+					   i810_vertex *tmp1,
+					   float width )
+{
+   i810_vertex *wv = i810AllocTriangles( imesa, 2 );
+      
    float dx, dy, ix, iy;
 
    dx = tmp0->x - tmp1->x;
@@ -151,8 +165,15 @@ static __inline__ void i810_draw_line( i810ContextPtr imesa,
    wv[5] = *tmp1;
    wv[5].x = tmp1->x + ix;
    wv[5].y = tmp1->y + iy;
+}
 
-   FINISH_PRIM();
+
+static __inline__ void i810_draw_line( i810ContextPtr imesa, 
+				       i810_vertex *tmp0, 
+				       i810_vertex *tmp1,
+				       float width )
+{
+   i810_draw_line_line( imesa, tmp0, tmp1 );
 }
 
 #endif
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810tritmp.h b/xc/lib/GL/mesa/src/drv/i810/i810tritmp.h
index 97c101c17..9d6b43d93 100644
--- a/xc/lib/GL/mesa/src/drv/i810/i810tritmp.h
+++ b/xc/lib/GL/mesa/src/drv/i810/i810tritmp.h
@@ -63,44 +63,37 @@ static __inline__ void TAG(triangle)( GLcontext *ctx, GLuint e0,
    }
 #endif  
 
-   i810glx.c_triangles++;
-
    
-   BEGIN_CLIP_LOOP(imesa)
-      {
-	 i810_vertex *wv = (i810_vertex *)i810AllocPrimitiveVerts( imesa, 30 );
-	 wv[0] = *v0;
+   {
+      i810_vertex *wv = i810AllocTriangles( imesa, 1 );
+      wv[0] = *v0;
 #if (IND & (I810_FLAT_BIT|I810_TWOSIDE_BIT))
-	 *((int *)(&wv[0].color)) = c0;
+      *((int *)(&wv[0].color)) = c0;
 #endif
 #if (IND & I810_OFFSET_BIT)
-	 wv[0].z = v0->z + offset;
+      wv[0].z = v0->z + offset;
 #endif
 
 
-	 wv[1] = *v1;
+      wv[1] = *v1;
 #if (IND & (I810_FLAT_BIT|I810_TWOSIDE_BIT))
-	 *((int *)(&wv[1].color)) = c1;
+      *((int *)(&wv[1].color)) = c1;
 #endif
 #if (IND & I810_OFFSET_BIT)
-	 wv[1].z = v1->z + offset;
+      wv[1].z = v1->z + offset;
 #endif
 
-	 wv[2] = *v2;
+      wv[2] = *v2;
 #if (IND & (I810_FLAT_BIT|I810_TWOSIDE_BIT))
-	 *((int *)(&wv[2].color)) = c2;
+      *((int *)(&wv[2].color)) = c2;
 #endif
 #if (IND & I810_OFFSET_BIT)
-	 wv[2].z = v2->z + offset;
+      wv[2].z = v2->z + offset;
 #endif
-
-	 FINISH_PRIM();
-      }
-   END_CLIP_LOOP(imesa);
+   }
 }
 
 
-
 static void TAG(quad)( GLcontext *ctx, GLuint v0,
 		       GLuint v1, GLuint v2, GLuint v3, 
 		       GLuint pv )
@@ -109,25 +102,35 @@ static void TAG(quad)( GLcontext *ctx, GLuint v0,
    TAG(triangle)( ctx, v1, v2, v3, pv );
 }
 
-
-#if ((IND & ~I810_FLAT_BIT) == 0)
-
 static void TAG(line)( GLcontext *ctx, GLuint v0, GLuint v1, GLuint pv )
 {
    i810ContextPtr imesa = I810_CONTEXT( ctx );
    i810VertexPtr i810VB = I810_DRIVER_DATA(ctx->VB)->verts;
-   i810_vertex tmp0 = i810VB[v0].v;
-   i810_vertex tmp1 = i810VB[v1].v;
-   float width = ctx->Line.Width;
+   int tmp0, tmp1;
+   (void) tmp0; (void) tmp1;
+
 
    if (IND & I810_FLAT_BIT) {
-      *(int *)&tmp1.color = *(int *)&tmp0.color = 
-	 *(int *)&i810VB[pv].v.color;
-   }
+      tmp0 = *(int *)&i810VB[v0].v.color;
+      tmp1 = *(int *)&i810VB[v1].v.color;
+      i810VB[v0].v.color = i810VB[pv].v.color;
+      i810VB[v1].v.color = i810VB[pv].v.color;
+   }      
 
-   BEGIN_CLIP_LOOP(imesa)
-      i810_draw_line( imesa, &tmp0, &tmp1, width );
-   END_CLIP_LOOP(imesa);
+   if (IND & I810_WIDE_LINE_BIT)
+   {
+      i810_draw_tri_line( imesa, &i810VB[v0].v, &i810VB[v1].v, 
+			  ctx->Line.Width );
+   } 
+   else 
+   {
+      i810_draw_line_line( imesa, &i810VB[v0].v, &i810VB[v1].v );
+   }      
+
+   if (IND & I810_FLAT_BIT) {
+      *(int *)&i810VB[v0].v.color = tmp0;
+      *(int *)&i810VB[v1].v.color = tmp1;
+   }      
 }
 
 
@@ -143,28 +146,22 @@ static void TAG(points)( GLcontext *ctx, GLuint first, GLuint last )
     * ctx->Driver.ReducedPrimitiveChange() callback.  
     */
    
-   BEGIN_CLIP_LOOP(imesa)
-      for(i=first;i<=last;i++) {
-	 if(VB->ClipMask[i]==0) {
-	    i810_vertex *tmp = &i810VB[i].v;
-	    i810_draw_point( imesa, tmp, sz );
-	 }
+   for(i=first;i<=last;i++) {
+      if(VB->ClipMask[i]==0) {
+	 i810_vertex *tmp = &i810VB[i].v;
+	 i810_draw_point( imesa, tmp, sz );
       }
-   END_CLIP_LOOP(imesa);
+   }
 }
 
-#endif
 
 
 static void TAG(init)( void )
 {
    tri_tab[IND] = TAG(triangle);
    quad_tab[IND] = TAG(quad);
-
-#if ((IND & ~I810_FLAT_BIT) == 0) 
    line_tab[IND] = TAG(line);
    points_tab[IND] = TAG(points);
-#endif
 }
 
 
diff --git a/xc/lib/GL/mesa/src/drv/i810/i810vb.c b/xc/lib/GL/mesa/src/drv/i810/i810vb.c
index 6ba7eb9a5..6f8d163c4 100644
--- a/xc/lib/GL/mesa/src/drv/i810/i810vb.c
+++ b/xc/lib/GL/mesa/src/drv/i810/i810vb.c
@@ -23,11 +23,9 @@
  *
  */
  
-#include "i810lib.h"
 #include "i810context.h"
 #include "i810vb.h"
 #include "i810log.h"
-#include "i810dma.h"
 
 #include "stages.h"
 
diff --git a/xc/lib/GL/mesa/src/drv/mga/Imakefile b/xc/lib/GL/mesa/src/drv/mga/Imakefile
index e4e39b747..cbcd09821 100644
--- a/xc/lib/GL/mesa/src/drv/mga/Imakefile
+++ b/xc/lib/GL/mesa/src/drv/mga/Imakefile
@@ -1,4 +1,6 @@
 
+#include <Threads.tmpl>
+
 #define DoNormalLib NormalLibGlx
 #define DoSharedLib SharedLibGlx
 #define DoExtraLib SharedLibGlx
@@ -9,38 +11,285 @@
 ALLOC_DEFINES = -DMALLOC_0_RETURNS_NULL
 #endif
 
+#if defined(LinuxArchitecture)
+OS_SUBDIR = linux
+#endif
+#if defined(FreeBSDArchitecture)
+OS_SUBDIR = bsd
+#endif
+
+#ifdef i386Architecture
+#ifdef MesaUse3DNow
+     ASM_DEFS = -DUSE_MMX_ASM -DUSE_X86_ASM -DUSE_3DNOW_ASM
+#else
+     ASM_DEFS = -DUSE_MMX_ASM -DUSE_X86_ASM
+#endif
+#endif
+
+
 #if BuildXF86DRI
   DRI_DEFINES = GlxDefines -DDRIVERTS
  DRI_INCLUDES = -I../../../../dri -I../../../../glx \
+		-I../../../dri \
 		-I$(TOP)/include -I$(TOP)/include/GL \
 		-I$(XF86OSSRC) -I$(SERVERSRC)/GL/dri \
 		-I$(XF86DRIVERSRC)/mga \
 		-I../../../include -I../.. -I../../X -I../common \
-	        -I$(XF86OSSRC)/linux/drm/kernel
+	        -I$(XF86OSSRC)/$(OS_SUBDIR)/drm/kernel
 #endif
 
 MESA_INCLUDES = -I. -I.. -I../../include
 
 
 
-      DEFINES = $(ALLOC_DEFINES) $(DRI_DEFINES)
-     INCLUDES = -I$(XLIBSRC) -I$(EXTINCSRC) $(MESA_INCLUDES) $(DRI_INCLUDES) \
-		-I/usr/include/glide
-      DRISRCS = mgaclear.c mgacnvtex.c mgadd.c mgadepth.c	\
-		mgafastpath.c					\
+      DEFINES = $(ALLOC_DEFINES) $(DRI_DEFINES) $(ASM_DEFS)
+     INCLUDES = -I$(XLIBSRC) -I$(EXTINCSRC) $(MESA_INCLUDES) $(DRI_INCLUDES) 
+
+#if 0
+	LOSRC = ../../../../lowpc.c
+	LOOBJ = ../../../../lowpc.o
+
+	HISRC = ../../../../highpc.c
+	HIOBJ = ../../../../highpc.o
+#endif
+
+      DRISRCS = ../../../dri/dri_mesa.c \
+		../../../../dri/dri_tmm.c
+
+      DRIOBJS = ../../../dri/dri_mesa.o \
+		../../../../dri/dri_tmm.o
+
+      DRMSRCS = ../../../../dri/drm/xf86drm.c \
+		../../../../dri/drm/xf86drmHash.c \
+		../../../../dri/drm/xf86drmRandom.c \
+		../../../../dri/drm/xf86drmSL.c
+
+      DRMOBJS = ../../../../dri/drm/xf86drm.o \
+		../../../../dri/drm/xf86drmHash.o \
+		../../../../dri/drm/xf86drmRandom.o \
+		../../../../dri/drm/xf86drmSL.o
+
+
+      MGASRCS = mgaclear.c mgacnvtex.c mgadd.c \
+		mgafastpath.c mgaeltpath.c \
 		mgapipeline.c	\
 		mgaspan.c mgastate.c mgatex.c		\
-		mgatris.c mgavb.c mgaioctl.c mga_xmesa.c
+		mgatris.c mgavb.c mgaioctl.c mga_xmesa.c mgabuffers.c
 
-      DRIOBJS = mgaclear.o mgacnvtex.o mgadd.o mgadepth.o	\
-		mgafastpath.o					\
+      MGAOBJS = mgaclear.o mgacnvtex.o mgadd.o \
+		mgafastpath.o mgaeltpath.o \
 		mgapipeline.o	\
 		mgaspan.o mgastate.o mgatex.o		\
-		mgatris.o mgavb.o mgaioctl.o mga_xmesa.o
+		mgatris.o mgavb.o mgaioctl.o mga_xmesa.o mgabuffers.o
+
+     MESASRCS = ../../aatriangle.c \
+		../../accum.c \
+		../../alpha.c \
+		../../alphabuf.c \
+		../../attrib.c \
+		../../bbox.c \
+		../../bitmap.c \
+		../../blend.c \
+		../../buffers.c \
+		../../clip.c \
+		../../colortab.c \
+		../../config.c \
+		../../context.c \
+		../../copypix.c \
+		../../cva.c \
+		../../debug_xform.c \
+		../../depth.c \
+		../../dlist.c \
+		../../drawpix.c \
+		../../enable.c \
+		../../enums.c \
+		../../eval.c \
+		../../extensions.c \
+		../../feedback.c \
+		../../fog.c \
+		../../get.c \
+		../../glapi.c \
+		../../glapinoop.c \
+		../../glthread.c \
+		../../hash.c \
+		../../image.c \
+		../../imaging.c \
+		../../light.c \
+		../../lines.c \
+		../../logic.c \
+		../../masking.c \
+		../../matrix.c \
+		../../mem.c \
+		../../mmath.c \
+		../../pb.c \
+		../../pipeline.c \
+		../../pixel.c \
+		../../pixeltex.c \
+		../../points.c \
+		../../polygon.c \
+		../../quads.c \
+		../../rastpos.c \
+		../../readpix.c \
+		../../rect.c \
+		../../scissor.c \
+		../../shade.c \
+		../../span.c \
+		../../stages.c \
+		../../state.c \
+		../../stencil.c \
+		../../teximage.c \
+		../../texobj.c \
+		../../texstate.c \
+		../../texture.c \
+		../../texutil.c \
+		../../translate.c \
+		../../triangle.c \
+		../../varray.c \
+		../../vb.c \
+		../../vbcull.c \
+		../../vbfill.c \
+		../../vbindirect.c \
+		../../vbrender.c \
+		../../vbxform.c \
+		../../vector.c \
+		../../vertices.c \
+		../../winpos.c \
+		../../xform.c \
+		../../zoom.c \
+		../../X86/common_x86.c
+
+     MESAOBJS = ../../aatriangle.o \
+		../../accum.o \
+		../../alpha.o \
+		../../alphabuf.o \
+		../../attrib.o \
+		../../bbox.o \
+		../../bitmap.o \
+		../../blend.o \
+		../../buffers.o \
+		../../clip.o \
+		../../colortab.o \
+		../../config.o \
+		../../context.o \
+		../../copypix.o \
+		../../cva.o \
+		../../debug_xform.o \
+		../../depth.o \
+		../../dlist.o \
+		../../drawpix.o \
+		../../enable.o \
+		../../enums.o \
+		../../eval.o \
+		../../extensions.o \
+		../../feedback.o \
+		../../fog.o \
+		../../get.o \
+		../../hash.o \
+		../../hint.o \
+		../../image.o \
+		../../imaging.o \
+		../../light.o \
+		../../lines.o \
+		../../logic.o \
+		../../masking.o \
+		../../matrix.o \
+		../../mem.o \
+		../../mmath.o \
+		../../pb.o \
+		../../pipeline.o \
+		../../pixel.o \
+		../../pixeltex.o \
+		../../points.o \
+		../../polygon.o \
+		../../quads.o \
+		../../rastpos.o \
+		../../readpix.o \
+		../../rect.o \
+		../../scissor.o \
+		../../shade.o \
+		../../span.o \
+		../../stages.o \
+		../../state.o \
+		../../stencil.o \
+		../../teximage.o \
+		../../texobj.o \
+		../../texstate.o \
+		../../texture.o \
+		../../texutil.o \
+		../../translate.o \
+		../../triangle.o \
+		../../varray.o \
+		../../vb.o \
+		../../vbcull.o \
+		../../vbfill.o \
+		../../vbindirect.o \
+		../../vbrender.o \
+		../../vbxform.o \
+		../../vector.o \
+		../../vertices.o \
+		../../winpos.o \
+		../../xform.o \
+		../../zoom.o
+
+#ifdef i386Architecture
+     X86_SRCS = ../../X86/x86.c \
+		../../X86/x86a.S \
+		../../X86/common_x86.c \
+		../../X86/common_x86asm.S \
+		../../X86/vertex.S
+
+     X86_OBJS = ../../X86/x86.o \
+		../../X86/x86a.o \
+		../../X86/common_x86.o \
+		../../X86/common_x86asm.o \
+		../../X86/vertex.o
+
+     MMX_SRCS = ../../X86/mmx_blend.S
+
+     MMX_OBJS = ../../X86/mmx_blend.o
+
+#ifdef MesaUse3DNow
+   3DNOW_SRCS = ../../X86/3dnow.c \
+		../../X86/3dnow_norm_raw.S \
+		../../X86/3dnow_xform_masked1.S \
+		../../X86/3dnow_xform_masked2.S \
+		../../X86/3dnow_xform_masked3.S \
+		../../X86/3dnow_xform_masked4.S \
+		../../X86/3dnow_xform_raw1.S \
+		../../X86/3dnow_xform_raw2.S \
+		../../X86/3dnow_xform_raw3.S \
+		../../X86/3dnow_xform_raw4.S \
+		../../X86/vertex_3dnow.S
 
+   3DNOW_OBJS = ../../X86/3dnow.o \
+		../../X86/3dnow_norm_raw.o \
+		../../X86/3dnow_xform_masked1.o \
+		../../X86/3dnow_xform_masked2.o \
+		../../X86/3dnow_xform_masked3.o \
+		../../X86/3dnow_xform_masked4.o \
+		../../X86/3dnow_xform_raw1.o \
+		../../X86/3dnow_xform_raw2.o \
+		../../X86/3dnow_xform_raw3.o \
+		../../X86/3dnow_xform_raw4.o \
+		../../X86/vertex_3dnow.o
+
+#endif
+#endif
+
+      ASMSRCS = $(X86_SRCS) $(MMX_SRCS) $(3DNOW_SRCS)
+      ASMOBJS = $(X86_OBJS) $(MMX_OBJS) $(3DNOW_OBJS)
+
+      COMMONSRCS = ../common/mm.c ../common/hwlog.c
+      COMMONOBJS = ../common/mm.o ../common/hwlog.o
+
+         SRCS = $(LOWSRC) $(DRISRCS) $(DRMSRCS) $(MESASRCS) $(ASMSRCS) $(COMMONSRCS) $(MGASRCS) $(HISRC)
+         OBJS = $(LOWOBJ) $(DRIOBJS) $(DRMOBJS) $(MESAOBJS) $(ASMOBJS) $(COMMONOBJS) $(MGAOBJS) $(HIOBJ)
+
+REQUIREDLIBS += -lm
+#if !GlxBuiltInMga
+REQUIREDLIBS += -L../../../.. -lGL
+#endif
 
-	 SRCS = $(DRISRCS) 
-	 OBJS = $(DRIOBJS) 
 
 #if !GlxUseBuiltInDRIDriver
 #undef DoNormalLib NormalLibGlx
@@ -61,7 +310,7 @@ LIBNAME = mga_dri.so
 ALL_OBJS = $(OBJS)
 ALL_DEPS = DONE
 SharedDepModuleTarget($(LIBNAME),$(ALL_DEPS),$(ALL_OBJS))
-InstallDynamicModule($(LIBNAME),$(MODULEDIR),.)
+InstallDynamicModule($(LIBNAME),$(MODULEDIR)/dri,.)
 #endif
 
 DependTarget()
diff --git a/xc/lib/GL/mesa/src/drv/mga/mga_xmesa.c b/xc/lib/GL/mesa/src/drv/mga/mga_xmesa.c
index 496c29a05..c6aadc144 100644
--- a/xc/lib/GL/mesa/src/drv/mga/mga_xmesa.c
+++ b/xc/lib/GL/mesa/src/drv/mga/mga_xmesa.c
@@ -27,7 +27,7 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 /*
  * Authors:
- *   Daryll Strauss <daryll@precisioninsight.com>
+ *   Keith Whitwell <keithw@precisioninsight.com>
  *
  */
 
@@ -36,10 +36,12 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include <X11/Xlibint.h>
 #include <stdio.h>
 
+#include "drm.h"
 #include "mga_xmesa.h"
 #include "context.h"
 #include "vbxform.h"
 #include "matrix.h"
+#include "mmath.h"
 #include "simple_list.h"
 
 #include "mgadd.h"
@@ -49,20 +51,25 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "mgadepth.h"
 #include "mgatris.h"
 #include "mgapipeline.h"
+#include "mgabuffers.h"
 
 #include "xf86dri.h"
-#include "mga_dri.h"
-#include "mga_drm_public.h"
 #include "mga_xmesa.h"
 
+#include "mga_dri.h"
+
 
 
 
 
 #ifndef MGA_DEBUG
 int MGA_DEBUG = (0
-/*  		  | MGA_DEBUG_ALWAYS_SYNC */
-/*  		  | MGA_DEBUG_VERBOSE_MSG */
+/*  		  | DEBUG_ALWAYS_SYNC */
+/*  		  | DEBUG_VERBOSE_MSG */
+/*   		  | DEBUG_VERBOSE_LRU */
+/*  		  | DEBUG_VERBOSE_DRI */
+/*  		  | DEBUG_VERBOSE_IOCTL */
+/*   		  | DEBUG_VERBOSE_2D */
 		  );
 #endif
 
@@ -71,18 +78,7 @@ static mgaContextPtr      mgaCtx = 0;
 
 mgaGlx_t	mgaglx;
 
-static int count_bits(unsigned int n)
-{
-   int bits = 0;
-
-   while (n > 0) {
-      if (n & 1) bits++;
-      n >>= 1;
-   }
-   return bits;
-}
-
-/* These functions are accessed by dlsym from dri_mesa_init.c:
+/* These functions are accessed externally to the driver:
  *
  * XMesaInitDriver
  * XMesaResetDriver
@@ -96,18 +92,12 @@ static int count_bits(unsigned int n)
  * XMesaSwapBuffers
  * XMesaMakeCurrent
  *
- * So this is kind of the public interface to the driver.  The driver
- * uses the X11 mesa driver context as a kind of wrapper around its
- * own driver context - but there isn't much justificiation for doing
- * it that way - the DRI might as well use a (void *) to refer to the
- * driver contexts.  Nothing in the X context really gets used.
  */
 
-
 GLboolean XMesaInitDriver(__DRIscreenPrivate *sPriv)
 {
    mgaScreenPrivate *mgaScreen;
-   MGADRIPtr         gDRIPriv = (MGADRIPtr)sPriv->pDevPriv;
+   MGADRIPtr         serverInfo = (MGADRIPtr)sPriv->pDevPriv;
 
    /* Allocate the private area */
    mgaScreen = (mgaScreenPrivate *)Xmalloc(sizeof(mgaScreenPrivate));
@@ -116,30 +106,84 @@ GLboolean XMesaInitDriver(__DRIscreenPrivate *sPriv)
    mgaScreen->sPriv = sPriv;
    sPriv->private = (void *)mgaScreen;
 
-   mgaScreen->chipset=gDRIPriv->chipset;
-   mgaScreen->width=gDRIPriv->width;
-   mgaScreen->height=gDRIPriv->height;
-   mgaScreen->mem=gDRIPriv->mem;
-   mgaScreen->cpp=gDRIPriv->cpp;
-   mgaScreen->frontPitch=gDRIPriv->frontPitch;
-   mgaScreen->frontOffset=gDRIPriv->frontOffset;
-   mgaScreen->backOffset=gDRIPriv->backOffset; 
-   mgaScreen->backPitch = gDRIPriv->backPitch;
-   mgaScreen->depthOffset=gDRIPriv->depthOffset;
-   mgaScreen->depthPitch = gDRIPriv->depthPitch;
-   mgaScreen->textureOffset=gDRIPriv->textureOffset;
-   mgaScreen->textureSize=gDRIPriv->textureSize;
-   mgaScreen->logTextureGranularity = gDRIPriv->logTextureGranularity;
+   /*
+   fprintf(stderr, "serverInfo->chipset: %d\n", serverInfo->chipset);
+   */
 
+   if (serverInfo->chipset != MGA_CARD_TYPE_G200 &&
+       serverInfo->chipset != MGA_CARD_TYPE_G400)
+	   return GL_FALSE;
+   
+   mgaScreen->chipset = serverInfo->chipset;
+   mgaScreen->width = serverInfo->width;
+   mgaScreen->height = serverInfo->height;
+   mgaScreen->mem = serverInfo->mem;
+   mgaScreen->cpp = serverInfo->cpp;
+   mgaScreen->frontPitch = serverInfo->frontPitch;
+   mgaScreen->frontOffset = serverInfo->frontOffset;
+   mgaScreen->backOffset = serverInfo->backOffset; 
+   mgaScreen->backPitch  =  serverInfo->backPitch;
+   mgaScreen->depthOffset = serverInfo->depthOffset;
+   mgaScreen->depthPitch  =  serverInfo->depthPitch;
+
+
+   mgaScreen->agp.handle = serverInfo->agp;
+   mgaScreen->agp.size = serverInfo->agpSize;
+
+   if (drmMap(sPriv->fd, 
+	      mgaScreen->agp.handle, 
+	      mgaScreen->agp.size, 
+	      (drmAddress *)&mgaScreen->agp.map) != 0) 
+   {
+      Xfree(mgaScreen);
+      return GL_FALSE;
+   }
 
-   mgaScreen->bufs = drmMapBufs(sPriv->fd);
+   mgaScreen->textureOffset[MGA_CARD_HEAP] = serverInfo->textureOffset;
+   mgaScreen->textureOffset[MGA_AGP_HEAP] = (serverInfo->agpTextureOffset |
+					     PDEA_pagpxfer_enable | 1);
 
+   /*
+   fprintf(stderr, "CARD texture size %x, granul %d --> %x\n", 
+	   serverInfo->textureSize, serverInfo->logTextureGranularity,
+	   1<<serverInfo->logTextureGranularity);
+   */
+
+   mgaScreen->textureSize[MGA_CARD_HEAP] = serverInfo->textureSize;
+   mgaScreen->textureSize[MGA_AGP_HEAP] = serverInfo->agpTextureSize;
+
+   mgaScreen->logTextureGranularity[MGA_CARD_HEAP] = 
+      serverInfo->logTextureGranularity;
+   mgaScreen->logTextureGranularity[MGA_AGP_HEAP] = 
+      serverInfo->logAgpTextureGranularity;
+
+   mgaScreen->texVirtual[MGA_CARD_HEAP] = (mgaScreen->sPriv->pFB + 
+					   serverInfo->textureOffset);
+   mgaScreen->texVirtual[MGA_AGP_HEAP] = (mgaScreen->agp.map + 
+					  serverInfo->agpTextureOffset);
+
+   mgaScreen->mAccess = serverInfo->mAccess;
+
+
+   /* For calculating setupdma addresses.
+    */
+   mgaScreen->dmaOffset = serverInfo->agpBufferOffset;   
+   
+   /*
+   fprintf(stderr, "\n\n\nbackOffset: %x pitch %x\n", 
+	   mgaScreen->backOffset,
+	   mgaScreen->backPitch);
+   */
+
+   mgaScreen->Attrib = MGA_PF_565;
+   mgaScreen->bufs = drmMapBufs(sPriv->fd);
 
    /* Other mgaglx stuff, too??
     */
    memset(&mgaglx, 0, sizeof(mgaglx));
 
    mgaDDFastPathInit();
+   mgaDDEltPathInit();
    mgaDDTrifuncInit();
    mgaDDSetupInit();
 
@@ -153,106 +197,53 @@ void XMesaResetDriver(__DRIscreenPrivate *sPriv)
    Xfree(sPriv->private);
 }
 
-/* Accessed by dlsym from dri_mesa_init.c
- */
-XMesaVisual XMesaCreateVisual(XMesaDisplay *display,
-			      XMesaVisualInfo visinfo,
-			      GLboolean rgb_flag,
-			      GLboolean alpha_flag,
-			      GLboolean db_flag,
-			      GLboolean stereo_flag,
-			      GLboolean ximage_flag,
-			      GLint depth_size,
-			      GLint stencil_size,
-			      GLint accum_size,
-			      GLint level)
-{
-   XMesaVisual v;
 
-   /* Only RGB visuals are supported on the MGA boards */
-   if (!rgb_flag) return 0;
-
-   v = (XMesaVisual)Xmalloc(sizeof(struct xmesa_visual));
-   if (!v) return 0;
-
-   v->visinfo = (XVisualInfo *)Xmalloc(sizeof(*visinfo));
-   if(!v->visinfo) {
-      Xfree(v);
-      return 0;
-   }
-   memcpy(v->visinfo, visinfo, sizeof(*visinfo));
-
-   v->display = display;
-   v->level = level;
-  
-   v->gl_visual = (GLvisual *)Xmalloc(sizeof(GLvisual));
-   if (!v->gl_visual) {
-      Xfree(v->visinfo);
-      XFree(v);
-      return 0;
-   }
-
-   v->gl_visual->RGBAflag   = rgb_flag;
-   v->gl_visual->DBflag     = db_flag;
-   v->gl_visual->StereoFlag = stereo_flag;
-
-   v->gl_visual->RedBits   = count_bits(visinfo->red_mask);
-   v->gl_visual->GreenBits = count_bits(visinfo->green_mask);
-   v->gl_visual->BlueBits  = count_bits(visinfo->blue_mask);
-   v->gl_visual->AlphaBits = 0; /* Not currently supported */
-
-   v->gl_visual->AccumBits   = accum_size;
-   v->gl_visual->DepthBits   = depth_size;
-   v->gl_visual->StencilBits = stencil_size;
-
-   return v;
-}
-
-void XMesaDestroyVisual(XMesaVisual v)
+GLvisual *XMesaCreateVisual(Display *dpy,
+                            __DRIscreenPrivate *driScrnPriv,
+                            const XVisualInfo *visinfo,
+                            const __GLXvisualConfig *config)
 {
-   Xfree(v->gl_visual);
-   Xfree(v->visinfo);
-   Xfree(v);
+   /* Drivers may change the args to _mesa_create_visual() in order to
+    * setup special visuals.
+    */
+   return _mesa_create_visual( config->rgba,
+                               config->doubleBuffer,
+                               config->stereo,
+                               _mesa_bitcount(visinfo->red_mask),
+                               _mesa_bitcount(visinfo->green_mask),
+                               _mesa_bitcount(visinfo->blue_mask),
+                               config->alphaSize,
+                               0, /* index bits */
+                               config->depthSize,
+                               config->stencilSize,
+                               config->accumRedSize,
+                               config->accumGreenSize,
+                               config->accumBlueSize,
+                               config->accumAlphaSize,
+                               0 /* num samples */ );
 }
 
-XMesaContext XMesaCreateContext(XMesaVisual v, XMesaContext share_list,
-				__DRIcontextPrivate *driContextPriv)
+
+GLboolean XMesaCreateContext( Display *dpy, GLvisual *mesaVis,
+                              __DRIcontextPrivate *driContextPriv )
 {
+   int i;
    GLcontext *ctx;
-   XMesaContext c;
    mgaContextPtr mmesa;
    __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
    mgaScreenPrivate *mgaScreen = (mgaScreenPrivate *)sPriv->private;
    drm_mga_sarea_t *saPriv=(drm_mga_sarea_t*)(((char*)sPriv->pSAREA)+
 					      sizeof(XF86DRISAREARec));
-   GLcontext *shareCtx = 0;
+   /*fprintf(stderr, "XMesaCreateContext\n");*/
 
-   c = (XMesaContext)Xmalloc(sizeof(struct xmesa_context));
-   if (!c) {
-      return 0;
-   }
-
-   mmesa = (mgaContextPtr)Xmalloc(sizeof(mgaContext));
+   mmesa = (mgaContextPtr)Xcalloc(sizeof(mgaContext), 1);
    if (!mmesa) {
-      Xfree(c);
-      return 0;
+      return GL_FALSE;
    }
 
-   c->driContextPriv = driContextPriv;
-   c->xm_visual = v;
-   c->xm_buffer = 0; /* Set by MakeCurrent */
-   c->display = v->display;
-   c->private = (void *)mmesa;
-
-   if (share_list)
-      shareCtx=((mgaContextPtr)(share_list->private))->glCtx;
-
-   ctx = mmesa->glCtx = gl_create_context(v->gl_visual, shareCtx, 
-					  (void*)mmesa, GL_TRUE);
+   ctx = driContextPriv->mesaContext;
 
-   /* Dri stuff
-    */
-   mmesa->display = v->display;
+   mmesa->display = dpy;
    mmesa->hHWContext = driContextPriv->hHWContext;
    mmesa->driFd = sPriv->fd;
    mmesa->driHwLock = &sPriv->pSAREA->lock;
@@ -260,41 +251,42 @@ XMesaContext XMesaCreateContext(XMesaVisual v, XMesaContext share_list,
    mmesa->mgaScreen = mgaScreen;
    mmesa->driScreen = sPriv;
    mmesa->sarea = saPriv;
+   mmesa->glBuffer = NULL;
 
-   mmesa->glBuffer=gl_create_framebuffer(v->gl_visual);
-
+   make_empty_list(&mmesa->SwappedOut);
 
-   mmesa->needClip=1;
+   mmesa->lastTexHeap = mgaScreen->texVirtual[MGA_AGP_HEAP] ? 2 : 1;
 
-   mmesa->texHeap = mmInit( 0, mgaScreen->textureSize );
+   for (i = 0 ; i < mmesa->lastTexHeap ; i++) {
+      mmesa->texHeap[i] = mmInit( 0, mgaScreen->textureSize[i]);
+      make_empty_list(&mmesa->TexObjList[i]);
+   }
 
-   /* Utah stuff
-    */
    mmesa->renderindex = -1;		/* impossible value */
    mmesa->new_state = ~0;
    mmesa->dirty = ~0;
-
-   mmesa->warp_pipe = 0;
-
-
-   make_empty_list(&mmesa->SwappedOut);
-   make_empty_list(&mmesa->TexObjList);
-   
+   mmesa->warp_pipe = 0;   
    mmesa->CurrentTexObj[0] = 0;
    mmesa->CurrentTexObj[1] = 0;
    
+   mmesa->texAge[0] = 0;
+   mmesa->texAge[1] = 0;
+   
+   ctx->DriverCtx = (void *) mmesa;
+   mmesa->glCtx = ctx;
+
    mgaDDExtensionsInit( ctx );
 
    mgaDDInitStateFuncs( ctx );
    mgaDDInitTextureFuncs( ctx );
    mgaDDInitSpanFuncs( ctx );
-   mgaDDInitDepthFuncs( ctx );
    mgaDDInitDriverFuncs( ctx );
    mgaDDInitIoctlFuncs( ctx );
 
 
    ctx->Driver.TriangleCaps = (DD_TRI_CULL|
 			       DD_TRI_LIGHT_TWOSIDE|
+			       DD_TRI_STIPPLE|
 			       DD_TRI_OFFSET);
 
    /* Ask mesa to clip fog coordinates for us.
@@ -312,19 +304,21 @@ XMesaContext XMesaCreateContext(XMesaVisual v, XMesaContext share_list,
 	 mgaDDRegisterPipelineStages(ctx->PipelineStage,
 				      ctx->PipelineStage,
 				      ctx->NrPipelineStages);
-   return c;
+
+   mgaInitState( mmesa );
+
+   driContextPriv->driverPrivate = (void *) mmesa;
+
+   return GL_TRUE;
 }
 
-void XMesaDestroyContext(XMesaContext c)
+void XMesaDestroyContext(__DRIcontextPrivate *driContextPriv)
 {
-   mgaContextPtr mmesa = (mgaContextPtr) c->private;
+   mgaContextPtr mmesa = (mgaContextPtr) driContextPriv->driverPrivate;
 
    if (mmesa) {
 /*        mgaTextureObjectPtr next_t, t; */
 
-      gl_destroy_context(mmesa->glCtx);
-      gl_destroy_framebuffer(mmesa->glBuffer);
-
 /*        foreach_s (t, next_t, &(mmesa->TexObjList)) */
 /*  	 mgaDestroyTexObj(mmesa, t); */
 
@@ -333,90 +327,62 @@ void XMesaDestroyContext(XMesaContext c)
 
       Xfree(mmesa);
 
-      c->private = 0;
+      driContextPriv->driverPrivate = NULL;
    }
 }
 
-XMesaBuffer XMesaCreateWindowBuffer(XMesaVisual v, XMesaWindow w,
-				    __DRIdrawablePrivate *driDrawPriv)
-{
-  return (XMesaBuffer)1;
-}
 
-XMesaBuffer XMesaCreatePixmapBuffer(XMesaVisual v, XMesaPixmap p,
-				    XMesaColormap c, 
-				    __DRIdrawablePrivate *driDrawPriv)
+GLframebuffer *XMesaCreateWindowBuffer( Display *dpy,
+                                        __DRIscreenPrivate *driScrnPriv,
+                                        __DRIdrawablePrivate *driDrawPriv,
+                                        GLvisual *mesaVis)
 {
-  return (XMesaBuffer)1;
+   return gl_create_framebuffer(mesaVis,
+                                GL_FALSE,  /* software depth buffer? */
+                                mesaVis->StencilBits > 0,
+                                mesaVis->AccumRedBits > 0,
+                                mesaVis->AlphaBits > 0
+                                );
 }
 
-void XMesaDestroyBuffer(XMesaBuffer b)
+
+GLframebuffer *XMesaCreatePixmapBuffer( Display *dpy,
+                                        __DRIscreenPrivate *driScrnPriv,
+                                        __DRIdrawablePrivate *driDrawPriv,
+                                        GLvisual *mesaVis)
 {
+#if 0
+   /* Different drivers may have different combinations of hardware and
+    * software ancillary buffers.
+    */
+   return gl_create_framebuffer(mesaVis,
+                                GL_FALSE,  /* software depth buffer? */
+                                mesaVis->StencilBits > 0,
+                                mesaVis->AccumRedBits > 0,
+                                mesaVis->AlphaBits > 0
+                                );
+#else
+   return NULL;  /* not implemented yet */
+#endif
 }
 
-void XMesaSwapBuffers(XMesaBuffer bogus_value_do_not_use)
+
+void XMesaSwapBuffers(__DRIdrawablePrivate *driDrawPriv)
 {
+   /* XXX should do swap according to the buffer, not the context! */
    mgaContextPtr mmesa = mgaCtx;    
    FLUSH_VB( mmesa->glCtx, "swap buffers" );
    mgaSwapBuffers(mmesa);
 }
 
 
-
-
-
-void mgaXMesaSetFrontClipRects( mgaContextPtr mmesa )
-{
-   __DRIdrawablePrivate *dPriv = mmesa->driDrawable;
-
-   mmesa->numClipRects = dPriv->numClipRects;
-   mmesa->pClipRects = dPriv->pClipRects;
-   mmesa->drawX = dPriv->x;
-   mmesa->drawY = dPriv->y;
-
-   mmesa->drawOffset = mmesa->mgaScreen->frontOffset;
-}
-
-
-void mgaXMesaSetBackClipRects( mgaContextPtr mmesa )
-{
-   __DRIdrawablePrivate *dPriv = mmesa->driDrawable;
-
-   if (dPriv->numAuxClipRects == 0) 
-   {
-      mmesa->numClipRects = dPriv->numClipRects;
-      mmesa->pClipRects = dPriv->pClipRects;
-      mmesa->drawX = dPriv->x;
-      mmesa->drawY = dPriv->y;
-   } else {
-      mmesa->numClipRects = dPriv->numAuxClipRects;
-      mmesa->pClipRects = dPriv->pAuxClipRects;
-      mmesa->drawX = dPriv->auxX;
-      mmesa->drawY = dPriv->auxY;
-   }
-
-   mmesa->drawOffset = mmesa->mgaScreen->backOffset;
-}
-
-
-static void mgaXMesaWindowMoved( mgaContextPtr mmesa ) 
+GLboolean XMesaUnbindContext(__DRIcontextPrivate *driContextPriv)
 {
-   /* Clear any contaminated CVA data.
-    */
-   mmesa->setupdone = 0;
-
-   switch (mmesa->glCtx->Color.DriverDrawBuffer) {
-   case GL_FRONT_LEFT:
-      mgaXMesaSetFrontClipRects( mmesa );
-      break;
-   case GL_BACK_LEFT:
-      mgaXMesaSetBackClipRects( mmesa );
-      break;
-   default:
-      fprintf(stderr, "fallback buffer\n");
-      break;
-   }
+   mgaContextPtr mmesa = (mgaContextPtr) driContextPriv->driverPrivate;
+   if (mmesa)
+      mmesa->dirty = ~0;
 
+   return GL_TRUE;
 }
 
 
@@ -426,27 +392,24 @@ static void mgaXMesaWindowMoved( mgaContextPtr mmesa )
  *
  * But why are we doing context initialization here???
  */
-GLboolean XMesaMakeCurrent(XMesaContext c, XMesaBuffer b)
+GLboolean XMesaMakeCurrent(__DRIcontextPrivate *driContextPriv,
+                           __DRIdrawablePrivate *driDrawPriv,
+                           __DRIdrawablePrivate *driReadPriv)
 {
+   if (driContextPriv) {
+      mgaCtx = (mgaContextPtr) driContextPriv->driverPrivate;
 
-   if (c->private==(void *)mgaCtx) return GL_TRUE;
-
-   if (c) {
-      __DRIdrawablePrivate *dPriv = c->driContextPriv->driDrawablePriv;
-
-      mgaCtx = (mgaContextPtr)c->private;
+      gl_make_current2(mgaCtx->glCtx, driDrawPriv->mesaBuffer, driReadPriv->mesaBuffer);
 
-
-      gl_make_current(mgaCtx->glCtx, mgaCtx->glBuffer);
-
-      mgaXMesaWindowMoved( mgaCtx );
-      mgaCtx->driDrawable = dPriv;
+      mgaCtx->driDrawable = driDrawPriv;
       mgaCtx->dirty = ~0;
+      mgaCtx->dirty_cliprects = (MGA_FRONT|MGA_BACK);
 
       if (!mgaCtx->glCtx->Viewport.Width)
-	 gl_Viewport(mgaCtx->glCtx, 0, 0, dPriv->w, dPriv->h);
+	 gl_Viewport(mgaCtx->glCtx, 0, 0, driDrawPriv->w, driDrawPriv->h);
 
-   } else {
+   }
+   else {
       gl_make_current(0,0);
       mgaCtx = NULL;
    }
@@ -455,58 +418,37 @@ GLboolean XMesaMakeCurrent(XMesaContext c, XMesaBuffer b)
 }
 
 
-void mgaXMesaUpdateState( mgaContextPtr mmesa ) 
+void mgaGetLock( mgaContextPtr mmesa, GLuint flags ) 
 {
    __DRIdrawablePrivate *dPriv = mmesa->driDrawable;
-   __DRIscreenPrivate *sPriv = mmesa->driScreen;
    drm_mga_sarea_t *sarea = mmesa->sarea;
-
    int me = mmesa->hHWContext;
-   int stamp = dPriv->lastStamp; 
-
-   /* If the window moved, may need to set a new cliprect now.
-    *
-    * NOTE: This releases and regains the hw lock, so all state
-    * checking must be done *after* this call:
-    */
-   XMESA_VALIDATE_DRAWABLE_INFO(mmesa->display, sPriv, dPriv);		
+   int i;
 
-   if (sarea->ctxOwner != me) {
-      mmesa->dirty |= MGA_UPLOAD_CTX;
-   }
+   drmGetLock(mmesa->driFd, mmesa->hHWContext, flags);	
 
-   if (sarea->texAge != mmesa->texAge) {
-      int sz = 1 << (mmesa->mgaScreen->logTextureGranularity);
-      int idx, nr = 0;
-
-      /* Have to go right round from the back to ensure stuff ends up
-       * LRU in our local list...
-       */
-      for (idx = sarea->texList[MGA_NR_TEX_REGIONS].prev ; 
-	   idx != MGA_NR_TEX_REGIONS && nr < MGA_NR_TEX_REGIONS ; 
-	   idx = sarea->texList[idx].prev, nr++)
-      {
-	 if (sarea->texList[idx].age > mmesa->texAge)
-	    mgaTexturesGone(mmesa, idx * sz, sz, 1);
-      }
-
-      if (nr == MGA_NR_TEX_REGIONS) {
-	 mgaTexturesGone(mmesa, 0, mmesa->mgaScreen->textureSize, 0);
-	 mgaResetGlobalLRU( mmesa );
-      }
-
-      mmesa->texAge = sarea->texAge;
-      mmesa->dirty |= MGA_UPLOAD_TEX0IMAGE | MGA_UPLOAD_TEX1IMAGE;
+   if (*(dPriv->pStamp) != dPriv->lastStamp) { 
+      mmesa->setupdone = 0;
+      mmesa->dirty_cliprects = (MGA_FRONT|MGA_BACK);
+      mgaUpdateRects( mmesa, (MGA_FRONT|MGA_BACK) );
    }
 
-   if (dPriv->lastStamp != stamp) 
-      mgaXMesaWindowMoved( mmesa );
+   mmesa->dirty |= MGA_UPLOAD_CTX | MGA_UPLOAD_CLIPRECTS;
 
-   sarea->ctxOwner=me;
-}
+    mmesa->sarea->dirty |= MGA_UPLOAD_CTX; 
 
+   if (sarea->ctxOwner != me) {
+      mmesa->dirty |= (MGA_UPLOAD_CTX | MGA_UPLOAD_TEX0 | 
+		       MGA_UPLOAD_TEX1 | MGA_UPLOAD_PIPE);
+      sarea->ctxOwner=me;
+   }
 
+   for (i = 0 ; i < mmesa->lastTexHeap ; i++)
+      if (sarea->texAge[i] != mmesa->texAge[i])
+	 mgaAgeTextures( mmesa, i );
 
+   sarea->last_quiescent = -1;	/* just kill it for now */
+}
 
 
 
diff --git a/xc/lib/GL/mesa/src/drv/mga/mga_xmesa.h b/xc/lib/GL/mesa/src/drv/mga/mga_xmesa.h
index d563bd429..489b71836 100644
--- a/xc/lib/GL/mesa/src/drv/mga/mga_xmesa.h
+++ b/xc/lib/GL/mesa/src/drv/mga/mga_xmesa.h
@@ -28,7 +28,6 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 /*
  * Authors:
  *   Keith Whitwell <keithw@precisioninsight.com>
- *   Daryll Strauss <daryll@precisioninsight.com> (Origninal tdfx driver).
  *
  */
 
@@ -42,7 +41,13 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "dri_mesaint.h"
 #include "dri_mesa.h"
 #include "types.h"
-#include "xmesaP.h"
+#include "mgaregs.h"
+
+typedef struct {
+   drmHandle handle;
+   drmSize size;
+   char *map;
+} mgaRegion, *mgaRegionPtr;
 
 typedef struct {
 
@@ -54,53 +59,69 @@ typedef struct {
    int cpp;			/* for front and back buffers */
 
    int Attrib;
+   unsigned int mAccess;
 
-   int frontOffset;
-   int frontPitch;
-   int backOffset;
-   int backPitch;
+   unsigned int frontOffset;
+   unsigned int frontPitch;
+   unsigned int backOffset;
+   unsigned int backPitch;
 
-   int depthOffset;
-   int depthPitch;
+   unsigned int depthOffset;
+   unsigned int depthPitch;
    int depthCpp;
 
-   int textureOffset;
-   int textureSize;
-   int logTextureGranularity;
+   unsigned int dmaOffset;		
+
+   unsigned int textureOffset[MGA_NR_TEX_HEAPS];
+   unsigned int textureSize[MGA_NR_TEX_HEAPS];
+   int logTextureGranularity[MGA_NR_TEX_HEAPS];
+   char *texVirtual[MGA_NR_TEX_HEAPS];
+
 
    __DRIscreenPrivate *sPriv;
    drmBufMapPtr  bufs;
 
+   /* Maps the dma buffers as well as textures ? 
+    */
+   mgaRegion agp;
+
 } mgaScreenPrivate;
 
 
 #include "mgalib.h"
 
-extern void mgaXMesaUpdateState( mgaContextPtr mmesa );
+extern void mgaGetLock( mgaContextPtr mmesa, GLuint flags );
 extern void mgaEmitHwStateLocked( mgaContextPtr mmesa );
 extern void mgaEmitScissorValues( mgaContextPtr mmesa, int box_nr, int emit );
-extern void mgaXMesaSetBackClipRects( mgaContextPtr mmesa );
-extern void mgaXMesaSetFrontClipRects( mgaContextPtr mmesa );
+
+#define GET_DISPATCH_AGE( mmesa ) mmesa->sarea->last_dispatch
+#define GET_ENQUEUE_AGE( mmesa ) mmesa->sarea->last_enqueue
 
 
 
 /* Lock the hardware and validate our state.  
  */
-#define LOCK_HARDWARE( mmesa )				\
-  do {							\
-    char __ret=0;					\
-    DRM_CAS(mmesa->driHwLock, mmesa->hHWContext,	\
-	    (DRM_LOCK_HELD|mmesa->hHWContext), __ret);	\
-    if (__ret) {					\
-        drmGetLock(mmesa->driFd, mmesa->hHWContext, 0);	\
-        mgaXMesaUpdateState( mmesa );			\
-    }							\
+#define LOCK_HARDWARE( mmesa )					\
+  do {								\
+    char __ret=0;						\
+    DRM_CAS(mmesa->driHwLock, mmesa->hHWContext,		\
+	    (DRM_LOCK_HELD|mmesa->hHWContext), __ret);		\
+    if (__ret)							\
+        mgaGetLock( mmesa, 0 );					\
   } while (0)
 
 
+/* 
+ */
+#define LOCK_HARDWARE_QUIESCENT( mmesa ) do {	                        \
+	LOCK_HARDWARE( mmesa );			                        \
+	mgaUpdateLock( mmesa, DRM_LOCK_QUIESCENT | DRM_LOCK_FLUSH );	\
+} while (0)
+
+
 /* Unlock the hardware using the global current context 
  */
-#define UNLOCK_HARDWARE(mmesa)					\
+#define UNLOCK_HARDWARE(mmesa) 				\
     DRM_UNLOCK(mmesa->driFd, mmesa->driHwLock, mmesa->hHWContext);	
 
 
@@ -109,6 +130,8 @@ extern void mgaXMesaSetFrontClipRects( mgaContextPtr mmesa );
 #define REFRESH_DRAWABLE_INFO( mmesa )		\
 do {						\
    LOCK_HARDWARE( mmesa );			\
+   mmesa->lastX = mmesa->drawX; 		\
+   mmesa->lastY = mmesa->drawY; 		\
    UNLOCK_HARDWARE( mmesa );			\
 } while (0)
 
diff --git a/xc/lib/GL/mesa/src/drv/mga/mgaeltpath.c b/xc/lib/GL/mesa/src/drv/mga/mgaeltpath.c
new file mode 100644
index 000000000..4751eb091
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/mga/mgaeltpath.c
@@ -0,0 +1,457 @@
+/*
+ * GLX Hardware Device Driver for Matrox G400
+ * Copyright (C) 1999 Keith Whitwell
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * KEITH WHITWELL, OR ANY OTHER CONTRIBUTORS 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 "types.h"
+#include "enums.h"
+#include "cva.h"
+#include "vertices.h"
+#include "mmath.h"
+#include "xform.h"
+
+#include "mgalib.h"
+#include "mgapipeline.h"
+#include "mgatris.h"
+#include "mgastate.h"
+#include "mgavb.h"
+
+
+/* Always use a full-sized stride for vertices. [FIXME]
+ * Stride in the buffers must be a quadword multiple.
+ */
+#define BUFFER_STRIDE 12 
+#define CLIP_STRIDE 10 
+
+
+static void fire_elts( mgaContextPtr mmesa )
+{
+   LOCK_HARDWARE( mmesa );
+
+   /* Fire queued elements and discard that buffer if its contents
+    * won't be referenced by future elements.
+    */
+   if (mmesa->elt_buf) {
+      GLuint retain = (mmesa->elt_buf == mmesa->retained_buf);
+
+      if (mmesa->first_elt != mmesa->next_elt) {
+	 mgaFireEltsLocked( mmesa, 
+			    ((GLuint)mmesa->first_elt - 
+			     (GLuint)mmesa->elt_buf->address),
+			    ((GLuint)mmesa->next_elt - 
+			     (GLuint)mmesa->elt_buf->address),
+			    !retain );
+      } else if (!retain) 
+	 mgaReleaseBufLocked( mmesa, mmesa->elt_buf );
+      
+      mmesa->elt_buf = 0;
+   } 
+   else if (mmesa->vertex_dma_buffer) 
+   {
+      mgaFlushVerticesLocked( mmesa );
+   }
+
+   mgaGetEltBufLocked( mmesa );
+      
+   UNLOCK_HARDWARE( mmesa );
+
+   mmesa->next_vert = (GLfloat *)((GLuint)mmesa->elt_buf->address + 
+				  mmesa->elt_buf->total - 
+				  BUFFER_STRIDE * sizeof(GLfloat));
+
+   mmesa->next_vert_phys = (mmesa->mgaScreen->dmaOffset + 
+			    mmesa->elt_buf->idx * MGA_DMA_BUF_SZ +
+			    mmesa->elt_buf->total -
+			    BUFFER_STRIDE * sizeof(GLfloat));
+
+   mmesa->first_elt = (GLuint *)mmesa->elt_buf->address;
+   mmesa->next_elt = (GLuint *)mmesa->elt_buf->address;
+
+}
+
+
+static void release_bufs( mgaContextPtr mmesa )
+{
+   if (mmesa->retained_buf && mmesa->retained_buf != mmesa->elt_buf) 
+   {
+      LOCK_HARDWARE( mmesa );
+      if (mmesa->first_elt != mmesa->next_elt) {
+	 mgaFireEltsLocked( mmesa, 
+			    ((GLuint)mmesa->first_elt - 
+			     (GLuint)mmesa->elt_buf->address),
+			    ((GLuint)mmesa->next_elt - 
+			     (GLuint)mmesa->elt_buf->address),
+			    0 );
+
+	 mmesa->first_elt = mmesa->next_elt;
+      }
+
+      mgaReleaseBufLocked( mmesa, mmesa->retained_buf );
+      UNLOCK_HARDWARE( mmesa );
+   }
+
+   mmesa->retained_buf = 0;
+}
+
+
+
+
+#define NEGATIVE(f)          (f < 0)
+#define DIFFERENT_SIGNS(a,b) ((a*b) < 0)
+#define LINTERP( T, A, B )   ( (A) + (T) * ( (B) - (A) ) )
+
+
+#define INTERP_RGBA(t, out, a, b) {			\
+   int i;						\
+   for (i = 0; i < 4; i++) {				\
+      GLfloat fa = UBYTE_COLOR_TO_FLOAT_COLOR(a[i]);	\
+      GLfloat fb = UBYTE_COLOR_TO_FLOAT_COLOR(b[i]);	\
+      GLfloat fo = LINTERP(t, fa, fb);			\
+      FLOAT_COLOR_TO_UBYTE_COLOR(out[i], fo);		\
+   }							\
+}
+
+
+#define CLIP(SGN,V,PLANE)			\
+if (mask & PLANE) {				\
+   GLuint *indata = inlist[in];			\
+   GLuint *outdata = inlist[in ^= 1];		\
+   GLuint nr = n;				\
+   GLfloat *J = verts[indata[nr-1]];		\
+   GLfloat dpJ = (SGN J[V]) + J[3];		\
+						\
+   for (i = n = 0 ; i < nr ; i++) {		\
+      GLuint elt_i = indata[i];			\
+      GLfloat *I = verts[elt_i];		\
+      GLfloat dpI = (SGN I[V]) + I[3];		\
+						\
+      if (DIFFERENT_SIGNS(dpI, dpJ)) {		\
+	 GLfloat *O = verts[next_vert];	\
+	 outdata[n++] = next_vert++;		\
+						\
+	 if (NEGATIVE(dpI)) {			\
+	    GLfloat t = dpI / (dpI - dpJ);	\
+	    interp(t, O, I, J);			\
+	 }					\
+         else					\
+	 {					\
+	    GLfloat t = dpJ / (dpJ - dpI);	\
+	    interp(t, O, J, I);			\
+	 }					\
+      }						\
+						\
+      if (!NEGATIVE(dpI))			\
+	 outdata[n++] = elt_i;			\
+						\
+      J = I;					\
+      dpJ = dpI;				\
+   }						\
+						\
+   if (n < 3) return;				\
+}
+
+
+static void mga_tri_clip( mgaContextPtr mmesa,
+			  struct vertex_buffer *VB,
+			  GLuint *elt,
+			  GLubyte mask )
+{
+   struct mga_elt_tab *tab = mmesa->elt_tab;
+   mga_interp_func interp = tab->interp;
+   GLuint inlist[2][VB_MAX_CLIPPED_VERTS];
+   GLuint in = 0;
+   GLuint n = 3, next_vert = 3;
+   GLuint i;
+   GLfloat verts[VB_MAX_CLIPPED_VERTS][CLIP_STRIDE];
+
+   /* Build temporary vertices in clipspace.  This is the potential
+    * downside to this path.
+    */
+   tab->build_tri_verts( mmesa, VB, (GLfloat *)verts, elt );
+
+   inlist[0][0] = 0;
+   inlist[0][1] = 1;
+   inlist[0][2] = 2;
+
+   CLIP(-,0,CLIP_RIGHT_BIT);
+   CLIP(+,0,CLIP_LEFT_BIT);
+   CLIP(-,1,CLIP_TOP_BIT);
+   CLIP(+,1,CLIP_BOTTOM_BIT);
+   CLIP(-,2,CLIP_FAR_BIT);
+   CLIP(+,2,CLIP_NEAR_BIT); 
+
+
+   {
+      GLuint *out = inlist[in];
+      GLuint space = (GLuint)mmesa->next_vert -	(GLuint)mmesa->next_elt;
+
+      if (space < n * (BUFFER_STRIDE + 3) * sizeof(GLuint))
+         fire_elts(mmesa); 
+
+      /* Project the new vertices and emit to dma buffers.  Translate
+       * out values to physical addresses for setup dma.  
+       */
+      tab->project_and_emit_verts( mmesa, (GLfloat *)verts, out, n );
+      
+      /* Convert the planar polygon to a list of triangles and emit to
+       * elt buffers. 
+       */
+      for (i = 2 ; i < n ; i++) {
+	 mmesa->next_elt[0] = out[0];
+	 mmesa->next_elt[1] = out[i-1];
+	 mmesa->next_elt[2] = out[i];
+	 mmesa->next_elt += 3;
+      }
+   }
+}
+
+
+
+
+/* Build a table of functions to clip each primitive type.  These
+ * produce a list of elements in the appropriate 'reduced' primitive,
+ * ie (points, lines, triangles) containing all the clipped and
+ * unclipped primitives from the original list.
+ */
+
+#define INIT(x)
+
+#define TRI_THRESHOLD         (3 * sizeof(GLuint))
+
+#define UNCLIPPED_VERT(x) (mmesa->first_vert_phys - x * BUFFER_STRIDE * 4)
+
+#define TRIANGLE( e2, e1, e0 )			\
+do {						\
+   if ((GLuint)mmesa->next_vert - 		\
+       (GLuint)mmesa->next_elt < TRI_THRESHOLD)	\
+      fire_elts(mmesa);			\
+   mmesa->next_elt[0] = UNCLIPPED_VERT(e2);	\
+   mmesa->next_elt[1] = UNCLIPPED_VERT(e1);	\
+   mmesa->next_elt[2] = UNCLIPPED_VERT(e0);	\
+   mmesa->next_elt+=3;				\
+} while (0)
+
+#define CLIP_TRIANGLE( e2, e1, e0 )				\
+do {								\
+   GLubyte ormask = mask[e2] | mask[e1] | mask[e0];		\
+   if (ormask == 0) {						\
+      TRIANGLE( e2, e1, e0 );					\
+   } else if ((mask[e2] & mask[e1] & mask[e0]) == 0) {		\
+      out[0] = e2;						\
+      out[1] = e1;						\
+      out[2] = e0;						\
+      mga_tri_clip( mmesa, VB, out, ormask );			\
+   }								\
+} while (0)
+
+#define LOCAL_VARS						\
+   mgaContextPtr mmesa = MGA_CONTEXT( VB->ctx );		\
+   GLuint *elt = VB->EltPtr->data;				\
+   GLuint out[VB_MAX_CLIPPED_VERTS];				\
+   GLubyte *mask = VB->ClipMask; \
+   (void) mask; (void) out; (void) elt; (void) mmesa;
+
+
+
+#define RENDER_POINTS(start, count) 
+#define RENDER_LINE(i1, i0) 
+#define RENDER_TRI(i2, i1, i0, pv, parity)		\
+do {							\
+   GLuint e2 = elt[i2], e1 = elt[i1], e0 = elt[i0];	\
+   if (parity) e2 = elt[i1], e1 = elt[i2];		\
+   CLIP_TRIANGLE( e2, e1, e0 );				\
+} while (0)
+
+#define RENDER_QUAD(i3, i2, i1, i0, pv )		\
+  CLIP_TRIANGLE(elt[i3], elt[i2], elt[i0]);     	\
+  CLIP_TRIANGLE(elt[i2], elt[i1], elt[i0])
+
+#define TAG(x) mga_##x##_elt
+#include "render_tmp.h"
+
+
+
+#define LOCAL_VARS						\
+   mgaContextPtr mmesa = MGA_CONTEXT( VB->ctx );		\
+   GLuint *elt = VB->EltPtr->data;				\
+   (void) elt; (void) mmesa;
+
+#define RENDER_POINTS(start, count) 
+#define RENDER_LINE(i1, i0) 
+#define RENDER_TRI(i2, i1, i0, pv, parity)		\
+do {							\
+   GLuint e2 = elt[i2], e1 = elt[i1], e0 = elt[i0];	\
+   if (parity) e2 = elt[i1], e1 = elt[i2];		\
+   TRIANGLE( e2, e1, e0 );				\
+} while (0)
+
+#define RENDER_QUAD(i3, i2, i1, i0, pv )	\
+  TRIANGLE(elt[i3], elt[i2], elt[i0]);		\
+  TRIANGLE(elt[i2], elt[i1], elt[i0])
+
+#define TAG(x) mga_##x##_elt_unclipped
+#include "render_tmp.h"
+
+
+
+
+static void refresh_projection_matrix( GLcontext *ctx )
+{
+   mgaContextPtr mmesa = MGA_CONTEXT(ctx);
+   GLfloat *m = mmesa->device_matrix;
+   GLmatrix *mat = &ctx->Viewport.WindowMap;
+
+   REFRESH_DRAWABLE_INFO(mmesa);
+
+   m[MAT_SX] =   mat->m[MAT_SX];
+   m[MAT_TX] =   mat->m[MAT_TX] + mmesa->drawX + .5;
+   m[MAT_SY] = (- mat->m[MAT_SY]);
+   m[MAT_TY] = (- mat->m[MAT_TY]) + mmesa->driDrawable->h + mmesa->drawY - .5;
+   m[MAT_SZ] =   mat->m[MAT_SZ] * (1.0 / 0x10000);
+   m[MAT_TZ] =   mat->m[MAT_TZ] * (1.0 / 0x10000);
+}
+
+#define CLIP_UBYTE_B 0   
+#define CLIP_UBYTE_G 1  
+#define CLIP_UBYTE_R 2
+#define CLIP_UBYTE_A 3   
+
+
+#define TYPE (0)
+#define TAG(x) x
+#include "mgaelttmp.h"
+
+#define TYPE (MGA_RGBA_BIT)
+#define TAG(x) x##_RGBA
+#include "mgaelttmp.h"
+
+#define TYPE (MGA_TEX0_BIT)
+#define TAG(x) x##_TEX0
+#include "mgaelttmp.h"
+
+#define TYPE (MGA_RGBA_BIT|MGA_TEX0_BIT)
+#define TAG(x) x##_RGBA_TEX0
+#include "mgaelttmp.h"
+
+#define TYPE (MGA_RGBA_BIT|MGA_TEX0_BIT|MGA_TEX1_BIT)
+#define TAG(x) x##_RGBA_TEX0_TEX1
+#include "mgaelttmp.h"
+
+#define TYPE (MGA_TEX0_BIT|MGA_TEX1_BIT)
+#define TAG(x) x##_TEX0_TEX1
+#include "mgaelttmp.h"
+
+
+/* Very sparsely popluated array - fix the indices.
+ */
+static struct mga_elt_tab mgaEltTab[0x80];
+
+void mgaDDEltPathInit( void )
+{
+   mga_render_init_elt();
+   mga_render_init_elt_unclipped();
+
+   mga_init_eltpath( &mgaEltTab[0] );
+   mga_init_eltpath_RGBA( &mgaEltTab[MGA_RGBA_BIT] );
+   mga_init_eltpath_TEX0( &mgaEltTab[MGA_TEX0_BIT] );
+   mga_init_eltpath_RGBA_TEX0( &mgaEltTab[MGA_RGBA_BIT|MGA_TEX0_BIT] );
+   mga_init_eltpath_TEX0_TEX1( &mgaEltTab[MGA_TEX0_BIT|MGA_TEX1_BIT] );
+   mga_init_eltpath_RGBA_TEX0_TEX1( &mgaEltTab[MGA_RGBA_BIT|MGA_TEX0_BIT|
+						  MGA_TEX1_BIT] );
+}
+
+#define VALID_SETUP (MGA_RGBA_BIT|MGA_TEX0_BIT|MGA_TEX1_BIT)
+
+
+
+/* Use a temporary array for device coordinates, so that we can easily
+ * tap into existing mesa assembly.  Otherwise consider emitting
+ * device coordinates to dma buffers directly from the project/cliptest
+ * routine.  (requires output stride, potential loss of writecombining 
+ * efficiency?)  
+ *
+ * This path is a lot closer to the standard vertex path in the
+ * initial stages than the original fastpath.  A slightly more optimal
+ * path could be constructed, but would require us to write new
+ * assembly.  
+ */
+void mgaDDEltPath( struct vertex_buffer *VB )
+{
+   GLcontext *ctx = VB->ctx;
+   GLenum prim = ctx->CVA.elt_mode;
+   mgaContextPtr mmesa = MGA_CONTEXT( ctx ); 
+   struct mga_elt_tab *tab = &mgaEltTab[mmesa->setupindex & VALID_SETUP];
+
+   VB->ClipPtr = TransformRaw(&VB->Clip, &ctx->ModelProjectMatrix, VB->ObjPtr );
+
+   refresh_projection_matrix( ctx );
+
+   VB->ClipAndMask = ~0; 
+   VB->ClipOrMask = 0;
+   VB->Projected = gl_clip_tab[VB->ClipPtr->size]( VB->ClipPtr,
+						   &VB->Win,
+						   VB->ClipMask,
+						   &VB->ClipOrMask,
+						   &VB->ClipAndMask );
+
+   if (VB->ClipAndMask) 
+      return;
+
+   if (mmesa->vertex_dma_buffer) 
+      mgaFlushVertices( mmesa );
+
+   if (mmesa->new_state)
+      mgaDDUpdateHwState( ctx );
+
+   /* Allocate a single buffer to hold unclipped vertices.  All
+    * unclipped vertices must be contiguous.  
+    */
+   if ((GLuint)mmesa->next_vert - (GLuint)mmesa->next_elt <  
+       VB->Count * BUFFER_STRIDE * sizeof(GLuint))	
+      fire_elts( mmesa );
+
+   mmesa->retained_buf = mmesa->elt_buf;
+   
+   /* Emit unclipped vertices to the buffer.
+    */
+   tab->emit_unclipped_verts( VB ); 
+	 
+   /* Emit indices and clipped vertices to one or more buffers.
+    */
+   if (VB->ClipOrMask) {
+      mmesa->elt_tab = tab;
+      mga_render_tab_elt[prim]( VB, 0, VB->EltPtr->count, 0 ); 
+   } else
+      mga_render_tab_elt_unclipped[prim]( VB, 0, VB->EltPtr->count, 0 );
+
+   /* Send to hardware and release any retained buffers.
+    */
+   release_bufs( mmesa );
+
+   /* This indicates that there is no cached data to reuse.  
+    */
+   VB->pipeline->data_valid = 0;
+   VB->pipeline->new_state = 0;
+}
+
diff --git a/xc/lib/GL/mesa/src/drv/mga/mgaelttmp.h b/xc/lib/GL/mesa/src/drv/mga/mgaelttmp.h
new file mode 100644
index 000000000..ce758a532
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/mga/mgaelttmp.h
@@ -0,0 +1,269 @@
+/*
+ * DRI Hardware Device Driver for G200/G400
+ * Copyright (C) 1999 Keith Whitwell
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * KEITH WHITWELL, OR ANY OTHER CONTRIBUTORS 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.
+ *
+ */
+
+/* Buffers fill from high addresses down with vertices and from low
+ * addresses up with elements.  
+ */
+
+
+/* Emit the bulk of the vertices to the first dma buffer.  Leave
+ * empty slots for clipped vertices so that we can still address
+ * vertices by index.
+ */
+static void TAG(emit_unclipped_verts)( struct vertex_buffer *VB )
+{
+   GLuint i;
+   mgaContextPtr mmesa = MGA_CONTEXT(VB->ctx);
+   GLfloat *dev = VB->Projected->start;
+   GLubyte *color = VB->ColorPtr->start;
+   GLfloat *tex0_data = VB->TexCoordPtr[0]->start;
+   GLfloat *tex1_data = VB->TexCoordPtr[1]->start;
+   GLuint color_stride = VB->ColorPtr->stride;
+   GLuint tex0_stride = VB->TexCoordPtr[0]->stride;
+   GLuint tex1_stride = VB->TexCoordPtr[1]->stride;
+
+   GLfloat *f = mmesa->next_vert;
+   GLuint count = VB->Count;
+   GLubyte *clipmask = VB->ClipMask;
+
+   const GLfloat *m = mmesa->device_matrix;
+   const GLfloat sx = m[0], sy = m[5], sz = m[10];
+   const GLfloat tx = m[12], ty = m[13], tz = m[14];
+
+   mmesa->retained_buf = mmesa->elt_buf;
+   mmesa->first_vert_phys = mmesa->next_vert_phys;
+
+   for (i = 0 ; i < count ; f -= BUFFER_STRIDE, i++) 
+   {
+      if (!clipmask[i]) 
+      {
+	 f[0] = sx * dev[0] + tx;
+	 f[1] = sy * dev[1] + ty;
+	 f[2] = sz * dev[2] + tz;
+	 f[3] = dev[3]; 
+
+	 if (TYPE & MGA_RGBA_BIT) {
+#if defined(USE_X86_ASM)
+	    __asm__ (
+	    "movl (%%edx),%%eax   \n"
+	    "bswap %%eax          \n"
+	    "rorl $8,%%eax        \n"
+	    "movl %%eax,16(%%edi) \n"
+	    :
+	    : "d" (color), "D" (f)
+	    : "%eax" );
+#else
+	    GLubyte *b = (GLubyte *)&f[4];
+	    b[CLIP_UBYTE_B] = color[2];
+	    b[CLIP_UBYTE_G] = color[1];
+	    b[CLIP_UBYTE_R] = color[0];
+	    b[CLIP_UBYTE_A] = color[3];
+#endif
+	 }
+
+	 if (TYPE & MGA_TEX0_BIT) {
+/*  	    fprintf(stderr, "i %d tex0 %f, %f\n", i,  */
+/*  		    tex0_data[0], tex0_data[1]); */
+	    *(int*)&f[6] = *(int*)&tex0_data[0];
+	    *(int*)&f[7] = *(int*)&tex0_data[1];
+	 }
+
+	 if (TYPE & MGA_TEX1_BIT) {
+	    *(int*)&f[8] = *(int*)&tex1_data[0];
+	    *(int*)&f[9] = *(int*)&tex1_data[1];
+	 }
+      }
+
+      STRIDE_F(dev, 16);
+      if (TYPE & MGA_RGBA_BIT) color += color_stride;
+      if (TYPE & MGA_TEX0_BIT) STRIDE_F(tex0_data, tex0_stride);
+      if (TYPE & MGA_TEX1_BIT) STRIDE_F(tex1_data, tex1_stride);
+   }
+
+   mmesa->next_vert = f;
+   mmesa->next_vert_phys -= count * BUFFER_STRIDE * sizeof(GLuint);
+}
+
+
+/* Build three temporary clipspace vertex for clipping a triangle.
+ * Recreate from the VB data rather than trying to read back from
+ * uncached memory.  
+ */
+static void TAG(build_tri_verts)( mgaContextPtr mmesa,
+				  struct vertex_buffer *VB, 
+				  GLfloat *O, 
+				  GLuint *elt ) 
+{
+   int i;
+
+   for (i = 0 ; i < 3 ; i++, O += CLIP_STRIDE) {
+      GLfloat *clip = VB->Clip.start + elt[i]*4;
+
+      O[0] = clip[0];
+      O[1] = clip[1];
+      O[2] = clip[2]; 
+      O[3] = clip[3];
+
+      if (TYPE & MGA_RGBA_BIT) {
+	 GLubyte *col = VEC_ELT(VB->ColorPtr, GLubyte, elt[i]);
+	 GLubyte *b = (GLubyte *)&O[4];
+	 b[CLIP_UBYTE_R] = col[0];
+	 b[CLIP_UBYTE_G] = col[1];
+	 b[CLIP_UBYTE_B] = col[2];
+	 b[CLIP_UBYTE_A] = col[3];
+      }
+
+      if (0)
+	 fprintf(stderr, 
+	      "build_tri_vert elt[%d]: %d phys: %x (first_phys %x elt_buf %x\n",
+	      i, elt[i], UNCLIPPED_VERT(elt[i]), 
+	      mmesa->first_vert_phys, (GLuint)mmesa->elt_buf);
+
+      *(GLuint *)&O[5] = UNCLIPPED_VERT(elt[i]);
+
+      if (TYPE & MGA_TEX0_BIT) {
+	 GLfloat *tex0_data = VEC_ELT(VB->TexCoordPtr[0], GLfloat, elt[i]);
+	 *(int*)&O[6] = *(int*)&tex0_data[0];
+	 *(int*)&O[7] = *(int*)&tex0_data[1];
+      }
+
+      if (TYPE & MGA_TEX1_BIT) {
+	 GLfloat *tex1_data = VEC_ELT(VB->TexCoordPtr[1], GLfloat, elt[i]);
+	 *(int*)&O[8] = *(int*)&tex1_data[0];
+	 *(int*)&O[9] = *(int*)&tex1_data[1];
+      }
+   }
+}
+
+
+/* Interpolate between two of the vertices constructed above.
+ */
+static void TAG(interp)( GLfloat t, 
+			 GLfloat *O,
+			 const GLfloat *I,
+			 const GLfloat *J )
+{
+   O[0] = LINTERP(t, I[0], J[0]);
+   O[1] = LINTERP(t, I[1], J[1]);
+   O[2] = LINTERP(t, I[2], J[2]);
+   O[3] = LINTERP(t, I[3], J[3]);
+
+   if (TYPE & MGA_RGBA_BIT) {
+      INTERP_RGBA(t,
+		  ((GLubyte *)&(O[4])),
+		  ((GLubyte *)&(I[4])),
+		  ((GLubyte *)&(J[4])));
+   }
+
+   if (0) fprintf(stderr, "setting 0x%x to ~0\n", (GLuint)&O[5]);
+
+   *(GLuint *)&O[5] = ~0;	/* note that this is a new vertex */
+ 
+   if (TYPE & MGA_TEX0_BIT) {
+      O[6] = LINTERP(t, I[6], J[6]);
+      O[7] = LINTERP(t, I[7], J[7]);
+   }
+
+   if (TYPE & MGA_TEX1_BIT) {
+      O[8] = LINTERP(t, I[8], J[8]);
+      O[9] = LINTERP(t, I[9], J[9]);
+   }
+}
+
+
+
+/* When clipping is complete, scan the final vertex list and emit any
+ * new ones to dma buffers.  Update the element list to a format
+ * suitable for sending to hardware.
+ */
+static void TAG(project_and_emit_verts)( mgaContextPtr mmesa,
+					 const GLfloat *verts,
+					 GLuint *elt,
+					 int nr)
+{
+   
+   GLfloat *O = mmesa->next_vert;
+   GLuint phys = mmesa->next_vert_phys;
+
+   const GLfloat *m = mmesa->device_matrix;
+   const GLfloat sx = m[0], sy = m[5], sz = m[10];
+   const GLfloat tx = m[12], ty = m[13], tz = m[14];
+   GLuint i;
+   
+   for (i = 0 ; i < nr ; i++) {
+      const GLfloat *I = &verts[elt[i] * CLIP_STRIDE];
+      GLuint tmp = *(GLuint *)&I[5];
+
+      if (0) fprintf(stderr, "elt[%d] (tmp 0x%x %d) %d --> ", i, (GLuint)&I[5],
+	      tmp, elt[i]);
+
+
+      if ((elt[i] = tmp) == ~0) 
+      {      
+	 GLfloat oow = 1.0/I[3];
+
+	 elt[i] = phys;
+	 phys -= BUFFER_STRIDE * sizeof(GLuint);
+	 
+	 O[0] = sx * I[0] * oow + tx;
+	 O[1] = sy * I[1] * oow + ty;
+	 O[2] = sz * I[2] * oow + tz;
+	 O[3] = oow; 
+
+	 if (TYPE & MGA_RGBA_BIT) {
+	    *(int*)&O[4] = *(int*)&I[4];
+	 }
+ 
+	 if (TYPE & MGA_TEX0_BIT) {
+	    *(int*)&O[6] = *(int*)&I[6];
+	    *(int*)&O[7] = *(int*)&I[7];
+	 }
+
+	 if (TYPE & MGA_TEX1_BIT) {
+	    *(int*)&O[8] = *(int*)&I[8];
+	    *(int*)&O[9] = *(int*)&I[9];
+	 }
+
+	 O -= BUFFER_STRIDE;
+      }
+      if (0) fprintf(stderr, "0x%x\n", elt[i]);
+   }
+
+   mmesa->next_vert = O;
+   mmesa->next_vert_phys = phys;
+}
+
+
+
+static void TAG(mga_init_eltpath)( struct mga_elt_tab *tab )
+{
+   tab->emit_unclipped_verts = TAG(emit_unclipped_verts);
+   tab->build_tri_verts = TAG(build_tri_verts);
+   tab->interp = TAG(interp);
+   tab->project_and_emit_verts = TAG(project_and_emit_verts);
+}
+
+#undef TYPE
+#undef TAG
+#undef STRIDE
diff --git a/xc/lib/GL/mesa/src/drv/mga/mgafastpath.c b/xc/lib/GL/mesa/src/drv/mga/mgafastpath.c
index 8017e7f02..76e0047f8 100644
--- a/xc/lib/GL/mesa/src/drv/mga/mgafastpath.c
+++ b/xc/lib/GL/mesa/src/drv/mga/mgafastpath.c
@@ -380,8 +380,10 @@ static void mga_project_vertices( struct vertex_buffer *VB )
    GLmatrix *mat = &ctx->Viewport.WindowMap;
    GLfloat m[16];
 
+   REFRESH_DRAWABLE_INFO(mmesa);
+
    m[MAT_SX] =   mat->m[MAT_SX];
-   m[MAT_TX] =   mat->m[MAT_TX] + mmesa->drawX - .5;
+   m[MAT_TX] =   mat->m[MAT_TX] + mmesa->drawX + .5;
    m[MAT_SY] = (- mat->m[MAT_SY]);
    m[MAT_TY] = (- mat->m[MAT_TY]) + mmesa->driDrawable->h + mmesa->drawY - .5;
    m[MAT_SZ] =   mat->m[MAT_SZ] * (1.0 / 0x10000);
@@ -401,8 +403,10 @@ static void mga_project_clipped_vertices( struct vertex_buffer *VB )
    GLmatrix *mat = &ctx->Viewport.WindowMap;
    GLfloat m[16];
 
+   REFRESH_DRAWABLE_INFO(mmesa);
+
    m[MAT_SX] =   mat->m[MAT_SX];
-   m[MAT_TX] =   mat->m[MAT_TX] + mmesa->drawX - .5;
+   m[MAT_TX] =   mat->m[MAT_TX] + mmesa->drawX + .5;
    m[MAT_SY] = (- mat->m[MAT_SY]);
    m[MAT_TY] = (- mat->m[MAT_TY]) + mmesa->driDrawable->h + mmesa->drawY - .5;
    m[MAT_SZ] =   mat->m[MAT_SZ] * (1.0 / 0x10000);
@@ -484,6 +488,16 @@ void mgaDDFastPath( struct vertex_buffer *VB )
 
    gl_prepare_arrays_cva( VB );	                 /* still need this */
 
+   if (gl_reduce_prim[prim] == GL_TRIANGLES && 
+       VB->Count < (MGA_DMA_BUF_SZ / 48) &&
+       (ctx->ModelProjectMatrix.flags & (MAT_FLAG_GENERAL|
+					 MAT_FLAG_PERSPECTIVE)) &&
+       mmesa->mgaScreen->chipset == MGA_CARD_TYPE_G400) 
+   {
+      mgaDDEltPath( VB );
+      return;
+   }
+
    /* Reserve enough space for the pathological case.
     */
    if (VB->EltPtr->count * 12 > MGA_DRIVER_DATA(VB)->size) {
@@ -507,18 +521,12 @@ void mgaDDFastPath( struct vertex_buffer *VB )
 	 ctx->CVA.elt_mode = gl_reduce_prim[prim];
 	 VB->EltPtr = &(MGA_DRIVER_DATA(VB)->clipped_elements);
 
-	 LOCK_HARDWARE( mmesa );
 	 mga_project_clipped_vertices( VB );    /* clip->device space */
 	 mga_render_elements_direct( VB );        /* render using new list */
-	 mgaFlushVerticesLocked( mmesa );
-	 UNLOCK_HARDWARE( mmesa );
       }
    } else {
-      LOCK_HARDWARE( mmesa );
       mga_project_vertices( VB );               /* clip->device space  */
       mga_render_elements_direct( VB );           /* render using orig list */
-      mgaFlushVerticesLocked( mmesa );
-      UNLOCK_HARDWARE( mmesa );
    }
 
    /* This indicates that there is no cached data to reuse.  
diff --git a/xc/lib/GL/mesa/src/drv/mga/mgafasttmp.h b/xc/lib/GL/mesa/src/drv/mga/mgafasttmp.h
index 43caddcac..683640409 100644
--- a/xc/lib/GL/mesa/src/drv/mga/mgafasttmp.h
+++ b/xc/lib/GL/mesa/src/drv/mga/mgafasttmp.h
@@ -39,7 +39,6 @@ static void TAG(mga_setup_full)( struct vertex_buffer *VB, GLuint do_cliptest )
    const GLfloat * const m = ctx->ModelProjectMatrix.m;
    GLuint start = VB->CopyStart;
    GLuint count = VB->Count;
-   GLuint i;
 
    gl_xform_points3_v16_general(MGA_DRIVER_DATA(VB)->verts[start].f,
 				m, 
@@ -71,26 +70,63 @@ static void TAG(mga_setup_full)( struct vertex_buffer *VB, GLuint do_cliptest )
       GLuint tex1_stride = VB->TexCoordPtr[1]->stride;
 
       GLfloat *f = MGA_DRIVER_DATA(VB)->verts[start].f;
-
-      for (i = start ; i < count ; i++, f += 16) {
+      GLfloat *end = f+(16*(count-start));
+      while (f != end) {
 	 if (TYPE & MGA_RGBA_BIT) {
+#if defined(USE_X86_ASM)
+	    __asm__ (
+	    "movl (%%edx),%%eax   \n"
+	    "bswap %%eax          \n"
+	    "rorl $8,%%eax        \n"
+	    "movl %%eax,16(%%edi) \n"
+	    :
+	    : "d" (color), "D" (f)
+	    : "%eax" );
+#else
+	    GLubyte *col = color;
 	    GLubyte *b = (GLubyte *)&f[CLIP_UBYTE_COLOR];
-	    GLubyte *col = color; color += color_stride;
-	    b[CLIP_UBYTE_R] = col[0];
-	    b[CLIP_UBYTE_G] = col[1];
 	    b[CLIP_UBYTE_B] = col[2];
+	    b[CLIP_UBYTE_G] = col[1];
+	    b[CLIP_UBYTE_R] = col[0];
 	    b[CLIP_UBYTE_A] = col[3];
+#endif
 	 }
 	 if (TYPE & MGA_TEX0_BIT) {
-	    f[CLIP_S0] = tex0_data[0];
-	    f[CLIP_T0] = tex0_data[1];
-	    STRIDE_F(tex0_data, tex0_stride);
+#if defined (USE_X86_ASM)
+	    __asm__ (
+	    "movl (%%ecx), %%eax    \n"
+	    "movl %%eax, 24(%%edi)  \n"
+	    "movl 4(%%ecx), %%eax   \n"
+	    "movl %%eax, 28(%%edi)"
+	    :
+	    : "c" (tex0_data), "D" (f)
+	    : "%eax");
+#else
+	    *(unsigned int *)(f+CLIP_S0) = *(unsigned int *)tex0_data;
+	    *(unsigned int *)(f+CLIP_T0) = *(unsigned int *)(tex0_data+1);
+#endif
 	 }
 	 if (TYPE & MGA_TEX1_BIT) {
-	    f[CLIP_S1] = tex1_data[0];
-	    f[CLIP_T1] = tex1_data[1];
-	    STRIDE_F(tex1_data, tex1_stride);
+	    /* Hits a second cache line.
+	     */
+#if defined (USE_X86_ASM)
+	    __asm__ (
+	    "movl (%%esi), %%eax    \n"
+	    "movl %%eax, 32(%%edi)  \n"
+	    "movl 4(%%esi), %%eax   \n"
+	    "movl %%eax, 36(%%edi)"
+	    :
+	    : "S" (tex1_data), "D" (f)
+	    : "%eax");
+#else
+	    *(unsigned int *)(f+CLIP_S1) = *(unsigned int *)tex1_data;
+	    *(unsigned int *)(f+CLIP_T1) = *(unsigned int *)(tex1_data+1);
+#endif	      
 	 }
+	 if(TYPE & MGA_RGBA_BIT)color += color_stride;
+	 if(TYPE & MGA_TEX0_BIT)STRIDE_F(tex0_data, tex0_stride);
+	 if(TYPE & MGA_TEX1_BIT)STRIDE_F(tex1_data, tex1_stride);
+	 f += 16;
       }
    }
 
diff --git a/xc/lib/GL/mesa/src/drv/mga/mgaioctl.c b/xc/lib/GL/mesa/src/drv/mga/mgaioctl.c
index c1819924d..640849e3f 100644
--- a/xc/lib/GL/mesa/src/drv/mga/mgaioctl.c
+++ b/xc/lib/GL/mesa/src/drv/mga/mgaioctl.c
@@ -14,114 +14,137 @@
 #include "mgalog.h"
 #include "mgavb.h"
 #include "mgatris.h"
+#include "mgabuffers.h"
+
 
 #include "drm.h"
 #include <sys/ioctl.h>
 
+#define DEPTH_SCALE 65535.0F
+
 static void mga_iload_dma_ioctl(mgaContextPtr mmesa,
-				int x1, int y1, int x2, int y2, 
-				unsigned long dest, unsigned int maccess)
+				unsigned long dest, 
+				int length)
 {
    int retcode;
    drm_mga_iload_t iload;
-   drmBufPtr buf = mmesa->dma_buffer;
+   drmBufPtr buf = mmesa->iload_buffer;
    
    iload.idx = buf->idx;
    iload.destOrg = dest;	
-   iload.mAccess = maccess;
-   iload.texture.x1 = x1;
-   iload.texture.y1 = y1;
-   iload.texture.y2 = x2;
-   iload.texture.x2 = y2;
+   iload.length = length;
+
+   if (MGA_DEBUG&DEBUG_VERBOSE_IOCTL) 
+      fprintf(stderr, "DRM_IOCTL_MGA_ILOAD idx %d dst %x length %d\n",
+	   iload.idx, iload.destOrg, iload.length);
+	   
    
    if ((retcode = ioctl(mmesa->driFd, DRM_IOCTL_MGA_ILOAD, &iload))) {
       printf("send iload retcode = %d\n", retcode);
       exit(1);
    }
-}
 
+   mmesa->iload_buffer = 0;
+
+   if (MGA_DEBUG&DEBUG_VERBOSE_IOCTL) 
+      fprintf(stderr, "finished iload dma put\n");
+
+}
 
-static void mga_vertex_dma_ioctl(mgaContextPtr mmesa)
+int mgaUpdateLock( mgaContextPtr mmesa, drmLockFlags flags )
 {
-   int retcode;
-   int size = MGA_DMA_BUF_SZ;
-   drmDMAReq dma;
-   drmBufPtr buf = mmesa->dma_buffer;
+   drm_lock_t lock;
+   
+   lock.flags = 0;
 
-   dma.context = mmesa->hHWContext;
-   dma.send_count = 1;
-   dma.send_list = &buf->idx;
-   dma.send_sizes = &size;
-   dma.flags = DRM_DMA_WAIT;
-   dma.request_count = 0;
-   dma.request_size = 0;
-   dma.request_list = 0;
-   dma.request_sizes = 0;
-
-   if ((retcode = drmDMA(mmesa->driFd, &dma))) {
-      printf("send iload retcode = %d\n", retcode);
-      exit(1);
+   if (mmesa->sarea->last_quiescent != mmesa->sarea->last_enqueue &&
+       flags & DRM_LOCK_QUIESCENT) {
+      if (MGA_DEBUG&DEBUG_VERBOSE_IOCTL)
+	fprintf(stderr, "mgaLockQuiescent\n");
+      lock.flags |= _DRM_LOCK_QUIESCENT;
+   }
+   
+   if (flags & DRM_LOCK_FLUSH)      lock.flags |= _DRM_LOCK_FLUSH;
+   if (flags & DRM_LOCK_FLUSH_ALL)  lock.flags |= _DRM_LOCK_FLUSH_ALL;
+ 
+   if (!lock.flags)
+      return 0;
+
+   if(ioctl(mmesa->driFd, DRM_IOCTL_MGA_FLUSH, &lock)) {
+      fprintf(stderr, "Lockupdate failed\n");
+      return -1;
    }
+   
+   if(flags & DRM_LOCK_QUIESCENT)
+     mmesa->sarea->last_quiescent = mmesa->sarea->last_enqueue;
+ 
+   return 0;
 }
 
-
-static void mga_get_buffer_ioctl( mgaContextPtr mmesa )
+static drmBufPtr mga_get_buffer_ioctl( mgaContextPtr mmesa )
 {
    int idx = 0;
    int size = 0;
    drmDMAReq dma;
    int retcode;
+   drmBufPtr buf;
    
-   fprintf(stderr,  "Getting dma buffer\n");
+   if (MGA_DEBUG&DEBUG_VERBOSE_IOCTL)
+      fprintf(stderr,  "Getting dma buffer\n");
    
    dma.context = mmesa->hHWContext;
    dma.send_count = 0;
    dma.send_list = NULL;
    dma.send_sizes = NULL;
-   dma.flags = DRM_DMA_WAIT;
+   dma.flags = 0;
    dma.request_count = 1;
    dma.request_size = MGA_DMA_BUF_SZ;
    dma.request_list = &idx;
    dma.request_sizes = &size;
-   
-   if ((retcode = drmDMA(mmesa->driFd, &dma))) {
-      fprintf(stderr, "request drmDMA retcode = %d\n", retcode);
-      exit(1);
-   }
+   dma.granted_count = 0;
 
-   mmesa->dma_buffer = &mmesa->mgaScreen->bufs->list[idx];
-}
 
-static void mga_swap_ioctl( mgaContextPtr mmesa )
-{
-   int retcode;
-   drm_mga_swap_t swap;
-      
-   if((retcode = ioctl(mmesa->driFd, DRM_IOCTL_MGA_SWAP, &swap))) {
-      printf("send swap retcode = %d\n", retcode);
-      exit(1);
-   }
-}
+   if (MGA_DEBUG&DEBUG_VERBOSE_IOCTL)
+      fprintf(stderr, "drmDMA (get) ctx %d count %d size 0x%x\n",
+	   dma.context, dma.request_count, 
+	   dma.request_size);
 
+   while (1) {
+      retcode = drmDMA(mmesa->driFd, &dma);
 
+      if (MGA_DEBUG&DEBUG_VERBOSE_IOCTL)
+	 fprintf(stderr, "retcode %d sz %d idx %d count %d\n", 
+		 retcode,
+		 dma.request_sizes[0],
+		 dma.request_list[0],
+		 dma.granted_count);
 
-static void mga_clear_ioctl( mgaContextPtr mmesa, int flags, int col, int zval )
-{
-   int retcode;
-   drm_mga_clear_t clear;
-   
-   clear.flags = flags;
-   clear.clear_color = col;
-   clear.clear_depth = zval;
-      
-   if((retcode = ioctl(mmesa->driFd, DRM_IOCTL_MGA_CLEAR, &clear))) {
-      printf("send clear retcode = %d\n", retcode);
-      exit(1);
+      if (retcode == 0 && 
+	  dma.request_sizes[0] &&
+	  dma.granted_count) 
+	 break;
+
+      if (MGA_DEBUG&DEBUG_VERBOSE_IOCTL)  
+	 fprintf(stderr, "\n\nflush");
+      mgaUpdateLock( mmesa, DRM_LOCK_FLUSH );
    }
-}
 
+   buf = &(mmesa->mgaScreen->bufs->list[idx]);
+   buf->used = 0;
 
+   if (MGA_DEBUG&DEBUG_VERBOSE_IOCTL)
+      fprintf(stderr, 
+	   "drmDMA (get) returns size[0] 0x%x idx[0] %d\n"
+	   "dma_buffer now: buf idx: %d size: %d used: %d\n",
+	   dma.request_sizes[0], dma.request_list[0],
+	   buf->idx, buf->total,
+	   buf->used);
 
+   if (MGA_DEBUG&DEBUG_VERBOSE_IOCTL)
+      fprintf(stderr, "finished getbuffer\n");
+
+   return buf;
+}
 
 
 
@@ -130,52 +153,59 @@ GLbitfield mgaClear( GLcontext *ctx, GLbitfield mask, GLboolean all,
 		     GLint cx, GLint cy, GLint cw, GLint ch ) 
 {
    mgaContextPtr mmesa = MGA_CONTEXT( ctx );
-   GLuint c = mmesa->ClearColor;
-   mgaUI32 zval = (mgaUI32) (ctx->Depth.Clear * DEPTH_SCALE);
    __DRIdrawablePrivate *dPriv = mmesa->driDrawable;
-   int flags = 0;
+   const GLuint colorMask = *((GLuint *) &ctx->Color.ColorMask);
+   drm_mga_clear_t clear;
+   int retcode;
    int i;
+   static int nrclears;
 
-   mgaMsg( 10, "mgaClear( %i, %i, %i, %i, %i )\n", 
-	   mask, x, y, width, height );
-
+   if (0) fprintf(stderr, "clear %d: %d,%d %dx%d\n", all,cx,cy,cw,ch);
 
-   mgaFlushVertices( mmesa );
+   clear.flags = 0;
+   clear.clear_color = mmesa->ClearColor;
+   clear.clear_depth = (mgaUI32) (ctx->Depth.Clear * DEPTH_SCALE);
 
+   FLUSH_BATCH( mmesa );
+	
+   if ((mask & DD_FRONT_LEFT_BIT) && colorMask == ~0) {
+      clear.flags |= MGA_FRONT;
+      mask &= ~DD_FRONT_LEFT_BIT;
+   }
 
-   if (mask & GL_COLOR_BUFFER_BIT) {
-      if (ctx->Color.DriverDrawBuffer == GL_FRONT_LEFT) {
-	 flags |= MGA_CLEAR_FRONT;
-	 mask &= ~GL_COLOR_BUFFER_BIT;
-      } else if (ctx->Color.DriverDrawBuffer == GL_BACK_LEFT) {
-	 flags |= MGA_CLEAR_BACK;
-	 mask &= ~GL_COLOR_BUFFER_BIT;
-      }
+   if ((mask & DD_BACK_LEFT_BIT) && colorMask == ~0) {
+      clear.flags |= MGA_BACK;
+      mask &= ~DD_BACK_LEFT_BIT;
    }
 
-   if ((flags & GL_DEPTH_BUFFER_BIT) && ctx->Depth.Mask) {
-      flags |= MGA_CLEAR_DEPTH;
-      mask &= ~GL_DEPTH_BUFFER_BIT;
+   if ((mask & DD_DEPTH_BIT) && ctx->Depth.Mask) {
+      clear.flags |= MGA_DEPTH;
+      mask &= ~DD_DEPTH_BIT;
    }
 
-   if (!flags)
+   if (!clear.flags)
       return mask;
 
    LOCK_HARDWARE( mmesa );
-   
+
+   if (mmesa->dirty_cliprects)
+      mgaUpdateRects( mmesa, (MGA_FRONT|MGA_BACK));
+
    /* flip top to bottom */
    cy = dPriv->h-cy-ch;
    cx += mmesa->drawX;
    cy += mmesa->drawY;
 
-   for (i = 0 ; i < dPriv->numClipRects ; ) { 	 
+   if (MGA_DEBUG&DEBUG_VERBOSE_IOCTL)
+      fprintf(stderr, "Clear, bufs %x nbox %d\n", 
+	      (int)clear.flags, (int)mmesa->numClipRects);
 
-      /* Use the cliprects for the current draw buffer
-       */
+   for (i = 0 ; i < mmesa->numClipRects ; ) 
+   { 	 
       int nr = MIN2(i + MGA_NR_SAREA_CLIPRECTS, mmesa->numClipRects);
       XF86DRIClipRectRec *box = mmesa->pClipRects;	 
-      xf86drmClipRectRec *b = mmesa->sarea->boxes;
-      mmesa->sarea->nbox = nr - i;
+      drm_clip_rect_t *b = mmesa->sarea->boxes;
+      int n = 0;
 
       if (!all) {
 	 for ( ; i < nr ; i++) {
@@ -196,16 +226,37 @@ GLbitfield mgaClear( GLcontext *ctx, GLbitfield mask, GLboolean all,
 	    b->x2 = x + w;
 	    b->y2 = y + h;
 	    b++;
+	    n++;
 	 }
       } else {
-	 for ( ; i < nr ; i++) 
-	    *b++ = *(xf86drmClipRectRec *)&box[i];
+	 for ( ; i < nr ; i++) {
+	    *b++ = *(drm_clip_rect_t *)&box[i];
+	    n++;
+	 }
       }
 
-      mga_clear_ioctl( mmesa, mask, c, zval );
+
+      if (MGA_DEBUG&DEBUG_VERBOSE_IOCTL)
+	 fprintf(stderr, 
+		 "DRM_IOCTL_MGA_CLEAR flag 0x%x color %x depth %x nbox %d\n",
+		 clear.flags, clear.clear_color, 
+		 clear.clear_depth, mmesa->sarea->nbox);
+		
+
+      mmesa->sarea->nbox = n;
+
+      retcode = ioctl(mmesa->driFd, DRM_IOCTL_MGA_CLEAR, &clear);
+      if (retcode) {
+	 printf("send clear retcode = %d\n", retcode);
+	 exit(1);
+      }
+      if (MGA_DEBUG&DEBUG_VERBOSE_IOCTL)
+	 fprintf(stderr, "finished clear %d\n", ++nrclears);
    }
 
    UNLOCK_HARDWARE( mmesa );
+   mmesa->dirty |= MGA_UPLOAD_CLIPRECTS;
+
    return mask;
 }
 
@@ -218,50 +269,61 @@ GLbitfield mgaClear( GLcontext *ctx, GLbitfield mask, GLboolean all,
 void mgaSwapBuffers( mgaContextPtr mmesa ) 
 {
    __DRIdrawablePrivate *dPriv = mmesa->driDrawable;
+   XF86DRIClipRectPtr pbox;
+   int nbox;
+   drm_mga_swap_t swap;
+   static int nrswaps;
+   int retcode;
    int i;
+   int tmp;
 
-   mgaFlushVertices( mmesa );
 
-   LOCK_HARDWARE( mmesa );
-   {
-      /* Use the frontbuffer cliprects
-       */
-      XF86DRIClipRectPtr pbox = dPriv->pClipRects;
-      int nbox = dPriv->numClipRects;
+   FLUSH_BATCH( mmesa );
 
-      for (i = 0 ; i < nbox ; )
-      {
-	 int nr = MIN2(i + MGA_NR_SAREA_CLIPRECTS, dPriv->numClipRects);
-	 XF86DRIClipRectRec *b = (XF86DRIClipRectRec *)mmesa->sarea->boxes;
-	 mmesa->sarea->nbox = nr - i;
+   LOCK_HARDWARE( mmesa );
+   
+   /* Use the frontbuffer cliprects
+    */
+   if (mmesa->dirty_cliprects & MGA_FRONT) 
+      mgaUpdateRects( mmesa, MGA_FRONT );
+   
 
-	 for ( ; i < nr ; i++) 
-	    *b++ = pbox[i];
-	 
-	 mga_swap_ioctl( mmesa );
-      }
-   }
+   pbox = dPriv->pClipRects;
+   nbox = dPriv->numClipRects;
 
+   if (0) fprintf(stderr, "swap, nbox %d\n", nbox);
 
-#if 1
-   UNLOCK_HARDWARE(mmesa);
-#else
+   for (i = 0 ; i < nbox ; )
    {
-      last_enqueue = mmesa->sarea->lastEnqueue;
-      last_dispatch = mmesa->sarea->lastDispatch;
-      UNLOCK_HARDWARE;
+      int nr = MIN2(i + MGA_NR_SAREA_CLIPRECTS, dPriv->numClipRects);
+      XF86DRIClipRectRec *b = (XF86DRIClipRectRec *)mmesa->sarea->boxes;
+
+      mmesa->sarea->nbox = nr - i;
+	 
+      for ( ; i < nr ; i++) 
+	 *b++ = pbox[i];
       
-      /* Throttle runaway apps - there should be an easier way to sleep
-       * on dma without locking out the rest of the system!
-       */
-      if (mmesa->lastSwap > last_dispatch) {
-	 drmGetLock(mmesa->driFd, mmesa->hHWContext, DRM_LOCK_QUIESCENT);	
-	 DRM_UNLOCK(mmesa->driFd, mmesa->driHwLock, mmesa->hHWContext);	
+      if (0)
+	 fprintf(stderr, "DRM_IOCTL_MGA_SWAP\n"); 
+
+      if((retcode = ioctl(mmesa->driFd, DRM_IOCTL_MGA_SWAP, &swap))) {
+	 printf("send swap retcode = %d\n", retcode);
+	 exit(1);
       }
 
-      mmesa->lastSwap = last_enqueue;
+      if (MGA_DEBUG&DEBUG_VERBOSE_IOCTL)
+	 fprintf(stderr, "finished swap %d\n", ++nrswaps);
    }
-#endif
+
+   tmp = GET_ENQUEUE_AGE(mmesa);
+
+   UNLOCK_HARDWARE( mmesa );
+
+   if (GET_DISPATCH_AGE(mmesa) < mmesa->lastSwap)
+      mgaWaitAge(mmesa, mmesa->lastSwap);
+
+   mmesa->lastSwap = tmp;
+   mmesa->dirty |= MGA_UPLOAD_CLIPRECTS;
 }
 
 
@@ -269,13 +331,165 @@ void mgaSwapBuffers( mgaContextPtr mmesa )
  */
 void mgaDDFinish( GLcontext *ctx  ) 
 {
-   mgaContextPtr mmesa = MGA_CONTEXT( ctx );
-   drmGetLock(mmesa->driFd, mmesa->hHWContext, DRM_LOCK_QUIESCENT);	
-   DRM_UNLOCK(mmesa->driFd, mmesa->driHwLock, mmesa->hHWContext);	
+   mgaContextPtr mmesa = MGA_CONTEXT(ctx);
+
+   FLUSH_BATCH( mmesa );
+
+   if (mmesa->sarea->last_quiescent != mmesa->sarea->last_enqueue) {
+      if (MGA_DEBUG&DEBUG_VERBOSE_IOCTL)
+	 fprintf(stderr, "mgaRegetLockQuiescent\n");
+
+      LOCK_HARDWARE( mmesa );
+      mgaUpdateLock( mmesa, DRM_LOCK_QUIESCENT | DRM_LOCK_FLUSH);
+      UNLOCK_HARDWARE( mmesa );
+
+      mmesa->sarea->last_quiescent = mmesa->sarea->last_enqueue;
+   }
+}
+
+void mgaWaitAgeLocked( mgaContextPtr mmesa, int age  ) 
+{
+   if (GET_DISPATCH_AGE(mmesa) < age) {
+      if (0) fprintf(stderr, "\n\n\nmgaWaitAgeLocked\n");
+      mgaUpdateLock( mmesa, DRM_LOCK_FLUSH );
+   }
+}
+
+
+void mgaWaitAge( mgaContextPtr mmesa, int age  ) 
+{
+   if (GET_DISPATCH_AGE(mmesa) < age) {
+      LOCK_HARDWARE(mmesa);
+      if (GET_DISPATCH_AGE(mmesa) < age) {
+	 if (0) fprintf(stderr, "\n\n\nmgaWaitAge\n");
+	 mgaUpdateLock( mmesa, DRM_LOCK_FLUSH );
+      }
+      UNLOCK_HARDWARE(mmesa);
+   }
+}
+
+
+static int intersect_rect( drm_clip_rect_t *out,
+			   drm_clip_rect_t *a,
+			   drm_clip_rect_t *b )
+{
+   *out = *a;
+   if (b->x1 > out->x1) out->x1 = b->x1;
+   if (b->y1 > out->y1) out->y1 = b->y1;
+   if (b->x2 < out->x2) out->x2 = b->x2;
+   if (b->y2 < out->y2) out->y2 = b->y2;
+   if (out->x1 >= out->x2) return 0;
+   if (out->y1 >= out->y2) return 0;
+   return 1;
+}
+
+
+
+
+static void age_mmesa( mgaContextPtr mmesa, int age )
+{
+   if (mmesa->CurrentTexObj[0]) mmesa->CurrentTexObj[0]->age = age;
+   if (mmesa->CurrentTexObj[1]) mmesa->CurrentTexObj[1]->age = age;
 }
 
+void mgaFlushVerticesLocked( mgaContextPtr mmesa )
+{
+   drm_clip_rect_t *pbox = (drm_clip_rect_t *)mmesa->pClipRects;
+   int nbox = mmesa->numClipRects;
+   drmBufPtr buffer = mmesa->vertex_dma_buffer;
+   drm_mga_vertex_t vertex;
+   int i;
+
+   mmesa->vertex_dma_buffer = 0;
+
+   if (!buffer)
+      return;
+
+   if (mmesa->dirty_cliprects & mmesa->draw_buffer)
+      mgaUpdateRects( mmesa, mmesa->draw_buffer );
+
+   if (mmesa->dirty & ~MGA_UPLOAD_CLIPRECTS) 
+      mgaEmitHwStateLocked( mmesa );
+
+   /* FIXME: Workaround bug in kernel module.
+    */
+   mmesa->sarea->dirty |= MGA_UPLOAD_CTX; 
+      
+   /* FIXME: dstorg bug
+    */
+   if (0)
+      if (mmesa->lastX != mmesa->drawX || mmesa->lastY != mmesa->drawY)
+	 fprintf(stderr, "****** last: %d,%d current: %d,%d\n",
+		 mmesa->lastX, mmesa->lastY,
+		 mmesa->drawX, mmesa->drawY);
+
+   vertex.idx = buffer->idx;
+   vertex.used = buffer->used;
+   vertex.discard = 0;
+
+   if (!nbox)
+      vertex.used = 0;
+
+   if (nbox >= MGA_NR_SAREA_CLIPRECTS)
+      mmesa->dirty |= MGA_UPLOAD_CLIPRECTS;
+
+   if (!vertex.used || !(mmesa->dirty & MGA_UPLOAD_CLIPRECTS)) 
+   {
+      if (nbox == 1) 
+	 mmesa->sarea->nbox = 0;
+      else
+	 mmesa->sarea->nbox = nbox;
+
+      if (MGA_DEBUG&DEBUG_VERBOSE_IOCTL)  
+	 fprintf(stderr, "Firing vertex -- case a nbox %d\n", nbox);
+
+      vertex.discard = 1;
+      ioctl(mmesa->driFd, DRM_IOCTL_MGA_VERTEX, &vertex);
+      age_mmesa(mmesa, mmesa->sarea->last_enqueue);
+   } 
+   else 
+   {      
+      for (i = 0 ; i < nbox ; )
+      {
+	 int nr = MIN2(i + MGA_NR_SAREA_CLIPRECTS, nbox);
+	 drm_clip_rect_t *b = mmesa->sarea->boxes;
+
+	 if (mmesa->scissor) {
+	    mmesa->sarea->nbox = 0;	 
+
+	    for ( ; i < nr ; i++) {	       
+	       *b = pbox[i];
+	       if (intersect_rect(b, b, &mmesa->scissor_rect)) {
+		  mmesa->sarea->nbox++;
+		  b++;
+	       } 
+	    }
+
+	    /* Culled?
+	     */
+	    if (!mmesa->sarea->nbox) {
+	       if (nr < nbox) continue;
+	       vertex.used = 0;
+	    }
+	 } else {
+	    mmesa->sarea->nbox = nr - i;
+	    for ( ; i < nr ; i++) 
+	       *b++ = pbox[i];
+	 }
+
+	 /* Finished with the buffer?
+	  */
+	 if (nr == nbox) 
+	    vertex.discard = 1;
 
+	 mmesa->sarea->dirty |= MGA_UPLOAD_CLIPRECTS;
+	 ioctl(mmesa->driFd, DRM_IOCTL_MGA_VERTEX, &vertex);
+	 age_mmesa(mmesa, mmesa->sarea->last_enqueue);
+      }
+   }
 
+   mmesa->dirty &= ~MGA_UPLOAD_CLIPRECTS;
+}
 
 void mgaFlushVertices( mgaContextPtr mmesa ) 
 {
@@ -284,38 +498,221 @@ void mgaFlushVertices( mgaContextPtr mmesa )
    UNLOCK_HARDWARE( mmesa );
 }
 
+void mgaFlushEltsLocked( mgaContextPtr mmesa ) 
+{
+   if (mmesa->first_elt != mmesa->next_elt) {
+      mgaFireEltsLocked( mmesa, 
+			 ((GLuint)mmesa->first_elt - 
+			  (GLuint)mmesa->elt_buf->address),
+			 ((GLuint)mmesa->next_elt - 
+			  (GLuint)mmesa->elt_buf->address),
+			 0 );
+      mmesa->first_elt = mmesa->next_elt;
+   }
+}
 
-void mgaFlushVerticesLocked( mgaContextPtr mmesa )
+void mgaFlushElts( mgaContextPtr mmesa ) 
+{
+   LOCK_HARDWARE( mmesa );
+   mgaFlushEltsLocked( mmesa );
+   UNLOCK_HARDWARE( mmesa );
+}
+
+
+mgaUI32 *mgaAllocVertexDwords( mgaContextPtr mmesa, int dwords )
+{
+   int bytes = dwords * 4;
+   mgaUI32 *head;
+
+   if (!mmesa->vertex_dma_buffer) {
+      LOCK_HARDWARE( mmesa );
+
+      if (mmesa->first_elt != mmesa->next_elt) 
+	 mgaFlushEltsLocked(mmesa);
+
+      mmesa->vertex_dma_buffer = mga_get_buffer_ioctl( mmesa );
+      UNLOCK_HARDWARE( mmesa );
+   } else if (mmesa->vertex_dma_buffer->used + bytes > 
+	      mmesa->vertex_dma_buffer->total) {
+      LOCK_HARDWARE( mmesa );
+      mgaFlushVerticesLocked( mmesa );
+      mmesa->vertex_dma_buffer = mga_get_buffer_ioctl( mmesa );
+      UNLOCK_HARDWARE( mmesa );
+   }
+
+   head = (mgaUI32 *)((char *)mmesa->vertex_dma_buffer->address + 
+		      mmesa->vertex_dma_buffer->used);
+
+   mmesa->vertex_dma_buffer->used += bytes;
+   return head;
+}
+
+
+void mgaFireILoadLocked( mgaContextPtr mmesa, 
+			 GLuint offset, GLuint length )
+{
+   if (!mmesa->iload_buffer) {
+      fprintf(stderr, "mgaFireILoad: no buffer\n");
+      return;
+   }
+
+   if (MGA_DEBUG&DEBUG_VERBOSE_IOCTL)
+      fprintf(stderr, "mgaFireILoad idx %d ofs 0x%x length %d\n",
+	      mmesa->iload_buffer->idx, (int)offset, (int)length );
+
+   mga_iload_dma_ioctl( mmesa, offset, length );
+}
+
+void mgaGetILoadBufferLocked( mgaContextPtr mmesa )
+{
+   if (MGA_DEBUG&DEBUG_VERBOSE_IOCTL)
+      fprintf(stderr, "mgaGetIloadBuffer (buffer now %p)\n",
+	   mmesa->iload_buffer);
+
+   mmesa->iload_buffer = mga_get_buffer_ioctl( mmesa );
+}
+
+
+ void mgaDDFlush( GLcontext *ctx )
 {
-   XF86DRIClipRectPtr pbox = mmesa->pClipRects;
+   mgaContextPtr mmesa = MGA_CONTEXT( ctx );
+
+
+   FLUSH_BATCH( mmesa );
+
+   /* This may be called redundantly - dispatch_age may trail what
+    * has actually been sent and processed by the hardware.
+    */
+   if (GET_DISPATCH_AGE( mmesa ) < mmesa->sarea->last_enqueue) {
+      LOCK_HARDWARE( mmesa );
+      if (0) fprintf(stderr, "mgaDDFlush %d %d\n", GET_DISPATCH_AGE( mmesa ),  mmesa->sarea->last_enqueue);
+      mgaUpdateLock( mmesa, DRM_LOCK_FLUSH );
+      UNLOCK_HARDWARE( mmesa );
+   }
+}
+
+
+
+void mgaFireEltsLocked( mgaContextPtr mmesa, 
+			GLuint start, 
+			GLuint end,
+			GLuint discard )
+{
+   drm_clip_rect_t *pbox = (drm_clip_rect_t *)mmesa->pClipRects;
    int nbox = mmesa->numClipRects;
+   drmBufPtr buffer = mmesa->elt_buf;
+   drm_mga_indices_t elts;
    int i;
 
-   if (mmesa->dirty)
+
+   if (0) fprintf(stderr, "FireElts %d %d\n", start/4, end/4);
+
+   if (!buffer)
+      return;
+
+   if (mmesa->dirty_cliprects & mmesa->draw_buffer)
+      mgaUpdateRects( mmesa, mmesa->draw_buffer );
+
+   if (mmesa->dirty & ~MGA_UPLOAD_CLIPRECTS) 
       mgaEmitHwStateLocked( mmesa );
 
-   for (i = 0 ; i < nbox ; )
+   /* FIXME: Workaround bug in kernel module.
+    */
+   mmesa->sarea->dirty |= MGA_UPLOAD_CTX; 
+      
+   elts.idx = buffer->idx;
+   elts.start = start;
+   elts.end = end;
+   elts.discard = 0;
+
+   if (!nbox)
+      elts.end = start;
+
+   if (nbox >= MGA_NR_SAREA_CLIPRECTS)
+      mmesa->dirty |= MGA_UPLOAD_CLIPRECTS;
+
+   if (elts.end == start || !(mmesa->dirty & MGA_UPLOAD_CLIPRECTS)) 
    {
-      int nr = MIN2(i + MGA_NR_SAREA_CLIPRECTS, nbox);
-      XF86DRIClipRectRec *b = (XF86DRIClipRectRec *)mmesa->sarea->boxes;
-      mmesa->sarea->nbox = nr - i;
+      if (nbox == 1) 
+	 mmesa->sarea->nbox = 0;
+      else
+	 mmesa->sarea->nbox = nbox;
+
+      if (0)
+	 fprintf(stderr, "Firing elts -- case a nbox %d\n", nbox);
+
+      elts.discard = discard;
+      ioctl(mmesa->driFd, DRM_IOCTL_MGA_INDICES, &elts);
+      age_mmesa(mmesa, mmesa->sarea->last_enqueue);
+   } 
+   else 
+   {      
+      for (i = 0 ; i < nbox ; )
+      {
+	 int nr = MIN2(i + MGA_NR_SAREA_CLIPRECTS, nbox);
+	 drm_clip_rect_t *b = mmesa->sarea->boxes;
+
+	 if (mmesa->scissor) {
+	    mmesa->sarea->nbox = 0;	 
+
+	    for ( ; i < nr ; i++) {	       
+	       *b = pbox[i];
+	       if (intersect_rect(b, b, &mmesa->scissor_rect)) {
+		  mmesa->sarea->nbox++;
+		  b++;
+	       } 
+	    }
+
+	    /* Culled?
+	     */
+	    if (!mmesa->sarea->nbox) {
+	       if (nr < nbox) continue;
+	       elts.end = start;
+	    }
+	 } else {
+	    mmesa->sarea->nbox = nr - i;
+	    for ( ; i < nr ; i++) 
+	       *b++ = pbox[i];
+	 }
 
-      for ( ; i < nr ; i++) 
-	 *b++ = pbox[i];
-	 
-      mga_vertex_dma_ioctl( mmesa );
+	 /* Potentially finished with the buffer?
+	  */
+	 if (nr == nbox) 
+	    elts.discard = discard;
 
-      break;			/* fix dma multiple dispatch */
+	 if (0)
+	    fprintf(stderr, "Firing elts -- case b nbox %d\n", nbox);
+
+	 mmesa->sarea->dirty |= MGA_UPLOAD_CLIPRECTS;
+	 ioctl(mmesa->driFd, DRM_IOCTL_MGA_INDICES, &elts);
+	 age_mmesa(mmesa, mmesa->sarea->last_enqueue);
+      }
    }
 
-   mga_get_buffer_ioctl( mmesa );
+   mmesa->dirty &= ~MGA_UPLOAD_CLIPRECTS;
+}
+
+void mgaGetEltBufLocked( mgaContextPtr mmesa )
+{
+   mmesa->elt_buf = mga_get_buffer_ioctl( mmesa );
 }
 
+void mgaReleaseBufLocked( mgaContextPtr mmesa, drmBufPtr buffer )
+{
+   drm_mga_vertex_t vertex;
+
+   if (!buffer) return;
+   
+   vertex.idx = buffer->idx;
+   vertex.used = 0;
+   vertex.discard = 1;
+   ioctl(mmesa->driFd, DRM_IOCTL_MGA_VERTEX, &vertex);
+}
 
 
 void mgaDDInitIoctlFuncs( GLcontext *ctx )
 {
    ctx->Driver.Clear = mgaClear;
-   ctx->Driver.Flush = 0;
+   ctx->Driver.Flush = mgaDDFlush;
    ctx->Driver.Finish = mgaDDFinish;
 }
diff --git a/xc/lib/GL/mesa/src/drv/mga/mgaioctl.h b/xc/lib/GL/mesa/src/drv/mga/mgaioctl.h
index 22ca26e17..900e4a4f8 100644
--- a/xc/lib/GL/mesa/src/drv/mga/mgaioctl.h
+++ b/xc/lib/GL/mesa/src/drv/mga/mgaioctl.h
@@ -9,9 +9,37 @@ GLbitfield mgaClear( GLcontext *ctx, GLbitfield mask, GLboolean all,
 
 void mgaSwapBuffers( mgaContextPtr mmesa ); 
 
+
+
+mgaUI32 *mgaAllocVertexDwords( mgaContextPtr mmesa, int dwords );
+
+
+void mgaGetILoadBufferLocked( mgaContextPtr mmesa );
+
+
+void mgaFireILoadLocked( mgaContextPtr mmesa, 
+			 GLuint offset, GLuint length );
+
+void mgaWaitAgeLocked( mgaContextPtr mmesa, int age );
+void mgaWaitAge( mgaContextPtr mmesa, int age );
+int mgaUpdateLock( mgaContextPtr mmesa, drmLockFlags flags );
+
 void mgaFlushVertices( mgaContextPtr mmesa ); 
 void mgaFlushVerticesLocked( mgaContextPtr mmesa );
 
+
+void mgaFireEltsLocked( mgaContextPtr mmesa, 
+			GLuint start, 
+			GLuint end,
+			GLuint discard );
+
+void mgaGetEltBufLocked( mgaContextPtr mmesa );
+void mgaReleaseBufLocked( mgaContextPtr mmesa, drmBufPtr buffer );
+
+void mgaFlushEltsLocked( mgaContextPtr mmesa );
+void mgaFlushElts( mgaContextPtr mmesa ) ;
+
+
 /* upload texture
  */
 
@@ -19,5 +47,12 @@ void mgaDDFinish( GLcontext *ctx );
 
 void mgaDDInitIoctlFuncs( GLcontext *ctx );
  
+#define FLUSH_BATCH(mmesa) do {						\
+        if (MGA_DEBUG&DEBUG_VERBOSE_IOCTL)  				\
+              fprintf(stderr, "FLUSH_BATCH in %s\n", __FUNCTION__);	\
+	if (mmesa->vertex_dma_buffer) mgaFlushVertices(mmesa);		\
+	else if (mmesa->next_elt != mmesa->first_elt) mgaFlushElts(mmesa);	        	\
+} while (0)
+
 
 #endif
diff --git a/xc/lib/GL/mesa/src/drv/mga/mgalib.h b/xc/lib/GL/mesa/src/drv/mga/mgalib.h
index 4600a69a0..7d086ae28 100644
--- a/xc/lib/GL/mesa/src/drv/mga/mgalib.h
+++ b/xc/lib/GL/mesa/src/drv/mga/mgalib.h
@@ -32,10 +32,10 @@
 #include "dri_tmm.h"
 #include "dri_mesaint.h"
 #include "dri_mesa.h"
-#include "xmesaP.h"
 
 #include "types.h"
 
+#include "drm.h"
 #include "mgacommon.h"
 #include "mm.h"
 #include "mgalog.h"
@@ -43,7 +43,6 @@
 #include "mgatex.h"
 #include "mgavb.h"
 
-#include "mga_drm_public.h"
 #include "mga_xmesa.h"
 
 
@@ -51,11 +50,8 @@
 #define MGA_FIELD(field,val) (((val) << (field ## _SHIFT)) & ~(field ## _MASK))
 #define MGA_GET_FIELD(field, val) ((val & ~(field ## _MASK)) >> (field ## _SHIFT))
 
-#define MGA_CHIP_MGAG200 0
-#define MGA_CHIP_MGAG400 1
-
-#define MGA_IS_G200(mmesa) (mmesa->mgaScreen->chipset == MGA_CHIP_MGAG200)
-#define MGA_IS_G400(mmesa) (mmesa->mgaScreen->chipset == MGA_CHIP_MGAG400)
+#define MGA_IS_G200(mmesa) (mmesa->mgaScreen->chipset == MGA_CARD_TYPE_G200)
+#define MGA_IS_G400(mmesa) (mmesa->mgaScreen->chipset == MGA_CARD_TYPE_G400)
 
 
 /* SoftwareFallback 
@@ -65,6 +61,7 @@
  */
 #define MGA_FALLBACK_TEXTURE   0x1
 #define MGA_FALLBACK_BUFFER    0x2
+#define MGA_FALLBACK_STIPPLE   0x3
 
 
 /* For mgaCtx->new_state.
@@ -97,17 +94,45 @@ typedef void (*mga_interp_func)( GLfloat t,
 #define MGA_PF_8888  (10 << 4)
 #define MGA_PF_HASALPHA (8 << 4)
 
+
+
+/* Reasons why the GL_BLEND fallback mightn't work:
+ */
+#define MGA_BLEND_ENV_COLOR 0x1
+#define MGA_BLEND_MULTITEX  0x2
+
+struct mga_elt_tab {
+   void (*emit_unclipped_verts)( struct vertex_buffer *VB );
+
+   void (*build_tri_verts)( mgaContextPtr mmesa, 
+			    struct vertex_buffer *VB, 
+			    GLfloat *O, GLuint *elt );
+
+   void (*interp)( GLfloat t, GLfloat *O, 
+		   const GLfloat *I, const GLfloat *J );
+
+   void (*project_and_emit_verts)( mgaContextPtr mmesa,
+				   const GLfloat *verts,
+				   GLuint *elts,
+				   int nr );
+};
+
 struct mga_context_t {
 
    GLcontext *glCtx;
 
-   /* Hardware state - moved from mgabuf.h
-    */
-   mgaUI32 Setup[MGA_CTX_SETUP_SIZE];
 
-   /* Variable sized vertices
+   /* Bookkeeping for texturing 
     */
-   mgaUI32 vertsize;
+   int lastTexHeap;
+   struct mga_texture_object_s TexObjList[MGA_NR_TEX_HEAPS];
+   struct mga_texture_object_s SwappedOut;
+   struct mga_texture_object_s *CurrentTexObj[2];
+   memHeap_t *texHeap[MGA_NR_TEX_HEAPS];
+   int c_texupload;
+   int c_texusage;
+   int tex_thrash;
+
 
    /* Map GL texture units onto hardware.
     */
@@ -116,72 +141,99 @@ struct mga_context_t {
    mgaUI32 tex_dest[2];
 
 
+   /* Manage fallbacks 
+    */
+   mgaUI32 IndirectTriangles;
+   int Fallback;  
 
-   /* bookkeeping for textureing */
-   struct mga_texture_object_s TexObjList;
-   struct mga_texture_object_s SwappedOut;
-   struct mga_texture_object_s *CurrentTexObj[2];
-
-
-   /* shared texture palette */
-   mgaUI16	GlobalPalette[256];
-  
-   int Fallback;  /* or'ed values of FALLBACK_* */
 
-   /* Support for CVA and the fast paths */
+   /* Support for CVA and the fastpath 
+    */
    unsigned int setupdone;
    unsigned int setupindex;
    unsigned int renderindex;
    unsigned int using_fast_path;
-   unsigned int using_immediate_fast_path;
    mga_interp_func interp;
 
-   /* Shortcircuit some state changes */
+
+   /* Support for limited GL_BLEND fallback
+    */
+   unsigned int blend_flags;
+   unsigned int envcolor;
+
+
+   /* Shortcircuit some state changes 
+    */
    points_func   PointsFunc;
    line_func     LineFunc;
    triangle_func TriangleFunc;
    quad_func     QuadFunc;
 
-   /* Manage our own state */
+
+   /* Manage driver and hardware state 
+    */
    GLuint        new_state; 
    GLuint        dirty;
-
-   GLubyte       clearcolor[4];
-   GLushort MonoColor;
-   GLushort ClearColor;
+   GLuint        Setup[MGA_CTX_SETUP_SIZE];
+   GLuint        warp_pipe;
+   GLuint        vertsize;
+   GLushort      MonoColor;
+   GLushort      ClearColor;
+   GLuint        poly_stipple;
 
    
-   /* DRI stuff
+   /* Dma buffers
     */
-   drmBufPtr  dma_buffer;
+   drmBufPtr  vertex_dma_buffer;
+   drmBufPtr  iload_buffer;
 
-   GLframebuffer *glBuffer;
-   memHeap_t *texHeap;
+   
+   /* Drawable, cliprect and scissor information
+    */
+   int dirty_cliprects;		/* which sets of cliprects are uptodate? */
+   int draw_buffer;		/* which buffer are we rendering to */
+   unsigned int drawOffset;		/* draw buffer address in  space */
+   int read_buffer;	
+   int readOffset;	
+   int drawX, drawY;		/* origin of drawable in draw buffer */
+   int lastX, lastY;		/* detect DSTORG bug */
+   GLuint numClipRects;		/* cliprects for the draw buffer */
+   XF86DRIClipRectPtr pClipRects;
+   XF86DRIClipRectRec draw_rect;
+   drm_clip_rect_t scissor_rect;
+   int scissor;
 
-   GLuint needClip;
-   GLuint warp_pipe;
 
-   /* These refer to the current draw (front vs. back) buffer:
+   /* Texture aging and DMA based aging.
     */
-   int drawOffset;		/* draw buffer address in agp space */
-   int drawX;			/* origin of drawable in draw buffer */
-   int drawY;
-   GLuint numClipRects;		/* cliprects for that buffer */
-   XF86DRIClipRectPtr pClipRects;
+   unsigned int texAge[MGA_NR_TEX_HEAPS];/* texture LRU age  */
+   unsigned int dirtyAge;		/* buffer age for synchronization */
+   unsigned int lastSwap;		/* throttling runaway apps */
 
-   int texAge;
 
-   XF86DRIClipRectRec draw_rect;
 
+   /* Mirrors of some DRI state.
+    */
+   GLframebuffer *glBuffer;
    drmContext hHWContext;
    drmLock *driHwLock;
    int driFd;
    Display *display;
-
    __DRIdrawablePrivate *driDrawable;
    __DRIscreenPrivate *driScreen;
    mgaScreenPrivate *mgaScreen; 
    drm_mga_sarea_t *sarea;
+
+
+   /* New setupdma path
+    */
+   drmBufPtr elt_buf, retained_buf;
+   GLuint *first_elt, *next_elt;
+   GLfloat *next_vert;
+   GLuint next_vert_phys;
+   GLuint first_vert_phys;
+   struct mga_elt_tab *elt_tab;
+   GLfloat device_matrix[16];
 };
 
 
@@ -191,16 +243,12 @@ typedef struct {
 
 	/* dma stuff */
         mgaUI32         systemTexture;
-        mgaUI32         noSetupDma;
 
         mgaUI32         default32BitTextures;
-        mgaUI32         swapBuffersCount;
         
 	/* options */
 	mgaUI32		nullprims;  /* skip all primitive generation */
-	mgaUI32		noFallback; /* don't fall back to software, do
-                                       best-effort rendering */
- 	mgaUI32		skipDma;    /* don't send anything to the hardware */
+        mgaUI32         noFallback;
  	
 	/* performance counters */	
 	mgaUI32		c_textureUtilization;
@@ -240,10 +288,12 @@ extern mgaGlx_t	mgaglx;
 extern int MGA_DEBUG;
 #endif
 
-#define MGA_DEBUG_ALWAYS_SYNC    0x1
-#define MGA_DEBUG_VERBOSE_MSG    0x2
-#define MGA_DEBUG_VERBOSE_LRU    0x4
-#define MGA_DEBUG_VERBOSE_DRI    0x8
+#define DEBUG_ALWAYS_SYNC    0x1
+#define DEBUG_VERBOSE_MSG    0x2
+#define DEBUG_VERBOSE_LRU    0x4
+#define DEBUG_VERBOSE_DRI    0x8
+#define DEBUG_VERBOSE_IOCTL  0x10
+#define DEBUG_VERBOSE_2D     0x20
 
 static __inline__ mgaUI32 mgaPackColor(mgaUI32 format,
 				       mgaUI8 r, mgaUI8 g, 
diff --git a/xc/lib/GL/mesa/src/drv/mga/mgapipeline.h b/xc/lib/GL/mesa/src/drv/mga/mgapipeline.h
index d42abddfe..dc0179227 100644
--- a/xc/lib/GL/mesa/src/drv/mga/mgapipeline.h
+++ b/xc/lib/GL/mesa/src/drv/mga/mgapipeline.h
@@ -12,4 +12,7 @@ extern GLboolean mgaDDBuildPrecalcPipeline( GLcontext *ctx );
 extern void mgaDDFastPath( struct vertex_buffer *VB );
 extern void mgaDDFastPathInit( void );
 
+extern void mgaDDEltPath( struct vertex_buffer *VB );
+extern void mgaDDEltPathInit( void );
+
 #endif
diff --git a/xc/lib/GL/mesa/src/drv/mga/mgarender.c b/xc/lib/GL/mesa/src/drv/mga/mgarender.c
deleted file mode 100644
index 6f406db2d..000000000
--- a/xc/lib/GL/mesa/src/drv/mga/mgarender.c
+++ /dev/null
@@ -1,334 +0,0 @@
-#include <stdio.h>
-#include "mgadd.h"
-#include "mgavb.h"
-#include "mgadma.h"
-#include "mgalib.h"
-#include "mgatris.h"
-#include "mgastate.h"
-#include "xsmesaP.h"
-#include "enums.h"
-
-#define POINT(x) mga_draw_point(&gWin[x], psize)
-#define LINE(x,y) mga_draw_line(&gWin[x], &gWin[y], lwidth)
-#define TRI(x,y,z) mga_draw_triangle(&gWin[x], &gWin[y], &gWin[z])
-
-
-/* Direct, and no clipping required.  I haven't written the clip funcs
- * yet, so this is only useful for the fast path, which does its own
- * clipping.
- */
-#define RENDER_POINTS( start, count )		\
-do {						\
-   GLuint e;					\
-   for(e=start;e<=count;e++)			\
-     POINT(elt[e]);				\
-} while (0)
-
-#define RENDER_LINE( i1, i )			\
-do {						\
-   GLuint e1 = elt[i1], e = elt[i];		\
-   LINE( e1, e );				\
-} while (0)
-
-
-#define RENDER_TRI( i2, i1, i, pv, parity)		\
-do {							\
-{  GLuint e2 = elt[i2], e1 = elt[i1], e = elt[i];	\
-  if (parity) {GLuint tmp = e2; e2 = e1; e1 = tmp;}	\
-  TRI(e2, e1, e);					\
-}} while (0)
-
-
-#define RENDER_QUAD( i3, i2, i1, i, pv)				\
-do {								\
-  GLuint e3 = elt[i3], e2 = elt[i2], e1 = elt[i1], e = elt[i];	\
-  TRI(e3, e2, e);						\
-  TRI(e2, e1, e);						\
-} while (0)
-
-#define LOCAL_VARS					\
-   mgaVertexPtr gWin = MGA_DRIVER_DATA(VB)->verts;	\
-   const GLuint *elt = VB->EltPtr->data;		\
-   const GLfloat lwidth = VB->ctx->Line.Width;		\
-   const GLfloat psize = VB->ctx->Point.Size;		\
-   GLcontext *ctx = VB->ctx; \
-   (void) lwidth; (void)psize; (void)ctx; (void) gWin;
-
-
-#define TAG(x) x##_mga_smooth_indirect
-#include "render_tmp.h"
-
-
-
-
-#define RENDER_POINTS( start, count )		\
-do {						\
-   GLuint e;					\
-   for(e=start;e<=count;e++)			\
-     POINT(e);					\
-} while (0)
-
-#define RENDER_LINE( i1, i )			\
-do {						\
-   LINE( i1, i );				\
-} while (0)
-
-
-#define RENDER_TRI( i2, i1, i, pv, parity)		\
-do {							\
-  GLuint e2 = i2, e1 = i1;				\
-  if (parity) {GLuint tmp = e2; e2 = e1; e1 = tmp;}	\
-  TRI(e2, e1, i);					\
-} while (0)
-
-
-#define RENDER_QUAD( i3, i2, i1, i, pv)		\
-do {						\
-  TRI(i3, i2, i);				\
-  TRI(i2, i1, i);				\
-} while (0)
-
-#define LOCAL_VARS					\
-   mgaVertexPtr gWin = MGA_DRIVER_DATA(VB)->verts;	\
-   const GLfloat lwidth = VB->ctx->Line.Width;		\
-   const GLfloat psize = VB->ctx->Point.Size;		\
-   GLcontext *ctx = VB->ctx; \
-   (void) lwidth; (void)psize; (void)ctx; (void) gWin;
-
-#define TAG(x) x##_mga_smooth_direct
-#include "render_tmp.h"
-
-
-
-
-
-/* Vertex array rendering via. the SetupDma function - no clipping required. 
- */
-#define RENDER_POINTS( start, count )		\
-do {						\
-    FatalError("Dead code in mgarender.c\n");	\
-} while (0)
-
-#define RENDER_LINE( i1, i )			\
-do {						\
-    FatalError("Dead code in mgarender.c\n");	\
-} while (0)
-
-
-#define RENDER_TRI( i2, i1, i, pv, parity)		\
-do {							\
-   GLuint e2 = elt[i2], e1 = elt[i1], e = elt[i];	\
-   if (/*parity*/0) {GLuint tmp = e2; e2 = e1; e1 = tmp;}	\
-   TRI(e2, e1, e);		\
-} while (0)
-
-#define RENDER_QUAD( i3, i2, i1, i, pv)				\
-do {								\
-  GLuint e3 = elt[i3], e2 = elt[i2], e1 = elt[i1], e = elt[i];	\
-  TRI(e3, e2, e);			\
-  TRI(e2, e1, e);			\
-} while (0)
-
-#define LOCAL_VARS					\
-   mgaVertexBufferPtr mgaVB = MGA_DRIVER_DATA(VB);	\
-   mgaUI32 phys = mgaVB->vert_phys_start;		\
-   const GLuint *elt = VB->EltPtr->data;		\
-   (void) phys; (void) elt;
-
-#define PRESERVE_VB_DEFS
-#undef TRI
-#define VERT_ADDR_10( phys, elt ) ( phys + elt * 48 )
-#define TRI( ee0, ee1, ee2 )				\
-{							\
-   mgaVB->elt_buf[0] = VERT_ADDR_10(phys, ee0);		\
-   mgaVB->elt_buf[1] = VERT_ADDR_10(phys, ee1);		\
-   mgaVB->elt_buf[2] = VERT_ADDR_10(phys, ee2);		\
-   mgaVB->elt_buf += 3;					\
-}
-#define TAG(x) x##_mga_elt_10
-#include "render_tmp.h"
-
-
-#define TAG(x) x##_mga_elt_8
-#undef TRI
-#define VERT_ADDR_8( phys, elt ) ( phys + elt * 32 )
-#define TRI( ee0, ee1, ee2 )			\
-{						\
-   mgaVB->elt_buf[0] = VERT_ADDR_8(phys, ee0);	\
-   mgaVB->elt_buf[1] = VERT_ADDR_8(phys, ee1);	\
-   mgaVB->elt_buf[2] = VERT_ADDR_8(phys, ee2);	\
-   mgaVB->elt_buf += 3;				\
-}
-#include "render_tmp.h"
-
-
-
-
-/* Currently only used on the fast path - need a set of render funcs
- * for clipped primitives.
- */
-void mgaDDRenderElementsDirect( struct vertex_buffer *VB )
-{
-   mgaVertexBufferPtr mgaVB = MGA_DRIVER_DATA(VB);
-   GLcontext *ctx = VB->ctx;
-   mgaContextPtr mmesa = MGA_CONTEXT( ctx );
-   GLenum prim = ctx->CVA.elt_mode;
-   GLuint nr = VB->EltPtr->count;
-   render_func func = render_tab_mga_smooth_indirect[prim];
-   GLuint p = 0;
-   mgaUI32 *start = mgaVB->elt_buf;
-   
-   struct vertex_buffer *saved_vb = ctx->VB;
-
-   if (mmesa->new_state)
-      mgaDDUpdateHwState( ctx );
-
-   /* Are we using Setup DMA?
-    */
-   if (start) {
-      switch (MGA_CONTEXT(ctx)->vertsize) {
-      case 10:
-	 func = render_tab_mga_elt_10[prim];
-	 break;
-      case 8:
-	 func = render_tab_mga_elt_8[prim];
-	 break;
-      default:
-      }
-   }
-
-   ctx->VB = VB;
-
-   do {
-      func( VB, 0, nr, 0 );
-   } while (ctx->Driver.MultipassFunc &&
-	    ctx->Driver.MultipassFunc( VB, ++p ));
-
-   ctx->VB = saved_vb;
-
-
-   if (start && nr) {
-      if (0)
-	 fprintf(stderr, "%d/%d elts\n", nr, mgaVB->elt_buf-start);
-      mgaSetupDma( start, mgaVB->elt_buf - start );
-      mgaVB->elt_buf = mgaVB->vert_buf = 0;
-   }
-}
-
-
-void mgaDDRenderElementsImmediate( struct vertex_buffer *VB )
-{
-   GLcontext *ctx = VB->ctx;
-   mgaContextPtr mmesa = MGA_CONTEXT( ctx );
-   GLuint nr = VB->EltPtr->count;
-   render_func *tab = render_tab_mga_smooth_indirect;
-   GLuint parity = VB->Parity;
-   GLuint p = 0;
-   GLuint i, next;
-
-   if (mmesa->new_state)
-      mgaDDUpdateHwState( ctx );
-
-   do {
-      for ( i= VB->CopyStart ; i < nr ; parity = 0, i = next ) 
-      {
-	 GLenum prim = VB->Primitive[i];
-	 next = VB->NextPrimitive[i];
-	 tab[prim]( VB, i, next, parity );
-      }
-   } while (ctx->Driver.MultipassFunc &&
-	    ctx->Driver.MultipassFunc( VB, ++p ));
-}
-
-
-void mgaDDRenderDirect( struct vertex_buffer *VB )
-{
-   GLcontext *ctx = VB->ctx;
-   mgaContextPtr mmesa = MGA_CONTEXT( ctx );
-   GLuint nr = VB->Count;
-   render_func *tab = render_tab_mga_smooth_direct;
-   GLuint parity = VB->Parity;
-   GLuint p = 0;
-   GLuint i, next;
-
-   if (mmesa->new_state)
-      mgaDDUpdateHwState( ctx );
-
-   do {
-      for ( i= VB->CopyStart ; i < nr ; parity = 0, i = next ) 
-      {
-	 GLenum prim = VB->Primitive[i];
-	 next = VB->NextPrimitive[i];
-	 tab[prim]( VB, i, next, parity );
-      }
-   } while (ctx->Driver.MultipassFunc &&
-	    ctx->Driver.MultipassFunc( VB, ++p ));
-}
-
-
-static void optimized_render_vb_triangle_mga_smooth_indirect(struct vertex_buffer *VB,
-							    GLuint start,
-							    GLuint count,
-							    GLuint parity)
-{
-   mgaVertexPtr gWin = MGA_DRIVER_DATA(VB)->verts;
-   const GLuint *elt = VB->EltPtr->data;
-   GLcontext *ctx = VB->ctx;
-   mgaContextPtr mmesa = MGA_CONTEXT( ctx );
-
-   mgaUI32 vertsize = mmesa->vertsize;
-
-   /* We know how many dwords we need so we can allocate all of it
-    * in one call */
-   mgaUI32 *wv = mgaAllocSecondaryBuffer( count * vertsize );
-
-   int vertex;
-
-   (void)ctx; (void) gWin; (void) parity;
-
-
-   for(vertex=start+2;vertex<count;vertex+=3){
-      GLuint e2=elt[vertex-2],e1=elt[vertex-1],e=elt[vertex];
-      mgaUI32 *src;
-
-      int j;
-
-      if(parity){
-         GLuint tmp=e2;
-         e2=e1;
-	 e1=tmp;
-      }
-
-      src=&gWin[e2].ui[0];
-      
-      for(j=vertsize;j;j--){
-	 *wv++=*src++;
-      }
-      
-      src=&gWin[e1].ui[0];
-      for(j=vertsize;j;j--){
-	 *wv++=*src++;
-      }
-      
-      src=&gWin[e].ui[0];
-      for(j=vertsize;j;j--){
-	 *wv++=*src++;
-      }
-   }
-   
-}
-
-
-void mgaDDRenderInit()
-{
-
-   render_init_mga_smooth_indirect();
-   render_init_mga_smooth_direct();
-   render_init_mga_elt_10();
-   render_init_mga_elt_8();
-   
-   render_tab_mga_smooth_indirect[GL_TRIANGLES] =
-      optimized_render_vb_triangle_mga_smooth_indirect; 
-}
-
-
diff --git a/xc/lib/GL/mesa/src/drv/mga/mgarender.h b/xc/lib/GL/mesa/src/drv/mga/mgarender.h
deleted file mode 100644
index c772ac459..000000000
--- a/xc/lib/GL/mesa/src/drv/mga/mgarender.h
+++ /dev/null
@@ -1,10 +0,0 @@
-
-#ifndef _MGA_RENDER_H
-#define _MGA_RENDER_H
-
-extern void mgaDDRenderElementsDirect( struct vertex_buffer *VB );
-extern void mgaDDRenderElementsImmediate( struct vertex_buffer *VB );
-extern void mgaDDRenderDirect( struct vertex_buffer *VB );
-extern void mgaDDRenderInit();
-
-#endif
diff --git a/xc/lib/GL/mesa/src/drv/mga/mgastate.c b/xc/lib/GL/mesa/src/drv/mga/mgastate.c
index ff9063e0b..7a3b82930 100644
--- a/xc/lib/GL/mesa/src/drv/mga/mgastate.c
+++ b/xc/lib/GL/mesa/src/drv/mga/mgastate.c
@@ -16,7 +16,7 @@
 #include "mgavb.h"
 #include "mgatris.h"
 #include "mgaregs.h"
-#include "mga_drm_public.h"
+#include "mgabuffers.h"
 
 static void mgaUpdateZMode(const GLcontext *ctx)
 {
@@ -60,17 +60,20 @@ static void mgaUpdateZMode(const GLcontext *ctx)
 
 static void mgaDDAlphaFunc(GLcontext *ctx, GLenum func, GLclampf ref)
 {
+   FLUSH_BATCH( MGA_CONTEXT(ctx) );
    MGA_CONTEXT(ctx)->new_state |= MGA_NEW_ALPHA;
 }
 
 
 static void mgaDDBlendEquation(GLcontext *ctx, GLenum mode) 
 {
+   FLUSH_BATCH( MGA_CONTEXT(ctx) );
    MGA_CONTEXT(ctx)->new_state |= MGA_NEW_ALPHA;
 }
 
 static void mgaDDBlendFunc(GLcontext *ctx, GLenum sfactor, GLenum dfactor)
 {
+   FLUSH_BATCH( MGA_CONTEXT(ctx) );
    MGA_CONTEXT(ctx)->new_state |= MGA_NEW_ALPHA;
 }
 
@@ -78,6 +81,7 @@ static void mgaDDBlendFuncSeparate( GLcontext *ctx, GLenum sfactorRGB,
 				    GLenum dfactorRGB, GLenum sfactorA,
 				    GLenum dfactorA )
 {
+   FLUSH_BATCH( MGA_CONTEXT(ctx) );
    MGA_CONTEXT(ctx)->new_state |= MGA_NEW_ALPHA;
 }
 
@@ -87,6 +91,7 @@ static void mgaDDLightModelfv(GLcontext *ctx, GLenum pname,
 			      const GLfloat *param)
 {
    if (pname == GL_LIGHT_MODEL_COLOR_CONTROL) {
+      FLUSH_BATCH( MGA_CONTEXT(ctx) );
       MGA_CONTEXT(ctx)->new_state |= MGA_NEW_TEXTURE;
    }
 }
@@ -95,6 +100,7 @@ static void mgaDDLightModelfv(GLcontext *ctx, GLenum pname,
 static void mgaDDShadeModel(GLcontext *ctx, GLenum mode)
 {
    if (1) {
+      FLUSH_BATCH( MGA_CONTEXT(ctx) );
       MGA_CONTEXT(ctx)->new_state |= MGA_NEW_TEXTURE; 
       mgaMsg(8, "mgaDDShadeModel: %x\n", mode);
    }
@@ -103,11 +109,13 @@ static void mgaDDShadeModel(GLcontext *ctx, GLenum mode)
 
 static void mgaDDDepthFunc(GLcontext *ctx, GLenum func)
 {
+   FLUSH_BATCH( MGA_CONTEXT(ctx) );
    MGA_CONTEXT(ctx)->new_state |= MGA_NEW_DEPTH;
 }
 
 static void mgaDDDepthMask(GLcontext *ctx, GLboolean flag)
 {
+   FLUSH_BATCH( MGA_CONTEXT(ctx) );
    MGA_CONTEXT(ctx)->new_state |= MGA_NEW_DEPTH;
 }
 
@@ -134,6 +142,7 @@ static void mgaUpdateFogAttrib( GLcontext *ctx )
 
 static void mgaDDFogfv(GLcontext *ctx, GLenum pname, const GLfloat *param)
 {
+   FLUSH_BATCH( MGA_CONTEXT(ctx) );
    MGA_CONTEXT(ctx)->new_state |= MGA_NEW_FOG;
 }
 
@@ -148,10 +157,11 @@ static void mgaDDFogfv(GLcontext *ctx, GLenum pname, const GLfloat *param)
 static void mgaUpdateAlphaMode(GLcontext *ctx)
 {
    mgaContextPtr mmesa = MGA_CONTEXT( ctx );
+   mgaScreenPrivate *mgaScreen = mmesa->mgaScreen;
    int a = 0;
 
    /* determine source of alpha for blending and testing */
-   if ( !ctx->Texture.Enabled || (mmesa->Fallback & MGA_FALLBACK_TEXTURE)) 
+   if ( !ctx->Texture.Enabled ) 
       a |= AC_alphasel_diffused;
    else {
       switch (ctx->Texture.Unit[0].EnvMode) {
@@ -168,7 +178,7 @@ static void mgaUpdateAlphaMode(GLcontext *ctx)
    }
 
 
-   /* alpha test control - disabled by default.
+   /* alpha test control.
     */
    if (ctx->Color.AlphaEnabled) {
       GLubyte ref = ctx->Color.AlphaRef;
@@ -219,13 +229,13 @@ static void mgaUpdateAlphaMode(GLcontext *ctx)
       case GL_ONE_MINUS_SRC_ALPHA:
 	 a |= AC_src_om_src_alpha; break;
       case GL_DST_ALPHA:
-	 if (0) /*(mgaScreen->Attrib & MGA_PF_HASALPHA)*/
+	 if (mgaScreen->Attrib & MGA_PF_HASALPHA)
 	    a |= AC_src_dst_alpha;
 	 else
 	    a |= AC_src_one;
 	 break;
       case GL_ONE_MINUS_DST_ALPHA:
-	 if (0) /*(mgaScreen->Attrib & MGA_PF_HASALPHA)*/
+	 if (mgaScreen->Attrib & MGA_PF_HASALPHA)
 	    a |= AC_src_om_dst_alpha; 
 	 else 
 	    a |= AC_src_zero;
@@ -250,13 +260,13 @@ static void mgaUpdateAlphaMode(GLcontext *ctx)
       case GL_ONE_MINUS_SRC_COLOR:
 	 a |= AC_dst_om_src_color; break;
       case GL_DST_ALPHA:
-	 if (0) /*(mgaDB->Attrib & MGA_PF_HASALPHA)*/
+	 if (mgaScreen->Attrib & MGA_PF_HASALPHA)
 	    a |= AC_dst_dst_alpha;
 	 else
 	    a |= AC_dst_one;
 	 break;
       case GL_ONE_MINUS_DST_ALPHA:
-	 if (0) /*(mgaScreen->Attrib & MGA_PF_HASALPHA)*/
+	 if (mgaScreen->Attrib & MGA_PF_HASALPHA)
 	    a |= AC_dst_om_dst_alpha;
 	 else
 	    a |= AC_dst_zero;
@@ -282,50 +292,44 @@ static void mgaUpdateAlphaMode(GLcontext *ctx)
  * Hardware clipping
  */
 
-static void mgaUpdateClipping(const GLcontext *ctx)
+void mgaUpdateClipping(const GLcontext *ctx)
 {
-#if 0
-   mgaContextPtr mmesa = MGA_CONTEXT( ctx );
-   __DRIdrawablePrivate *dPriv = mmesa->driDrawable;	
-   int x1,x2,y1,y2;
-
-   if ( ctx->Scissor.Enabled) {
-      x1 = ctx->Scissor.X;
-      x2 = ctx->Scissor.X + ctx->Scissor.Width - 1;
-      y1 = dPriv->Height - ctx->Scissor.Y - ctx->Scissor.Height;
-      y2 = dPriv->Height - ctx->Scissor.Y - 1;
-   } else {
-      x1 = 0;
-      y1 = 0;
-      x2 = mgaDB->Width-1;
-      y2 = mgaDB->Height-1;
-   }
-  
-   if (x1 < 0) x1 = 0;
-   if (y1 < 0) y1 = 0;
-   if (x2 >= mgaDB->Width) x2 = mgaDB->Width-1;
-   if (y2 >= mgaDB->Height) y2 = mgaDB->Height-1;
-
-   if (x1 > x2 || y1 > y2) {
-      x1 = 0; x2 = 0;
-      y2 = 0; y1 = 1;
-   }
-
-
-   mmesa->Setup[MGA_CTXREG_CXBNDRY] = (MGA_FIELD(CXB_cxright,x2) | 
-				      MGA_FIELD(CXB_cxleft,x1));
-   mmesa->Setup[MGA_CTXREG_YTOP] = y1*mgaDB->Pitch;
-   mmesa->Setup[MGA_CTXREG_YBOT] = y2*mgaDB->Pitch;
-
+   mgaContextPtr mmesa = MGA_CONTEXT(ctx);
 
-   mmesa->dirty |= MGA_UPLOAD_CTX;
-#endif
+   if (mmesa->driDrawable)
+   {
+      int x1 = mmesa->driDrawable->x + ctx->Scissor.X;
+      int y1 = mmesa->driDrawable->y + mmesa->driDrawable->h - (ctx->Scissor.Y+
+								ctx->Scissor.Height);
+      int x2 = mmesa->driDrawable->x + ctx->Scissor.X+ctx->Scissor.Width;
+      int y2 = mmesa->driDrawable->y + mmesa->driDrawable->h - ctx->Scissor.Y;
+
+      if (x1 < 0) x1 = 0;
+      if (y1 < 0) y1 = 0;
+      if (x2 < 0) x2 = 0;
+      if (y2 < 0) y2 = 0;
+
+      mmesa->scissor_rect.x1 = x1;
+      mmesa->scissor_rect.y1 = y1;
+      mmesa->scissor_rect.x2 = x2;
+      mmesa->scissor_rect.y2 = y2;
+
+      if (MGA_DEBUG&DEBUG_VERBOSE_2D)
+	 fprintf(stderr, "SET SCISSOR %d,%d-%d,%d\n", 
+		 mmesa->scissor_rect.x1,
+		 mmesa->scissor_rect.y1,
+		 mmesa->scissor_rect.x2,
+		 mmesa->scissor_rect.y2);
+
+      mmesa->dirty |= MGA_UPLOAD_CLIPRECTS;
+   }
 }
 
 
 static void mgaDDScissor( GLcontext *ctx, GLint x, GLint y, 
 			  GLsizei w, GLsizei h )
 {
+   FLUSH_BATCH( MGA_CONTEXT(ctx) );
    MGA_CONTEXT(ctx)->new_state |= MGA_NEW_CLIP;
 }
 
@@ -339,35 +343,6 @@ static void mgaDDDither(GLcontext *ctx, GLboolean enable)
 }
 
 
-static GLboolean mgaDDSetBuffer(GLcontext *ctx, GLenum mode )
-{
-   mgaContextPtr mmesa = MGA_CONTEXT(ctx);
-
-   mmesa->Fallback &= ~MGA_FALLBACK_BUFFER;
-
-   if (mode == GL_FRONT_LEFT) 
-   {
-      mmesa->drawOffset = mmesa->mgaScreen->fbOffset;
-      mmesa->Setup[MGA_CTXREG_DSTORG] = mmesa->mgaScreen->fbOffset;
-      mmesa->dirty |= MGA_UPLOAD_CTX;
-      mgaXMesaSetFrontClipRects( mmesa );
-      return GL_TRUE;
-   } 
-   else if (mode == GL_BACK_LEFT) 
-   {
-      mmesa->drawOffset = mmesa->mgaScreen->backOffset;
-      mmesa->Setup[MGA_CTXREG_DSTORG] = mmesa->mgaScreen->backOffset;
-      mmesa->dirty |= MGA_UPLOAD_CTX;
-      mgaXMesaSetBackClipRects( mmesa );
-      return GL_TRUE;
-   }
-   else 
-   {
-      mmesa->Fallback |= MGA_FALLBACK_BUFFER;
-      return GL_FALSE;
-   }
-}
-
 
 
 static void mgaDDSetColor(GLcontext *ctx, 
@@ -376,7 +351,7 @@ static void mgaDDSetColor(GLcontext *ctx,
 {
    mgaContextPtr mmesa = MGA_CONTEXT(ctx);
 
-   mmesa->MonoColor = mgaPackColor( mmesa->mgaScreen->fbFormat,
+   mmesa->MonoColor = mgaPackColor( mmesa->mgaScreen->Attrib,
 				      r, g, b, a );
 }
 
@@ -387,7 +362,7 @@ static void mgaDDClearColor(GLcontext *ctx,
 {
    mgaContextPtr mmesa = MGA_CONTEXT(ctx);
 
-   mmesa->ClearColor = mgaPackColor( mmesa->mgaScreen->fbFormat,
+   mmesa->ClearColor = mgaPackColor( mmesa->mgaScreen->Attrib,
 				      r, g, b, a );
 }
 
@@ -426,6 +401,7 @@ static void mgaUpdateCull( GLcontext *ctx )
 
 static void mgaDDCullFaceFrontFace(GLcontext *ctx, GLenum mode)
 {
+   FLUSH_BATCH( MGA_CONTEXT(ctx) );
    MGA_CONTEXT(ctx)->new_state |= MGA_NEW_CULL;
 }
 
@@ -461,6 +437,7 @@ static GLboolean mgaDDColorMask(GLcontext *ctx,
    }
 
    if (mmesa->Setup[MGA_CTXREG_PLNWT] != mask) {
+      FLUSH_BATCH( MGA_CONTEXT(ctx) );
       mmesa->Setup[MGA_CTXREG_PLNWT] = mask;      
       MGA_CONTEXT(ctx)->new_state |= MGA_NEW_MASK; 
       mmesa->dirty |= MGA_UPLOAD_CTX;
@@ -470,6 +447,83 @@ static GLboolean mgaDDColorMask(GLcontext *ctx,
 }
 
 /* =============================================================
+ * Polygon stipple 
+ * 
+ * The mga supports a subset of possible 4x4 stipples natively, GL
+ * wants 32x32.  Fortunately stipple is usually a repeating pattern.
+ */
+static int mgaStipples[16] = {
+   0xffff,
+   0xa5a5,
+   0x5a5a,
+   0xa0a0,
+   0x5050,
+   0x0a0a,
+   0x0505,
+   0x8020,
+   0x0401,
+   0x1040,
+   0x0208,
+   0x0802,
+   0x4010,
+   0x0104,
+   0x2080,
+   0x0000
+};
+
+static void mgaDDPolygonStipple( GLcontext *ctx, const GLubyte *mask )
+{
+   mgaContextPtr mmesa = MGA_CONTEXT(ctx);
+   const GLubyte *m = mask;
+   GLubyte p[4];
+   int i,j,k;
+   int active = (ctx->Polygon.StippleFlag && ctx->PB->primitive == GL_POLYGON);
+   GLuint stipple;
+
+   FLUSH_BATCH(mmesa);
+   ctx->Driver.TriangleCaps |= DD_TRI_STIPPLE;
+
+   if (active) {
+      mmesa->dirty |= MGA_UPLOAD_CTX;
+      mmesa->Setup[MGA_CTXREG_DWGCTL] &= ~(0xf<<20);
+   }
+
+   p[0] = mask[0] & 0xf; p[0] |= p[0] << 4;
+   p[1] = mask[4] & 0xf; p[1] |= p[1] << 4;
+   p[2] = mask[8] & 0xf; p[2] |= p[2] << 4;
+   p[3] = mask[12] & 0xf; p[3] |= p[3] << 4;
+
+   for (k = 0 ; k < 8 ; k++)
+      for (j = 0 ; j < 4; j++) 
+	 for (i = 0 ; i < 4 ; i++) 
+	    if (*m++ != p[j]) {
+	       ctx->Driver.TriangleCaps &= ~DD_TRI_STIPPLE;
+	       return;
+	    }
+
+   stipple = ( ((p[0] & 0xf) << 0) |
+	       ((p[1] & 0xf) << 4) |
+	       ((p[2] & 0xf) << 8) |
+	       ((p[3] & 0xf) << 12) );   
+
+   for (i = 0 ; i < 16 ; i++)
+      if (mgaStipples[i] == stipple) {
+	 mmesa->poly_stipple = i<<20;
+	 break;
+      }
+
+   if (i == 16) {
+      ctx->Driver.TriangleCaps &= ~DD_TRI_STIPPLE;
+      return;
+   }
+   
+   if (active) {
+      mmesa->Setup[MGA_CTXREG_DWGCTL] &= ~(0xf<<20);
+      mmesa->Setup[MGA_CTXREG_DWGCTL] |= mmesa->poly_stipple;
+   }
+}
+
+/* =============================================================
  */
 
 
@@ -477,15 +531,16 @@ static GLboolean mgaDDColorMask(GLcontext *ctx,
 
 static void mgaDDPrintDirty( const char *msg, GLuint state )
 {
-   fprintf(stderr, "%s (0x%x): %s%s%s%s%s%s\n",	   
+   fprintf(stderr, "%s (0x%x): %s%s%s%s%s%s%s\n",
 	   msg,
 	   (unsigned int) state,
-	   (state & MGA_REQUIRE_QUIESCENT) ? "req-quiescent, " : "",
+	   (state & MGA_WAIT_AGE) ? "wait-age, " : "",
 	   (state & MGA_UPLOAD_TEX0IMAGE)  ? "upload-tex0-img, " : "",
 	   (state & MGA_UPLOAD_TEX1IMAGE)  ? "upload-tex1-img, " : "",
 	   (state & MGA_UPLOAD_CTX)        ? "upload-ctx, " : "",
 	   (state & MGA_UPLOAD_TEX0)       ? "upload-tex0, " : "",
-	   (state & MGA_UPLOAD_TEX1)       ? "upload-tex1, " : ""
+	   (state & MGA_UPLOAD_TEX1)       ? "upload-tex1, " : "",
+	   (state & MGA_UPLOAD_PIPE)       ? "upload-pipe, " : ""
 	   );
 }
 
@@ -494,7 +549,7 @@ static void mgaDDPrintDirty( const char *msg, GLuint state )
  */
 void mgaEmitHwStateLocked( mgaContextPtr mmesa )
 {
-   if (MGA_DEBUG & MGA_DEBUG_VERBOSE_MSG)
+   if (MGA_DEBUG & DEBUG_VERBOSE_MSG)
       mgaDDPrintDirty( "mgaEmitHwStateLocked", mmesa->dirty );
 
    if ((mmesa->dirty & MGA_UPLOAD_TEX0IMAGE) && mmesa->CurrentTexObj[0])
@@ -503,27 +558,34 @@ void mgaEmitHwStateLocked( mgaContextPtr mmesa )
    if ((mmesa->dirty & MGA_UPLOAD_TEX1IMAGE) && mmesa->CurrentTexObj[1])
       mgaUploadTexImages(mmesa, mmesa->CurrentTexObj[1]);
   
-   if (mmesa->dirty & MGA_UPLOAD_CTX) {
+   if (mmesa->dirty & MGA_UPLOAD_CTX) 
       memcpy( mmesa->sarea->ContextState, 
 	      mmesa->Setup,
 	      sizeof(mmesa->Setup));
 
-      if (mmesa->CurrentTexObj[0])
+   if ((mmesa->dirty & MGA_UPLOAD_TEX0) && mmesa->CurrentTexObj[0])
 	 memcpy(mmesa->sarea->TexState[0],
 		mmesa->CurrentTexObj[0]->Setup,
 		sizeof(mmesa->sarea->TexState[0]));
 
-      if (mmesa->CurrentTexObj[1])
+   if ((mmesa->dirty & MGA_UPLOAD_TEX1) && mmesa->CurrentTexObj[1])
 	 memcpy(mmesa->sarea->TexState[1],
 		mmesa->CurrentTexObj[1]->Setup,
 		sizeof(mmesa->sarea->TexState[1]));
-   }
 
-   if (mmesa->dirty & MGA_UPLOAD_PIPE) 
-      mmesa->sarea->WarpPipe = mmesa->setupindex & MGA_WARP_T2GZSAF;
-      
+   mmesa->sarea->WarpPipe = ((mmesa->setupindex & MGA_WARP_T2GZSAF) | 
+			     MGA_ALPHA_BIT | MGA_SPEC_BIT | MGA_FOG_BIT);
+
+
    mmesa->sarea->dirty |= mmesa->dirty;
-   mmesa->dirty = 0;
+
+#if 0
+   mgaPrintSetupFlags("warp pipe", mmesa->sarea->WarpPipe);
+   fprintf(stderr, "in mgaEmitHwStateLocked: dirty now %x\n",
+	   mmesa->sarea->dirty);
+#endif
+
+   mmesa->dirty &= (MGA_UPLOAD_CLIPRECTS|MGA_WAIT_AGE);
 }
 
 
@@ -533,29 +595,50 @@ void mgaEmitHwStateLocked( mgaContextPtr mmesa )
 
 static void mgaDDEnable(GLcontext *ctx, GLenum cap, GLboolean state)
 {
+   mgaContextPtr mmesa = MGA_CONTEXT( ctx );
+
    switch(cap) {
    case GL_ALPHA_TEST:
-      MGA_CONTEXT(ctx)->new_state |= MGA_NEW_ALPHA;
+      FLUSH_BATCH( mmesa );
+      mmesa->new_state |= MGA_NEW_ALPHA;
       break;
    case GL_BLEND:
-      MGA_CONTEXT(ctx)->new_state |= MGA_NEW_ALPHA;
+      FLUSH_BATCH( mmesa );
+      mmesa->new_state |= MGA_NEW_ALPHA;
       break;
    case GL_DEPTH_TEST:
-      MGA_CONTEXT(ctx)->new_state |= MGA_NEW_DEPTH;
+      FLUSH_BATCH( mmesa );
+      mmesa->new_state |= MGA_NEW_DEPTH;
       break;
    case GL_SCISSOR_TEST:
-      MGA_CONTEXT(ctx)->new_state |= MGA_NEW_CLIP;
+      FLUSH_BATCH( mmesa );
+      mmesa->scissor = state;
+      mmesa->new_state |= MGA_NEW_CLIP;
       break;
    case GL_FOG:
-      MGA_CONTEXT(ctx)->new_state |= MGA_NEW_FOG;
+      FLUSH_BATCH( mmesa );
+      mmesa->new_state |= MGA_NEW_FOG;
       break;
    case GL_CULL_FACE:
-      MGA_CONTEXT(ctx)->new_state |= MGA_NEW_CULL;
+      FLUSH_BATCH( mmesa );
+      mmesa->new_state |= MGA_NEW_CULL;
       break;
    case GL_TEXTURE_1D:
    case GL_TEXTURE_2D:
    case GL_TEXTURE_3D:
-      MGA_CONTEXT(ctx)->new_state |= MGA_NEW_TEXTURE;
+      FLUSH_BATCH( mmesa );
+      mmesa->new_state |= MGA_NEW_TEXTURE;
+      break;
+   case GL_POLYGON_STIPPLE:
+      if ((ctx->Driver.TriangleCaps & DD_TRI_STIPPLE) &&
+	  ctx->PB->primitive == GL_POLYGON) 
+      {
+	 FLUSH_BATCH(mmesa);
+	 mmesa->dirty |= MGA_UPLOAD_CTX;
+	 mmesa->Setup[MGA_CTXREG_DWGCTL] &= ~(0xf<<20);
+	 if (state)
+	    mmesa->Setup[MGA_CTXREG_DWGCTL] |= mmesa->poly_stipple;
+      }
       break;
    default:
       ; 
@@ -581,6 +664,7 @@ static void mgaWarpUpdateState( GLcontext *ctx )
    {
       mmesa->warp_pipe = index;
       mmesa->new_state |= MGA_NEW_WARP;
+      mmesa->dirty |= MGA_UPLOAD_PIPE;
    }
 }
 
@@ -611,21 +695,15 @@ void mgaDDUpdateHwState( GLcontext *ctx )
 
    if (new_state) 
    {
-      mmesa->new_state = 0;
+      FLUSH_BATCH( mmesa );
 
-      /* Emit any vertices for the current state.  This will also
-       * push the current state into the sarea.
-       */
-/*        mgaFlushVertices( mmesa ); */
+      mmesa->new_state = 0;
 
       if (MESA_VERBOSE&VERBOSE_DRIVER)
 	 mgaDDPrintState("UpdateHwState", new_state);
 
       if (new_state & MGA_NEW_DEPTH)
-      {
 	 mgaUpdateZMode(ctx);
-	 mgaDDInitDepthFuncs(ctx);
-      }
 
       if (new_state & MGA_NEW_ALPHA)
 	 mgaUpdateAlphaMode(ctx);
@@ -641,8 +719,6 @@ void mgaDDUpdateHwState( GLcontext *ctx )
 
       if (new_state & (MGA_NEW_WARP|MGA_NEW_TEXTURE))
 	 mgaUpdateTextureState(ctx);
-
-      mmesa->new_state = 0; /* tex uploads scribble newstate */
    }
 }
 
@@ -654,8 +730,17 @@ void mgaDDUpdateHwState( GLcontext *ctx )
 
 void mgaDDReducedPrimitiveChange( GLcontext *ctx, GLenum prim )
 {
-   mgaFlushVertices( MGA_CONTEXT(ctx) );
-   mgaUpdateCull(ctx);
+   mgaContextPtr mmesa = MGA_CONTEXT(ctx);
+
+   FLUSH_BATCH( mmesa );
+   mgaUpdateCull(ctx);   
+
+   if (ctx->Polygon.StippleFlag && (ctx->Driver.TriangleCaps & DD_TRI_STIPPLE))
+   {
+      mmesa->Setup[MGA_CTXREG_DWGCTL] &= ~(0xf<<20);
+      if (ctx->PB->primitive == GL_POLYGON)
+	 mmesa->Setup[MGA_CTXREG_DWGCTL] |= mmesa->poly_stipple;
+   }
 }
 
 
@@ -677,10 +762,16 @@ void mgaDDUpdateState( GLcontext *ctx )
 
    /* Have to do this here to detect texture fallbacks in time:
     */
-   if (MGA_CONTEXT(ctx)->new_state & MGA_NEW_TEXTURE)
+   if (mmesa->new_state & MGA_NEW_TEXTURE)
       mgaDDUpdateHwState( ctx );
 
-   if (!mmesa->Fallback || mgaglx.noFallback) {
+   if (0) fprintf(stderr, "fallback %x indirect %x\n", mmesa->Fallback, 
+		  mmesa->IndirectTriangles);
+   
+   if (!mmesa->Fallback) {
+      ctx->IndirectTriangles &= ~DD_SW_RASTERIZE;
+      ctx->IndirectTriangles |= mmesa->IndirectTriangles;
+
       ctx->Driver.PointsFunc=mmesa->PointsFunc;
       ctx->Driver.LineFunc=mmesa->LineFunc;
       ctx->Driver.TriangleFunc=mmesa->TriangleFunc;
@@ -689,6 +780,60 @@ void mgaDDUpdateState( GLcontext *ctx )
 }
 
 
+
+void mgaInitState( mgaContextPtr mmesa )
+{
+   mgaScreenPrivate *mgaScreen = mmesa->mgaScreen;
+   GLcontext *ctx = mmesa->glCtx;
+
+   if (ctx->Color.DriverDrawBuffer == GL_BACK_LEFT) {
+      mmesa->draw_buffer = MGA_BACK;
+      mmesa->read_buffer = MGA_BACK;
+      mmesa->drawOffset = mmesa->mgaScreen->backOffset;
+      mmesa->readOffset = mmesa->mgaScreen->backOffset;
+      mmesa->Setup[MGA_CTXREG_DSTORG] = mgaScreen->backOffset;
+   } else {
+      mmesa->drawOffset = mmesa->mgaScreen->frontOffset;
+      mmesa->readOffset = mmesa->mgaScreen->frontOffset;
+      mmesa->draw_buffer = MGA_FRONT;
+      mmesa->read_buffer = MGA_FRONT;
+      mmesa->Setup[MGA_CTXREG_DSTORG] = mgaScreen->frontOffset;
+   }
+
+/*     mmesa->Setup[MGA_CTXREG_MACCESS] = mgaScreen->mAccess; */
+/*     mmesa->Setup[MGA_CTXREG_DWGCTL] = ( DC_clipdis_disable |  */
+/*  				       (0xC << DC_bop_SHIFT) | */
+/*  				       DC_shftzero_enable |  */
+/*  				       DC_zmode_nozcmp |  */
+/*  				       DC_atype_zi ); */
+
+
+   mmesa->Setup[MGA_CTXREG_MACCESS] = 0x1;
+   mmesa->Setup[MGA_CTXREG_DWGCTL] = 0xc4074;
+
+
+   mmesa->Setup[MGA_CTXREG_PLNWT] = ~0;
+   mmesa->Setup[MGA_CTXREG_ALPHACTRL] = ( AC_src_one |
+					  AC_dst_zero | 
+					  AC_amode_FCOL | 
+					  AC_astipple_disable | 
+					  AC_aten_disable | 
+					  AC_atmode_noacmp |
+					  AC_alphasel_fromtex );
+
+   mmesa->Setup[MGA_CTXREG_FOGCOLOR] = 
+      MGAPACKCOLOR888((GLubyte)(ctx->Fog.Color[0]*255.0F), 
+		      (GLubyte)(ctx->Fog.Color[1]*255.0F), 
+		      (GLubyte)(ctx->Fog.Color[2]*255.0F));
+
+   mmesa->Setup[MGA_CTXREG_WFLAG] = 0;
+   mmesa->Setup[MGA_CTXREG_TDUAL0] = 0;
+   mmesa->Setup[MGA_CTXREG_TDUAL1] = 0;
+   mmesa->Setup[MGA_CTXREG_FCOL] = 0;
+   mmesa->new_state = ~0;
+}
+
+
 void mgaDDInitStateFuncs( GLcontext *ctx )
 {
    ctx->Driver.UpdateState = mgaDDUpdateState;
@@ -710,11 +855,14 @@ void mgaDDInitStateFuncs( GLcontext *ctx )
    ctx->Driver.RenderStart = mgaDDUpdateHwState; 
    ctx->Driver.RenderFinish = 0; 
 
-   ctx->Driver.SetBuffer = mgaDDSetBuffer;
+   ctx->Driver.SetDrawBuffer = mgaDDSetDrawBuffer;
+   ctx->Driver.SetReadBuffer = mgaDDSetReadBuffer;
    ctx->Driver.Color = mgaDDSetColor;
    ctx->Driver.ClearColor = mgaDDClearColor;
    ctx->Driver.Dither = mgaDDDither;
 
+   ctx->Driver.PolygonStipple = mgaDDPolygonStipple;
+
    ctx->Driver.Index = 0;
    ctx->Driver.ClearIndex = 0;
    ctx->Driver.IndexMask = 0;
diff --git a/xc/lib/GL/mesa/src/drv/mga/mgatex.c b/xc/lib/GL/mesa/src/drv/mga/mgatex.c
index be7ad6018..eafccbaa2 100644
--- a/xc/lib/GL/mesa/src/drv/mga/mgatex.c
+++ b/xc/lib/GL/mesa/src/drv/mga/mgatex.c
@@ -47,22 +47,25 @@
  * to it.
  */
 static void mgaDestroyTexObj( mgaContextPtr mmesa, mgaTextureObjectPtr t ) {
-	int	i;
 
  	if ( !t ) return;
   	  	
 	/* free the texture memory */
-	mmFreeMem( t->MemBlock );
+	if (t->MemBlock) {
+		mmFreeMem( t->MemBlock );
+		t->MemBlock = 0;
+
+		if (t->age > mmesa->dirtyAge)
+			mmesa->dirtyAge = t->age;
+	}
  
  	/* free mesa's link */   
-	t->tObj->DriverData = NULL;
+	if (t->tObj) 
+		t->tObj->DriverData = NULL;
 
 	/* see if it was the driver's current object */
-	for ( i = 0 ; i < 2 ; i++ ) {
-		if ( mmesa->CurrentTexObj[i] == t ) {
-			mmesa->CurrentTexObj[i] = NULL;
-		}  
-	}
+	if (t->bound)
+		mmesa->CurrentTexObj[t->bound - 1] = 0; 
 	
 	remove_from_list(t);
 	free( t );
@@ -73,6 +76,9 @@ static void mgaSwapOutTexObj(mgaContextPtr mmesa, mgaTextureObjectPtr t)
 	if (t->MemBlock) {
 		mmFreeMem(t->MemBlock);
 		t->MemBlock = 0;      
+
+		if (t->age > mmesa->dirtyAge)
+			mmesa->dirtyAge = t->age;
 	}
 
 	t->dirty_images = ~0;
@@ -80,22 +86,76 @@ static void mgaSwapOutTexObj(mgaContextPtr mmesa, mgaTextureObjectPtr t)
 }
 
 
-void mgaResetGlobalLRU( mgaContextPtr mmesa )
+static void mgaPrintLocalLRU( mgaContextPtr mmesa, int heap ) 
+{
+   mgaTextureObjectPtr t;
+   int sz = 1 << (mmesa->mgaScreen->logTextureGranularity[heap]);
+
+   fprintf(stderr, "\nLocal LRU, heap %d:\n", heap);
+
+   foreach( t, &(mmesa->TexObjList[heap]) ) {
+      if (!t->tObj)
+	 fprintf(stderr, "Placeholder %d at %x sz %x\n", 
+		 t->MemBlock->ofs / sz,
+		 t->MemBlock->ofs,
+		 t->MemBlock->size);      
+      else
+	 fprintf(stderr, "Texture (bound %d) at %x sz %x\n", 
+		 t->bound,
+		 t->MemBlock->ofs,
+		 t->MemBlock->size);      
+
+   }
+
+   fprintf(stderr, "\n\n");
+}
+
+static void mgaPrintGlobalLRU( mgaContextPtr mmesa, int heap )
+{
+   int i, j;
+   drm_mga_tex_region_t *list = mmesa->sarea->texList[heap];
+
+   fprintf(stderr, "\nGlobal LRU, heap %d list %p:\n", heap, list);
+
+   for (i = 0, j = MGA_NR_TEX_REGIONS ; i < MGA_NR_TEX_REGIONS ; i++) {
+      fprintf(stderr, "list[%d] age %d next %d prev %d\n",
+	      j, list[j].age, list[j].next, list[j].prev);
+      j = list[j].next;
+      if (j == MGA_NR_TEX_REGIONS) break;
+   }
+   
+   if (j != MGA_NR_TEX_REGIONS) {
+      fprintf(stderr, "Loop detected in global LRU\n\n\n");
+      for (i = 0 ; i < MGA_NR_TEX_REGIONS ; i++) {
+	      fprintf(stderr, "list[%d] age %d next %d prev %d\n",
+		      i, list[i].age, list[i].next, list[i].prev);
+      }
+   }
+   
+   fprintf(stderr, "\n\n");
+}
+
+
+static void mgaResetGlobalLRU( mgaContextPtr mmesa, GLuint heap )
 {
-   mgaTexRegion *list = mmesa->sarea->texList;
-   int sz = 1 << mmesa->mgaScreen->logTextureGranularity;
+   drm_mga_tex_region_t *list = mmesa->sarea->texList[heap];
+   int sz = 1 << mmesa->mgaScreen->logTextureGranularity[heap];
    int i;
 
+   mmesa->texAge[heap] = ++mmesa->sarea->texAge[heap];
+
+   if (0) fprintf(stderr, "mgaResetGlobalLRU %d\n", (int)heap);
+
    /* (Re)initialize the global circular LRU list.  The last element
     * in the array (MGA_NR_TEX_REGIONS) is the sentinal.  Keeping it
     * at the end of the array allows it to be addressed rationally
     * when looking up objects at a particular location in texture
     * memory.  
     */
-   for (i = 0 ; (i+1) * sz < mmesa->mgaScreen->textureSize ; i++) {
+   for (i = 0 ; (i+1) * sz <= mmesa->mgaScreen->textureSize[heap] ; i++) {
       list[i].prev = i-1;
       list[i].next = i+1;
-      list[i].age = 0;
+      list[i].age = mmesa->sarea->texAge[heap];   
    }
 
    i--;
@@ -104,30 +164,41 @@ void mgaResetGlobalLRU( mgaContextPtr mmesa )
    list[i].next = MGA_NR_TEX_REGIONS;
    list[MGA_NR_TEX_REGIONS].prev = i;
    list[MGA_NR_TEX_REGIONS].next = 0;
-   mmesa->sarea->texAge = 0;
+
 }
 
 
 static void mgaUpdateTexLRU( mgaContextPtr mmesa, mgaTextureObjectPtr t ) 
 {
    int i;
-   int logsz = mmesa->mgaScreen->logTextureGranularity;
+   int heap = t->heap;
+   int logsz = mmesa->mgaScreen->logTextureGranularity[heap];
    int start = t->MemBlock->ofs >> logsz;
    int end = (t->MemBlock->ofs + t->MemBlock->size - 1) >> logsz;
-   mgaTexRegion *list = mmesa->sarea->texList;
+   drm_mga_tex_region_t *list = mmesa->sarea->texList[heap];
    
-   mmesa->texAge = ++mmesa->sarea->texAge;
+   mmesa->texAge[heap] = ++mmesa->sarea->texAge[heap];
+
+   if (!t->MemBlock) {
+	   fprintf(stderr, "no memblock\n\n");
+	   return;
+   }
 
    /* Update our local LRU
     */
-   move_to_head( &(mmesa->TexObjList), t );
+   move_to_head( &(mmesa->TexObjList[heap]), t );
+
+
+   if (0)
+	   fprintf(stderr, "mgaUpdateTexLRU heap %d list %p\n", heap, list);
+
 
    /* Update the global LRU
     */
    for (i = start ; i <= end ; i++) {
 
       list[i].in_use = 1;
-      list[i].age = mmesa->texAge;
+      list[i].age = mmesa->texAge[heap];
 
       /* remove_from_list(i)
        */
@@ -141,8 +212,12 @@ static void mgaUpdateTexLRU( mgaContextPtr mmesa, mgaTextureObjectPtr t )
       list[(unsigned)list[MGA_NR_TEX_REGIONS].next].prev = i;
       list[MGA_NR_TEX_REGIONS].next = i;
    }
-}
 
+   if (0) {
+	   mgaPrintGlobalLRU(mmesa, t->heap);
+	   mgaPrintLocalLRU(mmesa, t->heap);
+   }
+}
 
 /* Called for every shared texture region which has increased in age
  * since we last held the lock.
@@ -151,23 +226,29 @@ static void mgaUpdateTexLRU( mgaContextPtr mmesa, mgaTextureObjectPtr t )
  * and pushes a placeholder texture onto the LRU list to represent 
  * the other client's textures.  
  */
-void mgaTexturesGone( mgaContextPtr mmesa,
-		       GLuint offset, 
-		       GLuint size,
-		       GLuint in_use ) 
+static void mgaTexturesGone( mgaContextPtr mmesa,
+			     GLuint heap,
+			     GLuint offset, 
+			     GLuint size,
+			     GLuint in_use ) 
 {
    mgaTextureObjectPtr t, tmp;
-   
-   foreach_s ( t, tmp, &mmesa->TexObjList ) {
+
+
+
+   foreach_s ( t, tmp, &(mmesa->TexObjList[heap]) ) {
 
       if (t->MemBlock->ofs >= offset + size ||
 	  t->MemBlock->ofs + t->MemBlock->size <= offset)
 	 continue;
 
+
+
+
       /* It overlaps - kick it off.  Need to hold onto the currently bound
        * objects, however.
        */
-      if (t == mmesa->CurrentTexObj[0] || t == mmesa->CurrentTexObj[1]) 
+      if (t->bound) 
 	 mgaSwapOutTexObj( mmesa, t );
       else
 	 mgaDestroyTexObj( mmesa, t );
@@ -178,12 +259,54 @@ void mgaTexturesGone( mgaContextPtr mmesa,
       t = (mgaTextureObjectPtr) calloc(1,sizeof(*t));
       if (!t) return;
 
-      t->MemBlock = mmAllocMem( mmesa->texHeap, size, 0, offset);      
-      insert_at_head( &mmesa->TexObjList, t );
+      t->heap = heap;
+      t->MemBlock = mmAllocMem( mmesa->texHeap[heap], size, 0, offset);      
+      if (!t->MemBlock) {
+	      fprintf(stderr, "Couldn't alloc placeholder sz %x ofs %x\n",
+		      (int)size, (int)offset);
+	      mmDumpMemInfo( mmesa->texHeap[heap]);
+	      return;
+      }
+      insert_at_head( &(mmesa->TexObjList[heap]), t );
    }
 }
 
 
+void mgaAgeTextures( mgaContextPtr mmesa, int heap )
+{
+	drm_mga_sarea_t *sarea = mmesa->sarea;
+	int sz = 1 << (mmesa->mgaScreen->logTextureGranularity[heap]);
+	int idx, nr = 0;
+
+	/* Have to go right round from the back to ensure stuff ends up
+	 * LRU in our local list...  Fix with a cursor pointer.
+	 */
+	for (idx = sarea->texList[heap][MGA_NR_TEX_REGIONS].prev ; 
+	     idx != MGA_NR_TEX_REGIONS && nr < MGA_NR_TEX_REGIONS ; 
+	     idx = sarea->texList[heap][idx].prev, nr++)
+	{
+		if (sarea->texList[heap][idx].age > mmesa->texAge[heap]) {
+			mgaTexturesGone(mmesa, heap, idx * sz, sz, 1);
+		}
+	}
+
+	if (nr == MGA_NR_TEX_REGIONS) {
+		mgaTexturesGone(mmesa, heap, 0,
+				mmesa->mgaScreen->textureSize[heap], 0);
+		mgaResetGlobalLRU( mmesa, heap );
+	}
+
+	
+	if (0) {
+		mgaPrintGlobalLRU( mmesa, heap );
+		mgaPrintLocalLRU( mmesa, heap );
+	}
+
+	mmesa->texAge[heap] = sarea->texAge[heap];
+	mmesa->dirty |= MGA_UPLOAD_TEX0IMAGE | MGA_UPLOAD_TEX1IMAGE;
+}
+
+
 /*
  * mgaSetTexWrappings
  */
@@ -281,22 +404,22 @@ static void mgaSetTexBorderColor(mgaTextureObjectPtr t, GLubyte color[4]) {
 
 
 /*
- * mgaUploadSubImage
+ * mgaUploadSubImageLocked
  *
  * Perform an iload based update of a resident buffer.  This is used for
  * both initial loading of the entire image, and texSubImage updates.
  *
  * Performed with the hardware lock held.
  */
-static void mgaUploadSubImage( mgaContextPtr mmesa,
-			       mgaTextureObjectPtr t,
-			       int level,	     
-			       int x, int y, int width, int height ) {
+static void mgaUploadSubImageLocked( mgaContextPtr mmesa,
+				     mgaTextureObjectPtr t,
+				     int level,	     
+				     int x, int y, int width, int height ) {
 	int		x2;
 	int		dwords;
-        int		dstorg;
+        int		offset;
 	struct gl_texture_image *image;
-	int		texelBytes, texelsPerDword, texelMaccess;
+	int		texelBytes, texelsPerDword, texelMaccess, length;
 
 	if ( level < 0 || level >= MGA_TEX_MAXLEVELS ) {
 		mgaMsg( 1, "mgaUploadSubImage: bad level: %i\n", level );
@@ -310,15 +433,9 @@ static void mgaUploadSubImage( mgaContextPtr mmesa,
 	}
 	
 	/* find the proper destination offset for this level */
-   	dstorg = (mmesa->mgaScreen->textureOffset + t->MemBlock->ofs + 
+   	offset = (t->MemBlock->ofs + 
 		  t->offsets[level]);
 
-    	/* turn on PCI/AGP if needed 
-    	if ( textureHeapPhysical ) {
-    		dstorg |= 1 | mgaglx.use_agp;
-    	}
-	*/
-
 	texelBytes = t->texelBytes;	
 	switch( texelBytes ) {
 	case 1:
@@ -363,7 +480,7 @@ static void mgaUploadSubImage( mgaContextPtr mmesa,
 		x = (x + (texelsPerDword-1)) & ~(texelsPerDword-1);
 		width = x2 - x;
 	}
-    
+
 	/* we may not be able to upload the entire texture in one
 	   batch due to register limits or dma buffer limits.
 	   Recursively split it up. */
@@ -374,7 +491,8 @@ static void mgaUploadSubImage( mgaContextPtr mmesa,
 		}
 		mgaMsg(10, "mgaUploadSubImage: recursively subdividing\n" );
 
-		mgaUploadSubImage( mmesa, t, level, x, y, width, height >> 1 );
+		mgaUploadSubImageLocked( mmesa, t, level, x, y, 
+					 width, height >> 1 );
 		y += ( height >> 1 );
 		height -= ( height >> 1 );
 	}
@@ -385,31 +503,50 @@ static void mgaUploadSubImage( mgaContextPtr mmesa,
 	/* bump the performance counter */
 	mgaglx.c_textureSwaps += ( dwords << 2 );
 	
-	
-#if 0
-
-	/* fill in the secondary buffer with properly converted texels
-	from the mesa buffer */
-	mgaConvertTexture( dest, texelBytes, image, x, y, width, height );
-		
-	/* send the secondary data */	
-	mgaSecondaryDma( TT_BLIT, dest, dwords );
-#endif
-
+   	length = dwords * 4;
+
+	/* Fill in the secondary buffer with properly converted texels
+	 * from the mesa buffer. */
+   	if(t->heap == MGA_CARD_HEAP) {
+	   mgaGetILoadBufferLocked( mmesa );
+	   mgaConvertTexture( (mgaUI32 *)mmesa->iload_buffer->address, 
+			     texelBytes, image, x, y, width, height ); 
+	   if(length < 64) length = 64;
+	   mgaMsg(10, "TexelBytes : %d, offset: %d, length : %d\n",
+		  texelBytes,
+		  mmesa->mgaScreen->textureOffset[t->heap] +
+		  offset + 
+		  y * width * 4/texelsPerDword,
+		  length);
+	   
+	   mgaFireILoadLocked( mmesa, 
+			      mmesa->mgaScreen->textureOffset[t->heap] +
+			      offset + 
+			      y * width * 4/texelsPerDword,
+			      length);
+	} else {
+ 	/* This works, is slower for uploads to card space and needs
+	 * additional synchronization with the dma stream.
+	 */
+	   mgaConvertTexture( (mgaUI32 *)
+			      (mmesa->mgaScreen->texVirtual[t->heap] + 
+					  offset + 
+					  y * width * 4/texelsPerDword), 
+			     texelBytes, image, x, y, width, height ); 
+	}
 }
 
 
-
 static void mgaUploadTexLevel( mgaContextPtr mmesa, 
-			       mgaTextureObjectPtr t,
-			       int l )
+				     mgaTextureObjectPtr t,
+				     int l )
 {
-	mgaUploadSubImage( mmesa,
-			   t,
-			   l,
-			   0, 0,
-			   t->tObj->Image[l]->Width,
-			   t->tObj->Image[l]->Height);
+	mgaUploadSubImageLocked( mmesa,
+				 t,
+				 l,
+				 0, 0,
+				 t->tObj->Image[l]->Width,
+				 t->tObj->Image[l]->Height);
 }
 
 
@@ -527,9 +664,11 @@ static void mgaCreateTexObj(mgaContextPtr mmesa, struct gl_texture_object *tObj)
 	/* fill in our mga texture object */
 	t->tObj = tObj;
 	t->ctx = mmesa;
+	t->age = 0;
+	t->bound = 0;
 
 	
-	insert_at_tail(&(mmesa->TexObjList), t);
+	insert_at_tail(&(mmesa->SwappedOut), t);
 	
 	t->MemBlock = 0;
 
@@ -596,34 +735,61 @@ static void mgaCreateTexObj(mgaContextPtr mmesa, struct gl_texture_object *tObj)
   	tObj->DriverData = t;
 }
 
+static void mgaMigrateTexture( mgaContextPtr mmesa, mgaTextureObjectPtr t )
+{
+	/* NOT DONE */
+}
 
 
-int mgaUploadTexImages( mgaContextPtr mmesa, mgaTextureObjectPtr t )
+static int mgaChooseTexHeap( mgaContextPtr mmesa, mgaTextureObjectPtr t )
 {
+	return 0;
+}
+
 
+int mgaUploadTexImages( mgaContextPtr mmesa, mgaTextureObjectPtr t )
+{
+	int heap;
 	int i;
 	int ofs;
 	mgaglx.c_textureSwaps++;
 
+	heap = t->heap = mgaChooseTexHeap( mmesa, t );
+
 	/* Do we need to eject LRU texture objects?
 	 */
 	if (!t->MemBlock) {
 		while (1)
 		{
-			t->MemBlock = mmAllocMem( mmesa->texHeap, t->totalSize, 12, 0 ); 
+			mgaTextureObjectPtr tmp = mmesa->TexObjList[heap].prev;
+
+			t->MemBlock = mmAllocMem( mmesa->texHeap[heap], 
+						  t->totalSize,
+						  6, 0 ); 
 			if (t->MemBlock)
 				break;
 
-			if (mmesa->TexObjList.prev == &(mmesa->TexObjList)) {
-				fprintf(stderr, "Failed to upload texture, sz %d\n", t->totalSize);
-				mmDumpMemInfo( mmesa->texHeap );
+			if (mmesa->TexObjList[heap].prev->bound) {
+				fprintf(stderr, 
+					"Hit bound texture in upload\n"); 
 				return -1;
 			}
 
-			mgaDestroyTexObj( mmesa, mmesa->TexObjList.prev );
+			if (mmesa->TexObjList[heap].prev == 
+			    &(mmesa->TexObjList[heap])) 
+			{
+				fprintf(stderr, "Failed to upload texture, "
+					"sz %d\n", t->totalSize);
+				mmDumpMemInfo( mmesa->texHeap[heap] );
+				return -1;
+			}
+	 
+			mgaDestroyTexObj( mmesa, tmp );
 		}
  
-		ofs = t->MemBlock->ofs;
+		ofs = t->MemBlock->ofs 
+			+ mmesa->mgaScreen->textureOffset[heap]
+			;
 
 		t->Setup[MGA_TEXREG_ORG]  = ofs;
 		t->Setup[MGA_TEXREG_ORG1] = ofs + t->offsets[1];
@@ -638,8 +804,16 @@ int mgaUploadTexImages( mgaContextPtr mmesa, mgaTextureObjectPtr t )
 	 */
 	mgaUpdateTexLRU( mmesa, t );
 
+	
+	if (MGA_DEBUG&DEBUG_VERBOSE_LRU)
+		fprintf(stderr, "dispatch age: %d age freed memory: %d\n",
+			GET_DISPATCH_AGE(mmesa), mmesa->dirtyAge);
+
+	if (mmesa->dirtyAge >= GET_DISPATCH_AGE(mmesa)) 
+		mgaWaitAgeLocked( mmesa, mmesa->dirtyAge );
+
 	if (t->dirty_images) {
-		if (MGA_DEBUG & MGA_DEBUG_VERBOSE_MSG)
+  		if (MGA_DEBUG&DEBUG_VERBOSE_LRU)
 			fprintf(stderr, "*");
 
 		for (i = 0 ; i <= t->lastLevel ; i++)
@@ -692,12 +866,10 @@ static void mgaUpdateTextureEnvG200( GLcontext *ctx )
 	}
 }
 
-/* I don't have the alpha values correct yet:
- */
 static void mgaUpdateTextureStage( GLcontext *ctx, int unit )
 {
 	mgaContextPtr mmesa = MGA_CONTEXT( ctx );
-	mgaUI32 *reg = &mmesa->Setup[MGA_CTXREG_TDUAL0 + unit];
+	GLuint *reg = &mmesa->Setup[MGA_CTXREG_TDUAL0 + unit];
 	GLuint source = mmesa->tmu_source[unit];
 	struct gl_texture_object *tObj = ctx->Texture.Unit[source].Current;
 
@@ -725,13 +897,13 @@ static void mgaUpdateTextureStage( GLcontext *ctx, int unit )
 			*reg = ( TD0_color_arg2_diffuse |
 				 TD0_color_sel_mul | 
 				 TD0_alpha_arg2_diffuse |
-				 TD0_alpha_sel_arg1);
+				 TD0_alpha_sel_mul);
 		else
 			*reg = ( TD0_color_arg2_prevstage |
 				 TD0_color_alpha_prevstage |
 				 TD0_color_sel_mul | 
 				 TD0_alpha_arg2_prevstage |
-				 TD0_alpha_sel_arg1);
+				 TD0_alpha_sel_mul);
 		break;
 	case GL_DECAL:
 		*reg = (TD0_color_arg2_fcol | 
@@ -751,20 +923,49 @@ static void mgaUpdateTextureStage( GLcontext *ctx, int unit )
 				 TD0_color_add_add |
 				 TD0_color_sel_add | 
 				 TD0_alpha_arg2_diffuse |
-				 TD0_alpha_sel_arg1);
+				 TD0_alpha_sel_add);
 		else
 			*reg = ( TD0_color_arg2_prevstage |
 				 TD0_color_alpha_prevstage |
 				 TD0_color_add_add |
 				 TD0_color_sel_add | 
 				 TD0_alpha_arg2_prevstage |
-				 TD0_alpha_sel_arg1);
+				 TD0_alpha_sel_add);
 		break;
 
 	case GL_BLEND:
-		/* Use a multipass mechanism to do this:
+		if (0)
+			fprintf(stderr, "GL_BLEND unit %d flags %x\n", unit,
+				mmesa->blend_flags);
+
+		if (mmesa->blend_flags) 
+			mmesa->Fallback |= MGA_FALLBACK_TEXTURE;
+			return;
+
+		/* Do singletexture GL_BLEND with 'all ones' env-color
+		 * by using both texture units.  Multitexture gl_blend
+		 * is a fallback.  
 		 */
-		mmesa->Fallback |= MGA_FALLBACK_TEXTURE;
+		if (unit == 0) {
+			/* Part 1: R1 = Rf ( 1 - Rt )
+			 *         A1 = Af At
+			 */
+			*reg = ( TD0_color_arg2_diffuse |
+				 TD0_color_arg1_inv_enable |
+				 TD0_color_sel_mul | 
+				 TD0_alpha_arg2_diffuse |
+				 TD0_alpha_sel_arg1);
+		} else {
+			/* Part 2: R2 = R1 + Rt
+			 *         A2 = A1
+			 */
+			*reg = ( TD0_color_arg2_prevstage |
+				 TD0_color_add_add | 
+				 TD0_color_sel_add | 
+				 TD0_alpha_arg2_prevstage |
+				 TD0_alpha_sel_arg2);		
+		}	
+
 		break;
 	default:
 	}
@@ -782,9 +983,8 @@ static void mgaUpdateTextureObject( GLcontext *ctx, int unit ) {
 	mgaMsg(15,"mgaUpdateTextureState %d\n", unit);
 	
 	/* disable texturing until it is known to be good */
-	mmesa->Setup[MGA_CTXREG_DWGCTL] = 
-		(( mmesa->Setup[MGA_CTXREG_DWGCTL] & DC_opcod_MASK ) | 
-		 DC_opcod_trap);
+	mmesa->Setup[MGA_CTXREG_DWGCTL] &= DC_opcod_MASK;
+	mmesa->Setup[MGA_CTXREG_DWGCTL] |= DC_opcod_trap;
 
 	enabled = (ctx->Texture.Enabled>>(source*4))&TEXTURE0_ANY;
 	if (enabled != TEXTURE0_2D) {
@@ -797,6 +997,8 @@ static void mgaUpdateTextureObject( GLcontext *ctx, int unit ) {
 
 	if ( !tObj || tObj != ctx->Texture.Unit[source].CurrentD[2] ) 
 		return;
+
+/*  	fprintf(stderr, "unit %d: %d\n", unit, tObj->Name); */
 	
 	/* if the texture object doesn't exist at all (never used or
 	   swapped out), create it now, uploading all texture images */
@@ -804,7 +1006,6 @@ static void mgaUpdateTextureObject( GLcontext *ctx, int unit ) {
 	if ( !tObj->DriverData ) {
 		/* clear the current pointer so that texture object can be
 		   swapped out if necessary to make room */
-		mmesa->CurrentTexObj[source] = NULL;
 		mgaCreateTexObj( mmesa, tObj );
 
 		if ( !tObj->DriverData ) {
@@ -815,9 +1016,8 @@ static void mgaUpdateTextureObject( GLcontext *ctx, int unit ) {
 	}
 
 	/* we definately have a valid texture now */
-	mmesa->Setup[MGA_CTXREG_DWGCTL] = 
-		(( mmesa->Setup[MGA_CTXREG_DWGCTL] & DC_opcod_MASK ) | 
-		 DC_opcod_texture_trap);
+	mmesa->Setup[MGA_CTXREG_DWGCTL] &= DC_opcod_MASK;
+	mmesa->Setup[MGA_CTXREG_DWGCTL] |= DC_opcod_texture_trap;
 
 	t = (mgaTextureObjectPtr)tObj->DriverData;
 
@@ -825,6 +1025,11 @@ static void mgaUpdateTextureObject( GLcontext *ctx, int unit ) {
 		mmesa->dirty |= (MGA_UPLOAD_TEX0IMAGE << unit);
 
 	mmesa->CurrentTexObj[unit] = t;
+	t->bound = unit+1;
+
+	if (t->MemBlock)
+		mgaUpdateTexLRU( mmesa, t );
+
 
 	t->Setup[MGA_TEXREG_CTL2] &= ~TMC_dualtex_enable;
 	if (ctx->Texture.Enabled == (TEXTURE0_2D|TEXTURE1_2D))
@@ -848,23 +1053,31 @@ void mgaUpdateTextureState( GLcontext *ctx )
 	mgaContextPtr mmesa = MGA_CONTEXT( ctx );
 	mmesa->Fallback &= ~MGA_FALLBACK_TEXTURE;
 
+	if (mmesa->CurrentTexObj[0]) mmesa->CurrentTexObj[0]->bound = 0;
+	if (mmesa->CurrentTexObj[1]) mmesa->CurrentTexObj[1]->bound = 0;
+	mmesa->CurrentTexObj[0] = 0;
+	mmesa->CurrentTexObj[1] = 0;   
+
 	if (MGA_IS_G400(mmesa)) {
 		mgaUpdateTextureObject( ctx, 0 );		
 		mgaUpdateTextureStage( ctx, 0 );
 
-		mmesa->Setup[MGA_CTXREG_TDUAL1] = mmesa->Setup[MGA_CTXREG_TDUAL0];
+		mmesa->Setup[MGA_CTXREG_TDUAL1] = 
+			mmesa->Setup[MGA_CTXREG_TDUAL0];
 
 		if (mmesa->multitex) {
 			mgaUpdateTextureObject( ctx, 1 );	
 			mgaUpdateTextureStage( ctx, 1 );
 		}
+		
+		mmesa->dirty |= MGA_UPLOAD_TEX0 | MGA_UPLOAD_TEX1;
 	} else {
 		mgaUpdateTextureObject( ctx, 0 );
-		mgaUpdateTextureEnvG200( ctx );
+		mgaUpdateTextureEnvG200( ctx );		
 	}
 
 	/* schedule the register writes */
-	mmesa->dirty |= MGA_UPLOAD_CTX;
+	mmesa->dirty |= MGA_UPLOAD_CTX | MGA_UPLOAD_TEX0;
 }
 
 
@@ -880,14 +1093,57 @@ Driver functions called directly from mesa
 /*
  * mgaTexEnv
  */      
-void mgaTexEnv( GLcontext *ctx, GLenum pname, const GLfloat *param ) {
+void mgaTexEnv( GLcontext *ctx, GLenum pname, const GLfloat *param ) 
+{
+	mgaContextPtr mmesa = MGA_CONTEXT(ctx);
 	mgaMsg( 10, "mgaTexEnv( %i )\n", pname );
 
+
 	if (pname == GL_TEXTURE_ENV_MODE) {
 		/* force the texture state to be updated */
-		MGA_CONTEXT(ctx)->CurrentTexObj[0] = 0;  
+		FLUSH_BATCH( MGA_CONTEXT(ctx) );
 		MGA_CONTEXT(ctx)->new_state |= MGA_NEW_TEXTURE;
 	}
+	else if (pname == GL_TEXTURE_ENV_COLOR)
+	{
+		struct gl_texture_unit *texUnit = 
+			&ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+		GLfloat *fc = texUnit->EnvColor;
+		GLubyte c[4];
+		GLuint col;
+
+		
+		c[0] = fc[0];
+		c[1] = fc[1];
+		c[2] = fc[2];
+		c[3] = fc[3];
+
+		/* No alpha at 16bpp?
+		 */
+		col = mgaPackColor( mmesa->mgaScreen->Attrib,
+				    c[0], c[1], c[2], c[3] );
+
+  		mmesa->envcolor = (c[3]<<24) | (c[0]<<16) | (c[1]<<8) | (c[2]); 
+    
+		if (mmesa->Setup[MGA_CTXREG_FCOL] != col) {
+			FLUSH_BATCH(mmesa);	
+			mmesa->Setup[MGA_CTXREG_FCOL] = col;      
+			mmesa->dirty |= MGA_UPLOAD_CTX;
+
+			mmesa->blend_flags &= ~MGA_BLEND_ENV_COLOR;
+
+			/* Actually just require all four components to be
+			 * equal.  This permits a single-pass GL_BLEND.
+			 * 
+			 * More complex multitexture/multipass fallbacks
+			 * for blend can be done later.
+			 */
+			if (mmesa->envcolor != 0x0 &&
+			    mmesa->envcolor != 0xffffffff)
+				mmesa->blend_flags |= MGA_BLEND_ENV_COLOR;
+		}
+	}
+
 }
 
 /*
@@ -907,6 +1163,7 @@ void mgaTexImage( GLcontext *ctx, GLenum target,
 	mgaUpdateTextureState time. */  
 	t = (mgaTextureObjectPtr) tObj->DriverData;
 	if ( t ) {
+		if (t->bound) FLUSH_BATCH(mmesa);
 		/* if this is the current object, it will force an update */
  	 	mgaDestroyTexObj( mmesa, t );
 		mmesa->new_state |= MGA_NEW_TEXTURE;
@@ -936,6 +1193,7 @@ void mgaTexSubImage( GLcontext *ctx, GLenum target,
 	mgaUpdateTextureState time. */  
 	t = (mgaTextureObjectPtr) tObj->DriverData;
 	if ( t ) {
+		if (t->bound) FLUSH_BATCH(mmesa);
 		/* if this is the current object, it will force an update */
  	 	mgaDestroyTexObj( mmesa, t );
 		mmesa->new_state |= MGA_NEW_TEXTURE;
@@ -956,7 +1214,9 @@ void mgaTexSubImage( GLcontext *ctx, GLenum target,
  */
 void mgaTexParameter( GLcontext *ctx, GLenum target,
 		      struct gl_texture_object *tObj,
-		      GLenum pname, const GLfloat *params ) {
+		      GLenum pname, const GLfloat *params )
+{
+	mgaContextPtr mmesa = MGA_CONTEXT( ctx );
 	mgaTextureObjectPtr t;
 
 	mgaMsg( 10, "mgaTexParameter( %p, %i )\n", tObj, pname );
@@ -973,52 +1233,68 @@ void mgaTexParameter( GLcontext *ctx, GLenum target,
 	switch (pname) {
 	case GL_TEXTURE_MIN_FILTER:
 	case GL_TEXTURE_MAG_FILTER:
+		if (t->bound) FLUSH_BATCH(mmesa);
 		mgaSetTexFilter( t, tObj->MinFilter, tObj->MagFilter );
 		break;
 
 	case GL_TEXTURE_WRAP_S:
 	case GL_TEXTURE_WRAP_T:
+		if (t->bound) FLUSH_BATCH(mmesa);
 		mgaSetTexWrapping(t,tObj->WrapS,tObj->WrapT);
 		break;
   
 	case GL_TEXTURE_BORDER_COLOR:
+		if (t->bound) FLUSH_BATCH(mmesa);
 		mgaSetTexBorderColor(t,tObj->BorderColor);
 		break;
 
 	default:
 		return;
 	}
-	/* force the texture state to be updated */
-	MGA_CONTEXT(ctx)->CurrentTexObj[0] = NULL;  
-	MGA_CONTEXT(ctx)->new_state |= MGA_NEW_TEXTURE;
+
+	mmesa->new_state |= MGA_NEW_TEXTURE;
 }
 
 /*
  * mgaBindTexture
  */      
 void mgaBindTexture( GLcontext *ctx, GLenum target,
-		     struct gl_texture_object *tObj ) {
+		     struct gl_texture_object *tObj ) 
+{
+	mgaContextPtr mmesa = MGA_CONTEXT( ctx );
 
 	mgaMsg( 10, "mgaBindTexture( %p )\n", tObj );
-		
+
+	FLUSH_BATCH(mmesa);
+
+	if (mmesa->CurrentTexObj[ctx->Texture.CurrentUnit]) {
+		mmesa->CurrentTexObj[ctx->Texture.CurrentUnit]->bound = 0;
+		mmesa->CurrentTexObj[ctx->Texture.CurrentUnit] = 0;  
+	}
+
 	/* force the texture state to be updated 
 	 */
-	MGA_CONTEXT(ctx)->CurrentTexObj[0] = NULL;  
 	MGA_CONTEXT(ctx)->new_state |= MGA_NEW_TEXTURE;
 }
 
 /*
  * mgaDeleteTexture
  */      
-void mgaDeleteTexture( GLcontext *ctx, struct gl_texture_object *tObj ) {
+void mgaDeleteTexture( GLcontext *ctx, struct gl_texture_object *tObj ) 
+{
+	mgaContextPtr mmesa = MGA_CONTEXT( ctx );
+	mgaTextureObjectPtr t = (mgaTextureObjectPtr)tObj->DriverData;
 
 	mgaMsg( 10, "mgaDeleteTexture( %p )\n", tObj );
 	
-	/* delete our driver data */
-	if ( tObj->DriverData ) {
-		mgaContextPtr mmesa = MGA_CONTEXT( ctx );
-		mgaDestroyTexObj( mmesa,
-				  (mgaTextureObjectPtr)(tObj->DriverData) );
+	if ( t ) {
+		if (t->bound) {
+			FLUSH_BATCH(mmesa);
+			mmesa->CurrentTexObj[t->bound-1] = 0;
+			mmesa->new_state |= MGA_NEW_TEXTURE;
+		}
+
+		mgaDestroyTexObj( mmesa, t );
 		mmesa->new_state |= MGA_NEW_TEXTURE;
 	}
 }
diff --git a/xc/lib/GL/mesa/src/drv/mga/mgatris.c b/xc/lib/GL/mesa/src/drv/mga/mgatris.c
index bd8ef083f..387a79f3e 100644
--- a/xc/lib/GL/mesa/src/drv/mga/mgatris.c
+++ b/xc/lib/GL/mesa/src/drv/mga/mgatris.c
@@ -27,6 +27,7 @@
 #include <stdio.h>
 #include <math.h>
 
+#include "types.h"
 #include "vb.h"
 #include "pipeline.h"
 
@@ -122,34 +123,14 @@ void mgaDDTrifuncInit()
    init_twoside_offset();
    init_twoside_offset_flat();
 
-   /* Hmmm...
-    */
-   for (i = 0 ; i < 0x20 ; i++) {
-      if (i & ~MGA_FLAT_BIT) {
-	 points_tab[i] = points_tab[i&MGA_FLAT_BIT];
-	 line_tab[i] = line_tab[i&MGA_FLAT_BIT];
-      }
-   }
-
    for (i = 0 ; i < 0x20 ; i++) 
-      if ((i & (MGA_NODRAW_BIT|MGA_FALLBACK_BIT)) == MGA_NODRAW_BIT ||
-	  mgaglx.nullprims) 
+      if ((i & (MGA_NODRAW_BIT|MGA_FALLBACK_BIT)) == MGA_NODRAW_BIT)
       {
 	 quad_tab[i] = mga_null_quad; 
 	 tri_tab[i] = mga_null_triangle; 
 	 line_tab[i] = mga_null_line;
 	 points_tab[i] = mga_null_points;
       }
-
-   if (mgaglx.noFallback) {
-      for (i = 0 ; i < 0x10 ; i++) {
-	 points_tab[i|MGA_FALLBACK_BIT] = points_tab[i];
-	 line_tab[i|MGA_FALLBACK_BIT] = line_tab[i];
-	 tri_tab[i|MGA_FALLBACK_BIT] = tri_tab[i];
-	 quad_tab[i|MGA_FALLBACK_BIT] = quad_tab[i];
-      }
-   }
-
 }
 
 
@@ -162,50 +143,61 @@ void mgaDDChooseRenderState( GLcontext *ctx )
    mgaContextPtr mmesa = MGA_CONTEXT( ctx );
    GLuint flags = ctx->TriangleCaps;
 
-   ctx->IndirectTriangles &= ~DD_SW_RASTERIZE;
+   if (mmesa->Fallback)
+      return;
+
+   mmesa->IndirectTriangles = 0;
 
    if (flags) {
       GLuint ind = 0;
       GLuint shared = 0;
-      GLuint fallback = MGA_FALLBACK_BIT;
-
-      if (mgaglx.noFallback) fallback = 0;
 
       if (flags & DD_Z_NEVER)                      shared |= MGA_NODRAW_BIT;
       if (flags & DD_FLATSHADE)                    shared |= MGA_FLAT_BIT;
-      if (flags & DD_MULTIDRAW)                    shared |= fallback;
-      if (flags & (DD_SELECT|DD_FEEDBACK))         shared |= MGA_FALLBACK_BIT;
+      if (flags & (DD_MULTIDRAW|
+		   DD_SELECT|
+		   DD_FEEDBACK))                   shared |= MGA_FALLBACK_BIT;
       if (flags & DD_STENCIL)                      shared |= MGA_FALLBACK_BIT;
 
       ind = shared;
+#if 0
       if (flags & DD_POINT_SMOOTH)                 ind |= MGA_ANTIALIAS_BIT;
-      if (flags & DD_POINT_ATTEN)                  ind |= fallback;
+#else
+      if (flags & DD_POINT_SMOOTH)                 ind |= MGA_FALLBACK_BIT;
+#endif
 
       mmesa->renderindex = ind;
       mmesa->PointsFunc = points_tab[ind];
       if (ind & MGA_FALLBACK_BIT) 
-	 ctx->IndirectTriangles |= DD_POINT_SW_RASTERIZE;
+	 mmesa->IndirectTriangles |= DD_POINT_SW_RASTERIZE;
 
       ind = shared;
+#if 0
       if (flags & DD_LINE_SMOOTH)                   ind |= MGA_ANTIALIAS_BIT;
-      if (flags & DD_LINE_STIPPLE)                  ind |= fallback;
+#else
+      if (flags & DD_LINE_SMOOTH)                   ind |= MGA_FALLBACK_BIT;
+#endif
+      if (flags & DD_LINE_STIPPLE)                  ind |= MGA_FALLBACK_BIT;
 
       mmesa->renderindex |= ind;
       mmesa->LineFunc = line_tab[ind];
       if (ind & MGA_FALLBACK_BIT) 
-	 ctx->IndirectTriangles |= DD_LINE_SW_RASTERIZE;
+	 mmesa->IndirectTriangles |= DD_LINE_SW_RASTERIZE;
 
       ind = shared;
       if (flags & DD_TRI_SMOOTH)                    ind |= MGA_ANTIALIAS_BIT;
       if (flags & DD_TRI_OFFSET)                    ind |= MGA_OFFSET_BIT;
       if (flags & DD_TRI_LIGHT_TWOSIDE)             ind |= MGA_TWOSIDE_BIT;
-      if (flags & (DD_TRI_UNFILLED|DD_TRI_STIPPLE)) ind |= fallback;	
+      if (flags & DD_TRI_UNFILLED)                  ind |= MGA_FALLBACK_BIT;
+      if ((flags & DD_TRI_STIPPLE) &&
+	  (ctx->IndirectTriangles & DD_TRI_STIPPLE)) ind |= MGA_FALLBACK_BIT;	
 
       mmesa->renderindex |= ind;
       mmesa->TriangleFunc = tri_tab[ind];
       mmesa->QuadFunc = quad_tab[ind];
       if (ind & MGA_FALLBACK_BIT)
-	 ctx->IndirectTriangles |= (DD_TRI_SW_RASTERIZE | DD_QUAD_SW_RASTERIZE);
+	 mmesa->IndirectTriangles |= (DD_TRI_SW_RASTERIZE | 
+				      DD_QUAD_SW_RASTERIZE);
    } 
    else if (mmesa->renderindex)
    {
diff --git a/xc/lib/GL/mesa/src/drv/mga/mgatritmp.h b/xc/lib/GL/mesa/src/drv/mga/mgatritmp.h
index ac5a545d2..2a9d16b86 100644
--- a/xc/lib/GL/mesa/src/drv/mga/mgatritmp.h
+++ b/xc/lib/GL/mesa/src/drv/mga/mgatritmp.h
@@ -98,19 +98,31 @@ static void TAG(quad)( GLcontext *ctx, GLuint v0,
 }
 
 
-#if ((IND & ~MGA_FLAT_BIT) == 0)
-
 static void TAG(line)( GLcontext *ctx, GLuint v0, GLuint v1, GLuint pv )
 {
    mgaContextPtr mmesa = MGA_CONTEXT( ctx );
    mgaVertexPtr mgaVB = MGA_DRIVER_DATA(ctx->VB)->verts;
    float width = ctx->Line.Width;
 
-   if (IND & MGA_FLAT_BIT) {
+   if (IND & (MGA_TWOSIDE_BIT|MGA_FLAT_BIT)) {
       mgaVertex tmp0 = mgaVB[v0];
       mgaVertex tmp1 = mgaVB[v1];
-      *(int *)&tmp0.warp1.color = *(int *)&mgaVB[pv].warp1.color;
-      *(int *)&tmp1.warp1.color = *(int *)&mgaVB[pv].warp1.color;
+
+      if (IND & MGA_TWOSIDE_BIT) {
+	 GLubyte (*vbcolor)[4] = ctx->VB->ColorPtr->data;
+
+	 if (IND & MGA_FLAT_BIT) {
+ 	    MGA_COLOR((char *)&tmp0.warp1.color,vbcolor[pv]);
+	    *(int *)&tmp1.warp1.color = *(int *)&tmp0.warp1.color;
+	 } else {
+	    MGA_COLOR((char *)&tmp0.warp1.color,vbcolor[v0]);
+	    MGA_COLOR((char *)&tmp1.warp1.color,vbcolor[v1]);
+	 }
+
+      } else {
+	 *(int *)&tmp0.warp1.color = *(int *)&mgaVB[pv].warp1.color;
+	 *(int *)&tmp1.warp1.color = *(int *)&mgaVB[pv].warp1.color;
+      }
       mga_draw_line( mmesa, &tmp0, &tmp1, width );
    }
    else
@@ -125,13 +137,20 @@ static void TAG(points)( GLcontext *ctx, GLuint first, GLuint last )
    mgaVertexPtr mgaVB = MGA_DRIVER_DATA(VB)->verts;
    GLfloat sz = ctx->Point.Size * .5;
    int i;
-
+   
    for(i=first;i<=last;i++) 
-      if(VB->ClipMask[i]==0) 
-	 mga_draw_point( mmesa, &mgaVB[i], sz );
+      if(VB->ClipMask[i]==0) {
+	 if (IND & MGA_TWOSIDE_BIT) {	
+	    GLubyte (*vbcolor)[4] = VB->ColorPtr->data;
+	    mgaVertex tmp0 = mgaVB[i];
+	    MGA_COLOR((char *)&tmp0.warp1.color, vbcolor[i]);
+	    mga_draw_point( mmesa, &tmp0, sz );
+	 } else
+	    mga_draw_point( mmesa, &mgaVB[i], sz );
+      }
 }
 
-#endif
+
 
 
 static void TAG(init)( void )
@@ -139,10 +158,9 @@ static void TAG(init)( void )
    tri_tab[IND] = TAG(triangle);
    quad_tab[IND] = TAG(quad);
 
-#if ((IND & ~MGA_FLAT_BIT) == 0) 
    line_tab[IND] = TAG(line);
    points_tab[IND] = TAG(points);
-#endif
+
 }
 
 
diff --git a/xc/lib/GL/mesa/src/drv/mga/mgavb.c b/xc/lib/GL/mesa/src/drv/mga/mgavb.c
index f78c542c8..2d61f301d 100644
--- a/xc/lib/GL/mesa/src/drv/mga/mgavb.c
+++ b/xc/lib/GL/mesa/src/drv/mga/mgavb.c
@@ -104,7 +104,7 @@ static void name(struct vertex_buffer *VB, GLuint start, GLuint end)	\
    GLfloat yoffset = mmesa->driDrawable->h - 0.5 + mmesa->drawY;	\
    int i;								\
    (void) xoffset; (void) yoffset;					\
-									\
+   if (0) fprintf(stderr, "V");						\
    gl_import_client_data( VB, VB->ctx->RenderFlags,			\
 			  (VB->ClipOrMask				\
 			   ? VEC_WRITABLE|VEC_GOOD_STRIDE		\
@@ -260,11 +260,21 @@ void mgaChooseRasterSetupFunc(GLcontext *ctx)
    mmesa->tex_dest[0] = MGA_TEX0_BIT;
    mmesa->tex_dest[1] = MGA_TEX1_BIT;
    mmesa->multitex = 0;
+   mmesa->blend_flags &= ~MGA_BLEND_MULTITEX;
 
    if (ctx->Texture.Enabled & 0xf) {
       if (ctx->Texture.Unit[0].EnvMode == GL_REPLACE)
 	 funcindex &= ~MGA_RGBA_BIT;
 
+      if (ctx->Texture.Unit[0].EnvMode == GL_BLEND &&
+	  mmesa->envcolor) 
+      {
+	 mmesa->multitex = 1;
+	 mmesa->vertsize = 10;
+	 mmesa->tmu_source[1] = 0;
+	 funcindex |= MGA_TEX1_BIT;
+      }	 
+
       funcindex |= MGA_TEX0_BIT;
    }
 
@@ -272,7 +282,8 @@ void mgaChooseRasterSetupFunc(GLcontext *ctx)
       if (ctx->Texture.Enabled & 0xf) {
 	 mmesa->multitex = 1;
 	 mmesa->vertsize = 10;
-	 funcindex |= MGA_TEX1_BIT;
+	 mmesa->blend_flags |= MGA_BLEND_MULTITEX;
+	 funcindex |= MGA_TEX1_BIT;	 
       } else {
 	 /* Just a funny way of doing single texturing 
 	  */
@@ -281,11 +292,25 @@ void mgaChooseRasterSetupFunc(GLcontext *ctx)
 
 	 if (ctx->Texture.Unit[1].EnvMode == GL_REPLACE)
 	    funcindex &= ~MGA_RGBA_BIT;
+	 
+	 if (ctx->Texture.Unit[0].EnvMode == GL_BLEND &&
+	     mmesa->envcolor) 
+	 {
+	    mmesa->multitex = 1;
+	    mmesa->vertsize = 10;
+	    mmesa->tmu_source[1] = 1;
+	    funcindex |= MGA_TEX1_BIT;
+	 }	 
 
 	 funcindex |= MGA_TEX0_BIT;
       }
    }
 
+   /* Not really a good place to do this - need to make the mga state
+    * management code more event-driven so this can be calculated for
+    * free.
+    */
+
    if (ctx->Color.BlendEnabled)
       funcindex |= MGA_ALPHA_BIT;
 
@@ -298,6 +323,7 @@ void mgaChooseRasterSetupFunc(GLcontext *ctx)
    if (0)
       mgaPrintSetupFlags("xsmesa: full setup function", funcindex); 
 
+   mmesa->dirty |= MGA_UPLOAD_PIPE;
    mmesa->setupindex = funcindex;
 
    /* Called by mesa's clip functons:
@@ -477,17 +503,3 @@ void mgaDDUnregisterVB( struct vertex_buffer *VB )
 }
 
 
-mgaUI32 *mgaAllocVertexDwords( mgaContextPtr mmesa, int dwords )
-{
-   int bytes = dwords * 4;
-   mgaUI32 *head;
-
-   if (mmesa->dma_buffer->used + bytes > mmesa->dma_buffer->total)
-      mgaFlushVertices( mmesa );
-
-   head = (mgaUI32 *)((char *)mmesa->dma_buffer->address + 
-		      mmesa->dma_buffer->used);
-
-   mmesa->dma_buffer->used += bytes;
-   return head;
-}
diff --git a/xc/lib/GL/mesa/src/drv/r128/Imakefile b/xc/lib/GL/mesa/src/drv/r128/Imakefile
new file mode 100644
index 000000000..4cd07b8ff
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/Imakefile
@@ -0,0 +1,325 @@
+XCOMM $XFree86$
+
+#include <Threads.tmpl>
+
+#define DoNormalLib NormalLibGlx
+#define DoSharedLib SharedLibGlx
+#define DoExtraLib SharedLibGlx
+#define DoDebugLib DebugLibGlx
+#define DoProfileLib ProfileLibGlx
+
+#if Malloc0ReturnsNull
+ALLOC_DEFINES = -DMALLOC_0_RETURNS_NULL
+#endif
+
+#if BuildXF86DRI
+  DRI_DEFINES = GlxDefines -DDRIVERTS
+ DRI_INCLUDES = -I../../../../dri \
+		-I../../../../glx \
+		-I../../../dri \
+		-I$(TOP)/include \
+		-I$(TOP)/include/GL \
+		-I$(XF86OSSRC) \
+		-I$(XF86COMSRC) \
+		-I$(SERVERSRC)/GL/dri \
+		-I$(XF86DRIVERSRC)/r128 \
+		-I../../../include \
+		-I../.. \
+		-I../../X \
+		-I../common
+#endif
+
+LinkSourceFile(mm.c, ../common)
+LinkSourceFile(mm.h, ../common)
+LinkSourceFile(hwlog.c, ../common)
+LinkSourceFile(hwlog.h, ../common)
+
+MESA_INCLUDES = -I. -I.. -I../../include
+
+      DEFINES = $(ALLOC_DEFINES) $(DRI_DEFINES)
+     INCLUDES = -I$(XLIBSRC) -I$(EXTINCSRC) $(MESA_INCLUDES) $(DRI_INCLUDES)
+
+     R128SRCS = r128_cce.c \
+		r128_clear.c \
+		r128_context.c \
+		r128_dd.c \
+		r128_fastpath.c \
+		r128_pipeline.c \
+                r128_screen.c \
+		r128_span.c \
+		r128_state.c \
+		r128_swap.c \
+		r128_tex.c \
+		r128_tris.c \
+		r128_vb.c \
+		r128_xmesa.c
+
+     R128OBJS = r128_cce.o \
+		r128_clear.o \
+		r128_context.o \
+		r128_dd.o \
+		r128_fastpath.o \
+		r128_pipeline.o \
+                r128_screen.o \
+		r128_span.o \
+		r128_state.o \
+		r128_swap.o \
+		r128_tex.o \
+		r128_tris.o \
+		r128_vb.o \
+		r128_xmesa.o
+
+#if !GlxUseBuiltInDRIDriver
+      DRISRCS = ../../../dri/dri_mesa.c \
+		../../../../dri/dri_tmm.c
+
+      DRIOBJS = ../../../dri/dri_mesa.o \
+		../../../../dri/dri_tmm.o
+
+      DRMSRCS = ../../../../dri/drm/xf86drm.c \
+		../../../../dri/drm/xf86drmHash.c \
+		../../../../dri/drm/xf86drmRandom.c \
+		../../../../dri/drm/xf86drmSL.c \
+		../../../../dri/drm/xf86drmR128.c
+
+      DRMOBJS = ../../../../dri/drm/xf86drm.o \
+		../../../../dri/drm/xf86drmHash.o \
+		../../../../dri/drm/xf86drmRandom.o \
+		../../../../dri/drm/xf86drmSL.o \
+		../../../../dri/drm/xf86drmR128.o
+
+     MESASRCS = ../../aatriangle.c \
+		../../accum.c \
+		../../alpha.c \
+		../../alphabuf.c \
+		../../attrib.c \
+		../../bbox.c \
+		../../bitmap.c \
+		../../blend.c \
+		../../buffers.c \
+		../../clip.c \
+		../../colortab.c \
+		../../config.c \
+		../../context.c \
+		../../copypix.c \
+		../../cva.c \
+		../../debug_xform.c \
+		../../depth.c \
+		../../dlist.c \
+		../../drawpix.c \
+		../../enable.c \
+		../../enums.c \
+		../../eval.c \
+		../../extensions.c \
+		../../feedback.c \
+		../../fog.c \
+		../../get.c \
+		../../glapi.c \
+		../../glapinoop.c \
+		../../glthread.c \
+		../../hash.c \
+		../../image.c \
+		../../imaging.c \
+		../../light.c \
+		../../lines.c \
+		../../logic.c \
+		../../masking.c \
+		../../matrix.c \
+		../../mem.c \
+		../../mmath.c \
+		../../pb.c \
+		../../pipeline.c \
+		../../pixel.c \
+		../../pixeltex.c \
+		../../points.c \
+		../../polygon.c \
+		../../quads.c \
+		../../rastpos.c \
+		../../readpix.c \
+		../../rect.c \
+		../../scissor.c \
+		../../shade.c \
+		../../span.c \
+		../../stages.c \
+		../../state.c \
+		../../stencil.c \
+		../../teximage.c \
+		../../texobj.c \
+		../../texstate.c \
+		../../texture.c \
+		../../texutil.c \
+		../../translate.c \
+		../../triangle.c \
+		../../varray.c \
+		../../vb.c \
+		../../vbcull.c \
+		../../vbfill.c \
+		../../vbindirect.c \
+		../../vbrender.c \
+		../../vbxform.c \
+		../../vector.c \
+		../../vertices.c \
+		../../winpos.c \
+		../../xform.c \
+		../../zoom.c \
+		../../X86/common_x86.c
+
+     MESAOBJS = ../../aatriangle.o \
+		../../accum.o \
+		../../alpha.o \
+		../../alphabuf.o \
+		../../attrib.o \
+		../../bbox.o \
+		../../bitmap.o \
+		../../blend.o \
+		../../buffers.o \
+		../../clip.o \
+		../../colortab.o \
+		../../config.o \
+		../../context.o \
+		../../copypix.o \
+		../../cva.o \
+		../../debug_xform.o \
+		../../depth.o \
+		../../dlist.o \
+		../../drawpix.o \
+		../../enable.o \
+		../../enums.o \
+		../../eval.o \
+		../../extensions.o \
+		../../feedback.o \
+		../../fog.o \
+		../../get.o \
+		../../hash.o \
+		../../hint.o \
+		../../image.o \
+		../../imaging.o \
+		../../light.o \
+		../../lines.o \
+		../../logic.o \
+		../../masking.o \
+		../../matrix.o \
+		../../mem.o \
+		../../mmath.o \
+		../../pb.o \
+		../../pipeline.o \
+		../../pixel.o \
+		../../pixeltex.o \
+		../../points.o \
+		../../polygon.o \
+		../../quads.o \
+		../../rastpos.o \
+		../../readpix.o \
+		../../rect.o \
+		../../scissor.o \
+		../../shade.o \
+		../../span.o \
+		../../stages.o \
+		../../state.o \
+		../../stencil.o \
+		../../teximage.o \
+		../../texobj.o \
+		../../texstate.o \
+		../../texture.o \
+		../../texutil.o \
+		../../translate.o \
+		../../triangle.o \
+		../../varray.o \
+		../../vb.o \
+		../../vbcull.o \
+		../../vbfill.o \
+		../../vbindirect.o \
+		../../vbrender.o \
+		../../vbxform.o \
+		../../vector.o \
+		../../vertices.o \
+		../../winpos.o \
+		../../xform.o \
+		../../zoom.o
+
+#ifdef i386Architecture
+     X86_SRCS = ../../X86/x86.c \
+		../../X86/x86a.S \
+		../../X86/common_x86.c \
+		../../X86/common_x86asm.S \
+		../../X86/vertex.S
+
+     X86_OBJS = ../../X86/x86.o \
+		../../X86/x86a.o \
+		../../X86/common_x86.o \
+		../../X86/common_x86asm.o \
+		../../X86/vertex.o
+
+     MMX_SRCS = ../../X86/mmx_blend.S
+
+     MMX_OBJS = ../../X86/mmx_blend.o
+
+XCOMM Disabling 3Dnow code for the time being.
+#if 0
+   3DNOW_SRCS = ../../X86/3dnow.c \
+		../../X86/3dnow_norm_raw.S \
+		../../X86/3dnow_xform_masked1.S \
+		../../X86/3dnow_xform_masked2.S \
+		../../X86/3dnow_xform_masked3.S \
+		../../X86/3dnow_xform_masked4.S \
+		../../X86/3dnow_xform_raw1.S \
+		../../X86/3dnow_xform_raw2.S \
+		../../X86/3dnow_xform_raw3.S \
+		../../X86/3dnow_xform_raw4.S \
+		../../X86/vertex_3dnow.S
+
+   3DNOW_OBJS = ../../X86/3dnow.o \
+		../../X86/3dnow_norm_raw.o \
+		../../X86/3dnow_xform_masked1.o \
+		../../X86/3dnow_xform_masked2.o \
+		../../X86/3dnow_xform_masked3.o \
+		../../X86/3dnow_xform_masked4.o \
+		../../X86/3dnow_xform_raw1.o \
+		../../X86/3dnow_xform_raw2.o \
+		../../X86/3dnow_xform_raw3.o \
+		../../X86/3dnow_xform_raw4.o \
+		../../X86/vertex_3dnow.o
+#endif
+
+#endif
+#endif
+
+      ASMSRCS = $(X86_SRCS) $(MMX_SRCS) $(3DNOW_SRCS)
+      ASMOBJS = $(X86_OBJS) $(MMX_OBJS) $(3DNOW_OBJS)
+
+   COMMONSRCS = mm.c hwlog.c
+   COMMONOBJS = mm.o hwlog.o
+
+         SRCS = $(DRISRCS) $(DRMSRCS) $(MESASRCS) $(ASMSRCS) \
+		$(COMMONSRCS) $(R128SRCS)
+         OBJS = $(DRIOBJS) $(DRMOBJS) $(MESAOBJS) $(ASMOBJS) \
+		$(COMMONOBJS) $(R128OBJS)
+
+REQUIREDLIBS += -lm
+#if !GlxBuiltInMga
+REQUIREDLIBS += -L../../../.. -lGL
+#endif
+
+#if !GlxUseBuiltInDRIDriver
+#undef DoNormalLib NormalLibGlx
+#undef DoExtraLib SharedLibGlx
+#undef DoDebugLib DebugLibGlx
+#undef DoProfileLib ProfileLibGlx
+#endif
+
+#include <Library.tmpl>
+
+LibraryObjectRule()
+
+SubdirLibraryRule($(OBJS))
+NormalLintTarget($(SRCS))
+
+#if !GlxUseBuiltInDRIDriver
+LIBNAME = r128_dri.so
+ALL_OBJS = $(OBJS)
+ALL_DEPS = DONE
+SharedDepModuleTarget($(LIBNAME),$(ALL_DEPS),$(ALL_OBJS))
+InstallDynamicModule($(LIBNAME),$(MODULEDIR)/dri,.)
+#endif
+
+DependTarget()
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_cce.c b/xc/lib/GL/mesa/src/drv/r128/r128_cce.c
new file mode 100644
index 000000000..e949dd37c
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_cce.c
@@ -0,0 +1,389 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#include "r128_init.h"
+#include "r128_mesa.h"
+#include "r128_xmesa.h"
+#include "r128_context.h"
+#include "r128_lock.h"
+#include "r128_reg.h"
+#include "r128_cce.h"
+
+#if USE_USER_SPACE_RING
+
+#define USE_IOCTLS          0
+#define DO_SECURITY_CHECK   1
+#define R128_TIMEOUT        2000000
+
+#if DEBUG
+static void r128RingStatus(r128ScreenPtr pScrn)
+{
+    unsigned char *R128MMIO = pScrn->mmio;
+
+    R128_DEBUG(("GUI_STAT           = 0x%08x\n",
+		(unsigned int)INREG(R128_GUI_STAT)));
+    R128_DEBUG(("PM4_STAT           = 0x%08x\n",
+		(unsigned int)INREG(R128_PM4_STAT)));
+    R128_DEBUG(("PM4_BUFFER_DL_WPTR = 0x%08x\n",
+		(unsigned int)INREG(R128_PM4_BUFFER_DL_WPTR)));
+    R128_DEBUG(("PM4_BUFFER_DL_RPTR = 0x%08x\n",
+		(unsigned int)INREG(R128_PM4_BUFFER_DL_RPTR)));
+    R128_DEBUG(("ringWrite          = 0x%08x\n", pScrn->SAREA->ringWrite));
+    R128_DEBUG(("*ringReadPtr       = 0x%08x\n", *pScrn->ringReadPtr));
+}
+#endif
+
+/* FIXME: Requires access to secure registers */
+static int INPLL(r128ScreenPtr pScrn, int addr)
+{
+    unsigned char *R128MMIO = pScrn->mmio;
+
+    OUTREG8(R128_CLOCK_CNTL_INDEX, addr & 0x1f);
+    return INREG(R128_CLOCK_CNTL_DATA);
+}
+
+/* FIXME: Requires access to secure registers */
+static void r128EngineFlush(r128ScreenPtr pScrn)
+{
+    unsigned char *R128MMIO = pScrn->mmio;
+    int           i;
+
+    OUTREGP(R128_PC_NGUI_CTLSTAT, R128_PC_FLUSH_ALL, ~R128_PC_FLUSH_ALL);
+    for (i = 0; i < R128_TIMEOUT; i++) {
+	if (!(INREG(R128_PC_NGUI_CTLSTAT) & R128_PC_BUSY)) break;
+    }
+}
+
+/* FIXME: Requires access to secure registers */
+static void r128EngineReset(r128ScreenPtr pScrn)
+{
+#if USE_IOCTLS
+#if DEBUG
+    r128RingStatus(pScrn);
+#endif
+
+    (void)drmR128EngineReset(pScrn->driScreen->fd);
+#else
+    unsigned char *R128MMIO = pScrn->mmio;
+    CARD32        clock_cntl_index;
+    CARD32        mclk_cntl;
+    CARD32        gen_reset_cntl;
+	
+#if DEBUG
+    r128RingStatus(pScrn);
+#endif
+
+    r128EngineFlush(pScrn);
+
+    clock_cntl_index = INREG(R128_CLOCK_CNTL_INDEX);
+    mclk_cntl        = INPLL(pScrn, R128_MCLK_CNTL);
+
+    OUTPLL(R128_MCLK_CNTL, mclk_cntl | R128_FORCE_GCP | R128_FORCE_PIPE3D_CPP);
+
+    gen_reset_cntl   = INREG(R128_GEN_RESET_CNTL);
+
+    OUTREG(R128_GEN_RESET_CNTL, gen_reset_cntl | R128_SOFT_RESET_GUI);
+    INREG(R128_GEN_RESET_CNTL);
+    OUTREG(R128_GEN_RESET_CNTL, gen_reset_cntl & ~R128_SOFT_RESET_GUI);
+    INREG(R128_GEN_RESET_CNTL);
+
+    OUTPLL(R128_MCLK_CNTL,        mclk_cntl);
+    OUTREG(R128_CLOCK_CNTL_INDEX, clock_cntl_index);
+    OUTREG(R128_GEN_RESET_CNTL,   gen_reset_cntl);
+
+    /* Reset the ring buffer status */
+    if (R128CCE_USE_RING_BUFFER(pScrn->CCEMode)) {
+	OUTREG(R128_PM4_BUFFER_DL_WPTR, 0);
+	OUTREG(R128_PM4_BUFFER_DL_RPTR, 0);
+	pScrn->SAREA->ringWrite = 0;
+	*pScrn->ringReadPtr = 0;
+    }
+#endif
+    fprintf(stderr, "Error: Rage 128 timed out... exiting\n");
+    exit(-1);
+}
+
+#define r128CCEWaitForFifo(pScrn, entries)                                   \
+do {                                                                         \
+    if (pScrn->CCEFifoSlots < entries)                                       \
+	r128CCEWaitForFifoFunction(pScrn, entries);                          \
+    pScrn->CCEFifoSlots -= entries;                                          \
+} while (0)
+
+/* Wait for at least `entries' slots are free.  The actual number of
+   slots available is stored in info->CCEFifoSize. */
+static void r128CCEWaitForFifoFunction(r128ScreenPtr pScrn, int entries)
+{
+    unsigned char *R128MMIO = pScrn->mmio;
+    int           i;
+
+    for (;;) {
+	for (i = 0; i < R128_TIMEOUT; i++) {
+	    pScrn->CCEFifoSlots = INREG(R128_PM4_STAT) & R128_PM4_FIFOCNT_MASK;
+	    if (pScrn->CCEFifoSlots >= entries) return;
+	}
+	R128_DEBUG(("Reseting Engine: PM4_STAT = 0x%08x\n",
+		    (unsigned int)INREG(R128_PM4_STAT)));
+	r128EngineReset(pScrn);
+    }
+}
+
+/* Wait until the CCE is completely idle: the FIFO has drained and the
+   CCE is idle. */
+void r128CCEWaitForIdle(r128ScreenPtr pScrn)
+{
+#if USE_IOCTLS
+    if (drmR128WaitForIdle(pScrn->driScreen->fd) < 0) {
+	fprintf(stderr, "Error: Rage 128 timed out... exiting\n");
+	exit(-1);
+    }
+#else
+    unsigned char *R128MMIO = pScrn->mmio;
+    int           i;
+
+    if (R128CCE_USE_RING_BUFFER(pScrn->CCEMode)) {
+	OUTREGP(R128_PM4_BUFFER_DL_WPTR,
+		R128_PM4_BUFFER_DL_DONE, ~R128_PM4_BUFFER_DL_DONE);
+
+	for (;;) {
+	    for (i = 0; i < R128_TIMEOUT; i++) {
+		if (*pScrn->ringReadPtr == pScrn->SAREA->ringWrite) {
+		    int pm4stat = INREG(R128_PM4_STAT);
+		    if ((pm4stat & R128_PM4_FIFOCNT_MASK) == pScrn->CCEFifoSize
+			&& !(pm4stat & (R128_PM4_BUSY | R128_PM4_GUI_ACTIVE)))
+			return;
+		}
+	    }
+	    R128_DEBUG(("Reseting Engine: PM4_STAT = 0x%08x\n",
+			(unsigned int)INREG(R128_PM4_STAT)));
+	    r128EngineReset(pScrn);
+	}
+    } else {
+	r128CCEWaitForFifoFunction(pScrn, pScrn->CCEFifoSize);
+
+	for (;;) {
+	    for (i = 0; i < R128_TIMEOUT; i++) {
+		if (!(INREG(R128_PM4_STAT)
+		      & (R128_PM4_BUSY | R128_PM4_GUI_ACTIVE))) {
+		    r128EngineFlush(pScrn);
+		    return;
+		}
+	    }
+	    R128_DEBUG(("Reseting Engine: PM4_STAT = 0x%08x\n",
+			(unsigned int)INREG(R128_PM4_STAT)));
+	    r128EngineReset(pScrn);
+	}
+    }
+#endif
+}
+
+/* Send a packet to the CCE via the PIO registers */
+static void r128CCESubmitPacketPIO(r128ContextPtr r128ctx,
+				   CARD32 *buf, int count)
+{
+    unsigned char *R128MMIO     = r128ctx->r128Screen->mmio;
+#if DO_SECURITY_CHECK
+    CARD32         tmp          = 0;
+    int            psize        = 0;
+    int            writing      = 1;
+    int            addr         = R128_PM4_FIFO_DATA_EVEN;
+#endif
+
+#if DO_SECURITY_CHECK
+    while (count > 0) {
+	tmp = *buf++;
+	if (!psize) {
+	    writing = 1;
+
+	    if ((tmp & R128_CCE_PACKET_MASK) == R128_CCE_PACKET0) {
+		if ((tmp & R128_CCE_PACKET0_REG_MASK) <= (0x1004 >> 2)) {
+		    if ((tmp & R128_CCE_PACKET0_REG_MASK) !=
+			(R128_PM4_VC_FPU_SETUP >> 2)) {
+			writing = 0;
+		    }
+		}
+		psize = ((tmp & R128_CCE_PACKET_COUNT_MASK) >> 16) + 2;
+	    } else if ((tmp & R128_CCE_PACKET_MASK) == R128_CCE_PACKET1) {
+		if ((tmp & R128_CCE_PACKET1_REG0_MASK) <= (0x1004 >> 2)) {
+		    if ((tmp & R128_CCE_PACKET1_REG0_MASK) !=
+			(R128_PM4_VC_FPU_SETUP >> 2)) {
+			writing = 0;
+		    }
+		} else if ((tmp & R128_CCE_PACKET1_REG1_MASK) <=
+			   (0x1004 << 9)) {
+		    if ((tmp & R128_CCE_PACKET1_REG1_MASK) !=
+			(R128_PM4_VC_FPU_SETUP << 9)) {
+			writing = 0;
+		    }
+		}
+		psize = 3;
+	    } else {
+		psize = ((tmp & R128_CCE_PACKET_COUNT_MASK) >> 16) + 2;
+	    }
+	}
+	psize--;
+
+	if (writing) {
+	    r128CCEWaitForFifo(r128ctx->r128Screen, 1);
+	    OUTREG(addr, tmp);
+	    addr ^= 0x0004;
+	}
+
+	count--;
+    }
+
+    if (addr == R128_PM4_FIFO_DATA_ODD) {
+	    r128CCEWaitForFifo(r128ctx->r128Screen, 1);
+	    OUTREG(addr, R128_CCE_PACKET2);
+    }
+#else
+    while (count > 1) {
+	r128CCEWaitForFifo(r128ctx->r128Screen, 2);
+	OUTREG(R128_PM4_FIFO_DATA_EVEN, *buf++);
+	OUTREG(R128_PM4_FIFO_DATA_ODD,  *buf++);
+	count -= 2;
+    }
+
+    if (count) {
+	r128CCEWaitForFifo(r128ctx->r128Screen, 2);
+	OUTREG(R128_PM4_FIFO_DATA_EVEN, *buf);
+	OUTREG(R128_PM4_FIFO_DATA_ODD,  R128_CCE_PACKET2);
+    }
+#endif
+}
+
+/* Send a packet to the CCE via the ring */
+static void r128CCESubmitPacketRing(r128ContextPtr r128ctx,
+				    CARD32 *buf, int count)
+{
+    r128ScreenPtr     r128scrn  = r128ctx->r128Screen;
+    unsigned char    *R128MMIO  = r128ctx->r128Screen->mmio;
+    int            ringWrite    = r128scrn->SAREA->ringWrite;
+    int           *ringWritePtr = r128scrn->ringStartPtr + ringWrite;
+    int            timeout;
+#if DO_SECURITY_CHECK
+    CARD32         tmp          = 0;
+    int            psize        = 0;
+    int            writing      = 1;
+#endif
+
+    if (count > r128scrn->ringEntries) {
+	/* FIXME */
+	return;
+    }
+
+    while (count > 0) {
+#if DO_SECURITY_CHECK
+	tmp = *buf++;
+	if (!psize) {
+	    writing = 1;
+
+	    if ((tmp & R128_CCE_PACKET_MASK) == R128_CCE_PACKET0) {
+		if ((tmp & R128_CCE_PACKET0_REG_MASK) <= (0x1004 >> 2)) {
+		    if ((tmp & R128_CCE_PACKET0_REG_MASK) !=
+			(R128_PM4_VC_FPU_SETUP >> 2)) {
+			writing = 0;
+		    }
+		}
+		psize = ((tmp & R128_CCE_PACKET_COUNT_MASK) >> 16) + 2;
+	    } else if ((tmp & R128_CCE_PACKET_MASK) == R128_CCE_PACKET1) {
+		if ((tmp & R128_CCE_PACKET1_REG0_MASK) <= (0x1004 >> 2)) {
+		    if ((tmp & R128_CCE_PACKET1_REG0_MASK) !=
+			(R128_PM4_VC_FPU_SETUP >> 2)) {
+			writing = 0;
+		    }
+		} else if ((tmp & R128_CCE_PACKET1_REG1_MASK) <=
+			   (0x1004 << 9)) {
+		    if ((tmp & R128_CCE_PACKET1_REG1_MASK) !=
+			(R128_PM4_VC_FPU_SETUP << 9)) {
+			writing = 0;
+		    }
+		}
+		psize = 3;
+	    } else {
+		psize = ((tmp & R128_CCE_PACKET_COUNT_MASK) >> 16) + 2;
+	    }
+	}
+	psize--;
+
+	if (writing) {
+	    ringWrite++;
+	    *ringWritePtr++ = tmp;
+	}
+#else
+	ringWrite++;
+	*ringWritePtr++ = *buf++;
+#endif
+	if (ringWrite >= r128scrn->ringEntries) {
+	    ringWrite    = 0;
+	    ringWritePtr = r128scrn->ringStartPtr;
+	}
+	timeout = 0;
+	while (ringWrite == *r128scrn->ringReadPtr) {
+	    (void)INREG(R128_PM4_BUFFER_DL_RPTR);
+	    if (timeout++ >= R128_TIMEOUT) r128EngineReset(r128scrn);
+	}
+
+	count--;
+    }
+
+    if (ringWrite < 32) {
+	memcpy(r128scrn->ringEndPtr,
+	       r128scrn->ringStartPtr,
+	       ringWrite * sizeof(int));
+    }
+
+    r128scrn->SAREA->ringWrite = ringWrite;
+    OUTREG(R128_PM4_BUFFER_DL_WPTR, ringWrite);
+}
+
+void r128CCESubmitPackets(r128ContextPtr r128ctx, CARD32 *buf, int count)
+{
+#if USE_IOCTLS
+    if (drmR128SubmitPackets(r128ctx->r128Screen->driScreen->fd,
+			     r128ctx->CCEbuf, r128ctx->CCEcount,
+			     0) < 0) {
+	r128EngineReset(r128ctx->r128Screen);
+	fprintf(stderr, "Error: Illegal CCE packet... exiting\n");
+	exit(-1);
+    }
+#else
+    if (R128CCE_USE_RING_BUFFER(r128ctx->r128Screen->CCEMode))
+	r128CCESubmitPacketRing(r128ctx, buf, count);
+    else
+	r128CCESubmitPacketPIO(r128ctx, buf, count);
+#endif
+}
+
+#endif
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_cce.h b/xc/lib/GL/mesa/src/drv/r128/r128_cce.h
new file mode 100644
index 000000000..162686cb7
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_cce.h
@@ -0,0 +1,142 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_CCE_H_
+#define _R128_CCE_H_
+
+#ifdef GLX_DIRECT_RENDERING
+
+#include "r128_dri.h"
+#include "r128_reg.h"
+
+#include "xf86drmR128.h"
+
+typedef union {
+    float f;
+    int   i;
+} floatTOint;
+
+#define R128_DEFAULT_TOTAL_CCE_TIMEOUT 1000000 /* usecs */
+
+/* Insert an integer value into the CCE ring buffer. */
+#define R128CCE(v)                                                           \
+    do {                                                                     \
+	r128ctx->CCEbuf[r128ctx->CCEcount] = (v);                            \
+	r128ctx->CCEcount++;                                                 \
+    } while (0)
+
+/* Insert an floating point value into the CCE ring buffer. */
+#define R128CCEF(v)                                                          \
+    do {                                                                     \
+	floatTOint fTi;                                                      \
+	fTi.f = (v);                                                         \
+	r128ctx->CCEbuf[r128ctx->CCEcount] = fTi.i;                          \
+	r128ctx->CCEcount++;                                                 \
+    } while (0)
+
+#if USE_USER_SPACE_RING
+
+#define R128CCE_SUBMIT_PACKETS()                                             \
+    do {                                                                     \
+	r128CCESubmitPackets(r128ctx, r128ctx->CCEbuf, r128ctx->CCEcount);   \
+	r128ctx->CCEcount = 0;                                               \
+    } while (0)
+
+#define R128CCE_WAIT_FOR_IDLE(r128ctx)                                       \
+    r128CCEWaitForIdle(r128ctx->r128Screen)
+
+#else /* if !USE_USER_SPACE_RING */
+
+#define R128CCE_SUBMIT_PACKETS()                                             \
+    do {                                                                     \
+	CARD32 *_buf;                                                        \
+	int     _c   = r128ctx->CCEcount;                                    \
+	int     _fd  = r128ctx->r128Screen->driScreen->fd;                   \
+	int     _to  = 0;                                                    \
+	int     _ret;                                                        \
+                                                                             \
+	do {                                                                 \
+	    _buf = r128ctx->CCEbuf + (r128ctx->CCEcount - _c);               \
+	    _ret = drmR128SubmitPackets(_fd, _buf, &_c, 0);                  \
+	} while (_ret < 0 && _ret == -EBUSY && _to++ < r128ctx->CCEtimeout); \
+	if (_ret < 0) {                                                      \
+	    (void)drmR128EngineReset(_fd);                                   \
+	    fprintf(stderr, "Error: Could not submit packet... exiting\n");  \
+	    exit(-1);                                                        \
+	}                                                                    \
+	r128ctx->CCEcount = 0;                                               \
+    } while (0)
+
+#define R128CCE_WAIT_FOR_IDLE(r128ctx)                                       \
+    do {                                                                     \
+	int    _fd  = r128ctx->r128Screen->driScreen->fd;                    \
+	int    _to  = 0;                                                     \
+	int    _ret;                                                         \
+                                                                             \
+	(void)drmR128EngineFlush(_fd);                                       \
+	do {                                                                 \
+	    _ret = drmR128CCEWaitForIdle(_fd);                               \
+	} while (_ret < 0 && _ret == -EBUSY && _to++ < r128ctx->CCEtimeout); \
+	if (_ret < 0) {                                                      \
+	    (void)drmR128EngineReset(_fd);                                   \
+	    fprintf(stderr, "Error: Rage 128 timed out... exiting\n");       \
+	    exit(-1);                                                        \
+	}                                                                    \
+    } while (0)
+
+#endif
+
+#define R128CCE_WAIT_FOR_IDLE_LOCK(r128ctx)                                  \
+    do {                                                                     \
+	LOCK_HARDWARE(r128ctx);                                              \
+	R128CCE_WAIT_FOR_IDLE(r128ctx);                                      \
+	UNLOCK_HARDWARE(r128ctx);                                            \
+    } while (0)
+
+
+/* Insert a type-[0123] packet header into the ring buffer */
+#define R128CCE0(p,r,n)   R128CCE((p) | ((n) << 16) | ((r) >> 2))
+#define R128CCE1(p,r1,r2) R128CCE((p) | (((r2) >> 2) << 11) | ((r1) >> 2))
+#define R128CCE2(p)       R128CCE((p))
+#define R128CCE3(p,n)     R128CCE((p) | ((n) << 16))
+
+
+#if USE_USER_SPACE_RING
+extern void r128CCEWaitForIdle(r128ScreenPtr pScrn);
+extern void r128CCESubmitPackets(r128ContextPtr r128ctx,
+				 CARD32 *buf, int count);
+#endif
+
+#endif
+#endif /* _R128_CCE_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_ccevb.h b/xc/lib/GL/mesa/src/drv/r128/r128_ccevb.h
new file mode 100644
index 000000000..2084b8672
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_ccevb.h
@@ -0,0 +1,224 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_CCEVB_H_
+#define _R128_CCEVB_H_
+
+#ifdef GLX_DIRECT_RENDERING
+
+#include "xf86drmR128.h"
+
+typedef struct {
+    CARD32          *buf;   /* Pointer to current VB */
+    int              index; /* Index of current VB */
+    int              start; /* in bytes from the beginning of the VB map */
+    int              size;  /* in vertices */
+    int              count; /* in vertices */
+    drmR128PrimType  prim;  /* Primitive type for vertices in VB */
+    int              done;  /* VB has been sent to ring */
+} r128CCEVertBuf, *r128CCEVertBufPtr;
+
+
+
+/* Flush the CPU's write-combining cache */
+/* FIXME: This code is both processor and compiler specific */
+#define R128_FLUSH_WC_MEMORY()                                               \
+    do {                                                                     \
+	int xchangeDummy;                                                    \
+                                                                             \
+	__asm__ volatile("push %%eax ;"                                      \
+			 "xchg %%eax, %0 ;"                                  \
+			 "pop %%eax" : : "m" (xchangeDummy));                \
+	__asm__ volatile("push %%eax ;"                                      \
+			 "push %%ebx ;"                                      \
+			 "push %%ecx ;"                                      \
+			 "push %%edx ;"                                      \
+			 "movl $0,%%eax ;"                                   \
+			 "cpuid ;"                                           \
+			 "pop %%edx ;"                                       \
+			 "pop %%ecx ;"                                       \
+			 "pop %%ebx ;"                                       \
+			 "pop %%eax" : /* no outputs */ : /* no inputs */);  \
+    } while (0)
+
+
+/* Get a new VB from the pool of vertex buffers in AGP space */
+/* NOTE: By default the primitive type for a VB is set to individual tris */
+#define R128CCE_GET_NEW_VB(r128ctx)                                          \
+    do {                                                                     \
+	r128CCEVertBufPtr  vb    = &r128ctx->vb;                             \
+	int                to    = 0;                                        \
+	int                fd    = r128ctx->r128Screen->driScreen->fd;       \
+	int                num;                                              \
+	int                index;                                            \
+	int                size;                                             \
+                                                                             \
+	vb->buf = NULL;                                                      \
+                                                                             \
+	LOCK_HARDWARE(r128ctx);                                              \
+	while (!vb->buf && to++ < r128ctx->CCEtimeout) {                     \
+	    /* Ask the kernel a new vertex buffer */                         \
+	    num = drmR128GetVertexBuffers(fd, 1, &index, &size);             \
+                                                                             \
+	    if (num > 0) {                                                   \
+		vb->buf   = (CARD32 *)(r128ctx->r128Screen->                 \
+				       vbBufs->list[index].address);         \
+		vb->index = index;                                           \
+		vb->start = index*r128ctx->r128Screen->vbBufSize;            \
+		vb->size  = size/sizeof(r128_vertex);                        \
+		vb->count = 0;                                               \
+		vb->prim  = DRM_R128_PRIM_TRI_LIST;                          \
+		vb->done  = GL_FALSE;                                        \
+	    }                                                                \
+	}                                                                    \
+	UNLOCK_HARDWARE(r128ctx);                                            \
+                                                                             \
+	if (!vb->buf) {                                                      \
+	    (void)drmR128EngineReset(fd);                                    \
+	    fprintf(stderr, "Error: Could not get new VB... exiting\n");     \
+	    exit(-1);                                                        \
+	}                                                                    \
+    } while (0)
+
+/* Submit a VB to the hardware for processing */
+/* NOTE: This macro is only called while holding the hardware lock */
+#define R128CCE_SUBMIT_VB(r128ctx)                                           \
+    do {                                                                     \
+	int       index = r128ctx->vb.index;                                 \
+	int       size  = r128ctx->vb.count;                                 \
+	int       fd    = r128ctx->r128Screen->driScreen->fd;                \
+	int       to    = 0;                                                 \
+	int       ret;                                                       \
+	CARD32    prim;                                                      \
+                                                                             \
+	switch (r128ctx->vb.prim) {                                          \
+	case DRM_R128_PRIM_NONE:                                             \
+	    prim = R128_CCE_VC_CNTL_PRIM_TYPE_NONE;                          \
+	    break;                                                           \
+	case DRM_R128_PRIM_POINT:                                            \
+	    prim = R128_CCE_VC_CNTL_PRIM_TYPE_POINT;                         \
+	    break;                                                           \
+	case DRM_R128_PRIM_LINE:                                             \
+	    prim = R128_CCE_VC_CNTL_PRIM_TYPE_LINE;                          \
+	    break;                                                           \
+	case DRM_R128_PRIM_POLY_LINE:                                        \
+	    prim = R128_CCE_VC_CNTL_PRIM_TYPE_POLY_LINE;                     \
+	    break;                                                           \
+	case DRM_R128_PRIM_TRI_LIST:                                         \
+	    prim = R128_CCE_VC_CNTL_PRIM_TYPE_TRI_LIST;                      \
+	    break;                                                           \
+	case DRM_R128_PRIM_TRI_FAN:                                          \
+	    prim = R128_CCE_VC_CNTL_PRIM_TYPE_TRI_FAN;                       \
+	    break;                                                           \
+	case DRM_R128_PRIM_TRI_STRIP:                                        \
+	    prim = R128_CCE_VC_CNTL_PRIM_TYPE_TRI_STRIP;                     \
+	    break;                                                           \
+	case DRM_R128_PRIM_TRI_TYPE2:                                        \
+	    prim = R128_CCE_VC_CNTL_PRIM_TYPE_TRI_TYPE2;                     \
+	    break;                                                           \
+	default:                                                             \
+	    prim = R128_CCE_VC_CNTL_PRIM_TYPE_NONE;                          \
+	    break;                                                           \
+	}                                                                    \
+                                                                             \
+	/* Send the vertex buffer to the hardware */                         \
+	BEGIN_CLIP_LOOP(r128ctx);                                            \
+                                                                             \
+	R128CCE3(R128_CCE_PACKET3_3D_RNDR_GEN_INDX_PRIM, 3);                 \
+	R128CCE(r128ctx->r128Screen->vbOffset + r128ctx->vb.start);          \
+	R128CCE(r128ctx->vb.count);                                          \
+	R128CCE(R128_FULL_VERTEX_FORMAT);                                    \
+	R128CCE(prim | R128_CCE_VC_CNTL_PRIM_WALK_LIST |                     \
+		(r128ctx->vb.count << R128_CCE_VC_CNTL_NUM_SHIFT));          \
+	R128CCE_SUBMIT_PACKETS();                                            \
+                                                                             \
+	END_CLIP_LOOP(r128ctx);                                              \
+                                                                             \
+	/* Tell the kernel to release the vertex buffer */                   \
+	do {                                                                 \
+	    ret = drmR128FlushVertexBuffers(fd, 1, &index, &size,            \
+					    r128ctx->vb.prim);               \
+	} while (ret < 0 && ret == -EBUSY && to++ < r128ctx->CCEtimeout);    \
+	if (ret < 0) {                                                       \
+	    (void)drmR128EngineReset(fd);                                    \
+	    fprintf(stderr, "Error: Could not flush VB... exiting\n");       \
+	    exit(-1);                                                        \
+	}                                                                    \
+                                                                             \
+	r128ctx->vb.done = GL_TRUE;                                          \
+    } while (0)
+
+/* Flush the vertex buffer (assuming we already hold the lock) */
+#define R128CCE_FLUSH_VB(r128ctx)                                            \
+    do {                                                                     \
+	if (r128ctx->vb.count && !r128ctx->vb.done) {                        \
+	    /* FIXME: Is this _really_ needed? */                            \
+	    if (!R128CCE_USE_RING_BUFFER(r128ctx->r128Screen->CCEMode))      \
+		R128CCE_WAIT_FOR_IDLE(r128ctx);                              \
+                                                                             \
+	    R128CCE_SUBMIT_VB(r128ctx);                                      \
+	}                                                                    \
+    } while (0)
+
+/* Flush the vertex buffer */
+#define R128CCE_FLUSH_VB_LOCK(r128ctx)                                       \
+    do {                                                                     \
+	LOCK_HARDWARE(r128ctx);                                              \
+	R128CCE_FLUSH_VB(r128ctx);                                           \
+	UNLOCK_HARDWARE(r128ctx);                                            \
+    } while (0)
+
+/* Reserve space in a VB to store `n' vertices, and set the pointer `p'
+   to point to the next vertex in the current VB.  If adding these
+   vertices to the VB would exceed the size of the VB, then flush the
+   current VB, get a new one from the pool of VBs, reserve space in the
+   new VB for the `n' vertices, and set the pointer `p' to point to the
+   first vertex in the new VB. */
+#define R128CCE_ALLOC_VB_SPACE(r128ctx, p, n)                                \
+    do {                                                                     \
+	r128CCEVertBufPtr  vb = &r128ctx->vb;                                \
+                                                                             \
+	if (vb->done) R128CCE_GET_NEW_VB(r128ctx);                           \
+                                                                             \
+	if (vb->count + (n) > vb->size) {                                    \
+	    R128CCE_FLUSH_VB_LOCK(r128ctx);                                  \
+	    R128CCE_GET_NEW_VB(r128ctx);                                     \
+	}                                                                    \
+                                                                             \
+	(p)        = (r128_vertex *)vb->buf + vb->count;                     \
+	vb->count += (n);                                                    \
+    } while (0)
+
+#endif
+#endif /* _R128_CCEVB_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_clear.c b/xc/lib/GL/mesa/src/drv/r128/r128_clear.c
new file mode 100644
index 000000000..9ab0581b0
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_clear.c
@@ -0,0 +1,248 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#include "r128_init.h"
+#include "r128_mesa.h"
+#include "r128_xmesa.h"
+#include "r128_context.h"
+#include "r128_lock.h"
+#include "r128_reg.h"
+#include "r128_cce.h"
+#include "r128_state.h"
+#include "r128_vb.h"
+#include "r128_clear.h"
+
+/* Clear the depth buffer */
+void r128ClearDepthBuffer(r128ContextPtr r128ctx, GLboolean all,
+			  GLint cx, GLint cy, GLint cw, GLint ch)
+{
+    __DRIdrawablePrivate *dPriv    = r128ctx->driDrawable;
+    int                   nc;
+    XF86DRIClipRectPtr    c;
+    int                   dst_bpp;
+    CARD32                write_mask;
+
+    if (!(r128ctx->regs.tex_cntl_c & R128_Z_WRITE_ENABLE)) return;
+
+    switch (r128ctx->regs.z_sten_cntl_c & R128_Z_PIX_WIDTH_MASK) {
+    case R128_Z_PIX_WIDTH_16:
+	write_mask = 0x0000ffff;
+	dst_bpp    = R128_GMC_DST_16BPP;
+	break;
+    case R128_Z_PIX_WIDTH_24:
+	write_mask = 0x00ffffff;
+	dst_bpp    = R128_GMC_DST_32BPP;
+	break;
+    case R128_Z_PIX_WIDTH_32:
+	write_mask = 0xffffffff;
+	dst_bpp    = R128_GMC_DST_32BPP;
+	break;
+    default: return;
+    }
+
+    cx += dPriv->x;
+    cy  = dPriv->y + dPriv->h - cy - ch;
+
+    LOCK_HARDWARE(r128ctx);
+
+    /* Flush any outstanding vertex buffers */
+    R128CCE_FLUSH_VB(r128ctx);
+
+    /* Init the clip rects here in case they changed during the
+       LOCK_HARDWARE macro */
+    c  = dPriv->pClipRects;
+    nc = dPriv->numClipRects;
+
+    /* Set the write mask so that we _only_ clear the Z buffer */
+    R128CCE0(R128_CCE_PACKET0, R128_DP_WRITE_MASK, 0);
+    R128CCE(write_mask);
+
+    /* Temporarily disable Z and stencil buffer and texture mapping modes */
+    R128CCE0(R128_CCE_PACKET0, R128_TEX_CNTL_C, 0);
+    R128CCE(r128ctx->regs.tex_cntl_c & ~(R128_Z_ENABLE |
+					 R128_STENCIL_ENABLE |
+					 R128_TEXMAP_ENABLE));
+
+    /* Cycle through the clip rects */
+    while (nc--) {
+	int x = c[nc].x1;
+	int y = c[nc].y1;
+	int w = c[nc].x2 - x;
+	int h = c[nc].y2 - y;
+
+	if (!all) {
+	    if (x < cx) w -= cx - x, x = cx;
+	    if (y < cy) h -= cy - y, y = cy;
+
+	    if (x + w > cx + cw) w = cx + cw - x;
+	    if (y + h > cy + ch) h = cy + ch - y;
+
+	    if (w <= 0 || h <= 0) continue;
+	}
+
+	x += r128ctx->r128Screen->depthX;
+	y += r128ctx->r128Screen->depthY;
+
+	R128CCE3(R128_CCE_PACKET3_CNTL_PAINT_MULTI, 3);
+	R128CCE(R128_GMC_BRUSH_SOLID_COLOR
+		| dst_bpp
+		| R128_GMC_SRC_DATATYPE_COLOR
+		| R128_ROP3_P
+		| R128_GMC_3D_FCN_EN            /* FIXME?? */
+		| R128_GMC_CLR_CMP_CNTL_DIS     /* FIXME?? */
+		| R128_AUX_CLIP_DIS             /* FIXME?? */
+		| R128_GMC_WR_MSK_DIS);         /* FIXME?? */
+	R128CCE(r128ctx->ClearDepth);
+	R128CCE((x << 16) | y);
+	R128CCE((w << 16) | h);
+    }
+
+#if 0
+    /* Set the write mask so that we _only_ clear the Z buffer */
+    R128CCE0(R128_CCE_PACKET0, R128_DP_WRITE_MASK, 0);
+    R128CCE(0xffffffff);
+
+    /* Restore Z and stencil buffer and texture mapping modes */
+    R128CCE0(R128_CCE_PACKET0, R128_TEX_CNTL_C, 0);
+    R128CCE(r128ctx->regs.tex_cntl_c);
+#else
+    /* FIXME: We should be able to optimize this by restoring only the
+       registers that change (above) */
+    /* NOTE: The restore of TEX_CNTL_C and R128_DP_WRITE_MASK is handled by
+    vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv             */
+    r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+    r128ctx->dirty_context |= R128_CTX_ALL_DIRTY;
+#endif
+
+    R128CCE_SUBMIT_PACKETS();
+
+    UNLOCK_HARDWARE(r128ctx);
+}
+
+/* Clear a color buffer */
+void r128ClearColorBuffer(r128ContextPtr r128ctx, GLboolean all,
+			  GLint cx, GLint cy, GLint cw, GLint ch,
+			  GLint drawX, GLint drawY)
+{
+    __DRIdrawablePrivate *dPriv    = r128ctx->driDrawable;
+    int                   nc;
+    XF86DRIClipRectPtr    c;
+    int                   dst_bpp;
+
+    switch (r128ctx->r128Screen->bpp) {
+    case 8:
+	dst_bpp = R128_GMC_DST_8BPP_CI;
+	break;
+    case 16:
+	if (r128ctx->r128Screen->depth == 15) dst_bpp = R128_GMC_DST_15BPP;
+	else                                  dst_bpp = R128_GMC_DST_16BPP;
+	break;
+    case 24:
+	dst_bpp = R128_GMC_DST_24BPP;
+	break;
+    case 32:
+    default:
+	dst_bpp = R128_GMC_DST_32BPP;
+	break;
+    }
+
+    cx += dPriv->x;
+    cy  = dPriv->y + dPriv->h - cy - ch;
+
+    LOCK_HARDWARE(r128ctx);
+
+    /* Flush any outstanding vertex buffers */
+    R128CCE_FLUSH_VB(r128ctx);
+
+    /* Init the clip rects here in case they changed during the
+       LOCK_HARDWARE macro */
+    c  = dPriv->pClipRects;
+    nc = dPriv->numClipRects;
+
+    /* Temporarily disable Z and stencil buffer and texture mapping modes */
+    R128CCE0(R128_CCE_PACKET0, R128_TEX_CNTL_C, 0);
+    R128CCE(r128ctx->regs.tex_cntl_c & ~(R128_Z_ENABLE |
+					 R128_STENCIL_ENABLE |
+					 R128_TEXMAP_ENABLE));
+
+    /* Cycle through the clip rects */
+    while (nc--) {
+	int x = c[nc].x1;
+	int y = c[nc].y1;
+	int w = c[nc].x2 - x;
+	int h = c[nc].y2 - y;
+
+	if (!all) {
+	    if (x < cx) w -= cx - x, x = cx;
+	    if (y < cy) h -= cy - y, y = cy;
+
+	    if (x + w > cx + cw) w = cx + cw - x;
+	    if (y + h > cy + ch) h = cy + ch - h;
+
+	    if (w <= 0 || h <= 0) continue;
+	}
+
+	x += drawX;
+	y += drawY;
+
+	R128CCE3(R128_CCE_PACKET3_CNTL_PAINT_MULTI, 3);
+	R128CCE(R128_GMC_BRUSH_SOLID_COLOR
+		| dst_bpp
+		| R128_GMC_SRC_DATATYPE_COLOR
+		| R128_ROP3_P
+		| R128_GMC_3D_FCN_EN            /* FIXME?? */
+		| R128_GMC_CLR_CMP_CNTL_DIS     /* FIXME?? */
+		| R128_AUX_CLIP_DIS);           /* FIXME?? */
+	R128CCE(r128ctx->ClearColor);
+	R128CCE((x << 16) | y);
+	R128CCE((w << 16) | h);
+    }
+
+#if 0
+    /* Restore Z and stencil buffer and texture mapping modes */
+    R128CCE0(R128_CCE_PACKET0, R128_TEX_CNTL_C, 0);
+    R128CCE(r128ctx->regs.tex_cntl_c);
+#else
+    /* FIXME: We should be able to optimize this by restoring only the
+       registers that change (above) */
+    /* NOTE: The restore of TEX_CNTL_C is handled by
+    vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv             */
+    r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+    r128ctx->dirty_context |= R128_CTX_ALL_DIRTY;
+#endif
+
+    R128CCE_SUBMIT_PACKETS();
+
+    UNLOCK_HARDWARE(r128ctx);
+}
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_clear.h b/xc/lib/GL/mesa/src/drv/r128/r128_clear.h
new file mode 100644
index 000000000..7018b05b4
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_clear.h
@@ -0,0 +1,47 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_CLEAR_H_
+#define _R128_CLEAR_H_
+
+#ifdef GLX_DIRECT_RENDERING
+
+extern void r128ClearDepthBuffer(r128ContextPtr r128ctx, GLboolean all,
+				 GLint x, GLint y, GLint width, GLint height);
+extern void r128ClearColorBuffer(r128ContextPtr r128ctx, GLboolean all,
+				 GLint x, GLint y, GLint width, GLint height,
+				 GLint drawX, GLint drawY);
+
+#endif
+#endif /* _R128_CLEAR_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_context.c b/xc/lib/GL/mesa/src/drv/r128/r128_context.c
new file mode 100644
index 000000000..01573f559
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_context.c
@@ -0,0 +1,203 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#include <stdlib.h>
+
+#include "r128_init.h"
+#include "r128_mesa.h"
+#include "r128_xmesa.h"
+#include "r128_context.h"
+#include "r128_cce.h"
+#include "r128_dd.h"
+#include "r128_state.h"
+#include "r128_span.h"
+#include "r128_tex.h"
+#include "r128_vb.h"
+#include "r128_pipeline.h"
+
+#include "context.h"
+#include "simple_list.h"
+
+/* Create the device specific context */
+GLboolean r128CreateContext(Display *dpy, GLvisual *glVisual,
+			    __DRIcontextPrivate *driContextPriv)
+{
+    r128ContextPtr  r128ctx;
+    r128ScreenPtr   r128scrn;
+    GLcontext      *glCtx    = driContextPriv->mesaContext;
+    int             i;
+    char           *v;
+
+    r128ctx = (r128ContextPtr)Xmalloc(sizeof(*r128ctx));
+    if (!r128ctx) return GL_FALSE;
+
+    /* Initialize r128Context */
+    r128ctx->glCtx         = glCtx;
+    r128ctx->display       = dpy;
+    r128ctx->driContext    = driContextPriv;
+    r128ctx->driDrawable   = NULL; /* Set by XMesaMakeCurrent */
+
+    if (getenv("LIBGL_SOFTWARE_RENDERING"))
+	r128ctx->SWonly    = GL_TRUE;
+    else
+	r128ctx->SWonly    = GL_FALSE;
+    if (getenv("LIBGL_NO_SOFTWARE_FALLBACKS"))
+	r128ctx->SWfallbackDisable = GL_TRUE;
+    else
+	r128ctx->SWfallbackDisable = GL_FALSE;
+    r128scrn = r128ctx->r128Screen =
+	(r128ScreenPtr)(driContextPriv->driScreenPriv->private);
+
+    r128ctx->CurrentTexObj[0] = NULL;
+    r128ctx->CurrentTexObj[1] = NULL;
+    make_empty_list(&r128ctx->SwappedOut);
+    for (i = 0; i < r128scrn->NRTexHeaps; i++) {
+	make_empty_list(&r128ctx->TexObjList[i]);
+	r128ctx->texHeap[i] = mmInit(0, r128scrn->texSize[i]);
+	r128ctx->lastTexAge[i] = -1;
+    }
+
+    r128ctx->lastSwapAge      = 0;
+
+    r128ctx->useFastPath      = GL_FALSE;
+
+    r128ctx->vb.start         = 0;
+    r128ctx->vb.count         = 0;
+    r128ctx->vb.size          = 0;
+    r128ctx->vb.index         = 0;
+    r128ctx->vb.buf           = NULL;
+    r128ctx->vb.done          = GL_TRUE;
+
+    if (r128scrn->IsPCI || getenv("LIBGL_DISABLE_VERTEX_BUFFERS"))
+	r128ctx->disableVB    = GL_TRUE;
+    else
+	r128ctx->disableVB    = GL_FALSE;
+
+    r128ctx->CCEbuf           = (CARD32 *)malloc(sizeof(*r128ctx->CCEbuf) *
+						 r128scrn->ringEntries);
+    r128ctx->CCEcount         = 0;
+
+    if ((v = getenv("LIBGL_CCE_TIMEOUT")))
+	r128ctx->CCEtimeout   = strtoul(v, NULL, 10);
+    else
+	r128ctx->CCEtimeout   = (R128_DEFAULT_TOTAL_CCE_TIMEOUT /
+				 R128_DEFAULT_CCE_TIMEOUT);
+    if (r128ctx->CCEtimeout <= 0) r128ctx->CCEtimeout = 1;
+
+    /* Initialize GLcontext */
+    glCtx->DriverCtx = (void *)r128ctx;
+
+    r128DDInitExtensions(glCtx);
+
+    r128DDInitDriverFuncs(glCtx);
+    r128DDInitStateFuncs(glCtx);
+    r128DDInitSpanFuncs(glCtx);
+    r128DDInitTextureFuncs(glCtx);
+
+    glCtx->Driver.TriangleCaps = (DD_TRI_CULL
+				  | DD_TRI_LIGHT_TWOSIDE
+				  | DD_TRI_OFFSET);
+#if 0
+    /* FIXME */
+    glCtx->TriangleCaps |= DD_CLIP_FOG_COORD;
+#endif
+
+    /* Reset Mesa's current 2D texture pointers to the driver's textures */
+    glCtx->Shared->DefaultD[2][0].DriverData = NULL;
+    glCtx->Shared->DefaultD[2][1].DriverData = NULL;
+
+    /* If Mesa has current a vertex buffer, make sure the driver's VB
+       data is up to date */
+    if (glCtx->VB) r128DDRegisterVB(glCtx->VB);
+
+    /* Register the fast path */
+    if (glCtx->NrPipelineStages)
+	glCtx->NrPipelineStages =
+	    r128RegisterPipelineStages(glCtx->PipelineStage,
+				       glCtx->PipelineStage,
+				       glCtx->NrPipelineStages);
+
+    r128DDInitState(r128ctx);
+
+    driContextPriv->driverPrivate = (void *)r128ctx;
+
+    return GL_TRUE;
+}
+
+/* Destroy the device specific context */
+void r128DestroyContext(r128ContextPtr r128ctx)
+{
+    if (r128ctx) {
+	r128TexObjPtr t, next_t;
+	int           i;
+
+	free(r128ctx->CCEbuf);
+
+	for (i = 0; i < r128ctx->r128Screen->NRTexHeaps; i++) {
+	    foreach_s (t, next_t, &r128ctx->TexObjList[i])
+		r128DestroyTexObj(r128ctx, t);
+	}
+
+	foreach_s (t, next_t, &r128ctx->SwappedOut)
+	    r128DestroyTexObj(r128ctx, t);
+
+	gl_destroy_context(r128ctx->glCtx);
+	Xfree(r128ctx);
+    }
+}
+
+/* Load the device specific context into the hardware.  The actual
+   setting of the hardware state is done in the r128UpdateHWState(). */
+r128ContextPtr r128MakeCurrent(r128ContextPtr oldCtx,
+			       r128ContextPtr newCtx,
+			       __DRIdrawablePrivate *dPriv)
+{
+    if (oldCtx) {
+	if (!R128CCE_USE_RING_BUFFER(newCtx->r128Screen->CCEMode))
+	    newCtx->dirty         |= R128_REQUIRE_QUIESCENCE;
+	if (oldCtx != newCtx) {
+	    newCtx->dirty         |= R128_UPDATE_CONTEXT;
+	    newCtx->dirty_context |= R128_CTX_ALL_DIRTY;
+	}
+	if (oldCtx->driDrawable != dPriv)
+	    newCtx->dirty         |= R128_UPDATE_WINPOS;
+    } else {
+	newCtx->dirty             |= R128_UPDATE_CONTEXT;
+	newCtx->dirty_context     |= R128_CTX_ALL_DIRTY;
+    }
+
+    newCtx->driDrawable = dPriv;
+
+    return newCtx;
+}
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_context.h b/xc/lib/GL/mesa/src/drv/r128/r128_context.h
new file mode 100644
index 000000000..189a22cbe
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_context.h
@@ -0,0 +1,219 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_CONTEXT_H_
+#define _R128_CONTEXT_H_
+
+#ifdef GLX_DIRECT_RENDERING
+
+#include "r128_texobj.h"
+#include "r128_fastpath.h"
+#include "r128_ccevb.h"
+
+/* Flags for what needs to be updated before a new primitive is rendered */
+#define R128_CLEAN                  0x0000
+#define R128_REQUIRE_QUIESCENCE     0x0001
+#define R128_UPDATE_CONTEXT         0x0002
+#define R128_UPDATE_WINPOS          0x0004
+#define R128_UPDATE_TEX0IMAGES      0x0008
+#define R128_UPDATE_TEX1IMAGES      0x0010
+#define R128_UPDATE_TEXSTATE        0x0020
+#define R128_ALL_DIRTY              0xffff
+
+/* Flags for what context state needs to be updated */
+#define R128_CTX_CLEAN              0x0000
+#define R128_CTX_MISC               0x0001
+#define R128_CTX_ENGINESTATE        0x0002
+#define R128_CTX_TEX0STATE          0x0004
+#define R128_CTX_TEX1STATE          0x0008
+#define R128_CTX_TEXENVSTATE        0x0010
+#define R128_CTX_FOGSTATE           0x0020
+#define R128_CTX_FOGTABLE           0x0040
+#define R128_CTX_ZSTENSTATE         0x0080
+#define R128_CTX_SCISSORS           0x0100
+#define R128_CTX_ALPHASTATE         0x0200
+#define R128_CTX_SETUPSTATE         0x0400
+#define R128_CTX_WIN_Z_POS          0x0800
+#define R128_CTX_FLUSH_PIX_CACHE    0x1000
+#define R128_CTX_ALL_DIRTY          0xffff
+
+/* Flags for software fallback cases */
+#define R128_FALLBACK_TEXTURE       0x0001
+#define R128_FALLBACK_DRAW_BUFFER   0x0002
+#define R128_FALLBACK_READ_BUFFER   0x0004
+#define R128_FALLBACK_COLORMASK     0x0008
+#define R128_FALLBACK_STIPPLE       0x0010
+
+/* NOTE: The groups below need to be kept together so that a single
+   memcpy can be used to transfer data to the ring buffer */
+typedef struct {
+    CARD32  scale_3d_cntl;                        /* 0x1a00 */
+
+    CARD32  aux_sc_cntl;                          /* 0x1660 */
+    CARD32  aux1_sc_left;
+    CARD32  aux1_sc_right;
+    CARD32  aux1_sc_top;
+    CARD32  aux1_sc_bottom;
+    CARD32  aux2_sc_left;
+    CARD32  aux2_sc_right;
+    CARD32  aux2_sc_top;
+    CARD32  aux2_sc_bottom;
+    CARD32  aux3_sc_left;
+    CARD32  aux3_sc_right;
+    CARD32  aux3_sc_top;
+    CARD32  aux3_sc_bottom;                       /* 0x1690 */
+
+    CARD32  dst_pitch_offset_c;                   /* 0x1c80 */
+    CARD32  dp_gui_master_cntl;
+    CARD32  sc_top_left_c;
+    CARD32  sc_bottom_right_c;
+    CARD32  z_offset_c;
+    CARD32  z_pitch_c;
+    CARD32  z_sten_cntl_c;
+    CARD32  tex_cntl_c;
+    CARD32  misc_3d_state_cntl_reg;
+    CARD32  texture_clr_cmp_clr_c;
+    CARD32  texture_clr_cmp_msk_c;
+    CARD32  fog_color_c;
+    CARD32  prim_tex_cntl_c;
+    CARD32  prim_texture_combine_cntl_c;
+    CARD32  tex_size_pitch_c;
+    CARD32  prim_tex_offset[R128_TEX_MAXLEVELS];  /* 0x1ce4 */
+
+    CARD32  sec_tex_cntl_c;                       /* 0x1d00 */
+    CARD32  sec_tex_combine_cntl_c;
+    CARD32  sec_tex_offset[R128_TEX_MAXLEVELS];
+    CARD32  constant_color_c;
+    CARD32  prim_texture_border_color_c;
+    CARD32  sec_texture_border_color_c;
+    CARD32  sten_ref_mask_c;
+    CARD32  plane_3d_mask_c;                      /* 0x1d44 */
+
+    CARD32  setup_cntl;                           /* 0x1bc4 */
+
+    CARD32  pm4_vc_fpu_setup;                     /* 0x071c */
+
+    CARD32  fog_3d_table_start;                   /* 0x1810 */
+    CARD32  fog_3d_table_end;
+    CARD32  fog_3d_table_density;                 /* 0x181c */
+
+    CARD32  window_xy_offset;                     /* 0x1bcc */
+
+    CARD32  dp_write_mask;                        /* 0x16cc */
+
+    CARD32  pc_gui_ctlstat;                       /* 0x1748 */
+} r128ContextRegs;
+
+typedef struct {
+    GLcontext            *glCtx;	/* Mesa context */
+    int                   dirty;	/* Hardware state to be updated */
+    int                   dirty_context; /* Context state to be updated */
+
+    int                   SWonly;	/* Force software-only rendering */
+    int                   SWfallbackDisable; /* Disable software fallbacks */
+
+    r128TexObjPtr         CurrentTexObj[2]; /* Ptr to current texture
+					       object associated with
+					       each texture unit */
+					/* List of tex swapped in per heap*/
+    r128TexObj            TexObjList[R128_NR_TEX_HEAPS];
+    r128TexObj            SwappedOut;	/* List of textures swapped out */
+    memHeap_t            *texHeap[R128_NR_TEX_HEAPS]; /* Global tex heaps */
+					/* Last known global tex heap ages */
+    int                   lastTexAge[R128_NR_TEX_HEAPS];
+
+    CARD32                lastSwapAge;  /* Last known swap age */
+
+    GLenum                FogMode;      /* Current fog equation */
+
+    int                   Scissor;
+    XF86DRIClipRectRec    ScissorRect;  /* Current software scissor */
+
+    int                   useFastPath;	/* Currently using Fast Path code */
+    int                   SetupIndex;	/* Raster setup function index */
+    int                   SetupDone;	/* Partial raster setup done? */
+    int                   RenderIndex;	/* Render state function index */
+    r128InterpFunc        interp;	/* Current vert interp function */
+
+    r128CCEVertBuf        vb;		/* VB currently being filled */
+    int                   disableVB;    /* Disable the use of vertex buffers */
+
+    points_func           PointsFunc;	/* Current Points, Line, Triangle */
+    line_func             LineFunc;	/* and Quad rendering functions */
+    triangle_func         TriangleFunc;
+    quad_func             QuadFunc;
+
+    CARD32                IndirectTriangles; /* Flags for point, line,
+						tri and quad software
+						fallbacks */
+    CARD32                Fallback;     /* Need software fallback */
+
+    r128ContextRegs       regs;		/* Hardware state */
+    CARD32                Color;	/* Current draw color */
+    CARD32                ClearColor;	/* Color used to clear color buffer */
+    CARD32                ClearDepth;	/* Value used to clear depth buffer */
+
+    int                   drawX;	/* x-offset to current draw buffer */
+    int                   drawY;	/* y-offset to current draw buffer */
+
+    int                   readX;	/* x-offset to current read buffer */
+    int                   readY;	/* y-offset to current read buffer */
+
+    CARD32               *CCEbuf;	/* buffer to submit to CCE */
+    int                   CCEcount;	/* number of dwords in CCEbuf */
+
+    int                   CCEtimeout;   /* number of times to loop
+					   before exiting */
+
+    Display              *display;	/* X server display */
+
+    __DRIcontextPrivate  *driContext;	/* DRI context */
+    __DRIdrawablePrivate *driDrawable;	/* DRI drawable bound to this ctx */
+
+    r128ScreenPtr         r128Screen;	/* Screen private DRI data */
+} r128ContextRec, *r128ContextPtr;
+
+#define R128_MESACTX(r128ctx)     ((r128ctx)->glCtx)
+#define R128_DRIDRAWABLE(r128ctx) ((r128ctx)->driDrawable)
+#define R128_DRISCREEN(r128ctx)   ((r128ctx)->r128Screen->driScreen)
+
+extern GLboolean r128CreateContext(Display *dpy, GLvisual *glVisual,
+				   __DRIcontextPrivate *driContextPriv);
+extern void           r128DestroyContext(r128ContextPtr r128ctx);
+extern r128ContextPtr r128MakeCurrent(r128ContextPtr oldCtx,
+				      r128ContextPtr newCtx,
+				      __DRIdrawablePrivate *dPriv);
+
+#endif
+#endif /* _R128_CONTEXT_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_dd.c b/xc/lib/GL/mesa/src/drv/r128/r128_dd.c
new file mode 100644
index 000000000..0ed7ae471
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_dd.c
@@ -0,0 +1,182 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#include "r128_init.h"
+#include "r128_mesa.h"
+#include "r128_xmesa.h"
+#include "r128_context.h"
+#include "r128_lock.h"
+#include "r128_cce.h"
+#include "r128_clear.h"
+#include "r128_state.h"
+#include "r128_vb.h"
+#include "r128_pipeline.h"
+#include "r128_dd.h"
+
+/* Driver entry point for clearing color and ancillary buffers */
+static GLbitfield r128DDClear(GLcontext *ctx, GLbitfield mask, GLboolean all,
+			      GLint x, GLint y, GLint width, GLint height)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+
+    if (r128ctx->SWonly) {
+	/* FIXME: Provide software fallback for this case?? */
+    }
+
+    if (mask & DD_FRONT_LEFT_BIT) {
+	r128ClearColorBuffer(r128ctx, all, x, y, width, height,
+			     r128ctx->r128Screen->fbX,
+			     r128ctx->r128Screen->fbY);
+	mask &= ~DD_FRONT_LEFT_BIT;
+    }
+
+    if (mask & DD_BACK_LEFT_BIT) {
+	r128ClearColorBuffer(r128ctx, all, x, y, width, height,
+			     r128ctx->r128Screen->backX,
+			     r128ctx->r128Screen->backY);
+	mask &= ~DD_BACK_LEFT_BIT;
+    }
+
+    if (mask & DD_DEPTH_BIT) {
+	r128ClearDepthBuffer(r128ctx, all, x, y, width, height);
+	mask &= ~DD_DEPTH_BIT;
+    }
+
+#if 0
+    /* FIXME: Add stencil support */
+    if (mask & DD_STENCIL_BIT) {
+	r128ClearStencilBuffer(r128ctx, all, x, y, width, height);
+	mask &= ~DD_STENCIL_BIT;
+    }
+#endif
+
+    return mask;
+}
+
+/* Return the current color buffer size */
+static void r128DDGetBufferSize(GLcontext *ctx, GLuint *width, GLuint *height)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+
+    *width  = r128ctx->driDrawable->w;
+    *height = r128ctx->driDrawable->h;
+}
+
+/* Return various strings for glGetString() */
+static const GLubyte *r128DDGetString(GLcontext *ctx, GLenum name)
+{
+    switch (name) {
+    case GL_VENDOR:
+	return (GLubyte *)"Precision Insight, Inc.";
+    case GL_RENDERER:
+	return (GLubyte *)"Mesa DRI Rage128 20000320";
+    default:
+	return NULL;
+    }
+}
+
+/* Send all commands to the hardware.  If vertex buffers or indirect
+   buffers are in use, then we need to make sure they are sent to the
+   hardware.  All commands that are normally sent to the ring are
+   already considered `flushed'. */
+static void r128DDFlush(GLcontext *ctx)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+
+    R128CCE_FLUSH_VB_LOCK(r128ctx);
+}
+
+/* Make sure all commands have been sent to the hardware and have
+   completed processing. */
+static void r128DDFinish(GLcontext *ctx)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+
+    r128DDFlush(ctx);
+    R128CCE_WAIT_FOR_IDLE_LOCK(r128ctx);
+}
+
+/* Return various parameters requested by Mesa */
+static GLint r128DDGetParameteri(const GLcontext *ctx, GLint param)
+{
+    switch (param) {
+#if 0
+	/* FIXME: Support for these needs to be added to Mesa */
+    case DD_MAX_TEXTURE_SIZE:  return 1024;
+    case DD_MAX_TEXTURES:      return 2;
+#endif
+#if 0
+    case DD_HAVE_HARDWARE_FOG: return 1; /* FIXME: Add HW fog support */
+#endif
+    default:                   return 0;
+    }
+}
+
+/* Initialize the extensions supported by this driver */
+void r128DDInitExtensions(GLcontext *ctx)
+{
+    /* FIXME: Are there other extensions to enable/disable??? */
+    gl_extensions_disable(ctx, "GL_EXT_shared_texture_palette");
+    gl_extensions_disable(ctx, "GL_EXT_paletted_texture");
+    gl_extensions_disable(ctx, "GL_EXT_point_parameters");
+    gl_extensions_disable(ctx, "ARB_imaging");
+    gl_extensions_disable(ctx, "GL_EXT_blend_minmax");
+    gl_extensions_disable(ctx, "GL_EXT_blend_logic_op");
+    gl_extensions_disable(ctx, "GL_EXT_blend_subtract");
+    gl_extensions_disable(ctx, "GL_INGR_blend_func_separate");
+
+    if (getenv("LIBGL_NO_MULTITEXTURE"))
+	gl_extensions_disable(ctx, "GL_ARB_multitexture");
+}
+
+/* Initialize the driver's misc functions */
+void r128DDInitDriverFuncs(GLcontext *ctx)
+{
+    ctx->Driver.Clear                = r128DDClear;
+
+    ctx->Driver.GetBufferSize        = r128DDGetBufferSize;
+    ctx->Driver.GetString            = r128DDGetString;
+    ctx->Driver.Finish               = r128DDFinish;
+    ctx->Driver.Flush                = r128DDFlush;
+
+    ctx->Driver.Error                = NULL;
+    ctx->Driver.GetParameteri        = r128DDGetParameteri;
+
+    ctx->Driver.DrawPixels           = NULL;
+    ctx->Driver.Bitmap               = NULL;
+
+    ctx->Driver.RegisterVB           = r128DDRegisterVB;
+    ctx->Driver.UnregisterVB         = r128DDUnregisterVB;
+    ctx->Driver.BuildPrecalcPipeline = r128DDBuildPrecalcPipeline;
+}
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_dd.h b/xc/lib/GL/mesa/src/drv/r128/r128_dd.h
new file mode 100644
index 000000000..91aa02586
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_dd.h
@@ -0,0 +1,44 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_DD_H_
+#define _R128_DD_H_
+
+#ifdef GLX_DIRECT_RENDERING
+
+extern void r128DDInitExtensions(GLcontext *ctx);
+extern void r128DDInitDriverFuncs(GLcontext *ctx);
+
+#endif
+#endif /* _R128_DD_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_fastpath.c b/xc/lib/GL/mesa/src/drv/r128/r128_fastpath.c
new file mode 100644
index 000000000..330e5a23d
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_fastpath.c
@@ -0,0 +1,551 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#include "r128_init.h"
+#include "r128_mesa.h"
+#include "r128_state.h"
+#include "r128_vb.h"
+#include "r128_tris.h"
+#include "r128_fastpath.h"
+
+#include "mmath.h"
+#include "cva.h"
+#include "vertices.h"
+
+/* FIXME: These routines were copied from the i810 driver, and were only
+   slightly modified for the Rage 128.  They still need to be optmizied
+   and cleaned up.  Also, support for USE_RHW2 needs to be added. */
+
+typedef struct r128_fast_table {
+    r128BuildVerticesFunc build_vertices;
+    r128InterpFunc        interp;
+} r128FastPathTable;
+
+#define POINT(x)   r128DrawPointVB(r128ctx, &vert[x].v, psize)
+#define LINE(x,y)  r128DrawLineVB(r128ctx, &vert[x].v, &vert[y].v, lwidth)
+#define TRI(x,y,z) r128DrawTriangleVB(r128ctx, &vert[x].v, &vert[y].v, &vert[z].v)
+
+/* Direct, and no clipping required.  The clip funcs have not been
+  written yet, so this is only useful for the fast path. */
+#define RENDER_POINTS(start, count)                                       \
+do {                                                                      \
+    GLuint e;                                                             \
+    for (e = start; e <= count; e++)                                      \
+	POINT(elt[e]);                                                    \
+} while (0)
+
+#define RENDER_LINE(i1, i)                                                \
+do {                                                                      \
+    GLuint e1 = elt[i1], e = elt[i];                                      \
+    LINE(e1, e);                                                          \
+} while (0)
+
+#define RENDER_TRI(i2, i1, i, pv, parity)                                 \
+do {                                                                      \
+    GLuint e2 = elt[i2], e1 = elt[i1], e = elt[i];                        \
+    if (parity) {                                                         \
+	GLuint tmp = e2;                                                  \
+	e2 = e1;                                                          \
+	e1 = tmp;                                                         \
+    }                                                                     \
+    TRI(e2, e1, e);                                                       \
+} while (0)
+
+#define RENDER_QUAD(i3, i2, i1, i, pv)                                    \
+do {                                                                      \
+    GLuint e3 = elt[i3], e2 = elt[i2], e1 = elt[i1], e = elt[i];          \
+    TRI(e3, e2, e);                                                       \
+    TRI(e2, e1, e);                                                       \
+} while (0)
+
+#define LOCAL_VARS                                                        \
+    r128VertexPtr   vert    = R128_DRIVER_DATA(VB)->verts;                \
+    const GLuint   *elt     = VB->EltPtr->data;                           \
+    GLcontext      *ctx     = VB->ctx;                                    \
+    r128ContextPtr  r128ctx = R128_CONTEXT(ctx);                          \
+    const GLfloat   lwidth  = ctx->Line.Width;                            \
+    const GLfloat   psize   = ctx->Point.Size;                            \
+    (void) lwidth; (void)psize; (void) vert;
+
+#define TAG(x) x##_r128_smooth_indirect
+#include "render_tmp.h"
+
+
+
+#define NEGATIVE(f)          (f < 0)
+#define DIFFERENT_SIGNS(a,b) ((a*b) < 0)
+#define LINTERP(T, A, B)     ((A) + (T) * ((B) - (A)))
+
+
+#define INTERP_RGBA(t, out, a, b)                                         \
+do {                                                                      \
+    int i;                                                                \
+    for (i = 0; i < 4; i++) {                                             \
+	GLfloat fa = UBYTE_COLOR_TO_FLOAT_COLOR(a[i]);                    \
+	GLfloat fb = UBYTE_COLOR_TO_FLOAT_COLOR(b[i]);                    \
+	GLfloat fo = LINTERP(t, fa, fb);                                  \
+	FLOAT_COLOR_TO_UBYTE_COLOR(out[i], fo);                           \
+   }                                                                      \
+} while (0)
+
+
+#define CLIP(SGN, V, PLANE)                                               \
+do {                                                                      \
+    if (mask & PLANE) {                                                   \
+	GLuint   *indata  = inlist[in];                                   \
+	GLuint   *outdata = inlist[in ^= 1];                              \
+	GLuint   nr       = n;                                            \
+	GLfloat *J        = verts[indata[nr-1]].f;                        \
+	GLfloat  dpJ      = (SGN J[V]) + J[3];                            \
+                                                                          \
+	inlist[0] = vlist1;                                               \
+	for (i = n = 0 ; i < nr ; i++) {                                  \
+	    GLuint   elt_i = indata[i];                                   \
+	    GLfloat *I     = verts[elt_i].f;                              \
+	    GLfloat  dpI   = (SGN I[V]) + I[3];                           \
+                                                                          \
+	    if (DIFFERENT_SIGNS(dpI, dpJ)) {                              \
+		GLfloat *O = verts[next_vert].f;                          \
+		GLfloat t, *in, *out;                                     \
+                                                                          \
+		if (NEGATIVE(dpI)) {                                      \
+		    t = dpI / (dpI - dpJ);                                \
+		    in = I;                                               \
+		    out = J;                                              \
+		} else {                                                  \
+		    t = dpJ / (dpJ - dpI);                                \
+		    in = J;                                               \
+		    out = I;                                              \
+		}                                                         \
+                                                                          \
+		interp(t, O, in, out);                                    \
+                                                                          \
+		clipmask[next_vert] = 0;                                  \
+		outdata[n++] = next_vert++;                               \
+	    }                                                             \
+                                                                          \
+	    clipmask[elt_i] |= PLANE;      /* don't set up */             \
+                                                                          \
+	    if (!NEGATIVE(dpI)) {                                         \
+		outdata[n++] = elt_i;                                     \
+		clipmask[elt_i] &= ~PLANE; /* set up after all */         \
+	    }                                                             \
+                                                                          \
+	    J = I;                                                        \
+	    dpJ = dpI;                                                    \
+	}                                                                 \
+                                                                          \
+	if (n < 3) return;                                                \
+    }                                                                     \
+} while (0)
+
+#define LINE_CLIP(x,y,z,w,PLANE)                                          \
+do {                                                                      \
+    if (mask & PLANE) {                                                   \
+	GLfloat dpI = DOT4V(I,x,y,z,w);                                   \
+	GLfloat dpJ = DOT4V(J,x,y,z,w);                                   \
+                                                                          \
+	if (DIFFERENT_SIGNS(dpI, dpJ)) {                                  \
+	    GLfloat *O = verts[next_vert].f;                              \
+	    GLfloat  t = dpI / (dpI - dpJ);                               \
+                                                                          \
+	    interp(t, O, I, J);                                           \
+                                                                          \
+	    clipmask[next_vert] = 0;                                      \
+                                                                          \
+	    if (NEGATIVE(dpI)) {                                          \
+		clipmask[elts[0]] |= PLANE;                               \
+		I = O;                                                    \
+		elts[0] = next_vert++;                                    \
+	    } else {                                                      \
+		clipmask[elts[1]] |= PLANE;                               \
+		J = O;                                                    \
+		elts[1] = next_vert++;                                    \
+	    }                                                             \
+	} else if (NEGATIVE(dpI)) return;                                 \
+    }                                                                     \
+} while (0)
+
+
+static void r128TriClip(GLuint **p_elts,
+			r128Vertex *verts,
+			GLubyte *clipmask,
+			GLuint *p_next_vert,
+			GLubyte mask,
+			r128InterpFunc interp)
+{
+    GLuint *elts      = *p_elts;
+    GLuint  next_vert = *p_next_vert;
+    GLuint  in        = 0;
+    GLuint  n         = 3;
+    GLuint  vlist1[VB_MAX_CLIPPED_VERTS];
+    GLuint  vlist2[VB_MAX_CLIPPED_VERTS];
+    GLuint *inlist[2];
+    GLuint *out;
+    GLuint  i;
+
+    inlist[0] = elts;
+    inlist[1] = vlist2;
+
+    CLIP(-,0,CLIP_RIGHT_BIT);
+    CLIP(+,0,CLIP_LEFT_BIT);
+    CLIP(-,1,CLIP_TOP_BIT);
+    CLIP(+,1,CLIP_BOTTOM_BIT);
+    CLIP(-,2,CLIP_FAR_BIT);
+    CLIP(+,2,CLIP_NEAR_BIT);
+
+    /* Convert the planar polygon to a list of triangles */
+    out = inlist[in];
+   
+    for (i = 2 ; i < n ; i++) {
+	elts[0] = out[0];
+	elts[1] = out[i-1];
+	elts[2] = out[i];
+	elts += 3;
+    }
+
+    *p_next_vert = next_vert;
+    *p_elts      = elts;
+}
+
+
+static void r128LineClip(GLuint **p_elts,
+			 r128Vertex *verts,
+			 GLubyte *clipmask,
+			 GLuint *p_next_vert,
+			 GLubyte mask,
+			 r128InterpFunc interp)
+{
+    GLuint  *elts      = *p_elts;
+    GLfloat *I         = verts[elts[0]].f;
+    GLfloat *J         = verts[elts[1]].f;
+    GLuint   next_vert = *p_next_vert;
+
+    LINE_CLIP(1,0,0,-1,CLIP_LEFT_BIT);
+    LINE_CLIP(-1,0,0,1,CLIP_RIGHT_BIT);
+    LINE_CLIP(0,1,0,-1,CLIP_TOP_BIT);
+    LINE_CLIP(0,-1,0,1,CLIP_BOTTOM_BIT);
+    LINE_CLIP(0,0,1,-1,CLIP_FAR_BIT);
+    LINE_CLIP(0,0,-1,1,CLIP_NEAR_BIT);
+
+    *p_next_vert  = next_vert;
+    *p_elts      += 2;
+}
+
+
+
+#define CLIP_POINT(e)                                                     \
+do {                                                                      \
+   if (mask[e]) *out++ = e;                                               \
+} while (0)
+
+#define CLIP_LINE(e1, e0)                                                 \
+do {                                                                      \
+    GLubyte ormask = mask[e0] | mask[e1];                                 \
+    out[0] = e1;                                                          \
+    out[1] = e0;                                                          \
+    out += 2;                                                             \
+    if (ormask) {                                                         \
+	out-=2;                                                           \
+	if (!(mask[e0] & mask[e1])) {                                     \
+	    r128LineClip(&out, verts, mask, &next_vert, ormask, interp);  \
+	}                                                                 \
+    }                                                                     \
+} while (0)
+
+#define CLIP_TRIANGLE(e2, e1, e0)                                         \
+do {                                                                      \
+    GLubyte ormask;                                                       \
+    out[0] = e2;                                                          \
+    out[1] = e1;                                                          \
+    out[2] = e0;                                                          \
+    out += 3;                                                             \
+    ormask = mask[e2] | mask[e1] | mask[e0];                              \
+    if (ormask) {                                                         \
+	out -= 3;                                                         \
+	if (!(mask[e2] & mask[e1] & mask[e0])) {                          \
+	    r128TriClip(&out, verts, mask, &next_vert, ormask, interp);   \
+	}                                                                 \
+    }                                                                     \
+} while (0)
+
+
+
+/* Build a table of functions to clip each primitive type.  These
+ * produce a list of elements in the appropriate 'reduced' primitive,
+ * ie (points, lines, triangles) containing all the clipped and
+ * unclipped primitives from the original list.
+ */
+#define LOCAL_VARS                                                        \
+    r128ContextPtr       r128ctx   = R128_CONTEXT(VB->ctx);               \
+    r128VertexBufferPtr  r128VB    = R128_DRIVER_DATA(VB);                \
+    GLuint              *elt       = VB->EltPtr->data;                    \
+    r128Vertex          *verts     = r128VB->verts;                       \
+    GLuint               next_vert = r128VB->last_vert;                   \
+    GLuint              *out       = r128VB->clipped_elements.data;       \
+    GLubyte             *mask      = VB->ClipMask;                        \
+    r128InterpFunc       interp    = r128ctx->interp;                     \
+    (void) interp; (void) verts;
+
+#define POSTFIX                                                           \
+    r128VB->clipped_elements.count = out - r128VB->clipped_elements.data; \
+    r128VB->last_vert = next_vert;
+
+
+#define INIT(x)
+
+#define RENDER_POINTS(start, count)                                       \
+do {                                                                      \
+    GLuint i;                                                             \
+    for (i = start; i < count; i++)                                       \
+	CLIP_POINT(elt[i]);                                               \
+} while (0)
+
+#define RENDER_LINE(i1, i0)                                               \
+do {                                                                      \
+    CLIP_LINE(elt[i1], elt[i0]);                                          \
+} while (0)
+
+#define RENDER_TRI(i2, i1, i0, pv, parity)                                \
+do {                                                                      \
+    GLuint e2 = elt[i2], e1 = elt[i1], e0 = elt[i0];                      \
+    if (parity) e2 = elt[i1], e1 = elt[i2];                               \
+    CLIP_TRIANGLE(e2, e1, e0);                                            \
+} while (0)
+
+#define RENDER_QUAD(i3, i2, i1, i0, pv)                                   \
+do {                                                                      \
+  CLIP_TRIANGLE(elt[i3], elt[i2], elt[i0]);                               \
+  CLIP_TRIANGLE(elt[i2], elt[i1], elt[i0]);                               \
+} while (0)
+
+#define TAG(x) r128_clip_##x##_elt
+#include "render_tmp.h"
+
+
+/* Pack rgba and/or texture into the remaining half of a 32 byte vertex.
+ */
+#define CLIP_UBYTE_COLOR  4
+#define CLIP_UBYTE_B 0   
+#define CLIP_UBYTE_G 1  
+#define CLIP_UBYTE_R 2
+#define CLIP_UBYTE_A 3   
+#define CLIP_S0 6
+#define CLIP_T0 7
+#define CLIP_S1 8
+#define CLIP_T1 9
+
+#define TYPE (0)
+#define TAG(x) x
+#include "r128_fasttmp.h"
+
+#define TYPE (R128_RGBA_BIT)
+#define TAG(x) x##_RGBA
+#include "r128_fasttmp.h"
+
+#define TYPE (R128_TEX0_BIT)
+#define TAG(x) x##_TEX0
+#include "r128_fasttmp.h"
+
+#define TYPE (R128_RGBA_BIT | R128_TEX0_BIT)
+#define TAG(x) x##_RGBA_TEX0
+#include "r128_fasttmp.h"
+
+#define TYPE (R128_RGBA_BIT | R128_TEX0_BIT | R128_TEX1_BIT)
+#define TAG(x) x##_RGBA_TEX0_TEX1
+#include "r128_fasttmp.h"
+
+/* This one *could* get away with sneaking TEX1 into the color and
+ * specular slots, thus fitting inside a cache line.  Would be even
+ * better to switch to a smaller vertex.
+ */
+#define TYPE (R128_TEX0_BIT | R128_TEX1_BIT)
+#define TAG(x) x##_TEX0_TEX1
+#include "r128_fasttmp.h"
+
+
+
+/* Render elements directly from original list of vertices. */
+static void r128RenderElementsDirect(struct vertex_buffer *VB)
+{
+    GLcontext           *ctx     = VB->ctx;
+    r128ContextPtr       r128ctx = R128_CONTEXT(ctx);
+    GLenum               prim    = ctx->CVA.elt_mode;
+    GLuint               nr      = VB->EltPtr->count;
+    render_func          func    = render_tab_r128_smooth_indirect[prim];
+    GLuint               p       = 0;
+   
+    if (r128ctx->dirty) r128UpdateHWState(r128ctx);
+
+    do {
+	func(VB, 0, nr, 0);
+    } while (ctx->Driver.MultipassFunc &&
+	     ctx->Driver.MultipassFunc(VB, ++p));
+}
+
+/* Project vertices from clip to device space */
+static void r128ProjectVertices(struct vertex_buffer *VB)
+{
+    GLcontext           *ctx     = VB->ctx;
+    GLmatrix            *mat     = &ctx->Viewport.WindowMap;
+    r128ContextPtr       r128ctx = R128_CONTEXT(ctx);
+    r128VertexBufferPtr  r128VB  = R128_DRIVER_DATA(VB);
+    GLfloat              m[16];
+
+    m[MAT_SX] =  mat->m[MAT_SX];
+    m[MAT_TX] =  mat->m[MAT_TX];
+    m[MAT_SY] = -mat->m[MAT_SY];
+    m[MAT_TY] = -mat->m[MAT_TY] + r128ctx->driDrawable->h;
+    m[MAT_SZ] =  mat->m[MAT_SZ];
+    m[MAT_TZ] =  mat->m[MAT_TZ];
+
+    switch (ctx->Visual->DepthBits) {
+    case 16: m[MAT_SZ] /= 65536.0;      m[MAT_TZ] /= 65536.0;      break;
+    case 24: m[MAT_SZ] /= 16777216.0;   m[MAT_TZ] /= 16777216.0;   break;
+    case 32: m[MAT_SZ] /= 4294967296.0; m[MAT_TZ] /= 4294967296.0; break;
+    default: m[MAT_SZ] /= 65536.0;      m[MAT_TZ] /= 65536.0;      break;
+    }
+
+#if USE_RHW2
+    /* FIXME: Handle RHW2?? */
+    gl_project_v16(r128VB->verts[VB->CopyStart].f,
+		   r128VB->verts[r128VB->last_vert].f,
+		   m,
+		   16 * 4);
+#else
+    gl_project_v16(r128VB->verts[VB->CopyStart].f,
+		   r128VB->verts[r128VB->last_vert].f,
+		   m,
+		   16 * 4);
+#endif
+}
+
+/* Project clipped vertices from clip to device space */
+static void r128ProjectClippedVertices(struct vertex_buffer *VB)
+{
+    GLcontext           *ctx     = VB->ctx;
+    GLmatrix            *mat     = &ctx->Viewport.WindowMap;
+    r128ContextPtr       r128ctx = R128_CONTEXT(ctx);
+    r128VertexBufferPtr  r128VB  = R128_DRIVER_DATA(VB);
+    GLfloat              m[16];
+
+    m[MAT_SX] =  mat->m[MAT_SX];
+    m[MAT_TX] =  mat->m[MAT_TX];
+    m[MAT_SY] = -mat->m[MAT_SY];
+    m[MAT_TY] = -mat->m[MAT_TY] + r128ctx->driDrawable->h;
+    m[MAT_SZ] =  mat->m[MAT_SZ];
+    m[MAT_TZ] =  mat->m[MAT_TZ];
+
+    switch (ctx->Visual->DepthBits) {
+    case 16: m[MAT_SZ] /= 65536.0;      m[MAT_TZ] /= 65536.0;      break;
+    case 24: m[MAT_SZ] /= 16777216.0;   m[MAT_TZ] /= 16777216.0;   break;
+    case 32: m[MAT_SZ] /= 4294967296.0; m[MAT_TZ] /= 4294967296.0; break;
+    default: m[MAT_SZ] /= 65536.0;      m[MAT_TZ] /= 65536.0;      break;
+    }
+
+#if USE_RHW2
+    /* FIXME: Handle RHW2?? */
+    gl_project_clipped_v16(r128VB->verts[VB->CopyStart].f,
+			   r128VB->verts[r128VB->last_vert].f,
+			   m,
+			   16 * 4,
+			   VB->ClipMask + VB->CopyStart);
+#else
+    gl_project_clipped_v16(r128VB->verts[VB->CopyStart].f,
+			   r128VB->verts[r128VB->last_vert].f,
+			   m,
+			   16 * 4,
+			   VB->ClipMask + VB->CopyStart);
+#endif
+}
+
+static r128FastPathTable r128FastTab[0x80];
+
+/* Initialize the table of fast path support functions */
+void r128FastPathInit(void)
+{
+   r128_clip_render_init_elt();
+   render_init_r128_smooth_indirect();
+
+   r128_init_fastpath(&r128FastTab[0]);
+   r128_init_fastpath_RGBA(&r128FastTab[R128_RGBA_BIT]);
+   r128_init_fastpath_TEX0(&r128FastTab[R128_TEX0_BIT]);
+   r128_init_fastpath_RGBA_TEX0(&r128FastTab[R128_RGBA_BIT|R128_TEX0_BIT]);
+   r128_init_fastpath_TEX0_TEX1(&r128FastTab[R128_TEX0_BIT|R128_TEX1_BIT]);
+   r128_init_fastpath_RGBA_TEX0_TEX1(&r128FastTab[R128_RGBA_BIT|R128_TEX0_BIT|
+						  R128_TEX1_BIT]);
+}
+
+#define VALID_SETUP (R128_RGBA_BIT | R128_TEX0_BIT | R128_TEX1_BIT)
+
+void r128FastPath(struct vertex_buffer *VB)
+{
+    GLcontext         *ctx  = VB->ctx;
+    GLenum             prim = ctx->CVA.elt_mode;
+    r128ContextPtr     r128ctx = R128_CONTEXT(ctx);
+    r128FastPathTable *tab  = &r128FastTab[r128ctx->SetupIndex & VALID_SETUP];
+    GLuint             do_cliptest = 1;
+
+    gl_prepare_arrays_cva(VB);   /* still need this */
+
+    /* Reserve enough space for the pathological case */
+    if (VB->EltPtr->count * 12 > R128_DRIVER_DATA(VB)->size) {
+	r128ResizeVB(VB, VB->EltPtr->count * 12);
+	do_cliptest = 1;
+    }
+
+    tab->build_vertices(VB, do_cliptest);         /* object->clip space */
+
+    if (r128ctx->dirty) r128UpdateHWState(r128ctx);
+
+    if (VB->ClipOrMask) {
+	if (!VB->ClipAndMask) {
+	    render_func *clip = r128_clip_render_tab_elt;
+
+	    r128ctx->interp = tab->interp;
+	    clip[prim](VB, 0, VB->EltPtr->count, 0); /* build new elts */
+	    ctx->CVA.elt_mode = gl_reduce_prim[prim];
+	    VB->EltPtr = &(R128_DRIVER_DATA(VB)->clipped_elements);
+	    r128ProjectClippedVertices(VB);    /* clip->device space */
+	    r128RenderElementsDirect(VB);      /* render using new list */
+	}
+    } else {
+	r128ProjectVertices(VB);               /* clip->device space  */
+	r128RenderElementsDirect(VB);          /* render using orig list */
+    }
+
+    /* This indicates that there is no cached data to reuse */
+    VB->pipeline->data_valid = 0;
+    VB->pipeline->new_state = 0;
+}
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_fastpath.h b/xc/lib/GL/mesa/src/drv/r128/r128_fastpath.h
new file mode 100644
index 000000000..47bb92049
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_fastpath.h
@@ -0,0 +1,48 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_FASTPATH_H_
+#define _R128_FASTPATH_H_
+
+typedef void (*r128BuildVerticesFunc)(struct vertex_buffer *VB,
+				      GLuint do_cliptest);
+typedef void (*r128InterpFunc)(GLfloat t,
+			       GLfloat *result,
+			       const GLfloat *in,
+			       const GLfloat *out);
+
+extern void r128FastPathInit(void);
+extern void r128FastPath(struct vertex_buffer *VB);
+
+#endif /* _R128_FASTPATH_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_fasttmp.h b/xc/lib/GL/mesa/src/drv/r128/r128_fasttmp.h
new file mode 100644
index 000000000..e76dd52b4
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_fasttmp.h
@@ -0,0 +1,155 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+/* FIXME: These routines were copied from the i810 driver, and were only
+   slightly modified for the Rage 128.  They still need to be optmizied
+   and cleaned up.  Also, support for USE_RHW2 needs to be added. */
+
+/* The first part of setup is applied to all vertices, clipped or
+ * unclipped.  This data w!ill be used for clipping, and then all
+ * vertices with a zero clipmask will be projected to device space.
+ *
+ * This could be split into several loops, but - it seems that the
+ * large stride of the fxVertices makes cache issues the big
+ * performance factor, and that multiple loops mean multiple cache
+ * misses....
+ */
+static void TAG(r128_setup_full)(struct vertex_buffer *VB, GLuint do_cliptest)
+{
+    GLcontext           *ctx    = VB->ctx;
+    r128VertexBufferPtr  r128VB = R128_DRIVER_DATA(VB);
+    const GLfloat       *m      = ctx->ModelProjectMatrix.m;
+    GLuint               start  = VB->CopyStart;
+    GLuint               count  = VB->Count;
+    GLuint               i;
+
+    gl_xform_points3_v16_general(r128VB->verts[start].f,
+				 m, 
+				 VB->ObjPtr->start,
+				 VB->ObjPtr->stride,
+				 count - start);
+
+    if (do_cliptest) {
+	VB->ClipAndMask = ~0; 
+	VB->ClipOrMask = 0;
+	gl_cliptest_points4_v16(r128VB->verts[start].f,
+				r128VB->verts[count].f,
+				&(VB->ClipOrMask),
+				&(VB->ClipAndMask),
+				VB->ClipMask + start);
+    }
+
+    /* These branches are all resolved at compile time.  Hopefully all
+     * the pointers are valid addresses even when not enabled.
+     */
+    if (TYPE) {
+	GLubyte *color        = VB->ColorPtr->start;
+	GLfloat *tex0_data    = VB->TexCoordPtr[0]->start;
+	GLfloat *tex1_data    = VB->TexCoordPtr[1]->start;
+
+	GLuint   color_stride = VB->ColorPtr->stride;
+	GLuint   tex0_stride  = VB->TexCoordPtr[0]->stride;
+	GLuint   tex1_stride  = VB->TexCoordPtr[1]->stride;
+
+	GLfloat *f            = r128VB->verts[start].f;
+
+	for (i = start ; i < count ; i++, f += 16) {
+	    if (TYPE & R128_RGBA_BIT) {
+		GLubyte *b = (GLubyte *)&f[CLIP_UBYTE_COLOR];
+		GLubyte *col = color; color += color_stride;
+		b[CLIP_UBYTE_R] = col[0];
+		b[CLIP_UBYTE_G] = col[1];
+		b[CLIP_UBYTE_B] = col[2];
+		b[CLIP_UBYTE_A] = col[3];
+	    }
+	    if (TYPE & R128_TEX0_BIT) {
+		f[CLIP_S0] = tex0_data[0];
+		f[CLIP_T0] = tex0_data[1];
+		STRIDE_F(tex0_data, tex0_stride);
+	    }
+	    if (TYPE & R128_TEX1_BIT) {
+		f[CLIP_S1] = tex1_data[0];
+		f[CLIP_T1] = tex1_data[1];
+		STRIDE_F(tex1_data, tex1_stride);
+	    }
+	}
+    }
+
+    r128VB->clipped_elements.count = start;
+    r128VB->last_vert = count;
+}
+
+
+/* Changed to just put the interp func instead of the whole clip
+ * routine into the header.  Less code and better chance of doing some
+ * of this stuff in assembly.
+ */
+static void TAG(r128_interp_vert)(GLfloat t, 
+				  GLfloat *O,
+				  const GLfloat *I,
+				  const GLfloat *J)
+{
+    O[0] = LINTERP(t, I[0], J[0]);
+    O[1] = LINTERP(t, I[1], J[1]);
+    O[2] = LINTERP(t, I[2], J[2]);
+    O[3] = LINTERP(t, I[3], J[3]);
+
+    if (TYPE & R128_RGBA_BIT) {
+	INTERP_RGBA(t,
+		    ((GLubyte *)&(O[4])),
+		    ((GLubyte *)&(I[4])),
+		    ((GLubyte *)&(J[4])));
+    }
+
+    if (TYPE & R128_TEX0_BIT) {
+	O[6] = LINTERP(t, I[6], J[6]);
+	O[7] = LINTERP(t, I[7], J[7]);
+    }
+
+    if (TYPE & R128_TEX1_BIT) {
+	O[8] = LINTERP(t, I[8], J[8]);
+	O[9] = LINTERP(t, I[9], J[9]);
+    }
+}
+
+
+static void TAG(r128_init_fastpath)(r128FastPathTable *tab)
+{
+   tab->build_vertices = TAG(r128_setup_full);
+   tab->interp         = TAG(r128_interp_vert);
+}
+
+#undef TYPE
+#undef TAG
+#undef SIZE
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_init.h b/xc/lib/GL/mesa/src/drv/r128/r128_init.h
new file mode 100644
index 000000000..fcaf70a51
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_init.h
@@ -0,0 +1,92 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_INIT_H_
+#define _R128_INIT_H_
+
+#ifdef GLX_DIRECT_RENDERING
+
+#include <X11/Xlibint.h>
+
+#include "types.h"
+#include "xf86drm.h"
+#include "dri_tmm.h"
+#include "dri_mesaint.h"
+
+#include "r128_screen.h"
+#include "r128_context.h"
+
+/* NOTE: The vertex buffer code is currently unstable because of the
+   switches between CCE and MMIO mode in the X server.  Fixing the X
+   server to use the CCE should fix this problem.  So, for now, it is
+   recommended that you do not use it. */
+
+#define DEBUG                       1
+#define DEBUG_LOCKING               1
+#define USE_FAST_PATH               1
+#define USE_USER_SPACE_RING         0
+
+
+#if DEBUG
+
+#include <stdio.h>
+#define R128_DEBUG(p)                                                       \
+    do {                                                                    \
+	printf p ;                                                          \
+	fflush(stdout);                                                     \
+    } while (0)
+
+extern int R128_DEBUG_FLAGS;
+
+#define DEBUG_VERBOSE_2D            0x0001
+#define DEBUG_VERBOSE_CCE           0x0008
+#define DEBUG_VERBOSE_OUTREG        0x0010
+#define DEBUG_ALWAYS_SYNC           0x0040
+#define DEBUG_VERBOSE_MSG           0x0080
+#define DEBUG_NO_OUTRING            0x0100
+#define DEBUG_NO_OUTREG             0x0200
+#define DEBUG_VERBOSE_API           0x0400
+#define DEBUG_VALIDATE_RING         0x0800
+#define DEBUG_VERBOSE_LRU           0x1000
+#define DEBUG_VERBOSE_DRI           0x2000
+#define DEBUG_VERBOSE_IOCTL         0x4000
+
+#else
+
+#define R128_DEBUG_FLAGS            0
+
+#endif
+
+#endif
+#endif /* _R128_INIT_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_lock.h b/xc/lib/GL/mesa/src/drv/r128/r128_lock.h
new file mode 100644
index 000000000..7fd81f807
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_lock.h
@@ -0,0 +1,134 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_LOCK_H_
+#define _R128_LOCK_H_
+
+#ifdef GLX_DIRECT_RENDERING
+
+/* Turn DEBUG_LOCKING on to find locking conflicts (see r128_init.h) */
+#if DEBUG_LOCKING
+extern char *prevLockFile;
+extern int   prevLockLine;
+
+#define DEBUG_LOCK()                                                    \
+    do {                                                                \
+	prevLockFile = (__FILE__);                                      \
+	prevLockLine = (__LINE__);                                      \
+    } while (0)
+
+#define DEBUG_RESET()                                                   \
+    do {                                                                \
+	prevLockFile = 0;                                               \
+	prevLockLine = 0;                                               \
+    } while (0)
+
+#define DEBUG_CHECK_LOCK()                                              \
+    do {                                                                \
+	if (prevLockFile) {                                             \
+	    fprintf(stderr,                                             \
+		    "LOCK SET!\n\tPrevious %s:%d\n\tCurrent: %s:%d\n",  \
+		    prevLockFile, prevLockLine, __FILE__, __LINE__);    \
+	    exit(1);                                                    \
+	}                                                               \
+    } while (0)
+
+#else
+
+#define DEBUG_LOCK()
+#define DEBUG_RESET()
+#define DEBUG_CHECK_LOCK()
+
+#endif
+
+/*
+ * !!! We may want to separate locks from locks with validation.  This
+ * could be used to improve performance for those things commands that
+ * do not do any drawing !!!
+ */
+
+/* Lock the hardware using the current context */
+#define LOCK_HARDWARE(CC)                                               \
+    do {                                                                \
+	char                  __ret = 0;                                \
+	__DRIcontextPrivate  *cPriv = CC->driContext;                   \
+	__DRIscreenPrivate   *sPriv = CC->r128Screen->driScreen;        \
+                                                                        \
+	DEBUG_CHECK_LOCK();                                             \
+	DRM_CAS(&sPriv->pSAREA->lock, cPriv->hHWContext,                \
+		DRM_LOCK_HELD|cPriv->hHWContext, __ret);                \
+	if (__ret) {                                                    \
+	    /* We lost the context, so we need to request the lock from \
+               the kernel and update our state. */                      \
+	    drmGetLock(sPriv->fd, cPriv->hHWContext, 0);                \
+	    XMesaUpdateState(cPriv);                                    \
+	}                                                               \
+	DEBUG_LOCK();                                                   \
+    } while (0)
+
+/* Unlock the hardware using the current context */
+#define UNLOCK_HARDWARE(CC)                                             \
+    do {                                                                \
+	__DRIcontextPrivate  *cPriv = CC->driContext;                   \
+	__DRIscreenPrivate   *sPriv = CC->r128Screen->driScreen;        \
+                                                                        \
+	DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, cPriv->hHWContext); \
+	DEBUG_RESET();                                                  \
+    } while (0)
+
+/*
+ * This pair of macros makes a loop over the drawing operations, so it
+ * is not self contained and does not have the nice single statement
+ * semantics of most macros.
+ */
+#define BEGIN_CLIP_LOOP(CC)                                             \
+    do {                                                                \
+	__DRIdrawablePrivate *_dPriv = CC->driDrawable;                 \
+	XF86DRIClipRectPtr    _pc = _dPriv->pClipRects;                 \
+	int                   _nc, _sc;                                 \
+                                                                        \
+	for (_nc = _dPriv->numClipRects; _nc > 0; _nc -= 3, _pc += 3) { \
+	    _sc = (_nc <= 3) ? _nc : 3;                                 \
+	    r128SetClipRects(CC, _pc, _sc)
+
+/* FIXME: This should be a function call to turn off aux clipping */
+#define END_CLIP_LOOP(CC)                                               \
+	    R128CCE0(R128_CCE_PACKET0, R128_AUX_SC_CNTL, 0);            \
+	    R128CCE(0x00000000);                                        \
+	    R128CCE_SUBMIT_PACKETS();                                   \
+	}                                                               \
+    } while (0)
+
+#endif
+#endif /* _R128_LOCK_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_mesa.h b/xc/lib/GL/mesa/src/drv/r128/r128_mesa.h
new file mode 100644
index 000000000..3a697b466
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_mesa.h
@@ -0,0 +1,43 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_MESA_H_
+#define _R128_MESA_H_
+
+#ifdef GLX_DIRECT_RENDERING
+
+#define R128_CONTEXT(ctx) ((r128ContextPtr)(ctx->DriverCtx))
+
+#endif
+#endif /* _R128_MESA_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_pipeline.c b/xc/lib/GL/mesa/src/drv/r128/r128_pipeline.c
new file mode 100644
index 000000000..55487ffcd
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_pipeline.c
@@ -0,0 +1,126 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#include "r128_init.h"
+#include "r128_mesa.h"
+#include "r128_vb.h"
+#include "r128_fastpath.h"
+#include "r128_pipeline.h"
+
+#include "types.h"
+
+static struct gl_pipeline_stage r128FastStage = {
+    "R128 Fast Path",
+    (PIPE_OP_VERT_XFORM   |
+     PIPE_OP_RAST_SETUP_0 |
+     PIPE_OP_RAST_SETUP_1 |
+     PIPE_OP_RENDER),
+    PIPE_PRECALC,
+    0, 0, 0, 0, 0, 0, 0, 0, 0,   
+    r128FastPath
+};
+
+/* Build the PRECALC pipeline with our stage, if possible.  Otherwise,
+   return GL_FALSE */
+GLboolean r128DDBuildPrecalcPipeline(GLcontext *ctx)
+{
+#if USE_FAST_PATH
+    r128ContextPtr      r128ctx = R128_CONTEXT(ctx);
+    struct gl_pipeline *pipe    = &ctx->CVA.pre;
+
+    if (r128ctx->RenderIndex == 0 &&
+	!(ctx->Enabled & (TEXTURE0_3D     |
+			  TEXTURE1_3D     |
+			  ENABLE_TEXMAT0  |
+			  ENABLE_TEXMAT1  |
+			  ENABLE_TEXGEN0  |
+			  ENABLE_TEXGEN1  |
+			  ENABLE_USERCLIP |
+			  ENABLE_LIGHT    |
+			  ENABLE_FOG)) &&
+	(ctx->Array.Flags & (VERT_OBJ_234 |
+			     VERT_TEX0_4  |
+			     VERT_TEX1_4  |
+			     VERT_ELT)) == (VERT_OBJ_23 |
+					    VERT_ELT)) {
+	pipe->stages[0]  = &r128FastStage;
+	pipe->stages[1]  = 0;
+	pipe->new_inputs = ctx->RenderFlags & VERT_DATA;
+	pipe->ops        = pipe->stages[0]->ops;
+
+	r128ctx->useFastPath = GL_TRUE;
+	return GL_TRUE;
+    }
+
+    if (r128ctx->useFastPath) {
+	r128ctx->useFastPath = GL_FALSE;
+
+	ctx->CVA.VB->ClipOrMask   = 0;
+	ctx->CVA.VB->ClipAndMask  = CLIP_ALL_BITS;
+	ctx->Array.NewArrayState |= ctx->Array.Summary;
+    }
+#endif
+
+    return GL_FALSE;
+}
+
+/* Register the pipeline with our stages included */
+GLuint r128RegisterPipelineStages(struct gl_pipeline_stage *out,
+				  const struct gl_pipeline_stage *in,
+				  GLuint nr)
+{
+#if USE_FAST_PATH
+    int i;
+
+    for (i = 0; i < nr; i++) {
+	out[i] = in[i];
+	switch (in[i].ops) {
+	case PIPE_OP_RAST_SETUP_0:
+	    out[i].cva_state_change = (NEW_LIGHTING  |
+				       NEW_TEXTURING |
+				       NEW_RASTER_OPS);
+	    out[i].state_change     = ~0;
+	    out[i].check            = r128CheckPartialRasterSetup;
+	    out[i].run              = r128PartialRasterSetup;
+	    break;
+
+	case PIPE_OP_RAST_SETUP_0|PIPE_OP_RAST_SETUP_1:
+	    out[i].run              = r128DoRasterSetup;
+	    break;
+	}
+    }
+#endif
+
+    return nr;
+}
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_pipeline.h b/xc/lib/GL/mesa/src/drv/r128/r128_pipeline.h
new file mode 100644
index 000000000..280e6b943
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_pipeline.h
@@ -0,0 +1,43 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_PIPELINE_H_
+#define _R128_PIPELINE_H_
+
+extern GLboolean r128DDBuildPrecalcPipeline(GLcontext *ctx);
+extern GLuint    r128RegisterPipelineStages(struct gl_pipeline_stage *out,
+					    const struct gl_pipeline_stage *in,
+					    GLuint nr);
+
+#endif /* _R128_PIPELINE_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_screen.c b/xc/lib/GL/mesa/src/drv/r128/r128_screen.c
new file mode 100644
index 000000000..b385271f4
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_screen.c
@@ -0,0 +1,258 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#include "r128_dri.h"
+#include "r128_reg.h"
+
+#include "r128_init.h"
+#include "r128_context.h"
+#include "r128_xmesa.h"
+#include "r128_cce.h"
+#include "r128_tris.h"
+#include "r128_vb.h"
+#include "r128_fastpath.h"
+
+#include <sys/mman.h>
+
+/* Create the device specific screen private data struct */
+r128ScreenPtr r128CreateScreen(__DRIscreenPrivate *sPriv)
+{
+    r128ScreenPtr    r128Screen;
+    R128DRIPtr       r128DRIPriv = (R128DRIPtr)sPriv->pDevPriv;
+
+    /* Allocate the private area */
+    r128Screen = (r128ScreenPtr)Xmalloc(sizeof(*r128Screen));
+    if (!r128Screen) return NULL;
+
+    /* This is first since which regions we map depends on whether or
+       not we are using a PCI card */
+    r128Screen->IsPCI          = r128DRIPriv->IsPCI;
+
+    r128Screen->mmioRgn.handle = r128DRIPriv->registerHandle;
+    r128Screen->mmioRgn.size   = r128DRIPriv->registerSize;
+    if (drmMap(sPriv->fd,
+	       r128Screen->mmioRgn.handle,
+	       r128Screen->mmioRgn.size,
+	       (drmAddressPtr)&r128Screen->mmio)) {
+	Xfree(r128Screen);
+	return NULL;
+    }
+
+    if (!r128Screen->IsPCI) {
+	r128Screen->ringRgn.handle = r128DRIPriv->ringHandle;
+	r128Screen->ringRgn.size   = r128DRIPriv->ringMapSize;
+	if (drmMap(sPriv->fd,
+		   r128Screen->ringRgn.handle,
+		   r128Screen->ringRgn.size,
+		   (drmAddressPtr)&r128Screen->ring)) {
+	    drmUnmap((drmAddress)r128Screen->mmio, r128Screen->mmioRgn.size);
+	    Xfree(r128Screen);
+	    return NULL;
+	}
+
+	r128Screen->ringReadRgn.handle = r128DRIPriv->ringReadPtrHandle;
+	r128Screen->ringReadRgn.size   = r128DRIPriv->ringReadMapSize;
+	if (drmMap(sPriv->fd,
+		   r128Screen->ringReadRgn.handle,
+		   r128Screen->ringReadRgn.size,
+		   (drmAddressPtr)&r128Screen->ringReadPtr)) {
+	    drmUnmap((drmAddress)r128Screen->ring, r128Screen->ringRgn.size);
+	    drmUnmap((drmAddress)r128Screen->mmio, r128Screen->mmioRgn.size);
+	    Xfree(r128Screen);
+	    return NULL;
+	}
+
+	r128Screen->vbRgn.handle = r128DRIPriv->vbHandle;
+	r128Screen->vbRgn.size   = r128DRIPriv->vbMapSize;
+	r128Screen->vbOffset     = r128DRIPriv->vbOffset;
+	if (drmMap(sPriv->fd,
+		   r128Screen->vbRgn.handle,
+		   r128Screen->vbRgn.size,
+		   (drmAddressPtr)&r128Screen->vb)) {
+	    drmUnmap((drmAddress)r128Screen->ringReadPtr,
+		     r128Screen->ringReadRgn.size);
+	    drmUnmap((drmAddress)r128Screen->ring, r128Screen->ringRgn.size);
+	    drmUnmap((drmAddress)r128Screen->mmio, r128Screen->mmioRgn.size);
+	    Xfree(r128Screen);
+	    return NULL;
+	}
+	r128Screen->vbOffset     = r128DRIPriv->vbOffset;
+	r128Screen->vbBufSize    = r128DRIPriv->vbBufSize;
+
+	r128Screen->indRgn.handle = r128DRIPriv->indHandle;
+	r128Screen->indRgn.size   = r128DRIPriv->indMapSize;
+	if (drmMap(sPriv->fd,
+		   r128Screen->indRgn.handle,
+		   r128Screen->indRgn.size,
+		   (drmAddressPtr)&r128Screen->ind)) {
+	    drmUnmap((drmAddress)r128Screen->vb, r128Screen->vbRgn.size);
+	    drmUnmap((drmAddress)r128Screen->ringReadPtr,
+		     r128Screen->ringReadRgn.size);
+	    drmUnmap((drmAddress)r128Screen->ring, r128Screen->ringRgn.size);
+	    drmUnmap((drmAddress)r128Screen->mmio, r128Screen->mmioRgn.size);
+	    Xfree(r128Screen);
+	    return NULL;
+	}
+
+	r128Screen->agpTexRgn.handle = r128DRIPriv->agpTexHandle;
+	r128Screen->agpTexRgn.size   = r128DRIPriv->agpTexMapSize;
+	if (drmMap(sPriv->fd,
+		   r128Screen->agpTexRgn.handle,
+		   r128Screen->agpTexRgn.size,
+		   (drmAddressPtr)&r128Screen->agpTex)) {
+	    drmUnmap((drmAddress)r128Screen->ind,  r128Screen->indRgn.size);
+	    drmUnmap((drmAddress)r128Screen->vb,   r128Screen->vbRgn.size);
+	    drmUnmap((drmAddress)r128Screen->ringReadPtr,
+		     r128Screen->ringReadRgn.size);
+	    drmUnmap((drmAddress)r128Screen->ring, r128Screen->ringRgn.size);
+	    drmUnmap((drmAddress)r128Screen->mmio, r128Screen->mmioRgn.size);
+	    Xfree(r128Screen);
+	    return NULL;
+	}
+	r128Screen->agpTexOffset   = r128DRIPriv->agpTexOffset;
+
+	if (!(r128Screen->vbBufs = drmMapBufs(sPriv->fd))) {
+	    drmUnmap((drmAddress)r128Screen->agpTex,
+		     r128Screen->agpTexRgn.size);
+	    drmUnmap((drmAddress)r128Screen->ind,  r128Screen->indRgn.size);
+	    drmUnmap((drmAddress)r128Screen->vb,   r128Screen->vbRgn.size);
+	    drmUnmap((drmAddress)r128Screen->ringReadPtr,
+		     r128Screen->ringReadRgn.size);
+	    drmUnmap((drmAddress)r128Screen->ring, r128Screen->ringRgn.size);
+	    drmUnmap((drmAddress)r128Screen->mmio, r128Screen->mmioRgn.size);
+	    Xfree(r128Screen);
+	    return NULL;
+	}
+    }
+
+    r128Screen->deviceID         = r128DRIPriv->deviceID;
+
+    r128Screen->width            = r128DRIPriv->width;
+    r128Screen->height           = r128DRIPriv->height;
+    r128Screen->depth            = r128DRIPriv->depth;
+    r128Screen->bpp              = r128DRIPriv->bpp;
+    r128Screen->pixel_code       = (r128Screen->bpp != 16 ?
+				    r128Screen->bpp :
+				    r128Screen->depth);
+
+    r128Screen->fb               = sPriv->pFB;
+    r128Screen->fbOffset         = sPriv->fbOrigin;
+    r128Screen->fbStride         = sPriv->fbStride;
+    r128Screen->fbSize           = sPriv->fbSize;
+
+    r128Screen->fbX              = r128DRIPriv->fbX;
+    r128Screen->fbY              = r128DRIPriv->fbY;
+    r128Screen->backX            = r128DRIPriv->backX;
+    r128Screen->backY            = r128DRIPriv->backY;
+    r128Screen->depthX           = r128DRIPriv->depthX;
+    r128Screen->depthY           = r128DRIPriv->depthY;
+
+    r128Screen->texOffset[R128_LOCAL_TEX_HEAP]     = (r128DRIPriv->textureY *
+						      r128Screen->fbStride +
+						      r128DRIPriv->textureX *
+						      (r128Screen->bpp/8));
+    r128Screen->texSize[R128_LOCAL_TEX_HEAP]       = r128DRIPriv->textureSize;
+    r128Screen->log2TexGran[R128_LOCAL_TEX_HEAP]   = r128DRIPriv->log2TexGran;
+
+    if (r128Screen->IsPCI) {
+	r128Screen->texOffset[R128_AGP_TEX_HEAP]   = 0;
+	r128Screen->texSize[R128_AGP_TEX_HEAP]     = 0;
+	r128Screen->log2TexGran[R128_AGP_TEX_HEAP] = 0;
+	r128Screen->NRTexHeaps                     = R128_NR_TEX_HEAPS-1;
+    } else {
+	r128Screen->texOffset[R128_AGP_TEX_HEAP]   = 0;
+	r128Screen->texSize[R128_AGP_TEX_HEAP]     =
+	    r128DRIPriv->agpTexMapSize;
+	r128Screen->log2TexGran[R128_AGP_TEX_HEAP] =
+	    r128DRIPriv->log2AGPTexGran;
+	r128Screen->NRTexHeaps                     = R128_NR_TEX_HEAPS;
+    }
+
+#if 1
+    /* FIXME: For testing only */
+    if (getenv("LIBGL_SHOW_BUFFERS")) {
+	r128Screen->backX        = 0;
+	r128Screen->backY        = r128DRIPriv->height/2;
+	r128Screen->depthX       = r128DRIPriv->width/2;
+	r128Screen->depthY       = r128DRIPriv->height/2;
+    }
+#endif
+
+    r128Screen->CCEMode          = r128DRIPriv->CCEMode;
+    r128Screen->CCEFifoSize      = r128DRIPriv->CCEFifoSize;
+
+    r128Screen->ringEntries      = r128DRIPriv->ringSize/sizeof(CARD32);
+    if (!r128Screen->IsPCI) {
+	r128Screen->ringStartPtr = (int *)r128Screen->ring;
+	r128Screen->ringEndPtr   = (int *)(r128Screen->ring
+					   + r128DRIPriv->ringSize);
+    }
+
+    r128Screen->MMIOFifoSlots    = 0;
+    r128Screen->CCEFifoSlots     = 0;
+
+    r128Screen->CCEFifoAddr      = R128_PM4_FIFO_DATA_EVEN;
+
+    r128Screen->SAREA            = (R128SAREAPrivPtr)((char *)sPriv->pSAREA +
+						      sizeof(XF86DRISAREARec));
+
+    r128Screen->driScreen        = sPriv;
+
+    r128FastPathInit();
+    r128TriangleFuncsInit();
+    r128SetupInit();
+
+    return r128Screen;
+}
+
+/* Destroy the device specific screen private data struct */
+void r128DestroyScreen(__DRIscreenPrivate *sPriv)
+{
+    r128ScreenPtr r128Screen = (r128ScreenPtr)sPriv->private;
+
+    if (!r128Screen->IsPCI) {
+	drmUnmapBufs(r128Screen->vbBufs);
+
+	drmUnmap((drmAddress)r128Screen->agpTex, r128Screen->agpTexRgn.size);
+	drmUnmap((drmAddress)r128Screen->ind,    r128Screen->indRgn.size);
+	drmUnmap((drmAddress)r128Screen->vb,     r128Screen->vbRgn.size);
+	drmUnmap((drmAddress)r128Screen->ringReadPtr,
+		 r128Screen->ringReadRgn.size);
+	drmUnmap((drmAddress)r128Screen->ring,   r128Screen->ringRgn.size);
+    }
+    drmUnmap((drmAddress)r128Screen->mmio,   r128Screen->mmioRgn.size);
+
+    Xfree(r128Screen);
+    sPriv->private = NULL;
+}
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_screen.h b/xc/lib/GL/mesa/src/drv/r128/r128_screen.h
new file mode 100644
index 000000000..02deb64e9
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_screen.h
@@ -0,0 +1,129 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_SCREEN_H_
+#define _R128_SCREEN_H_
+
+#ifdef GLX_DIRECT_RENDERING
+
+#include "r128_sarea.h"
+
+typedef struct {
+    drmHandle  handle;			/* Handle to the DRM region */
+    drmSize    size;			/* Size of the DRM region */
+} r128RegionRec, *r128RegionPtr;
+
+typedef struct {
+    /* MMIO register data */
+    r128RegionRec       mmioRgn;
+    unsigned char      *mmio;
+
+    /* CCE ring buffer data */
+    r128RegionRec       ringRgn;
+    unsigned char      *ring;
+
+    /* CCE ring read pointer data */
+    r128RegionRec       ringReadRgn;
+
+    /* CCE vertex buffer data */
+    r128RegionRec       vbRgn;
+    unsigned char      *vb;
+    int                 vbOffset;
+    int                 vbBufSize;
+    drmBufMapPtr        vbBufs;
+
+    /* CCE indirect buffer data */
+    r128RegionRec       indRgn;
+    unsigned char      *ind;
+
+    /* CCE AGP Texture data */
+    r128RegionRec       agpTexRgn;
+    unsigned char      *agpTex;
+    int                 agpTexOffset;
+
+    /* Frame buffer data */
+    unsigned char      *fb;
+    unsigned long       fbOffset;
+    int                 fbStride;
+    int                 fbSize;
+
+    int                 IsPCI;		/* Current card is a PCI card */
+
+    int                 CCEMode;	/* CCE mode that server/clients use */
+    int                 CCEFifoSize;	/* Size of the CCE command FIFO */
+
+    /* CCE ring buffer data */
+    int                 ringEntries;
+
+    volatile int       *ringReadPtr;	/* Pointer to current read addr */
+    int                *ringStartPtr;   /* Pointer to end of ring buffer */
+    int                *ringEndPtr;     /* Pointer to end of ring buffer */
+
+    /* DRI screen private data */
+    int                 deviceID;	/* PCI device ID */
+    int                 width;		/* Width in pixels of display */
+    int                 height;		/* Height in scanlines of display */
+    int                 depth;		/* Depth of display (8, 15, 16, 24) */
+    int                 bpp;		/* Bit depth of disp (8, 16, 24, 32) */
+    int                 pixel_code;	/* 8, 15, 16, 24, 32 */
+
+    int                 fbX;		/* Start of frame buffer */
+    int                 fbY;
+    int                 backX;		/* Start of shared back buffer */
+    int                 backY;
+    int                 depthX;		/* Start of shared depth buffer */
+    int                 depthY;
+
+    /* Shared texture data */
+    int                 NRTexHeaps;
+    int                 texOffset[R128_NR_TEX_HEAPS];
+    int                 texSize[R128_NR_TEX_HEAPS];
+    int                 log2TexGran[R128_NR_TEX_HEAPS];
+
+    int                 MMIOFifoSlots;	/* Free slots in the FIFO (64 max) */
+    int                 CCEFifoSlots;	/* Free slots in the CCE FIFO */
+
+    int                 CCEFifoAddr;    /* MMIO offset to write next CCE
+					   value (only used when CCE is
+					   in PIO mode). */
+    R128SAREAPrivPtr    SAREA;		/* Pointer to SAREA private data */
+
+    __DRIscreenPrivate *driScreen;
+} r128ScreenRec, *r128ScreenPtr;
+
+r128ScreenPtr r128CreateScreen(__DRIscreenPrivate *sPriv);
+void          r128DestroyScreen(__DRIscreenPrivate *sPriv);
+
+#endif
+#endif /* _R128_SCREEN_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_span.c b/xc/lib/GL/mesa/src/drv/r128/r128_span.c
new file mode 100644
index 000000000..b5d1774fd
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_span.c
@@ -0,0 +1,329 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *   Keith Whitwell <keith@precisioninsight.com>
+ *   Gareth Hughes <gareth@precisioninsight.com>
+ *
+ */
+
+#include "r128_init.h"
+#include "r128_mesa.h"
+#include "r128_xmesa.h"
+#include "r128_context.h"
+#include "r128_lock.h"
+#include "r128_state.h"
+#include "r128_reg.h"
+#include "r128_cce.h"
+#include "r128_span.h"
+
+#define DBG 0
+
+#define LOCAL_VARS                                                            \
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);                               \
+    r128ScreenPtr r128scrn = r128ctx->r128Screen;                             \
+    __DRIdrawablePrivate *dPriv = r128ctx->driDrawable;                       \
+    GLuint pitch = r128scrn->fbStride;                                        \
+    GLuint height = dPriv->h;                                                 \
+    char *buf = (char *)(r128scrn->fb +                                       \
+			 (r128ctx->drawX + dPriv->x) * (r128scrn->bpp/8) +    \
+			 (r128ctx->drawY + dPriv->y) * pitch);                \
+    char *read_buf = (char *)(r128scrn->fb +                                  \
+			      (r128ctx->readX + dPriv->x) * (r128scrn->bpp/8)+\
+			      (r128ctx->readY + dPriv->y) * pitch);           \
+    GLushort p;                                                               \
+    (void) read_buf; (void) buf; (void) p
+
+#define LOCAL_DEPTH_VARS                                                      \
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);                               \
+    r128ScreenPtr r128scrn = r128ctx->r128Screen;                             \
+    __DRIdrawablePrivate *dPriv = r128ctx->driDrawable;                       \
+    GLuint pitch = r128scrn->fbStride;                                        \
+    GLuint height = dPriv->h;                                                 \
+    char *buf = (char *)(r128scrn->fb +                                       \
+			 (r128scrn->depthX + dPriv->x) * (r128scrn->bpp/8) +  \
+			 (r128scrn->depthY + dPriv->y) * pitch);              \
+    (void) buf
+
+#define INIT_MONO_PIXEL(p)                                                    \
+    p = R128_CONTEXT(ctx)->Color
+
+#define CLIPPIXEL(_x, _y)                                                     \
+    ((_x >= minx) && (_x < maxx) && (_y >= miny) && (_y < maxy))
+
+
+#define CLIPSPAN(_x, _y, _n, _x1, _n1, _i)                                    \
+    if (( _y < miny) || (_y >= maxy)) {                                       \
+	_n1 = 0, _x1 = x;                                                     \
+    } else {                                                                  \
+	_n1 = _n;                                                             \
+	_x1 = _x;                                                             \
+	if (_x1 < minx) _i += (minx - _x1), _x1 = minx;                       \
+	if (_x1 + _n1 >= maxx) n1 -= (_x1 + n1 - maxx) + 1;                   \
+    }
+
+#define Y_FLIP(_y) (height - _y - 1)
+
+
+#define HW_LOCK()                                                             \
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);                               \
+    LOCK_HARDWARE(r128ctx);                                                   \
+    R128CCE_WAIT_FOR_IDLE(r128ctx);
+
+#define HW_CLIPLOOP()                                                         \
+    do {                                                                      \
+	__DRIdrawablePrivate *dPriv = r128ctx->driDrawable;                   \
+	int _nc = dPriv->numClipRects;                                        \
+                                                                              \
+	while (_nc--) {                                                       \
+	    int minx = dPriv->pClipRects[_nc].x1 - dPriv->x;                  \
+	    int miny = dPriv->pClipRects[_nc].y1 - dPriv->y;                  \
+	    int maxx = dPriv->pClipRects[_nc].x2 - dPriv->x;                  \
+	    int maxy = dPriv->pClipRects[_nc].y2 - dPriv->y;
+
+#define HW_ENDCLIPLOOP()                                                      \
+	}                                                                     \
+    } while (0)
+
+#define HW_UNLOCK()                                                           \
+    UNLOCK_HARDWARE(r128ctx)
+
+
+
+/* 16 bit, RGB565 color spanline and pixel functions */
+#define WRITE_RGBA(_x, _y, r, g, b, a)                                        \
+    *(GLushort *)(buf + _x*2 + _y*pitch) = ((((int)r & 0xf8) << 8) |          \
+					    (((int)g & 0xfc) << 3) |          \
+					    (((int)b & 0xf8) >> 3) )
+
+#define WRITE_PIXEL(_x, _y, p)                                                \
+    *(GLushort *)(buf + _x*2 + _y*pitch) = p
+
+#define READ_RGBA(rgba, _x, _y)                                               \
+    do {                                                                      \
+	GLushort p = *(GLushort *)(read_buf + _x*2 + _y*pitch);               \
+	rgba[0] = (p >> 8) & 0xf8;                                            \
+	rgba[1] = (p >> 3) & 0xfc;                                            \
+	rgba[2] = (p << 3) & 0xf8;                                            \
+	rgba[3] = 0xff;                                                       \
+    } while (0)
+
+#define TAG(x) r128##x##_RGB565
+#include "spantmp.h"
+
+
+
+/* 15 bit, ARGB1555 color spanline and pixel functions */
+#define WRITE_RGBA(_x, _y, r, g, b, a)                                        \
+    *(GLushort *)(buf + _x*2 + _y*pitch) = (((r & 0xf8) << 7) |               \
+					    ((g & 0xf8) << 2) |               \
+					    ((b & 0xf8) >> 3) |               \
+					    ((a) ? 0x8000 : 0))
+
+#define WRITE_PIXEL(_x, _y, p)                                                \
+    *(GLushort *)(buf + _x*2 + _y*pitch) = p
+
+#define READ_RGBA(rgba, _x, _y)                                               \
+    do {                                                                      \
+	GLushort p = *(GLushort *)(read_buf + _x*2 + _y*pitch);               \
+	rgba[0] = (p >> 7) & 0xf8;                                            \
+	rgba[1] = (p >> 2) & 0xf8;                                            \
+	rgba[2] = (p << 3) & 0xf8;                                            \
+	rgba[3] = (p & 0x8000) ? 0xff : 0;                                    \
+    } while (0)
+
+#define TAG(x) r128##x##_ARGB1555
+#include "spantmp.h"
+
+
+
+/* 16 bit depthbuffer functions */
+#define WRITE_DEPTH(_x, _y, d)                                                \
+    *(GLdepth *)(buf + _x*2 + _y*pitch) = d
+
+#define READ_DEPTH(d, _x, _y)                                                 \
+    d = *(GLdepth *)(buf + _x*2 + _y*pitch)
+
+#define TAG(x) r128##x##_16
+#include "depthtmp.h"
+
+
+
+/* 24 bit, RGB888 color spanline and pixel functions */
+#define WRITE_RGBA(_x, _y, r, g, b, a)                                        \
+    *(GLuint *)(buf + _x*3 + _y*pitch) = ((r << 16) |                         \
+					  (g << 8)  |                         \
+					  (b << 0))
+
+#define WRITE_PIXEL(_x, _y, p)                                                \
+    *(GLuint *)(buf + _x*3 + _y*pitch) = p
+
+#define READ_RGBA(rgba, _x, _y)                                               \
+    do {                                                                      \
+	GLuint p = *(GLuint *)(read_buf + _x*3 + _y*pitch);                   \
+	rgba[0] = (p >> 16) & 0xff;                                           \
+	rgba[1] = (p >> 8)  & 0xff;                                           \
+	rgba[2] = (p >> 0)  & 0xff;                                           \
+	rgba[3] = 0xff;                                                       \
+    } while (0)
+
+#define TAG(x) r128##x##_RGB888
+#include "spantmp.h"
+
+
+
+/* 24 bit depthbuffer functions */
+#define WRITE_DEPTH(_x, _y, d)                                                \
+    *(GLdepth *)(buf + _x*3 + _y*pitch) = d
+
+#define READ_DEPTH(d, _x, _y)                                                 \
+    d = *(GLdepth *)(buf + _x*3 + _y*pitch)
+
+#define TAG(x) r128##x##_24
+#include "depthtmp.h"
+
+
+
+/* 32 bit, ARGB8888 color spanline and pixel functions */
+#define WRITE_RGBA(_x, _y, r, g, b, a)                                        \
+    *(GLuint *)(buf + _x*4 + _y*pitch) = ((r << 16) |                         \
+					  (g << 8)  |                         \
+					  (b << 0)  |                         \
+					  (a << 24) )
+
+#define WRITE_PIXEL(_x, _y, p)                                                \
+    *(GLuint *)(buf + _x*4 + _y*pitch) = p
+
+#define READ_RGBA(rgba, _x, _y)                                               \
+    do {                                                                      \
+	GLuint p = *(GLuint *)(read_buf + _x*4 + _y*pitch);                   \
+	rgba[0] = (p >> 16) & 0xff;                                           \
+	rgba[1] = (p >> 8)  & 0xff;                                           \
+	rgba[2] = (p >> 0)  & 0xff;                                           \
+	rgba[3] = (p >> 24) & 0xff;                                           \
+    } while (0)
+
+#define TAG(x) r128##x##_ARGB8888
+#include "spantmp.h"
+
+
+
+/* 32 bit depthbuffer functions */
+#define WRITE_DEPTH(_x, _y, d)                                                \
+    *(GLdepth *)(buf + _x*4 + _y*pitch) = d
+
+#define READ_DEPTH(d, _x, _y)                                                 \
+    d = *(GLdepth *)(buf + _x*4 + _y*pitch)
+
+#define TAG(x) r128##x##_32
+#include "depthtmp.h"
+
+
+
+void r128DDInitSpanFuncs(GLcontext *ctx)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+
+    switch (r128ctx->r128Screen->pixel_code) {
+    case 8: /* Color Index mode not supported */
+	break;
+
+    case 15:
+	ctx->Driver.WriteRGBASpan       = r128WriteRGBASpan_ARGB1555;
+	ctx->Driver.WriteRGBSpan        = r128WriteRGBSpan_ARGB1555;
+	ctx->Driver.WriteMonoRGBASpan   = r128WriteMonoRGBASpan_ARGB1555;
+	ctx->Driver.WriteRGBAPixels     = r128WriteRGBAPixels_ARGB1555;
+	ctx->Driver.WriteMonoRGBAPixels = r128WriteMonoRGBAPixels_ARGB1555;
+	ctx->Driver.ReadRGBASpan        = r128ReadRGBASpan_ARGB1555;
+	ctx->Driver.ReadRGBAPixels      = r128ReadRGBAPixels_ARGB1555;
+
+	ctx->Driver.ReadDepthSpan       = r128ReadDepthSpan_16;
+	ctx->Driver.WriteDepthSpan      = r128WriteDepthSpan_16;
+	ctx->Driver.ReadDepthPixels     = r128ReadDepthPixels_16;
+	ctx->Driver.WriteDepthPixels    = r128WriteDepthPixels_16;
+	break;
+
+    case 16:
+	ctx->Driver.WriteRGBASpan       = r128WriteRGBASpan_RGB565;
+	ctx->Driver.WriteRGBSpan        = r128WriteRGBSpan_RGB565;
+	ctx->Driver.WriteMonoRGBASpan   = r128WriteMonoRGBASpan_RGB565;
+	ctx->Driver.WriteRGBAPixels     = r128WriteRGBAPixels_RGB565;
+	ctx->Driver.WriteMonoRGBAPixels = r128WriteMonoRGBAPixels_RGB565;
+	ctx->Driver.ReadRGBASpan        = r128ReadRGBASpan_RGB565;
+	ctx->Driver.ReadRGBAPixels      = r128ReadRGBAPixels_RGB565;
+
+	ctx->Driver.ReadDepthSpan       = r128ReadDepthSpan_16;
+	ctx->Driver.WriteDepthSpan      = r128WriteDepthSpan_16;
+	ctx->Driver.ReadDepthPixels     = r128ReadDepthPixels_16;
+	ctx->Driver.WriteDepthPixels    = r128WriteDepthPixels_16;
+	break;
+
+    case 24:
+	ctx->Driver.WriteRGBASpan       = r128WriteRGBASpan_RGB888;
+	ctx->Driver.WriteRGBSpan        = r128WriteRGBSpan_RGB888;
+	ctx->Driver.WriteMonoRGBASpan   = r128WriteMonoRGBASpan_RGB888;
+	ctx->Driver.WriteRGBAPixels     = r128WriteRGBAPixels_RGB888;
+	ctx->Driver.WriteMonoRGBAPixels = r128WriteMonoRGBAPixels_RGB888;
+	ctx->Driver.ReadRGBASpan        = r128ReadRGBASpan_RGB888;
+	ctx->Driver.ReadRGBAPixels      = r128ReadRGBAPixels_RGB888;
+
+	ctx->Driver.ReadDepthSpan       = r128ReadDepthSpan_24;
+	ctx->Driver.WriteDepthSpan      = r128WriteDepthSpan_24;
+	ctx->Driver.ReadDepthPixels     = r128ReadDepthPixels_24;
+	ctx->Driver.WriteDepthPixels    = r128WriteDepthPixels_24;
+	break;
+
+    case 32:
+	ctx->Driver.WriteRGBASpan       = r128WriteRGBASpan_ARGB8888;
+	ctx->Driver.WriteRGBSpan        = r128WriteRGBSpan_ARGB8888;
+	ctx->Driver.WriteMonoRGBASpan   = r128WriteMonoRGBASpan_ARGB8888;
+	ctx->Driver.WriteRGBAPixels     = r128WriteRGBAPixels_ARGB8888;
+	ctx->Driver.WriteMonoRGBAPixels = r128WriteMonoRGBAPixels_ARGB8888;
+	ctx->Driver.ReadRGBASpan        = r128ReadRGBASpan_ARGB8888;
+	ctx->Driver.ReadRGBAPixels      = r128ReadRGBAPixels_ARGB8888;
+
+	ctx->Driver.ReadDepthSpan       = r128ReadDepthSpan_32;
+	ctx->Driver.WriteDepthSpan      = r128WriteDepthSpan_32;
+	ctx->Driver.ReadDepthPixels     = r128ReadDepthPixels_32;
+	ctx->Driver.WriteDepthPixels    = r128WriteDepthPixels_32;
+	break;
+
+    default:
+	break;
+    }
+
+    ctx->Driver.WriteCI8Span      = NULL;
+    ctx->Driver.WriteCI32Span     = NULL;
+    ctx->Driver.WriteMonoCISpan   = NULL;
+    ctx->Driver.WriteCI32Pixels   = NULL;
+    ctx->Driver.WriteMonoCIPixels = NULL;
+    ctx->Driver.ReadCI32Span      = NULL;
+    ctx->Driver.ReadCI32Pixels    = NULL;
+}
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_span.h b/xc/lib/GL/mesa/src/drv/r128/r128_span.h
new file mode 100644
index 000000000..f83a4da9f
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_span.h
@@ -0,0 +1,44 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *   Gareth Hughes <gareth@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_SPAN_H_
+#define _R128_SPAN_H_
+
+#ifdef GLX_DIRECT_RENDERING
+
+extern void r128DDInitSpanFuncs(GLcontext *ctx);
+
+#endif
+#endif /* _R128_SPAN_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_state.c b/xc/lib/GL/mesa/src/drv/r128/r128_state.c
new file mode 100644
index 000000000..3d1ceaa24
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_state.c
@@ -0,0 +1,1309 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#include "r128_init.h"
+#include "r128_mesa.h"
+#include "r128_xmesa.h"
+#include "r128_context.h"
+#include "r128_lock.h"
+#include "r128_state.h"
+#include "r128_reg.h"
+#include "r128_cce.h"
+#include "r128_tris.h"
+#include "r128_vb.h"
+#include "r128_tex.h"
+
+#include "mmath.h"
+
+#define INTERESTED (~(NEW_MODELVIEW      | \
+		      NEW_PROJECTION     | \
+		      NEW_TEXTURE_MATRIX | \
+		      NEW_USER_CLIP      | \
+		      NEW_CLIENT_STATE   | \
+		      NEW_TEXTURE_ENABLE))
+
+static void r128DDUpdateState(GLcontext *ctx)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+
+    if (R128_DEBUG_FLAGS & DEBUG_VERBOSE_API) {
+	fprintf(stderr, "r128DDUpdateState(%p)\n", ctx);
+    }
+
+    if (ctx->NewState & INTERESTED) {
+	r128ChooseRenderState(ctx);
+	r128ChooseRasterSetupFunc(ctx);
+    }
+
+    if (!r128ctx->Fallback) {
+	ctx->IndirectTriangles   &= ~DD_SW_RASTERIZE;
+	ctx->IndirectTriangles   |= r128ctx->IndirectTriangles;
+
+	ctx->Driver.PointsFunc    = r128ctx->PointsFunc;
+	ctx->Driver.LineFunc      = r128ctx->LineFunc;
+	ctx->Driver.TriangleFunc  = r128ctx->TriangleFunc;
+	ctx->Driver.QuadFunc      = r128ctx->QuadFunc;
+	ctx->Driver.RectFunc      = NULL;
+    }
+}
+
+static void r128DDUpdateHWState(GLcontext *ctx)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+
+    /* FIXME: state is being updated too often */
+    if (r128ctx->dirty)
+	r128UpdateHWState(r128ctx);
+}
+
+static void r128DDReducedPrimitiveChange(GLcontext *ctx, GLenum prim)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+
+    /* FIXME: Also need to flush between tris and tristrips/fans when we
+       support them directly */
+    R128CCE_FLUSH_VB_LOCK(r128ctx);
+}
+
+static void r128DDClearColor(GLcontext *ctx, 
+                             GLubyte r, GLubyte g, GLubyte b, GLubyte a)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+
+    r128ctx->ClearColor = r128PackColor(r128ctx->r128Screen->depth,
+					r, g, b, a);
+}
+
+static void r128DDColor(GLcontext *ctx,
+			GLubyte r, GLubyte g, GLubyte b, GLubyte a)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+
+    r128ctx->Color = r128PackColor(r128ctx->r128Screen->depth, r, g, b, a);
+}
+
+static GLboolean r128DDSetDrawBuffer(GLcontext *ctx, GLenum mode)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+    int            x       = r128ctx->driDrawable->x;
+    int            y       = r128ctx->driDrawable->y;
+    int            found;
+
+    r128ctx->Fallback &= ~R128_FALLBACK_DRAW_BUFFER;
+
+    switch (mode) {
+    case GL_FRONT_LEFT:
+	r128ctx->drawX = r128ctx->r128Screen->fbX;
+	r128ctx->drawY = r128ctx->r128Screen->fbY;
+	found = GL_TRUE;
+	break;
+    case GL_BACK_LEFT:
+	r128ctx->drawX = r128ctx->r128Screen->backX;
+	r128ctx->drawY = r128ctx->r128Screen->backY;
+	found = GL_TRUE;
+	break;
+    default:
+	r128ctx->Fallback |= R128_FALLBACK_DRAW_BUFFER;
+	found = GL_FALSE;
+	break;
+    }
+
+    x += r128ctx->drawX;
+    y += r128ctx->drawY;
+
+    R128CCE_FLUSH_VB_LOCK(r128ctx);
+    r128ctx->regs.window_xy_offset = ((y << R128_WINDOW_Y_SHIFT) |
+				      (x << R128_WINDOW_X_SHIFT));
+
+    /* Recalculate the Z buffer offset since we might be drawing to the
+       back buffer and window_xy_offset affects both color buffer and
+       depth drawing */
+    r128ctx->regs.z_offset_c        = ((r128ctx->r128Screen->depthX -
+					r128ctx->drawX) *
+				       (r128ctx->r128Screen->bpp/8) +
+				       (r128ctx->r128Screen->depthY -
+					r128ctx->drawY) *
+				       r128ctx->r128Screen->fbStride);
+
+    r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+    r128ctx->dirty_context |= R128_CTX_WIN_Z_POS;
+    return found;
+}
+
+static void r128DDSetReadBuffer(GLcontext *ctx,
+				GLframebuffer *colorBuffer,
+				GLenum mode)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+
+    r128ctx->Fallback &= ~R128_FALLBACK_READ_BUFFER;
+
+    switch (mode) {
+    case GL_FRONT_LEFT:
+	r128ctx->readX = r128ctx->r128Screen->fbX;
+	r128ctx->readY = r128ctx->r128Screen->fbY;
+	break;
+    case GL_BACK_LEFT:
+	r128ctx->readX = r128ctx->r128Screen->backX;
+	r128ctx->readY = r128ctx->r128Screen->backY;
+	break;
+    default:
+	r128ctx->Fallback |= R128_FALLBACK_READ_BUFFER;
+	break;
+    }
+}
+
+static GLboolean r128DDColorMask(GLcontext *ctx,
+				 GLboolean r, GLboolean g,
+				 GLboolean b, GLboolean a)
+{
+    r128ContextPtr r128ctx  = R128_CONTEXT(ctx);
+
+    GLuint mask = r128PackColor(r128ctx->r128Screen->pixel_code,
+				ctx->Color.ColorMask[RCOMP],
+				ctx->Color.ColorMask[GCOMP],
+				ctx->Color.ColorMask[BCOMP],
+				ctx->Color.ColorMask[ACOMP]);
+
+    if (r128ctx->regs.plane_3d_mask_c != mask) {
+	R128CCE_FLUSH_VB_LOCK(r128ctx);
+	r128ctx->regs.plane_3d_mask_c = mask;
+
+	/* FIXME: Load into hardware now??? */
+	r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+	r128ctx->dirty_context |= R128_CTX_MISC;
+    }
+
+    return GL_TRUE;
+}
+
+static void r128DDDither(GLcontext *ctx, GLboolean enable)
+{
+}
+
+static void r128DDAlphaFunc(GLcontext *ctx, GLenum func, GLclampf ref)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+    CARD32         a       = r128ctx->regs.misc_3d_state_cntl_reg;
+
+    a &= ~(R128_ALPHA_TEST_MASK | R128_REF_ALPHA_MASK);
+    a |= ctx->Color.AlphaRef & R128_REF_ALPHA_MASK;
+
+    switch (func) {
+    case GL_NEVER:    a |= R128_ALPHA_TEST_NEVER;        break;
+    case GL_LESS:     a |= R128_ALPHA_TEST_LESS;         break;
+    case GL_LEQUAL:   a |= R128_ALPHA_TEST_LESSEQUAL;    break;
+    case GL_EQUAL:    a |= R128_ALPHA_TEST_EQUAL;        break;
+    case GL_GEQUAL:   a |= R128_ALPHA_TEST_GREATEREQUAL; break;
+    case GL_GREATER:  a |= R128_ALPHA_TEST_GREATER;      break;
+    case GL_NOTEQUAL: a |= R128_ALPHA_TEST_NEQUAL;       break;
+    case GL_ALWAYS:   a |= R128_ALPHA_TEST_ALWAYS;
+	break;
+    default:
+	/* ERROR!!! */
+	return;
+    }
+
+    if (r128ctx->regs.misc_3d_state_cntl_reg != a) {
+	R128CCE_FLUSH_VB_LOCK(r128ctx);
+	r128ctx->regs.misc_3d_state_cntl_reg = a;
+
+	/* FIXME: Load into hardware now??? */
+	r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+	r128ctx->dirty_context |= R128_CTX_ALPHASTATE;
+    }
+}
+
+static void r128DDBlendFunc(GLcontext *ctx, GLenum sfactor, GLenum dfactor)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+    CARD32         b       = r128ctx->regs.misc_3d_state_cntl_reg;
+
+    b &= ~(R128_ALPHA_BLEND_SRC_MASK | R128_ALPHA_BLEND_DST_MASK);
+
+    switch (sfactor) {
+    case GL_ZERO:                     b |= R128_ALPHA_BLEND_SRC_ZERO;
+	break;
+    case GL_ONE:                      b |= R128_ALPHA_BLEND_SRC_ONE;
+	break;
+    case GL_DST_COLOR:                b |= R128_ALPHA_BLEND_SRC_DESTCOLOR;
+	break;
+    case GL_ONE_MINUS_DST_COLOR:      b |= R128_ALPHA_BLEND_SRC_INVDESTCOLOR;
+	break;
+    case GL_SRC_ALPHA:                b |= R128_ALPHA_BLEND_SRC_SRCALPHA;
+	break;
+    case GL_ONE_MINUS_SRC_ALPHA:      b |= R128_ALPHA_BLEND_SRC_INVSRCALPHA;
+	break;
+    case GL_DST_ALPHA:                b |= R128_ALPHA_BLEND_SRC_DESTALPHA;
+	break;
+    case GL_ONE_MINUS_DST_ALPHA:      b |= R128_ALPHA_BLEND_SRC_INVDESTALPHA;
+	break;
+    case GL_SRC_ALPHA_SATURATE:       b |= R128_ALPHA_BLEND_SRC_SRCALPHASAT;
+	break;
+#if 0
+	/* FIXME: These are not supported directly by the Rage 128.
+           They could be emulated using something like the TexEnv
+           modes. */
+    case GL_CONSTANT_COLOR:           b |= 0;
+	break;
+    case GL_ONE_MINUS_CONSTANT_COLOR: b |= 0;
+	break;
+    case GL_CONSTANT_ALPHA:           b |= 0;
+	break;
+    case GL_ONE_MINUS_CONSTANT_ALPHA: b |= 0;
+	break;
+#endif
+    default:
+	/* ERROR!!! */
+	return;
+    }
+
+    switch (dfactor) {
+    case GL_ZERO:                     b |= R128_ALPHA_BLEND_DST_ZERO;
+	break;
+    case GL_ONE:                      b |= R128_ALPHA_BLEND_DST_ONE;
+	break;
+    case GL_SRC_COLOR:                b |= R128_ALPHA_BLEND_DST_SRCCOLOR;
+	break;
+    case GL_ONE_MINUS_SRC_COLOR:      b |= R128_ALPHA_BLEND_DST_INVSRCCOLOR;
+	break;
+    case GL_SRC_ALPHA:                b |= R128_ALPHA_BLEND_DST_SRCALPHA;
+	break;
+    case GL_ONE_MINUS_SRC_ALPHA:      b |= R128_ALPHA_BLEND_DST_INVSRCALPHA;
+	break;
+    case GL_DST_ALPHA:                b |= R128_ALPHA_BLEND_DST_DESTALPHA;
+	break;
+    case GL_ONE_MINUS_DST_ALPHA:      b |= R128_ALPHA_BLEND_DST_INVDESTALPHA;
+	break;
+#if 0
+	/* FIXME: These are not supported directly by the Rage 128.
+           They could be emulated using something like the TexEnv
+           modes. */
+    case GL_CONSTANT_COLOR:           b |= 0;
+	break;
+    case GL_ONE_MINUS_CONSTANT_COLOR: b |= 0;
+	break;
+    case GL_CONSTANT_ALPHA:           b |= 0;
+	break;
+    case GL_ONE_MINUS_CONSTANT_ALPHA: b |= 0;
+	break;
+#endif
+    default:
+	/* ERROR!!! */
+	return;
+    }
+
+    if (r128ctx->regs.misc_3d_state_cntl_reg != b) {
+	R128CCE_FLUSH_VB_LOCK(r128ctx);
+	r128ctx->regs.misc_3d_state_cntl_reg = b;
+
+	/* FIXME: Load into hardware now??? */
+	r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+	r128ctx->dirty_context |= R128_CTX_ALPHASTATE;
+    }
+}
+
+static void r128DDBlendFuncSeparate(GLcontext *ctx, GLenum sfactorRGB,
+				    GLenum dfactorRGB, GLenum sfactorA,
+				    GLenum dfactorA)
+{
+    if (sfactorRGB != sfactorA || dfactorRGB != dfactorA) {
+	/* ERROR!!! */
+	return;
+    }
+
+    r128DDBlendFunc(ctx, sfactorRGB, dfactorRGB);
+}
+
+static void r128DDClearDepth(GLcontext *ctx, GLclampd d)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+
+    switch (r128ctx->regs.z_sten_cntl_c & R128_Z_PIX_WIDTH_MASK) {
+    case R128_Z_PIX_WIDTH_16: r128ctx->ClearDepth = d * 0x0000ffff; break;
+    case R128_Z_PIX_WIDTH_24: r128ctx->ClearDepth = d * 0x00ffffff; break;
+    case R128_Z_PIX_WIDTH_32: r128ctx->ClearDepth = d * 0xffffffff; break;
+    default: return;
+    }
+}
+
+static void r128DDCullFace(GLcontext *ctx, GLenum mode)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+    CARD32         f       = r128ctx->regs.pm4_vc_fpu_setup;
+
+    if (!ctx->Polygon.CullFlag) return;
+
+    f &= ~(R128_BACKFACE_MASK | R128_FRONTFACE_MASK);
+
+    switch (mode) {
+    case GL_FRONT:          f |= R128_BACKFACE_SOLID;  break;
+    case GL_BACK:           f |= R128_FRONTFACE_SOLID; break;
+    case GL_FRONT_AND_BACK:                            break;
+    default: return;
+    }
+
+    if (r128ctx->regs.pm4_vc_fpu_setup != f) {
+	R128CCE_FLUSH_VB_LOCK(r128ctx);
+	r128ctx->regs.pm4_vc_fpu_setup = f;
+
+	/* FIXME: Load into hardware now??? */
+	r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+	r128ctx->dirty_context |= R128_CTX_SETUPSTATE;
+    }
+}
+
+static void r128DDFrontFace(GLcontext *ctx, GLenum mode)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+    CARD32         f       = r128ctx->regs.pm4_vc_fpu_setup;
+
+    f &= ~R128_FRONT_DIR_MASK;
+
+    switch (mode) {
+    case GL_CW:  f |= R128_FRONT_DIR_CW;  break;
+    case GL_CCW: f |= R128_FRONT_DIR_CCW; break;
+    default: return;
+    }
+
+    if (r128ctx->regs.pm4_vc_fpu_setup != f) {
+	R128CCE_FLUSH_VB_LOCK(r128ctx);
+	r128ctx->regs.pm4_vc_fpu_setup = f;
+
+	/* FIXME: Load into hardware now??? */
+	r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+	r128ctx->dirty_context |= R128_CTX_SETUPSTATE;
+    }
+}
+
+static void r128DDDepthFunc(GLcontext *ctx, GLenum func)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+    CARD32         z       = r128ctx->regs.z_sten_cntl_c;
+
+    z &= ~R128_Z_TEST_MASK;
+
+    switch (func) {
+    case GL_NEVER:    z |= R128_Z_TEST_NEVER;        break;
+    case GL_LESS:     z |= R128_Z_TEST_LESS;         break;
+    case GL_LEQUAL:   z |= R128_Z_TEST_LESSEQUAL;    break;
+    case GL_EQUAL:    z |= R128_Z_TEST_EQUAL;        break;
+    case GL_GEQUAL:   z |= R128_Z_TEST_GREATEREQUAL; break;
+    case GL_GREATER:  z |= R128_Z_TEST_GREATER;      break;
+    case GL_NOTEQUAL: z |= R128_Z_TEST_NEQUAL;       break;
+    case GL_ALWAYS:   z |= R128_Z_TEST_ALWAYS;       break;
+    default: return;
+    }
+
+    if (r128ctx->regs.z_sten_cntl_c != z) {
+	R128CCE_FLUSH_VB_LOCK(r128ctx);
+	r128ctx->regs.z_sten_cntl_c = z;
+
+	/* FIXME: Load into hardware now??? */
+	r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+	r128ctx->dirty_context |= R128_CTX_ZSTENSTATE;
+    }
+}
+
+static void r128DDDepthMask(GLcontext *ctx, GLboolean flag)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+    CARD32         t       = r128ctx->regs.tex_cntl_c;
+
+    if (flag) t |=  R128_Z_WRITE_ENABLE;
+    else      t &= ~R128_Z_WRITE_ENABLE;
+
+    if (r128ctx->regs.tex_cntl_c != t) {
+	R128CCE_FLUSH_VB_LOCK(r128ctx);
+	r128ctx->regs.tex_cntl_c = t;
+
+	/* FIXME: Load into hardware now??? */
+	r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+	r128ctx->dirty_context |= R128_CTX_ENGINESTATE;
+    }
+}
+
+static void r128DDEnable(GLcontext *ctx, GLenum cap, GLboolean state)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+    CARD32         t       = r128ctx->regs.tex_cntl_c;
+    CARD32         f       = r128ctx->regs.pm4_vc_fpu_setup;
+
+    if (R128_DEBUG_FLAGS & DEBUG_VERBOSE_API) {
+	fprintf(stderr, "r128DDEnable( %p, 0x%x = %s )\n",
+		ctx, cap, state ? "GL_TRUE" : "GL_FALSE");
+    }
+
+    switch (cap) {
+    case GL_ALPHA_TEST:
+	if (state) t |=  R128_ALPHA_TEST_ENABLE;
+	else       t &= ~R128_ALPHA_TEST_ENABLE;
+	break;
+
+    case GL_AUTO_NORMAL: return;
+
+    case GL_BLEND:
+	if (state) t |=  R128_ALPHA_ENABLE;
+	else       t &= ~R128_ALPHA_ENABLE;
+	break;
+
+    case GL_CLIP_PLANE0:
+    case GL_CLIP_PLANE1:
+    case GL_CLIP_PLANE2:
+    case GL_CLIP_PLANE3:
+    case GL_CLIP_PLANE4:
+    case GL_CLIP_PLANE5:
+    case GL_COLOR_MATERIAL: return;
+
+    case GL_CULL_FACE:
+	f &= ~(R128_BACKFACE_MASK | R128_FRONTFACE_MASK);
+	if (state) {
+	    switch (ctx->Polygon.CullFaceMode) {
+	    case GL_FRONT:          f |= R128_BACKFACE_SOLID;  break;
+	    case GL_BACK:           f |= R128_FRONTFACE_SOLID; break;
+	    case GL_FRONT_AND_BACK:                            break;
+	    default: return;
+	    }
+	} else {
+	    f |= R128_BACKFACE_SOLID | R128_FRONTFACE_SOLID;
+	}
+	break;
+
+    case GL_DEPTH_TEST:
+	if (state) t |=  R128_Z_ENABLE;
+	else       t &= ~R128_Z_ENABLE;
+	break;
+
+    case GL_DITHER:
+	if (state) t |=  R128_DITHER_ENABLE;
+	else       t &= ~R128_DITHER_ENABLE;
+	break;
+
+    case GL_FOG:
+	if (state) t |=  R128_FOG_ENABLE;
+	else       t &= ~R128_FOG_ENABLE;
+	break;
+
+    case GL_LIGHT0:
+    case GL_LIGHT1:
+    case GL_LIGHT2:
+    case GL_LIGHT3:
+    case GL_LIGHT4:
+    case GL_LIGHT5:
+    case GL_LIGHT6:
+    case GL_LIGHT7:
+    case GL_LIGHTING:
+    case GL_LINE_SMOOTH:
+    case GL_LINE_STIPPLE:
+    case GL_INDEX_LOGIC_OP:
+    case GL_COLOR_LOGIC_OP:
+    case GL_MAP1_COLOR_4:
+    case GL_MAP1_INDEX:
+    case GL_MAP1_NORMAL:
+    case GL_MAP1_TEXTURE_COORD_1:
+    case GL_MAP1_TEXTURE_COORD_2:
+    case GL_MAP1_TEXTURE_COORD_3:
+    case GL_MAP1_TEXTURE_COORD_4:
+    case GL_MAP1_VERTEX_3:
+    case GL_MAP1_VERTEX_4:
+    case GL_MAP2_COLOR_4:
+    case GL_MAP2_INDEX:
+    case GL_MAP2_NORMAL:
+    case GL_MAP2_TEXTURE_COORD_1:
+    case GL_MAP2_TEXTURE_COORD_2:
+    case GL_MAP2_TEXTURE_COORD_3:
+    case GL_MAP2_TEXTURE_COORD_4:
+    case GL_MAP2_VERTEX_3:
+    case GL_MAP2_VERTEX_4:
+    case GL_NORMALIZE:
+    case GL_POINT_SMOOTH:
+    case GL_POLYGON_SMOOTH:
+    case GL_POLYGON_STIPPLE:
+    case GL_POLYGON_OFFSET_POINT:
+    case GL_POLYGON_OFFSET_LINE:
+    case GL_POLYGON_OFFSET_FILL:
+    case GL_RESCALE_NORMAL_EXT: return;
+
+    case GL_SCISSOR_TEST:
+	/* FIXME: Hook up the software scissor */
+#if 0
+	r128ctx->Scissor = state;
+#endif
+	break;
+
+    case GL_SHARED_TEXTURE_PALETTE_EXT:
+    case GL_STENCIL_TEST: return;
+
+    case GL_TEXTURE_1D:
+    case GL_TEXTURE_2D:
+	/* This is handled in r128UpdateTex[01]State() */
+	r128ctx->dirty |= R128_UPDATE_TEXSTATE;
+	break;
+
+    case GL_TEXTURE_3D:
+    case GL_TEXTURE_GEN_Q:
+    case GL_TEXTURE_GEN_R:
+    case GL_TEXTURE_GEN_S:
+    case GL_TEXTURE_GEN_T: return;
+
+	/* Client state */
+    case GL_VERTEX_ARRAY:
+    case GL_NORMAL_ARRAY:
+    case GL_COLOR_ARRAY:
+    case GL_INDEX_ARRAY:
+    case GL_TEXTURE_COORD_ARRAY:
+    case GL_EDGE_FLAG_ARRAY: return;
+
+    default: return;
+    }
+
+    if (r128ctx->regs.tex_cntl_c != t) {
+	R128CCE_FLUSH_VB_LOCK(r128ctx);
+	r128ctx->regs.tex_cntl_c = t;
+
+	/* FIXME: Load into hardware now??? */
+	r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+	r128ctx->dirty_context |= R128_CTX_ENGINESTATE;
+    }
+    if (r128ctx->regs.pm4_vc_fpu_setup != f) {
+	R128CCE_FLUSH_VB_LOCK(r128ctx);
+	r128ctx->regs.pm4_vc_fpu_setup = f;
+
+	/* FIXME: Load into hardware now??? */
+	r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+	r128ctx->dirty_context |= R128_CTX_SETUPSTATE;
+    }
+
+}
+
+static void r128DDFogfv(GLcontext *ctx, GLenum pname, const GLfloat *param)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+    GLubyte        c[4];
+    CARD32         col;
+    floatTOint     fog;
+    GLenum         mode;
+
+    if (R128_DEBUG_FLAGS & DEBUG_VERBOSE_API) {
+	fprintf(stderr, "r128DDFogfv(%p, 0x%x)\n", ctx, pname);
+    }
+
+    switch (pname) {
+    case GL_FOG_MODE:
+	mode = (GLenum)(GLint)*param;
+	if (r128ctx->FogMode != mode) {
+	    R128CCE_FLUSH_VB_LOCK(r128ctx);
+	    r128ctx->FogMode = mode;
+
+	    /* FIXME: Load into hardware now??? */
+	    r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+	    r128ctx->dirty_context |= R128_CTX_FOGTABLE;
+	}
+	break;
+
+    case GL_FOG_DENSITY:
+	fog.f = *param;
+	if (r128ctx->regs.fog_3d_table_density != fog.i) {
+	    R128CCE_FLUSH_VB_LOCK(r128ctx);
+	    r128ctx->regs.fog_3d_table_density = fog.i;
+
+	    /* FIXME: Load into hardware now??? */
+	    r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+	    r128ctx->dirty_context |= R128_CTX_FOGSTATE;
+	}
+	break;
+
+    case GL_FOG_START:
+	fog.f = *param;
+	if (r128ctx->regs.fog_3d_table_start != fog.i) {
+	    R128CCE_FLUSH_VB_LOCK(r128ctx);
+	    r128ctx->regs.fog_3d_table_start = fog.i;
+
+	    /* FIXME: Load into hardware now??? */
+	    r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+	    r128ctx->dirty_context |= R128_CTX_FOGSTATE;
+	}
+	break;
+
+    case GL_FOG_END:
+	fog.f = *param;
+	if (r128ctx->regs.fog_3d_table_end != fog.i) {
+	    R128CCE_FLUSH_VB_LOCK(r128ctx);
+	    r128ctx->regs.fog_3d_table_end = fog.i;
+
+	    /* FIXME: Load into hardware now??? */
+	    r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+	    r128ctx->dirty_context |= R128_CTX_FOGSTATE;
+	}
+	break;
+
+    case GL_FOG_COLOR:
+	FLOAT_RGBA_TO_UBYTE_RGBA(c, ctx->Fog.Color);
+	col = r128PackColor(32, c[0], c[1], c[2], c[3]);
+	if (r128ctx->regs.fog_color_c != col) {
+	    R128CCE_FLUSH_VB_LOCK(r128ctx);
+	    r128ctx->regs.fog_color_c = col;
+
+	    /* FIXME: Load into hardware now??? */
+	    r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+	    r128ctx->dirty_context |= R128_CTX_FOGSTATE;
+	}
+	break;
+
+    default:
+	return;
+    }
+}
+
+static void r128DDScissor(GLcontext *ctx,
+			  GLint x, GLint y, GLsizei w, GLsizei h)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+
+    r128ctx->ScissorRect.x1 = x;
+    r128ctx->ScissorRect.y1 = r128ctx->driDrawable->h - (y + h);
+    r128ctx->ScissorRect.x2 = x + w;
+    r128ctx->ScissorRect.y2 = r128ctx->driDrawable->h - y;
+}
+
+/* Initialize the driver's state functions */
+void r128DDInitStateFuncs(GLcontext *ctx)
+{
+    ctx->Driver.UpdateState            = r128DDUpdateState;
+
+    ctx->Driver.ClearIndex             = NULL;
+    ctx->Driver.ClearColor             = r128DDClearColor;
+    ctx->Driver.Index                  = NULL;
+    ctx->Driver.Color                  = r128DDColor;
+    ctx->Driver.SetDrawBuffer          = r128DDSetDrawBuffer;
+    ctx->Driver.SetReadBuffer          = r128DDSetReadBuffer;
+
+    ctx->Driver.IndexMask              = NULL;
+    ctx->Driver.ColorMask              = r128DDColorMask;
+    ctx->Driver.LogicOp                = NULL;
+    ctx->Driver.Dither                 = r128DDDither;
+
+    ctx->Driver.NearFar                = NULL;
+
+    ctx->Driver.RenderStart            = r128DDUpdateHWState;
+    ctx->Driver.RenderFinish           = NULL;
+    ctx->Driver.RasterSetup            = NULL;
+
+    ctx->Driver.RenderVBClippedTab     = NULL;
+    ctx->Driver.RenderVBCulledTab      = NULL;
+    ctx->Driver.RenderVBRawTab         = NULL;
+
+    ctx->Driver.ReducedPrimitiveChange = r128DDReducedPrimitiveChange;
+    ctx->Driver.MultipassFunc          = NULL;
+
+    ctx->Driver.AlphaFunc              = r128DDAlphaFunc;
+    ctx->Driver.BlendEquation          = NULL;
+    ctx->Driver.BlendFunc              = r128DDBlendFunc;
+    ctx->Driver.BlendFuncSeparate      = r128DDBlendFuncSeparate;
+    ctx->Driver.ClearDepth             = r128DDClearDepth;
+    ctx->Driver.CullFace               = r128DDCullFace;
+    ctx->Driver.FrontFace              = r128DDFrontFace;
+    ctx->Driver.DepthFunc              = r128DDDepthFunc;
+    ctx->Driver.DepthMask              = r128DDDepthMask;
+    ctx->Driver.DepthRange             = NULL;
+    ctx->Driver.Enable                 = r128DDEnable;
+    ctx->Driver.Fogfv                  = r128DDFogfv;
+    ctx->Driver.Hint                   = NULL;
+    ctx->Driver.Lightfv                = NULL;
+    ctx->Driver.LightModelfv           = NULL;
+    ctx->Driver.PolygonMode            = NULL;
+    ctx->Driver.Scissor                = r128DDScissor;
+    ctx->Driver.ShadeModel             = NULL;
+    ctx->Driver.ClearStencil           = NULL;
+    ctx->Driver.StencilFunc            = NULL;
+    ctx->Driver.StencilMask            = NULL;
+    ctx->Driver.StencilOp              = NULL;
+    ctx->Driver.Viewport               = NULL;
+}
+
+/* Initialize the context's hardware state */
+void r128DDInitState(r128ContextPtr r128ctx)
+{
+    int    dst_bpp, depth_bpp, pitch, i;
+    CARD32 depthClear;
+
+    pitch = r128ctx->r128Screen->fbStride / r128ctx->r128Screen->bpp;
+
+    switch (r128ctx->r128Screen->pixel_code) {
+    case  8: dst_bpp = R128_GMC_DST_8BPP_CI; break;
+    case 15: dst_bpp = R128_GMC_DST_15BPP;   break;
+    case 16: dst_bpp = R128_GMC_DST_16BPP;   break;
+    case 24: dst_bpp = R128_GMC_DST_24BPP;   break;
+    case 32: dst_bpp = R128_GMC_DST_32BPP;   break;
+    default:
+	fprintf(stderr, "Error: Unsupported pixel depth %d... exiting\n",
+		r128ctx->r128Screen->pixel_code);
+	exit(-1);
+    }
+
+    /* FIXME: Figure out how to use 16bpp depth buffer in 32bpp mode */
+    switch (r128ctx->glCtx->Visual->DepthBits) {
+    case 16: depthClear = 0x0000ffff; depth_bpp  = R128_Z_PIX_WIDTH_16; break;
+    case 24: depthClear = 0x00ffffff; depth_bpp  = R128_Z_PIX_WIDTH_24; break;
+    case 32: depthClear = 0xffffffff; depth_bpp  = R128_Z_PIX_WIDTH_32; break;
+    default:
+	fprintf(stderr, "Error: Unsupported depth %d... exiting\n",
+		r128ctx->r128Screen->bpp);
+	exit(-1);
+	break;
+    }
+
+    r128ctx->dirty             = R128_ALL_DIRTY;
+    r128ctx->dirty_context     = R128_CTX_ALL_DIRTY;
+
+    r128ctx->RenderIndex       = R128_FALLBACK_BIT;
+    r128ctx->PointsFunc        = NULL;
+    r128ctx->LineFunc          = NULL;
+    r128ctx->TriangleFunc      = NULL;
+    r128ctx->QuadFunc          = NULL;
+
+    r128ctx->IndirectTriangles = 0;
+    r128ctx->Fallback          = 0;
+
+    if (r128ctx->glCtx->Visual->DBflag) {
+	r128ctx->drawX         = r128ctx->r128Screen->backX;
+	r128ctx->drawY         = r128ctx->r128Screen->backY;
+	r128ctx->readX         = r128ctx->r128Screen->backX;
+	r128ctx->readY         = r128ctx->r128Screen->backY;
+    } else {
+	r128ctx->drawX         = r128ctx->r128Screen->fbX;
+	r128ctx->drawY         = r128ctx->r128Screen->fbY;
+	r128ctx->readX         = r128ctx->r128Screen->fbX;
+	r128ctx->readY         = r128ctx->r128Screen->fbY;
+    }
+
+    r128ctx->ClearColor = 0x00000000;
+    r128ctx->ClearDepth = depthClear;
+
+    r128ctx->regs.scale_3d_cntl =
+	R128_SCALE_DITHER_TABLE         |
+	R128_TEX_CACHE_SIZE_FULL        |
+	R128_DITHER_INIT_RESET          |
+	R128_SCALE_3D_TEXMAP_SHADE      |
+	R128_SCALE_PIX_REPLICATE        |
+     /* R128_TEX_CACHE_SPLIT            | */
+	R128_ALPHA_COMB_ADD_CLAMP       |
+	R128_FOG_TABLE                  |
+	R128_ALPHA_BLEND_SRC_ONE        |
+	R128_ALPHA_BLEND_DST_ZERO       |
+	R128_ALPHA_TEST_ALWAYS          |
+	R128_COMPOSITE_SHADOW_CMP_EQUAL |
+	R128_TEX_MAP_ALPHA_IN_TEXTURE   |
+	R128_TEX_CACHE_LINE_SIZE_8QW;
+
+    r128ctx->regs.dst_pitch_offset_c = pitch << R128_PITCH_SHIFT;
+
+    r128ctx->regs.dp_gui_master_cntl =
+	R128_GMC_DST_PITCH_OFFSET_CNTL |
+	R128_GMC_DST_CLIPPING          |
+	R128_GMC_BRUSH_SOLID_COLOR     |
+	dst_bpp                        |
+	R128_GMC_SRC_DATATYPE_COLOR    |
+	R128_GMC_BYTE_MSB_TO_LSB       |
+	R128_GMC_CONVERSION_TEMP_6500  |
+	R128_ROP3_S                    |
+	R128_DP_SRC_SOURCE_MEMORY      |
+	R128_GMC_3D_FCN_EN             |
+	R128_GMC_CLR_CMP_CNTL_DIS      |
+	R128_AUX_CLIP_DIS              |
+	R128_GMC_WR_MSK_DIS;
+
+    r128ctx->regs.sc_top_left_c     = 0x00000000;
+    r128ctx->regs.sc_bottom_right_c = 0x1fff1fff;
+
+    r128ctx->regs.aux_sc_cntl       = 0x00000000;
+
+    r128ctx->regs.aux1_sc_left      = 0x00000000;
+    r128ctx->regs.aux1_sc_right     = 0x00001fff;
+    r128ctx->regs.aux1_sc_top       = 0x00000000;
+    r128ctx->regs.aux1_sc_bottom    = 0x00001fff;
+
+    r128ctx->regs.aux2_sc_left      = 0x00000000;
+    r128ctx->regs.aux2_sc_right     = 0x00001fff;
+    r128ctx->regs.aux2_sc_top       = 0x00000000;
+    r128ctx->regs.aux2_sc_bottom    = 0x00001fff;
+
+    r128ctx->regs.aux3_sc_left      = 0x00000000;
+    r128ctx->regs.aux3_sc_right     = 0x00001fff;
+    r128ctx->regs.aux3_sc_top       = 0x00000000;
+    r128ctx->regs.aux3_sc_bottom    = 0x00001fff;
+
+    r128ctx->regs.z_offset_c        = (r128ctx->r128Screen->depthX *
+				       (r128ctx->r128Screen->bpp/8) +
+				       r128ctx->r128Screen->depthY *
+				       r128ctx->r128Screen->fbStride);
+    r128ctx->regs.z_pitch_c         = pitch;
+
+    r128ctx->regs.z_sten_cntl_c =
+	depth_bpp                |
+	R128_Z_TEST_LESS         |
+	R128_STENCIL_TEST_ALWAYS |
+	R128_STENCIL_S_FAIL_KEEP |
+	R128_STENCIL_ZPASS_KEEP  |
+	R128_STENCIL_ZFAIL_KEEP;
+
+    r128ctx->regs.tex_cntl_c =
+	R128_Z_WRITE_ENABLE          |
+	R128_SHADE_ENABLE            |
+	R128_DITHER_ENABLE           |
+	R128_ALPHA_IN_TEX_COMPLETE_A |
+	R128_LIGHT_DIS               |
+	R128_ALPHA_LIGHT_DIS         |
+	R128_TEX_CACHE_FLUSH         |
+	(0x0f << R128_LOD_BIAS_SHIFT);
+
+    r128ctx->regs.misc_3d_state_cntl_reg =
+	R128_MISC_SCALE_3D_TEXMAP_SHADE |
+	R128_MISC_SCALE_PIX_REPLICATE   |
+	R128_ALPHA_COMB_ADD_CLAMP       |
+	R128_FOG_TABLE                  |
+	R128_ALPHA_BLEND_SRC_ONE        |
+	R128_ALPHA_BLEND_DST_ZERO       |
+	R128_ALPHA_TEST_ALWAYS;
+
+    r128ctx->regs.texture_clr_cmp_clr_c = 0x00000000;
+    r128ctx->regs.texture_clr_cmp_msk_c = 0xffffffff;
+
+    r128ctx->regs.prim_tex_cntl_c =
+	R128_MIN_BLEND_NEAREST |
+	R128_MAG_BLEND_NEAREST |
+	R128_MIP_MAP_DISABLE   |
+	R128_TEX_CLAMP_S_WRAP  |
+	R128_TEX_CLAMP_T_WRAP;
+
+    r128ctx->regs.prim_texture_combine_cntl_c =
+	R128_COMB_MODULATE          |
+	R128_COLOR_FACTOR_TEX       |
+	R128_INPUT_FACTOR_INT_COLOR |
+	R128_COMB_ALPHA_COPY        |
+	R128_ALPHA_FACTOR_TEX_ALPHA |
+	R128_INP_FACTOR_A_INT_ALPHA;
+
+    r128ctx->regs.tex_size_pitch_c     =
+	(0 << R128_TEX_PITCH_SHIFT)      |
+	(0 << R128_TEX_SIZE_SHIFT)       |
+	(0 << R128_TEX_HEIGHT_SHIFT)     |
+	(0 << R128_TEX_MIN_SIZE_SHIFT)   |
+	(0 << R128_SEC_TEX_PITCH_SHIFT)  |
+	(0 << R128_SEC_TEX_SIZE_SHIFT)   |
+	(0 << R128_SEC_TEX_HEIGHT_SHIFT) |
+	(0 << R128_SEC_TEX_MIN_SIZE_SHIFT);
+
+    for (i = 0; i < R128_TEX_MAXLEVELS; i++)
+	r128ctx->regs.prim_tex_offset[i]  = 0x00000000;
+
+    r128ctx->regs.sec_tex_cntl_c =
+	R128_SEC_SELECT_PRIM_ST;
+
+    r128ctx->regs.sec_tex_combine_cntl_c =
+	R128_COMB_DIS                |
+	R128_COLOR_FACTOR_TEX        |
+	R128_INPUT_FACTOR_PREV_COLOR |
+	R128_COMB_ALPHA_DIS          |
+	R128_ALPHA_FACTOR_TEX_ALPHA  |
+	R128_INP_FACTOR_A_PREV_ALPHA;
+
+    for (i = 0; i < R128_TEX_MAXLEVELS; i++)
+	r128ctx->regs.sec_tex_offset[i] = 0x00000000;
+
+    r128ctx->regs.constant_color_c            = 0x00ffffff;
+    r128ctx->regs.prim_texture_border_color_c = 0x00ffffff;
+    r128ctx->regs.sec_texture_border_color_c  = 0x00ffffff;
+    r128ctx->regs.sten_ref_mask_c             = 0xffff0000;
+    r128ctx->regs.plane_3d_mask_c             = 0xffffffff;
+
+    r128ctx->regs.setup_cntl =
+	R128_COLOR_GOURAUD         |
+	R128_PRIM_TYPE_TRI         |
+#if 1
+	/* FIXME: Let r128 multiply? */
+	R128_TEXTURE_ST_MULT_W     |
+#else
+	/* FIXME: Or, pre multiply? */
+	R128_TEXTURE_ST_DIRECT     |
+#endif
+	R128_STARTING_VERTEX_1     |
+	R128_ENDING_VERTEX_3       |
+	R128_SU_POLY_LINE_NOT_LAST |
+	R128_SUB_PIX_4BITS;
+
+    r128ctx->regs.pm4_vc_fpu_setup =
+	R128_FRONT_DIR_CCW         |
+	R128_BACKFACE_SOLID        |
+	R128_FRONTFACE_SOLID       |
+	R128_FPU_COLOR_GOURAUD     |
+	R128_FPU_SUB_PIX_4BITS     |
+	R128_FPU_MODE_3D           |
+	R128_TRAP_BITS_DISABLE     |
+	R128_XFACTOR_2             |
+	R128_YFACTOR_2             |
+	R128_FLAT_SHADE_VERTEX_OGL |
+	R128_FPU_ROUND_TRUNCATE    |
+	R128_WM_SEL_8DW;
+
+    r128ctx->FogMode                   = GL_EXP;
+    r128ctx->regs.fog_color_c          = 0x00808080;
+    r128ctx->regs.fog_3d_table_start   = 0x00000000;
+    r128ctx->regs.fog_3d_table_end     = 0xffffffff;
+    r128ctx->regs.fog_3d_table_density = 0x00000000;
+
+    r128ctx->regs.window_xy_offset     = 0x00000000;
+
+    r128ctx->regs.dp_write_mask        = 0xffffffff;
+
+    r128ctx->regs.pc_gui_ctlstat       = R128_PC_FLUSH_GUI;
+
+    r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+    r128ctx->dirty_context |= R128_CTX_ALL_DIRTY;
+}
+
+/* Upload the fog table for the current fog mode */
+static void r128UploadFogTable(r128ContextPtr r128ctx)
+{
+    int i;
+
+    R128CCE0(R128_CCE_PACKET0, R128_FOG_TABLE_INDEX, 0);
+    R128CCE(0x00000000);
+
+    R128CCE0(R128_CCE_PACKET0_ONE_REG_WR, R128_FOG_TABLE_DATA, 255);
+
+    switch (r128ctx->FogMode) {
+    case GL_LINEAR:
+	for (i = 0; i < 256; i++) {
+	    R128CCE(255 - i);
+	}
+	break;
+    case GL_EXP:
+	for (i = 0; i < 256; i++) {
+	    R128CCE(255 - FLOAT_TO_UBYTE(exp(i - 255)));
+	}
+	break;
+    case GL_EXP2:
+	for (i = 0; i < 256; i++) {
+	    R128CCE(255 - FLOAT_TO_UBYTE(exp((i - 255) * (i - 255))));
+	}
+	break;
+    }
+}
+
+/* Load the current context's state into the hardware */
+/* NOTE: This function is only called while holding the hardware lock */
+static void r128LoadContext(r128ContextPtr r128ctx)
+{
+    int i;
+    int tex_size_pitch_done = GL_FALSE;
+
+    if (R128_DEBUG_FLAGS & DEBUG_VERBOSE_API) {
+	fprintf(stderr, "r128LoadContext(%p)\n", r128ctx->glCtx);
+    }
+
+#if 0
+    r128ctx->dirty_context = R128_CTX_ALL_DIRTY;
+#endif
+
+#if 1
+    /* FIXME: Why do these need to be updated even when they don't change? */
+    r128ctx->dirty_context |= (R128_CTX_MISC        |
+			       R128_CTX_ENGINESTATE |
+			       R128_CTX_ALPHASTATE);
+#endif
+
+#if 1
+    /* FIXME: Is this _really_ needed? */
+    if (r128ctx->dirty_context)
+	if (!R128CCE_USE_RING_BUFFER(r128ctx->r128Screen->CCEMode))
+	    R128CCE_WAIT_FOR_IDLE(r128ctx);
+#endif
+
+    if (r128ctx->dirty_context & R128_CTX_MISC) {
+	R128CCE1(R128_CCE_PACKET1, R128_SCALE_3D_CNTL, R128_DP_WRITE_MASK);
+	R128CCE(r128ctx->regs.scale_3d_cntl);
+	R128CCE(r128ctx->regs.dp_write_mask);
+
+	R128CCE0(R128_CCE_PACKET0, R128_DST_PITCH_OFFSET_C, 1);
+	R128CCE(r128ctx->regs.dst_pitch_offset_c);
+	R128CCE(r128ctx->regs.dp_gui_master_cntl);
+
+	R128CCE0(R128_CCE_PACKET0, R128_TEXTURE_CLR_CMP_CLR_C, 1);
+	R128CCE(r128ctx->regs.texture_clr_cmp_clr_c);
+	R128CCE(r128ctx->regs.texture_clr_cmp_msk_c);
+
+	R128CCE0(R128_CCE_PACKET0, R128_STEN_REF_MASK_C, 1);
+	R128CCE(r128ctx->regs.sten_ref_mask_c);
+	R128CCE(r128ctx->regs.plane_3d_mask_c);
+    }
+
+    if (r128ctx->dirty_context & R128_CTX_ENGINESTATE) {
+	R128CCE0(R128_CCE_PACKET0, R128_TEX_CNTL_C, 0);
+	R128CCE(r128ctx->regs.tex_cntl_c);
+    }
+
+    if (r128ctx->dirty_context & R128_CTX_TEX0STATE) {
+	R128CCE0(R128_CCE_PACKET0, R128_PRIM_TEX_CNTL_C, 2+R128_TEX_MAXLEVELS);
+	R128CCE(r128ctx->regs.prim_tex_cntl_c);
+	R128CCE(r128ctx->regs.prim_texture_combine_cntl_c);
+	R128CCE(r128ctx->regs.tex_size_pitch_c);
+	for (i = 0; i < R128_TEX_MAXLEVELS; i++)
+	    R128CCE(r128ctx->regs.prim_tex_offset[i]);
+
+	R128CCE0(R128_CCE_PACKET0, R128_PRIM_TEXTURE_BORDER_COLOR_C, 0);
+	R128CCE(r128ctx->regs.prim_texture_border_color_c);
+
+	tex_size_pitch_done = GL_TRUE;
+    }
+
+    if (r128ctx->dirty_context & R128_CTX_TEX1STATE) {
+	if (!tex_size_pitch_done) {
+	    R128CCE0(R128_CCE_PACKET0, R128_TEX_SIZE_PITCH_C, 0);
+	    R128CCE(r128ctx->regs.tex_size_pitch_c);
+	}
+
+	R128CCE0(R128_CCE_PACKET0, R128_SEC_TEX_CNTL_C, 1+R128_TEX_MAXLEVELS);
+	R128CCE(r128ctx->regs.sec_tex_cntl_c);
+	R128CCE(r128ctx->regs.sec_tex_combine_cntl_c);
+	for (i = 0; i < R128_TEX_MAXLEVELS; i++)
+	    R128CCE(r128ctx->regs.sec_tex_offset[i]);
+
+	R128CCE0(R128_CCE_PACKET0, R128_SEC_TEXTURE_BORDER_COLOR_C, 0);
+	R128CCE(r128ctx->regs.sec_texture_border_color_c);
+    }
+
+    if (r128ctx->dirty_context & R128_CTX_TEXENVSTATE) {
+	R128CCE0(R128_CCE_PACKET0, R128_CONSTANT_COLOR_C, 0);
+	R128CCE(r128ctx->regs.constant_color_c);
+    }
+
+    if (r128ctx->dirty_context & R128_CTX_FOGSTATE) {
+	R128CCE0(R128_CCE_PACKET0, R128_FOG_3D_TABLE_START, 1);
+	R128CCE(r128ctx->regs.fog_3d_table_start);
+	R128CCE(r128ctx->regs.fog_3d_table_end);
+
+	R128CCE1(R128_CCE_PACKET1,
+		 R128_FOG_COLOR_C, R128_FOG_3D_TABLE_DENSITY);
+	R128CCE(r128ctx->regs.fog_color_c);
+	R128CCE(r128ctx->regs.fog_3d_table_density);
+    }
+
+    if (r128ctx->dirty_context & R128_CTX_FOGTABLE) {
+	r128UploadFogTable(r128ctx);
+    }
+
+    if (r128ctx->dirty_context & R128_CTX_ZSTENSTATE) {
+	R128CCE0(R128_CCE_PACKET0, R128_Z_STEN_CNTL_C, 0);
+	R128CCE(r128ctx->regs.z_sten_cntl_c);
+    }
+
+    if (r128ctx->dirty_context & R128_CTX_SCISSORS) {
+	R128CCE0(R128_CCE_PACKET0, R128_SC_TOP_LEFT_C, 1);
+	R128CCE(r128ctx->regs.sc_top_left_c);
+	R128CCE(r128ctx->regs.sc_bottom_right_c);
+    }
+
+    if (r128ctx->dirty_context & (R128_CTX_ALPHASTATE |
+				  R128_CTX_FOGSTATE)) {
+	R128CCE0(R128_CCE_PACKET0, R128_MISC_3D_STATE_CNTL_REG, 0);
+	R128CCE(r128ctx->regs.misc_3d_state_cntl_reg);
+    }
+
+    if (r128ctx->dirty_context & R128_CTX_SETUPSTATE) {
+	R128CCE1(R128_CCE_PACKET1, R128_SETUP_CNTL, R128_PM4_VC_FPU_SETUP);
+	R128CCE(r128ctx->regs.setup_cntl);
+	R128CCE(r128ctx->regs.pm4_vc_fpu_setup);
+    }
+
+    if (r128ctx->dirty_context & R128_CTX_WIN_Z_POS) {
+	R128CCE0(R128_CCE_PACKET0, R128_WINDOW_XY_OFFSET, 0);
+	R128CCE(r128ctx->regs.window_xy_offset);
+
+	R128CCE0(R128_CCE_PACKET0, R128_Z_OFFSET_C, 1);
+	R128CCE(r128ctx->regs.z_offset_c);
+	R128CCE(r128ctx->regs.z_pitch_c);
+    }
+
+#if 0
+    if (r128ctx->dirty_context & R128_CTX_FLUSH_PIX_CACHE) {
+	R128CCE0(R128_CCE_PACKET0, R128_PC_GUI_CTLSTAT, 0);
+	R128CCE(r128ctx->regs.pc_gui_ctlstat);
+    }
+#endif
+
+    R128CCE_SUBMIT_PACKETS();
+
+    /* Turn off the texture cache flushing */
+    r128ctx->regs.tex_cntl_c     &= ~R128_TEX_CACHE_FLUSH;
+
+    /* Turn off the pixel cache flushing */
+    r128ctx->regs.pc_gui_ctlstat &= ~R128_PC_FLUSH_ALL;
+
+    r128ctx->dirty_context = R128_CTX_CLEAN;
+}
+
+/* Set the hardware clip rects for drawing to the current color buffer */
+/* NOTE: This function is only called while holding the hardware lock */
+void r128SetClipRects(r128ContextPtr r128ctx,
+		      XF86DRIClipRectPtr pc, int nc)
+{
+    if (!pc) return;
+
+    /* Clear any previous auxiliary scissors */
+    r128ctx->regs.aux_sc_cntl = 0x00000000;
+
+    switch (nc) {
+    case 3:
+	R128CCE0(R128_CCE_PACKET0, R128_AUX3_SC_LEFT, 3);
+	R128CCE(pc[2].x1   + r128ctx->drawX);
+	R128CCE(pc[2].x2-1 + r128ctx->drawX);
+	R128CCE(pc[2].y1   + r128ctx->drawY);
+	R128CCE(pc[2].y2-1 + r128ctx->drawY);
+
+	r128ctx->regs.aux_sc_cntl |= R128_AUX3_SC_EN | R128_AUX3_SC_MODE_OR;
+
+    case 2:
+	R128CCE0(R128_CCE_PACKET0, R128_AUX2_SC_LEFT, 3);
+	R128CCE(pc[1].x1   + r128ctx->drawX);
+	R128CCE(pc[1].x2-1 + r128ctx->drawX);
+	R128CCE(pc[1].y1   + r128ctx->drawY);
+	R128CCE(pc[1].y2-1 + r128ctx->drawY);
+
+	r128ctx->regs.aux_sc_cntl |= R128_AUX2_SC_EN | R128_AUX2_SC_MODE_OR;
+
+    case 1:
+	R128CCE0(R128_CCE_PACKET0, R128_AUX1_SC_LEFT, 3);
+	R128CCE(pc[0].x1   + r128ctx->drawX);
+	R128CCE(pc[0].x2-1 + r128ctx->drawX);
+	R128CCE(pc[0].y1   + r128ctx->drawY);
+	R128CCE(pc[0].y2-1 + r128ctx->drawY);
+
+	r128ctx->regs.aux_sc_cntl |= R128_AUX1_SC_EN | R128_AUX1_SC_MODE_OR;
+	break;
+
+    default:
+	return;
+    }
+
+    R128CCE0(R128_CCE_PACKET0, R128_AUX_SC_CNTL, 0);
+    R128CCE(r128ctx->regs.aux_sc_cntl);
+
+    R128CCE_SUBMIT_PACKETS();
+}
+
+/* Update the driver's notion of the window position */
+/* NOTE: This function is only called while holding the hardware lock */
+static void r128UpdateWindowPosition(r128ContextPtr r128ctx)
+{
+    int x = r128ctx->driDrawable->x + r128ctx->drawX;
+    int y = r128ctx->driDrawable->y + r128ctx->drawY;
+
+#if 0
+    /* FIXME: Is this _really_ needed? */
+    R128CCE_FLUSH_VB(r128ctx);
+#endif
+    r128ctx->regs.window_xy_offset = ((y << R128_WINDOW_Y_SHIFT) |
+				      (x << R128_WINDOW_X_SHIFT));
+
+    /* Recalculate the Z buffer offset since we might be drawing to the
+       back buffer and window_xy_offset affects both color buffer and
+       depth drawing */
+    r128ctx->regs.z_offset_c        = ((r128ctx->r128Screen->depthX -
+					r128ctx->drawX) *
+				       (r128ctx->r128Screen->bpp/8) +
+				       (r128ctx->r128Screen->depthY -
+					r128ctx->drawY) *
+				       r128ctx->r128Screen->fbStride);
+
+    r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+    r128ctx->dirty_context |= R128_CTX_WIN_Z_POS;
+}
+
+/* Update the hardware state */
+/* NOTE: This function is only called while holding the hardware lock */
+static void r128UpdateHWStateLocked(r128ContextPtr r128ctx)
+{
+    if (r128ctx->dirty & R128_REQUIRE_QUIESCENCE)
+	R128CCE_WAIT_FOR_IDLE(r128ctx);
+
+    /* Update any state that might have changed recently */
+
+    /* Update the clip rects */
+    if (r128ctx->dirty & R128_UPDATE_WINPOS)
+	r128UpdateWindowPosition(r128ctx);
+
+    /* Update texture state and then upload the images */
+    /* Note: Texture images can only be updated after the state has been set */
+    if (r128ctx->dirty & R128_UPDATE_TEXSTATE)
+	r128UpdateTextureState(r128ctx);
+    if (r128ctx->dirty & R128_UPDATE_TEX0IMAGES)
+        r128UploadTexImages(r128ctx, r128ctx->CurrentTexObj[0]);
+    if (r128ctx->dirty & R128_UPDATE_TEX1IMAGES)
+        r128UploadTexImages(r128ctx, r128ctx->CurrentTexObj[1]);
+
+    /* Load the state into the hardware */
+    /* Note: This must be done after all other state has been set */
+    if (r128ctx->dirty & R128_UPDATE_CONTEXT)
+	r128LoadContext(r128ctx);
+
+    r128ctx->dirty = R128_CLEAN;
+}
+
+/* Update the hardware state */
+void r128UpdateHWState(r128ContextPtr r128ctx)
+{
+    LOCK_HARDWARE(r128ctx);
+    r128UpdateHWStateLocked(r128ctx);
+    UNLOCK_HARDWARE(r128ctx);
+}
+
+/* Update the driver's state */
+/* NOTE: This function is only called while holding the hardware lock */
+void r128UpdateState(r128ContextPtr r128ctx, int winMoved)
+{
+    R128SAREAPrivPtr sarea = r128ctx->r128Screen->SAREA;
+    int              i;
+
+    if (sarea->ctxOwner != r128ctx->driContext->hHWContext) {
+	sarea->ctxOwner = r128ctx->driContext->hHWContext;
+	r128ctx->dirty_context = R128_CTX_ALL_DIRTY;
+	r128LoadContext(r128ctx);
+    }
+
+    for (i = 0; i < r128ctx->r128Screen->NRTexHeaps; i++)
+	r128AgeTextures(r128ctx, i);
+
+    if (winMoved) r128ctx->dirty |= R128_UPDATE_WINPOS;
+}
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_state.h b/xc/lib/GL/mesa/src/drv/r128/r128_state.h
new file mode 100644
index 000000000..f847452f0
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_state.h
@@ -0,0 +1,50 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_STATE_H_
+#define _R128_STATE_H_
+
+#ifdef GLX_DIRECT_RENDERING
+
+extern void   r128DDInitState(r128ContextPtr r128ctx);
+extern void   r128DDInitStateFuncs(GLcontext *ctx);
+
+extern void   r128UpdateState(r128ContextPtr r128ctx, int winMoved);
+extern void   r128UpdateHWState(r128ContextPtr r128ctx);
+
+extern void   r128SetClipRects(r128ContextPtr r128ctx,
+			       XF86DRIClipRectPtr pc, int nc);
+
+#endif
+#endif /* _R128_STATE_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_swap.c b/xc/lib/GL/mesa/src/drv/r128/r128_swap.c
new file mode 100644
index 000000000..6b347023d
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_swap.c
@@ -0,0 +1,125 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#include "r128_init.h"
+#include "r128_mesa.h"
+#include "r128_xmesa.h"
+#include "r128_context.h"
+#include "r128_lock.h"
+#include "r128_reg.h"
+#include "r128_cce.h"
+#include "r128_state.h"
+#include "r128_vb.h"
+#include "r128_swap.h"
+
+/* Copy the back color buffer to the front color buffer */
+void r128SwapBuffers(r128ContextPtr r128ctx)
+{
+    unsigned char *R128MMIO        = r128ctx->r128Screen->mmio;
+    __DRIdrawablePrivate *dPriv    = r128ctx->driDrawable;
+    int                   nc;
+    XF86DRIClipRectPtr    c;
+    int                   dst_bpp;
+    CARD32                swapAge;
+
+    if (r128ctx->SWonly) {
+	/* FIXME: Provide software fallback for this case?? */
+    }
+
+    switch (r128ctx->r128Screen->bpp) {
+    case 8:
+	dst_bpp = R128_GMC_DST_8BPP_CI;
+	break;
+    case 16:
+	if (r128ctx->r128Screen->depth == 15) dst_bpp = R128_GMC_DST_15BPP;
+	else                                  dst_bpp = R128_GMC_DST_16BPP;
+	break;
+    case 24:
+	dst_bpp = R128_GMC_DST_24BPP;
+	break;
+    case 32:
+    default:
+	dst_bpp = R128_GMC_DST_32BPP;
+	break;
+    }
+
+    LOCK_HARDWARE(r128ctx);
+
+    /* Flush any outstanding vertex buffers */
+    R128CCE_FLUSH_VB(r128ctx);
+
+    /* Throttle the frame rate -- only allow one pending swap buffers
+       request at a time */
+    while (r128ctx->lastSwapAge > (swapAge = INREG(R128_SWAP_AGE_REG)));
+
+    /* Init the clip rects here in case they changed during the
+       LOCK_HARDWARE macro */
+    c  = dPriv->pClipRects;
+    nc = dPriv->numClipRects;
+
+    /* Cycle through the clip rects */
+    while (nc--) {
+	int fx = c[nc].x1;
+	int fy = c[nc].y1;
+	int fw = c[nc].x2 - fx;
+	int fh = c[nc].y2 - fy;
+	int bx = fx + r128ctx->r128Screen->backX;
+	int by = fy + r128ctx->r128Screen->backY;
+
+	fx += r128ctx->r128Screen->fbX;
+	fy += r128ctx->r128Screen->fbY;
+
+	R128CCE3(R128_CCE_PACKET3_CNTL_BITBLT_MULTI, 3);
+	R128CCE(R128_GMC_BRUSH_NONE
+		| R128_GMC_SRC_DATATYPE_COLOR
+		| R128_DP_SRC_SOURCE_MEMORY
+		| dst_bpp
+		| R128_ROP3_S);
+	R128CCE((bx << 16) | by);
+	R128CCE((fx << 16) | fy);
+	R128CCE((fw << 16) | fh);
+    }
+
+    ++swapAge;
+    R128CCE0(R128_CCE_PACKET0, R128_SWAP_AGE_REG, 0);
+    R128CCE(swapAge);
+    r128ctx->lastSwapAge = swapAge;
+
+    r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+    r128ctx->dirty_context |= R128_CTX_ALL_DIRTY;
+
+    R128CCE_SUBMIT_PACKETS();
+
+    UNLOCK_HARDWARE(r128ctx);
+}
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_swap.h b/xc/lib/GL/mesa/src/drv/r128/r128_swap.h
new file mode 100644
index 000000000..32851041c
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_swap.h
@@ -0,0 +1,43 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_SWAP_H_
+#define _R128_SWAP_H_
+
+#ifdef GLX_DIRECT_RENDERING
+
+extern void r128SwapBuffers(r128ContextPtr r128ctx);
+
+#endif
+#endif /* _R128_SWAP_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_tex.c b/xc/lib/GL/mesa/src/drv/r128/r128_tex.c
new file mode 100644
index 000000000..d75345404
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_tex.c
@@ -0,0 +1,1851 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *   Gareth Hughes <gareth@precisioninsight.com>
+ *
+ */
+
+#include "r128_init.h"
+#include "r128_mesa.h"
+#include "r128_xmesa.h"
+#include "r128_context.h"
+#include "r128_lock.h"
+#include "r128_state.h"
+#include "r128_reg.h"
+#include "r128_cce.h"
+#include "r128_vb.h"
+#include "r128_tex.h"
+
+#include "mmath.h"
+#include "simple_list.h"
+
+static void r128SetTexWrap(r128TexObjPtr t, GLenum srwap, GLenum twrap);
+static void r128SetTexFilter(r128TexObjPtr t, GLenum minf, GLenum magf);
+static void r128SetTexBorderColor(r128TexObjPtr t, GLubyte c[4]);
+
+/* Allocate and initialize hardware state associated with texture `t' */
+/* NOTE: This function is only called while holding the hardware lock */
+static r128TexObjPtr r128CreateTexObj(r128ContextPtr r128ctx,
+				      struct gl_texture_object *tObj)
+{
+    r128TexObjPtr            t;
+    struct gl_texture_image *image;
+    int                      log2Pitch, log2Height, log2Size, log2MinSize;
+    int                      totalSize;
+    int                      i;
+
+    image = tObj->Image[0];
+    if (!image) return NULL; /* ERROR!!! */
+
+    t = (r128TexObjPtr)calloc(1,sizeof(*t));
+    if (!t) return NULL;     /* ERROR!!! */
+
+    if (R128_DEBUG_FLAGS & DEBUG_VERBOSE_API)
+	fprintf(stderr, "r128CreateTexObj(%p)\n", tObj);
+
+    switch (image->Format) {
+    case GL_RGBA:
+    case GL_ALPHA:
+    case GL_LUMINANCE_ALPHA:
+    case GL_INTENSITY:
+	if (r128ctx->r128Screen->bpp == 32) {
+	    t->texelBytes    = 4;
+	    t->textureFormat = R128_DATATYPE_ARGB8888;
+	} else {
+	    t->texelBytes    = 2;
+	    t->textureFormat = R128_DATATYPE_ARGB4444;
+	}
+	break;
+
+    case GL_RGB:
+	if (r128ctx->r128Screen->bpp == 32) {
+	    t->texelBytes    = 4;
+	    t->textureFormat = R128_DATATYPE_ARGB8888;
+	} else {
+	    t->texelBytes    = 2;
+	    t->textureFormat = R128_DATATYPE_RGB565;
+	}
+	break;
+
+    case GL_LUMINANCE:
+	if (r128ctx->r128Screen->bpp == 32) {
+	    t->texelBytes    = 4;
+	    t->textureFormat = R128_DATATYPE_ARGB8888;
+	} else {
+	    t->texelBytes    = 2;
+					/* Use this to get true greys */
+	    t->textureFormat = R128_DATATYPE_ARGB1555;
+	}
+	break;
+
+    case GL_COLOR_INDEX:
+	t->texelBytes    = 1;
+	t->textureFormat = R128_DATATYPE_CI8;
+	break;
+
+    default:
+	/* ERROR!!! */
+	fprintf(stderr, "r128CreateTexObj: bad image->Format\n");
+	free(t);
+	return NULL;
+    }
+
+    /* Calculate dimensions in log domain */
+    for (i = 1, log2Height = 0; i < image->Height; i *= 2) log2Height++;
+    for (i = 1, log2Pitch = 0;  i < image->Width;  i *= 2) log2Pitch++;
+    if (image->Width > image->Height) log2Size = log2Pitch;
+    else                              log2Size = log2Height;
+
+    t->dirty_images = 0;
+
+    /* Calculate mipmap offsets and dimensions */
+    totalSize = 0;
+    for (i = 0; i <= log2Size && tObj->Image[i]; i++) {
+	t->image[i].offset  = totalSize;
+	t->image[i].width   = tObj->Image[i]->Width;
+	t->image[i].height  = tObj->Image[i]->Height;
+	t->dirty_images    |= 1 << i;
+	totalSize          += (tObj->Image[i]->Height *
+			       tObj->Image[i]->Width *
+			       t->texelBytes);
+
+	/* Offsets must be 32-byte aligned for host data blits */
+	totalSize           = (totalSize + 31) & ~31;
+    }
+    log2MinSize = log2Size - i + 1;
+
+    t->totalSize    = totalSize;
+    t->internFormat = image->IntFormat;
+
+    t->bound        = 0;
+    t->heap         = 0; /* This is set in r128UploadTexImages */
+    t->tObj         = tObj;
+
+    t->memBlock     = NULL;
+    t->bufAddr      = NULL;
+
+    t->regs.tex_cntl         = t->textureFormat;
+    t->regs.size_pitch       = ((log2Pitch   << R128_TEX_PITCH_SHIFT)  |
+				(log2Size    << R128_TEX_SIZE_SHIFT)   |
+				(log2Height  << R128_TEX_HEIGHT_SHIFT) |
+				(log2MinSize << R128_TEX_MIN_SIZE_SHIFT));
+    t->regs.border_color     = 0x00000000;
+
+    if (log2MinSize == log2Size ||
+	log2MinSize != 0)
+	t->regs.tex_cntl |= R128_MIP_MAP_DISABLE;
+
+    r128SetTexWrap(t, tObj->WrapS, tObj->WrapT);
+    r128SetTexFilter(t, tObj->MinFilter, tObj->MagFilter);
+    r128SetTexBorderColor(t, tObj->BorderColor);
+
+    tObj->DriverData = t;
+
+    make_empty_list(t);
+
+    return t;
+}
+
+/* Destroy hardware state associated with texture `t' */
+/* NOTE: This function can be called while holding the hardware lock and
+   while not holding the lock*/
+void r128DestroyTexObj(r128ContextPtr r128ctx, r128TexObjPtr t)
+{
+    if (!t) return;
+
+    if (t->memBlock) {
+	mmFreeMem(t->memBlock);
+	t->memBlock = NULL;
+    }
+
+    if (t->tObj)  t->tObj->DriverData                = NULL;
+    if (t->bound) r128ctx->CurrentTexObj[t->bound-1] = NULL;
+
+    remove_from_list(t);
+    free(t);
+}
+
+/* Keep track of swapped out texture objects */
+/* NOTE: This function is only called while holding the hardware lock */
+static void r128SwapOutTexObj(r128ContextPtr r128ctx, r128TexObjPtr t)
+{
+    if (t->memBlock) {
+	mmFreeMem(t->memBlock);
+	t->memBlock = NULL;
+    }
+
+    t->dirty_images = ~0;
+    move_to_tail(&r128ctx->SwappedOut, t);
+}
+
+/* Print out debugging information about texture LRU */
+void r128PrintLocalLRU(r128ContextPtr r128ctx, int heap)
+{
+    r128TexObjPtr t;
+    int           sz = 1 << (r128ctx->r128Screen->log2TexGran[heap]);
+
+    foreach(t, &r128ctx->TexObjList[heap]) {
+	if (!t->tObj) {
+	    fprintf(stderr, "Placeholder %d at 0x%x sz 0x%x\n",
+		    t->memBlock->ofs / sz,
+		    t->memBlock->ofs,
+		    t->memBlock->size);
+	} else {
+	    fprintf(stderr, "Texture (bound %d) at 0x%x sz 0x%x\n",
+		    t->bound,
+		    t->memBlock->ofs,
+		    t->memBlock->size);
+	}
+    }
+}
+
+void r128PrintGlobalLRU(r128ContextPtr r128ctx, int heap)
+{
+    R128TexRegion *list = r128ctx->r128Screen->SAREA->texList[heap];
+    int            i, j;
+
+    for (i = 0, j = R128_NR_TEX_REGIONS ; i < R128_NR_TEX_REGIONS ; i++) {
+	fprintf(stderr, "list[%d] age %d next %d prev %d\n",
+		j, list[j].age, list[j].next, list[j].prev);
+	j = list[j].next;
+	if (j == R128_NR_TEX_REGIONS) break;
+    }
+
+    if (j != R128_NR_TEX_REGIONS) {
+	fprintf(stderr, "Loop detected in global LRU\n");
+    }
+}
+
+/* Reset the global texture LRU */
+/* NOTE: This function is only called while holding the hardware lock */
+static void r128ResetGlobalLRU(r128ContextPtr r128ctx, int heap)
+{
+    R128TexRegion *list   = r128ctx->r128Screen->SAREA->texList[heap];
+    int            log2sz = 1 << r128ctx->r128Screen->log2TexGran[heap];
+    int            i;
+
+    /*
+     * (Re)initialize the global circular LRU list.  The last element in
+     * the array (R128_NR_TEX_REGIONS) is the sentinal.  Keeping it at
+     * the end of the array allows it to be addressed rationally when
+     * looking up objects at a particular location in texture memory.
+     */
+
+    for (i = 0; (i+1) * log2sz <= r128ctx->r128Screen->texSize[heap]; i++) {
+	list[i].prev = i-1;
+	list[i].next = i+1;
+	list[i].age = 0;
+    }
+
+    i--;
+    list[0].prev = R128_NR_TEX_REGIONS;
+    list[i].prev = i-1;
+    list[i].next = R128_NR_TEX_REGIONS;
+    list[R128_NR_TEX_REGIONS].prev = i;
+    list[R128_NR_TEX_REGIONS].next = 0;
+    r128ctx->r128Screen->SAREA->texAge[heap] = 0;
+}
+
+/* Update the local and glock texture LRUs */
+/* NOTE: This function is only called while holding the hardware lock */
+static void r128UpdateTexLRU(r128ContextPtr r128ctx, r128TexObjPtr t)
+{
+    int            heap   = t->heap;
+    R128TexRegion *list = r128ctx->r128Screen->SAREA->texList[heap];
+    int            log2sz = r128ctx->r128Screen->log2TexGran[heap];
+
+    int            start  = t->memBlock->ofs >> log2sz;
+    int            end    = (t->memBlock->ofs + t->memBlock->size-1) >> log2sz;
+    int            i;
+
+    r128ctx->lastTexAge[heap] = ++r128ctx->r128Screen->SAREA->texAge[heap];
+
+    /* Update our local LRU */
+    move_to_head(&r128ctx->TexObjList[heap], t);
+
+    /* Update the global LRU */
+    for (i = start ; i <= end ; i++) {
+	list[i].in_use = 1;
+	list[i].age = r128ctx->lastTexAge[heap];
+
+	/* remove_from_list(i) */
+	list[(CARD32)list[i].next].prev = list[i].prev;
+	list[(CARD32)list[i].prev].next = list[i].next;
+
+	/* insert_at_head(list, i) */
+	list[i].prev = R128_NR_TEX_REGIONS;
+	list[i].next = list[R128_NR_TEX_REGIONS].next;
+	list[(CARD32)list[R128_NR_TEX_REGIONS].next].prev = i;
+	list[R128_NR_TEX_REGIONS].next = i;
+    }
+}
+
+/* Update our notion of what textures have been changed since we last
+   held the lock.  This pertains to both our local textures and the
+   textures belonging to other clients.  Keep track of other client's
+   textures by pushing a placeholder texture onto the LRU list -- these
+   are denoted by (tObj == NULL). */
+/* NOTE: This function is only called while holding the hardware lock */
+static void r128TexturesGone(r128ContextPtr r128ctx, int heap,
+			     int offset, int size, int in_use)
+{
+    r128TexObjPtr t, tmp;
+   
+    foreach_s (t, tmp, &r128ctx->TexObjList[heap]) {
+	if (t->memBlock->ofs >= offset + size ||
+	    t->memBlock->ofs + t->memBlock->size <= offset)
+	    continue;
+
+	/* It overlaps - kick it out.  Need to hold onto the currently
+	   bound objects, however. */
+	if (t->bound) r128SwapOutTexObj(r128ctx, t);
+	else          r128DestroyTexObj(r128ctx, t);
+    }
+
+    if (in_use) {
+	t = (r128TexObjPtr) calloc(1,sizeof(*t));
+	if (!t) return;
+
+	t->memBlock = mmAllocMem(r128ctx->texHeap[heap], size, 0, offset);
+	insert_at_head(&r128ctx->TexObjList[heap], t);
+    }
+}
+
+/* Update our client's shared texture state.  If another client has
+   modified a region in which we have textures, then we need to figure
+   out which of our textures has been removed, and update our global
+   LRU. */
+void r128AgeTextures(r128ContextPtr r128ctx, int heap)
+{
+    R128SAREAPrivPtr sarea = r128ctx->r128Screen->SAREA;
+
+    if (sarea->texAge[heap] != r128ctx->lastTexAge[heap]) {
+	int log2sz = 1 << r128ctx->r128Screen->log2TexGran[heap];
+	int nr     = 0;
+	int idx;
+
+	for (idx = sarea->texList[heap][R128_NR_TEX_REGIONS].prev;
+	     idx != R128_NR_TEX_REGIONS && nr < R128_NR_TEX_REGIONS;
+	     idx = sarea->texList[heap][idx].prev, nr++) {
+
+	    /* If switching texturing schemes, then the SAREA might not
+               have been properly cleared, so we need to reset the
+               global texture LRU. */
+	    if (idx * log2sz > r128ctx->r128Screen->texSize[heap]) {
+		nr = R128_NR_TEX_REGIONS;
+		break;
+	    }
+
+	    if (sarea->texList[heap][idx].age > r128ctx->lastTexAge[heap])
+		r128TexturesGone(r128ctx, heap, idx * log2sz, log2sz,
+				 sarea->texList[heap][idx].in_use);
+	}
+
+	if (nr == R128_NR_TEX_REGIONS) {
+	    r128TexturesGone(r128ctx, heap,
+			     0, r128ctx->r128Screen->texSize[heap], 0);
+	    r128ResetGlobalLRU(r128ctx, heap);
+	}
+
+	r128ctx->dirty |= R128_UPDATE_TEX0IMAGES;
+	r128ctx->dirty |= R128_UPDATE_TEX1IMAGES;
+	r128ctx->lastTexAge[heap] = sarea->texAge[heap];
+    }
+}
+
+/* Convert a block of Mesa-formatted texture to an 8bpp hardware format */
+static void r128ConvertTexture8bpp(r128ContextPtr r128ctx,
+				   struct gl_texture_image *image,
+				   int x, int y, int width, int height,
+				   int pitch)
+{
+    CARD8  *src;
+    CARD32  pix;
+    int     i, j;
+
+    switch (image->Format) {
+    case GL_RGB:
+	for (i = 0; i < height; i++) {
+	    src = (CARD8 *)image->Data + ((y + i) * pitch + x) * 3;
+	    for (j = width >> 2; j; j--) {
+		pix = ((R128PACKCOLOR332( src[0], src[1],  src[2])       ) |
+		       (R128PACKCOLOR332( src[3], src[4],  src[5])  <<  8) |
+		       (R128PACKCOLOR332( src[6], src[7],  src[8])  << 16) |
+		       (R128PACKCOLOR332( src[9], src[10], src[11]) << 24));
+		R128CCE(pix);
+		src += 12;
+	    }
+	}
+	break;
+
+    case GL_ALPHA:
+    case GL_LUMINANCE:
+    case GL_INTENSITY:
+    case GL_COLOR_INDEX:
+	for (i = 0; i < height; i++) {
+	    src = (CARD8 *)image->Data + ((y + i) * pitch + x);
+	    for (j = width >> 2; j; j--) {
+		pix = src[0] | (src[1] << 8) | (src[2] << 16) | (src[3] << 24);
+		R128CCE(pix);
+		src += 4;
+	    }
+	}
+	break;
+
+    default:
+	fprintf(stderr, "r128ConvertTexture8bpp: unsupported format 0x%x\n",
+		image->Format);
+    }
+}
+
+/* Convert a block of Mesa-formatted texture to a 16bpp hardware format */
+static void r128ConvertTexture16bpp(r128ContextPtr r128ctx,
+				    struct gl_texture_image *image,
+				    int x, int y, int width, int height,
+				    int pitch)
+{
+    CARD8  *src;
+    CARD32  pix;
+    int     i, j;
+
+    switch (image->Format) {
+    case GL_RGB:
+	for (i = 0; i < height; i++) {
+	    src = (CARD8 *)image->Data + ((y + i) * pitch + x) * 3;
+	    for (j = width >> 1; j; j--) {
+		pix = ((R128PACKCOLOR565(src[0], src[1], src[2])      ) |
+		       (R128PACKCOLOR565(src[3], src[4], src[5]) << 16));
+		R128CCE(pix);
+		src += 6;
+	    }
+	}
+	break;
+
+    case GL_RGBA:
+	for (i = 0; i < height; i++) {
+	    src = (CARD8 *)image->Data + ((y + i) * pitch + x) * 4;
+	    for (j = width >> 1; j; j--) {
+		pix =
+		    ((R128PACKCOLOR4444(src[0], src[1], src[2], src[3])    ) |
+		     (R128PACKCOLOR4444(src[4], src[5], src[6], src[7])<<16));
+		R128CCE(pix);
+		src += 8;
+	    }
+	}
+	break;
+
+    case GL_ALPHA:
+	for (i = 0; i < height; i++) {
+	    src = (CARD8 *)image->Data + ((y + i) * pitch + x);
+	    for (j = width >> 1; j; j--) {
+		pix = ((R128PACKCOLOR4444(0xff, 0xff, 0xff, src[0])      ) |
+		       (R128PACKCOLOR4444(0xff, 0xff, 0xff, src[1]) << 16));
+		R128CCE(pix);
+		src += 2;
+	    }
+	}
+	break;
+
+    case GL_LUMINANCE:
+	for (i = 0; i < height; i++) {
+	    src = (CARD8 *)image->Data + ((y + i) * pitch + x);
+	    for (j = width >> 1; j; j--) {
+		pix = ((R128PACKCOLOR1555(src[0], src[0], src[0], 0xff)    ) |
+		       (R128PACKCOLOR1555(src[1], src[1], src[1], 0xff)<<16));
+		R128CCE(pix);
+		src += 2;
+	    }
+	}
+	break;
+
+    case GL_LUMINANCE_ALPHA:
+	for (i = 0; i < height; i++) {
+	    src = (CARD8 *)image->Data + ((y + i) * pitch + x) * 2;
+	    for (j = width >> 1; j; j--) {
+		pix =
+		    ((R128PACKCOLOR4444(src[0], src[0], src[0], src[1])    ) |
+		     (R128PACKCOLOR4444(src[2], src[2], src[2], src[3])<<16));
+		R128CCE(pix);
+		src += 4;
+	    }
+	}
+	break;
+
+    case GL_INTENSITY:
+	for (i = 0; i < height; i++) {
+	    src = (CARD8 *)image->Data + ((y + i) * pitch + x);
+	    for (j = width >> 1; j; j--) {
+		pix =
+		    ((R128PACKCOLOR4444(src[0], src[0], src[0], src[0])    ) |
+		     (R128PACKCOLOR4444(src[1], src[1], src[1], src[1])<<16));
+		R128CCE(pix);
+		src += 2;
+	    }
+	}
+	break;
+
+    default:
+	fprintf(stderr, "r128ConvertTexture16bpp: unsupported format 0x%x\n",
+		image->Format);
+    }
+}
+
+/* Convert a block of Mesa-formatted texture to a 32bpp hardware format */
+static void r128ConvertTexture32bpp(r128ContextPtr r128ctx,
+				    struct gl_texture_image *image,
+				    int x, int y, int width, int height,
+				    int pitch)
+{
+    CARD8  *src;
+    CARD32  pix;
+    int     i, j;
+
+    switch (image->Format) {
+    case GL_RGB:
+	for (i = 0; i < height; i++) {
+	    src = (CARD8 *)image->Data + ((y + i) * pitch + x) * 3;
+	    for (j = width; j; j--) {
+		pix = R128PACKCOLOR8888(src[0], src[1], src[2], 0xff);
+		R128CCE(pix);
+		src += 3;
+	    }
+	}
+	break;
+
+    case GL_RGBA:
+	for (i = 0; i < height; i++) {
+	    src = (CARD8 *)image->Data + ((y + i) * pitch + x) * 4;
+	    for (j = width; j; j--) {
+		pix = R128PACKCOLOR8888(src[0], src[1], src[2], src[3]);
+		R128CCE(pix);
+		src += 4;
+	    }
+	}
+	break;
+
+    case GL_ALPHA:
+	for (i = 0; i < height; i++) {
+	    src = (CARD8 *)image->Data + ((y + i) * pitch + x);
+	    for (j = width; j; j--) {
+		pix = R128PACKCOLOR8888(0xff, 0xff, 0xff, src[0]);
+		R128CCE(pix);
+		src += 1;
+	    }
+	}
+	break;
+
+    case GL_LUMINANCE:
+	for (i = 0 ; i < height ; i++) {
+	    src = (CARD8 *)image->Data + ((y + i) * pitch + x);
+	    for (j = width; j; j--) {
+		pix = R128PACKCOLOR8888(src[0], src[0], src[0], 0xff);
+		R128CCE(pix);
+		src += 1;
+	    }
+	}
+	break;
+
+    case GL_LUMINANCE_ALPHA:
+	for (i = 0; i < height; i++) {
+	    src = (CARD8 *)image->Data + ((y + i) * pitch + x) * 2;
+	    for (j = width; j; j-- ) {
+		pix = R128PACKCOLOR8888(src[0], src[0], src[0], src[1]);
+		R128CCE(pix);
+		src += 2;
+	    }
+	}
+	break;
+
+    case GL_INTENSITY:
+	for (i = 0; i < height; i++) {
+	    src = (CARD8 *)image->Data + ((y + i) * pitch + x);
+	    for (j = width; j; j--) {
+		pix = R128PACKCOLOR8888(src[0], src[0], src[0], src[0]);
+		R128CCE(pix);
+		src += 1;
+	    }
+	}
+	break;
+
+    default:
+	fprintf(stderr, "r128ConvertTexture32bpp: unsupported format 0x%x\n",
+		image->Format);
+    }
+}
+
+/* Upload the texture image associated with texture `t' at level `level'
+   at the address relative to `start'. */
+/* NOTE: This function is only called while holding the hardware lock */
+static void r128UploadSubImage(r128ContextPtr r128ctx,
+			       r128TexObjPtr t, int level,
+			       int x, int y, int width, int height)
+{
+    struct gl_texture_image *image;
+    int                      texelsPerDword = 0;
+    int                      imageWidth, imageHeight;
+    int                      remaining, rows;
+    int                      format, pitch, dwords;
+    CARD32                   offset;
+
+    /* Ensure we have a valid texture to upload */
+    if (level < 0 || level > R128_TEX_MAXLEVELS) return;
+    if (!(image = t->tObj->Image[level]))        return;
+
+    /* FIXME: support RGB888 (i.e., 24bpp) textures? */
+    switch (t->texelBytes) {
+    case 1: texelsPerDword = 4; break;
+    case 2: texelsPerDword = 2; break;
+    case 4: texelsPerDword = 1; break;
+    }
+
+    imageWidth  = image->Width;
+    imageHeight = image->Height;
+
+    format = t->textureFormat >> 16;
+
+    /* The texel upload routines have a minimum width, so force the size
+       if needed */
+    if (imageWidth < texelsPerDword) {
+	int factor;
+
+	factor        = texelsPerDword / imageWidth;
+	imageWidth    = texelsPerDword;
+	imageHeight /= factor;
+	if (imageHeight == 0) {
+	    /* In this case, the texel converter will actually walk a
+	       texel or two off the end of the image, but normal malloc
+	       alignment should prevent it from ever causing a fault. */
+	    imageHeight = 1;
+	}
+    }
+
+    /* We can't upload to a pitch less than 8 texels so we will need to
+       linearly upload all modified rows for textures smaller than this.
+       This makes the x/y/width/height different for the blitter and the
+       texture walker. */
+    if (imageWidth >= 8) {
+	/* The texture walker and the blitter look identical */
+	pitch      = imageWidth >> 3;
+    } else {
+	int factor;
+	int y2;
+	int start, end;
+
+	start = (y * imageWidth) & ~7;
+	end   = (y + height) * imageWidth;
+
+	if (end - start < 8) {
+	    /* Handle the case where the total number of texels uploaded
+               is < 8 */
+	    x       = 0;
+	    y       = start / 8;
+	    width   = end - start;
+	    height  = 1;
+	} else {
+	    /* Upload some number of full 8 texel blit rows */
+	    factor  = 8 / imageWidth;
+
+	    y2      = y + height - 1;
+	    y      /= factor;
+	    y2     /= factor;
+
+	    x       = 0;
+	    width   = 8;
+	    height  = y2 - y + 1;
+	}
+
+	/* Fixed pitch of 8 */
+	pitch = 1;
+    }
+
+    dwords = width * height / texelsPerDword;
+    offset = (CARD32)(t->bufAddr + t->image[level].offset);
+
+    /* Fix offset for AGP textures */
+    if (t->heap == R128_AGP_TEX_HEAP)
+	offset += R128_AGP_TEX_OFFSET + r128ctx->r128Screen->agpTexOffset;
+
+    if (R128_DEBUG_FLAGS & DEBUG_VERBOSE_API) {
+	fprintf(stderr, "r128UploadSubImage: %d,%d of %d,%d at %d,%d\n",
+		width, height, image->Width, image->Height, x, y);
+	fprintf(stderr, "          blit ofs: 0x%08x pitch: 0x%x dwords: %d "
+				  "level: %d format: %x\n",
+		(int)offset, pitch, dwords, level, format);
+    }
+
+    /* Subdivide the texture if required */
+    if (dwords < R128_CCE_PACKET_MAX_DWORDS) {
+	rows = height;
+    } else {
+	rows = (R128_CCE_PACKET_MAX_DWORDS * texelsPerDword) / (2 * width);
+    }
+
+    /* Flush the pixel cache, and mark the contents as Read Invalid */
+    R128CCE0(R128_CCE_PACKET0, R128_PC_GUI_CTLSTAT, 0);
+    R128CCE(r128ctx->regs.pc_gui_ctlstat |
+	    R128_PC_RI_GUI               |
+	    R128_PC_FLUSH_GUI);
+    R128CCE_SUBMIT_PACKETS();
+
+    /* Build the CCE host data blit header */
+    R128CCE3(R128_CCE_PACKET3_CNTL_HOSTDATA_BLT, 0);
+
+    /* DP_GUI_MASTER_CNTL */
+    R128CCE(R128_GMC_DST_PITCH_OFFSET_CNTL |
+	    R128_GMC_BRUSH_NONE            |
+	    (format << 8)                  |
+	    R128_GMC_SRC_DATATYPE_COLOR    |
+	    R128_ROP3_S                    |
+	    R128_DP_SRC_SOURCE_HOST_DATA   |
+	    R128_GMC_CLR_CMP_CNTL_DIS      |
+	    R128_AUX_CLIP_DIS              |
+	    R128_GMC_WR_MSK_DIS           );
+
+    /* DST_OFFSET_PITCH - fixed at the moment until we get better ring
+       control */
+    R128CCE((pitch << 21) | (offset>>5));
+
+    /* FRGD_COLOR, BKGD_COLOR */
+    R128CCE(0xffffffff);
+    R128CCE(0xffffffff);
+
+    for (remaining = height; remaining > 0; remaining -= rows, y += rows) {
+	height = (remaining >= rows) ? rows : remaining;
+	dwords = width * height / texelsPerDword;
+
+	if (R128_DEBUG_FLAGS & DEBUG_VERBOSE_API) {
+	    fprintf(stderr, "          blitting: %d,%d at %d,%d - %d dwords\n",
+		    width, height, x, y, dwords);
+	}
+
+	r128ctx->CCEbuf[0] &= ~R128_CCE_PACKET_COUNT_MASK;
+	r128ctx->CCEbuf[0] |= (dwords + 6) << 16;
+
+	/* Blit coords, size */
+	R128CCE((y      << 16) | x);
+	R128CCE((height << 16) | width);
+	R128CCE(dwords);
+
+	/* Actually do the texture conversion */
+	switch (t->texelBytes) {
+	case 1:
+	    r128ConvertTexture8bpp(r128ctx, image,
+				   x, y, width, height, width);
+	    break;
+	case 2:
+	    r128ConvertTexture16bpp(r128ctx, image,
+				    x, y, width, height, width);
+	    break;
+	case 4:
+	    r128ConvertTexture32bpp(r128ctx,
+				    image, x, y, width, height, width);
+	    break;
+	}
+
+	/* Flush the pixel cache */
+	R128CCE0(R128_CCE_PACKET0, R128_PC_GUI_CTLSTAT, 0 );
+	R128CCE(r128ctx->regs.pc_gui_ctlstat | R128_PC_FLUSH_GUI);
+
+	/* Save the partial blit header */
+	R128CCE_SUBMIT_PACKETS();
+	r128ctx->CCEcount = 5;
+    }
+    
+    /* Clean up CCE ring buffer */
+    r128ctx->CCEcount = 0;
+}
+
+/* Upload the texture images associated with texture `t'.  This might
+   require removing our own and/or other client's texture objects to
+   make room for these images. */
+/* NOTE: This function is only called while holding the hardware lock */
+int r128UploadTexImages(r128ContextPtr r128ctx, r128TexObjPtr t)
+{
+    int i;
+    int minLevel;
+    int maxLevel;
+    int heap;
+
+    if (!t) return 0;
+
+    /* Choose the heap appropriately */
+    heap = t->heap = R128_LOCAL_TEX_HEAP;
+    if (!r128ctx->r128Screen->IsPCI &&
+	t->totalSize > r128ctx->r128Screen->texSize[heap])
+	heap = t->heap = R128_AGP_TEX_HEAP;
+
+    /* Do we need to eject LRU texture objects? */
+    if (!t->memBlock) {
+	/* Allocate a memory block on a 4k boundary (1<<12 == 4096) */
+	t->memBlock = mmAllocMem(r128ctx->texHeap[heap], t->totalSize, 12, 0);
+
+	/* Try AGP before kicking anything out of local mem */
+	if (!t->memBlock && heap == R128_LOCAL_TEX_HEAP) {
+	    t->memBlock = mmAllocMem(r128ctx->texHeap[R128_AGP_TEX_HEAP],
+				     t->totalSize, 12, 0);
+
+	    if (t->memBlock) heap = t->heap = R128_AGP_TEX_HEAP;
+	}
+
+	/* Kick out textures until the requested texture fits */
+	while (!t->memBlock) {
+	    if (r128ctx->TexObjList[heap].prev->bound) {
+		fprintf(stderr,
+			"r128UploadTexImages: ran into bound texture\n");
+		return -1;
+	    }
+	    if (r128ctx->TexObjList[heap].prev ==
+		&(r128ctx->TexObjList[heap])) {
+		if (r128ctx->r128Screen->IsPCI) {
+		    fprintf(stderr, "r128UploadTexImages: upload texture "
+			    "failure on local texture heaps, sz=%d\n",
+			    t->totalSize);
+		    return -1;
+		} else if (heap == R128_LOCAL_TEX_HEAP) {
+		    heap = t->heap = R128_AGP_TEX_HEAP;
+		    continue;
+		} else {
+		    fprintf(stderr, "r128UploadTexImages: upload texture "
+			    "failure on both local and AGP texture heaps, "
+			    "sz=%d\n",
+			    t->totalSize);
+		    return -1;
+		}
+	    }
+
+	    r128DestroyTexObj(r128ctx, r128ctx->TexObjList[heap].prev);
+
+	    t->memBlock = mmAllocMem(r128ctx->texHeap[heap],
+				     t->totalSize, 12, 0);
+	}
+
+	/* Set the base offset of the texture image */
+	t->bufAddr = (unsigned char *)r128ctx->r128Screen->texOffset[heap];
+	t->bufAddr += t->memBlock->ofs;
+
+	maxLevel = ((t->regs.size_pitch & R128_TEX_SIZE_MASK) >>
+		    R128_TEX_SIZE_SHIFT);
+	minLevel = ((t->regs.size_pitch & R128_TEX_MIN_SIZE_MASK) >>
+		    R128_TEX_MIN_SIZE_SHIFT);
+
+	/* Update the hardware's texture image addresses */
+	switch (t->bound) {
+	case 1:
+	    /* Set texture offsets */
+	    if (t->regs.tex_cntl & R128_MIP_MAP_DISABLE) {
+		for (i = 0; i < R128_TEX_MAXLEVELS; i++)
+		    r128ctx->regs.prim_tex_offset[i] = (CARD32)t->bufAddr;
+	    } else {
+		for (i = maxLevel; i >= minLevel; i--)
+		    r128ctx->regs.prim_tex_offset[i] =
+			t->image[maxLevel-i].offset + (CARD32)t->bufAddr;
+	    }
+	    /* Fix AGP texture offsets */
+	    if (heap == R128_AGP_TEX_HEAP)
+		for (i = 0; i < R128_TEX_MAXLEVELS; i++)
+		    r128ctx->regs.prim_tex_offset[i] +=
+			R128_AGP_TEX_OFFSET +
+			r128ctx->r128Screen->agpTexOffset;
+	    /* Force loading the new state into the hardware */
+	    r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+	    r128ctx->dirty_context |= R128_CTX_TEX0STATE;
+	    break;
+
+	case 2:
+	    /* Set texture offsets */
+	    if (t->regs.tex_cntl & R128_MIP_MAP_DISABLE) {
+		for (i = 0; i < R128_TEX_MAXLEVELS; i++)
+		    r128ctx->regs.sec_tex_offset[i] = (CARD32)t->bufAddr;
+	    } else {
+		for (i = maxLevel; i >= minLevel; i--)
+		    r128ctx->regs.sec_tex_offset[i] = 
+			t->image[maxLevel-i].offset + (CARD32)t->bufAddr;
+	    }
+	    /* Fix AGP texture offsets */
+	    if (heap == R128_AGP_TEX_HEAP)
+		for (i = 0; i < R128_TEX_MAXLEVELS; i++)
+		    r128ctx->regs.sec_tex_offset[i] +=
+			R128_AGP_TEX_OFFSET +
+			r128ctx->r128Screen->agpTexOffset;
+	    /* Force loading the new state into the hardware */
+	    r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+	    r128ctx->dirty_context |= R128_CTX_TEX1STATE;
+	    break;
+
+	default:
+	    return -1;
+	}
+    }
+
+    /* Let the world know we've used this memory recently */
+    r128UpdateTexLRU(r128ctx, t);
+
+    /* Upload any images that are new */
+    if (t->dirty_images) {
+	int num_levels = (((t->regs.size_pitch & R128_TEX_SIZE_MASK) >>
+			   R128_TEX_SIZE_SHIFT) -
+			  ((t->regs.size_pitch & R128_TEX_MIN_SIZE_MASK) >>
+			   R128_TEX_MIN_SIZE_SHIFT));
+
+	for (i = 0; i <= num_levels; i++) {
+	    if (t->dirty_images & (1<<i)) {
+		r128UploadSubImage(r128ctx, t, i, 0, 0,
+				   t->image[i].width, t->image[i].height);
+	    }
+	}
+
+	r128ctx->regs.tex_cntl_c |= R128_TEX_CACHE_FLUSH;
+
+	r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+	r128ctx->dirty_context |= R128_CTX_ENGINESTATE;
+    }
+
+    t->dirty_images = 0;
+    return 0;
+}
+
+/* Update the hardware state for texture unit 0 */
+/* NOTE: This function is only called while holding the hardware lock */
+static void r128UpdateTex0State(r128ContextPtr r128ctx)
+{
+    GLcontext                *ctx = r128ctx->glCtx;
+    r128TexObjPtr             t;
+    struct gl_texture_object *tObj;
+    int                       i;
+    CARD32                    tex_size_pitch, tex_combine_cntl;
+
+    /* Only update the hardware texture state if the texture is current,
+       complete and enabled. */
+    if (!(tObj = ctx->Texture.Unit[0].Current))       return;
+    if ((tObj != ctx->Texture.Unit[0].CurrentD[2]) &&
+	(tObj != ctx->Texture.Unit[0].CurrentD[1]))   return;
+    if (!tObj->Complete)                              return;
+
+    /* If neither tex0 nor tex1 is enabled, then disable tex0.  However,
+       if tex1 is enabled but tex0 is disabled, then we need to enable
+       tex0 and have it to copy the input (see how tex_combine_cntl is
+       setup below). */
+    if (!(ctx->Texture.Enabled & (ENABLE_TEX0 | ENABLE_TEX1))) {
+	r128ctx->regs.tex_cntl_c &= ~R128_TEXMAP_ENABLE;
+	return;
+    }
+
+    /* If this is the first time the texture has been used, then create
+       a new texture object for it. */
+    t = tObj->DriverData;
+    if (!t) t = r128CreateTexObj(r128ctx, tObj);
+    if (!t) return;
+
+    if (R128_DEBUG_FLAGS & DEBUG_VERBOSE_API)
+	fprintf(stderr, "r128UpdateTex0State(%p, 0x%08x)\n",
+		tObj, (int)t->dirty_images);
+
+    /* Force any texture images to be loaded into the hardware */
+    if (t->dirty_images) r128ctx->dirty |= R128_UPDATE_TEX0IMAGES;
+
+    /* Bind texture to texture 0 unit */
+    r128ctx->CurrentTexObj[0] = t;
+    t->bound = 1;
+
+    if (t->memBlock) r128UpdateTexLRU(r128ctx, t);
+
+    /* Set the texture environment state */
+    switch (ctx->Texture.Unit[0].EnvMode) {
+    case GL_REPLACE:
+	switch (tObj->Image[0]->Format) {
+	case GL_RGBA:
+	case GL_LUMINANCE_ALPHA:
+	case GL_INTENSITY:
+	    tex_combine_cntl = (R128_COMB_DIS                |
+				R128_COLOR_FACTOR_TEX        |
+				R128_INPUT_FACTOR_INT_COLOR  |
+				R128_COMB_ALPHA_DIS          |
+				R128_ALPHA_FACTOR_TEX_ALPHA  |
+				R128_INP_FACTOR_A_INT_ALPHA);
+	    break;
+	case GL_RGB:
+	case GL_LUMINANCE:
+	    tex_combine_cntl = (R128_COMB_DIS                |
+				R128_COLOR_FACTOR_TEX        |
+				R128_INPUT_FACTOR_INT_COLOR  |
+				R128_COMB_ALPHA_COPY_INP     |
+				R128_ALPHA_FACTOR_TEX_ALPHA  |
+				R128_INP_FACTOR_A_INT_ALPHA);
+	    break;
+	case GL_ALPHA:
+	    tex_combine_cntl = (R128_COMB_COPY_INP           |
+				R128_COLOR_FACTOR_TEX        |
+				R128_INPUT_FACTOR_INT_COLOR  |
+				R128_COMB_ALPHA_DIS          |
+				R128_ALPHA_FACTOR_TEX_ALPHA  |
+				R128_INP_FACTOR_A_INT_ALPHA);
+	    break;
+	case GL_COLOR_INDEX:
+	default:
+	    return;
+	}
+	break;
+
+    case GL_BLEND:
+	switch (tObj->Image[0]->Format) {
+	case GL_RGBA:
+	case GL_LUMINANCE_ALPHA:
+	    switch (r128ctx->regs.constant_color_c &
+		    R128_CONSTANT_COLOR_MASK) {
+	    case R128_CONSTANT_COLOR_ZERO:
+		tex_combine_cntl = (R128_COMB_MODULATE           |
+				    R128_COLOR_FACTOR_NTEX       |
+				    R128_INPUT_FACTOR_INT_COLOR  |
+				    R128_COMB_ALPHA_MODULATE     |
+				    R128_ALPHA_FACTOR_TEX_ALPHA  |
+				    R128_INP_FACTOR_A_INT_ALPHA);
+		break;
+	    case R128_CONSTANT_COLOR_ONE:
+	    default:
+		tex_combine_cntl = (R128_COMB_MODULATE           |
+				    R128_COLOR_FACTOR_TEX        |
+				    R128_INPUT_FACTOR_INT_COLOR  |
+				    R128_COMB_ALPHA_MODULATE     |
+				    R128_ALPHA_FACTOR_TEX_ALPHA  |
+				    R128_INP_FACTOR_A_INT_ALPHA);
+		r128ctx->Fallback |= R128_FALLBACK_TEXTURE;
+		break;
+	    }
+	    break;
+	case GL_RGB:
+	case GL_LUMINANCE:
+	    switch (r128ctx->regs.constant_color_c &
+		    R128_CONSTANT_COLOR_MASK) {
+	    case R128_CONSTANT_COLOR_ZERO:
+		tex_combine_cntl = (R128_COMB_MODULATE           |
+				    R128_COLOR_FACTOR_NTEX       |
+				    R128_INPUT_FACTOR_INT_COLOR  |
+				    R128_COMB_ALPHA_COPY_INP     |
+				    R128_ALPHA_FACTOR_TEX_ALPHA  |
+				    R128_INP_FACTOR_A_INT_ALPHA);
+		break;
+	    case R128_CONSTANT_COLOR_ONE:
+	    default:
+		tex_combine_cntl = (R128_COMB_MODULATE           |
+				    R128_COLOR_FACTOR_TEX        |
+				    R128_INPUT_FACTOR_INT_COLOR  |
+				    R128_COMB_ALPHA_COPY_INP     |
+				    R128_ALPHA_FACTOR_TEX_ALPHA  |
+				    R128_INP_FACTOR_A_INT_ALPHA);
+		r128ctx->Fallback |= R128_FALLBACK_TEXTURE;
+		break;
+	    }
+	    break;
+	case GL_ALPHA:
+	    tex_combine_cntl = (R128_COMB_COPY_INP           |
+				R128_COLOR_FACTOR_TEX        |
+				R128_INPUT_FACTOR_INT_COLOR  |
+				R128_COMB_ALPHA_MODULATE     |
+				R128_ALPHA_FACTOR_TEX_ALPHA  |
+				R128_INP_FACTOR_A_INT_ALPHA);
+	    break;
+	case GL_INTENSITY:
+	    switch (r128ctx->regs.constant_color_c &
+		    R128_CONSTANT_COLOR_MASK) {
+	    case R128_CONSTANT_COLOR_ZERO:
+		tex_combine_cntl = (R128_COMB_MODULATE           |
+				    R128_COLOR_FACTOR_NTEX       |
+				    R128_INPUT_FACTOR_INT_COLOR  |
+				    R128_COMB_ALPHA_MODULATE     |
+				    R128_ALPHA_FACTOR_NTEX_ALPHA |
+				    R128_INP_FACTOR_A_INT_ALPHA);
+		break;
+	    case R128_CONSTANT_COLOR_ONE:
+	    default:
+		tex_combine_cntl = (R128_COMB_MODULATE           |
+				    R128_COLOR_FACTOR_TEX        |
+				    R128_INPUT_FACTOR_INT_COLOR  |
+				    R128_COMB_ALPHA_MODULATE     |
+				    R128_ALPHA_FACTOR_TEX_ALPHA  |
+				    R128_INP_FACTOR_A_INT_ALPHA);
+		r128ctx->Fallback |= R128_FALLBACK_TEXTURE;
+		break;
+	    }
+	    break;
+	case GL_COLOR_INDEX:
+	default:
+	    return;
+	}
+	break;
+
+    case GL_MODULATE:
+	switch (tObj->Image[0]->Format) {
+	case GL_RGBA:
+	case GL_LUMINANCE_ALPHA:
+	case GL_INTENSITY:
+	    tex_combine_cntl = (R128_COMB_MODULATE           |
+				R128_COLOR_FACTOR_TEX        |
+				R128_INPUT_FACTOR_INT_COLOR  |
+				R128_COMB_ALPHA_MODULATE     |
+				R128_ALPHA_FACTOR_TEX_ALPHA  |
+				R128_INP_FACTOR_A_INT_ALPHA);
+	    break;
+	case GL_RGB:
+	case GL_LUMINANCE:
+	    tex_combine_cntl = (R128_COMB_MODULATE           |
+				R128_COLOR_FACTOR_TEX        |
+				R128_INPUT_FACTOR_INT_COLOR  |
+				R128_COMB_ALPHA_COPY_INP     |
+				R128_ALPHA_FACTOR_TEX_ALPHA  |
+				R128_INP_FACTOR_A_INT_ALPHA);
+	    break;
+	case GL_ALPHA:
+	    tex_combine_cntl = (R128_COMB_COPY_INP           |
+				R128_COLOR_FACTOR_TEX        |
+				R128_INPUT_FACTOR_INT_COLOR  |
+				R128_COMB_ALPHA_MODULATE     |
+				R128_ALPHA_FACTOR_TEX_ALPHA  |
+				R128_INP_FACTOR_A_INT_ALPHA);
+	    break;
+	case GL_COLOR_INDEX:
+	default:
+	    return;
+	}
+	break;
+
+    case GL_DECAL:
+	switch (tObj->Image[0]->Format) {
+	case GL_RGBA:
+	    tex_combine_cntl = (R128_COMB_BLEND_TEXTURE      |
+				R128_COLOR_FACTOR_TEX        |
+				R128_INPUT_FACTOR_INT_COLOR  |
+				R128_COMB_ALPHA_COPY_INP     |
+				R128_ALPHA_FACTOR_TEX_ALPHA  |
+				R128_INP_FACTOR_A_INT_ALPHA);
+	    break;
+	case GL_RGB:
+	    tex_combine_cntl = (R128_COMB_DIS                |
+				R128_COLOR_FACTOR_TEX        |
+				R128_INPUT_FACTOR_INT_COLOR  |
+				R128_COMB_ALPHA_COPY_INP     |
+				R128_ALPHA_FACTOR_TEX_ALPHA  |
+				R128_INP_FACTOR_A_INT_ALPHA);
+	    break;
+	case GL_ALPHA:
+	case GL_LUMINANCE:
+	case GL_LUMINANCE_ALPHA:
+	case GL_INTENSITY:
+	    /* Undefined behaviour - just copy the input */
+	    tex_combine_cntl = (R128_COMB_COPY_INP           |
+				R128_COLOR_FACTOR_TEX        |
+				R128_INPUT_FACTOR_INT_COLOR  |
+				R128_COMB_ALPHA_COPY_INP     |
+				R128_ALPHA_FACTOR_TEX_ALPHA  |
+				R128_INP_FACTOR_A_INT_ALPHA);
+	    break;
+	case GL_COLOR_INDEX:
+	default:
+	    return;
+	}
+	break;
+
+    case GL_ADD:
+	switch (tObj->Image[0]->Format) {
+	case GL_RGBA:
+	case GL_LUMINANCE_ALPHA:
+	case GL_INTENSITY:
+	    tex_combine_cntl = (R128_COMB_ADD                |
+				R128_COLOR_FACTOR_TEX        |
+				R128_INPUT_FACTOR_INT_COLOR  |
+				R128_COMB_ALPHA_MODULATE     |
+				R128_ALPHA_FACTOR_TEX_ALPHA  |
+				R128_INP_FACTOR_A_INT_ALPHA);
+	    break;
+	case GL_RGB:
+	case GL_LUMINANCE:
+	    tex_combine_cntl = (R128_COMB_ADD                |
+				R128_COLOR_FACTOR_TEX        |
+				R128_INPUT_FACTOR_INT_COLOR  |
+				R128_COMB_ALPHA_COPY_INP     |
+				R128_ALPHA_FACTOR_TEX_ALPHA  |
+				R128_INP_FACTOR_A_INT_ALPHA);
+	    break;
+	case GL_ALPHA:
+	    tex_combine_cntl = (R128_COMB_COPY_INP           |
+				R128_COLOR_FACTOR_TEX        |
+				R128_INPUT_FACTOR_INT_COLOR  |
+				R128_COMB_ALPHA_MODULATE     |
+				R128_ALPHA_FACTOR_TEX_ALPHA  |
+				R128_INP_FACTOR_A_INT_ALPHA);
+	    break;
+	case GL_COLOR_INDEX:
+	default:
+	    return;
+	}
+	break;
+
+    default:
+	return;
+    }
+
+    /* If tex0 is disabled, then make sure it just copies the input */
+    if (!(ctx->Texture.Enabled & ENABLE_TEX0))
+	tex_combine_cntl = (R128_COMB_COPY_INP           |
+			    R128_COLOR_FACTOR_TEX        |
+			    R128_INPUT_FACTOR_INT_COLOR  |
+			    R128_COMB_ALPHA_COPY_INP     |
+			    R128_ALPHA_FACTOR_TEX_ALPHA  |
+			    R128_INP_FACTOR_A_INT_ALPHA);
+
+    /* Enable tex0 */
+    r128ctx->regs.tex_cntl_c |= R128_TEXMAP_ENABLE;
+
+    tex_size_pitch  = r128ctx->regs.tex_size_pitch_c;
+    tex_size_pitch &= ~R128_TEX_SIZE_PITCH_MASK;
+    tex_size_pitch |= t->regs.size_pitch << R128_TEX_SIZE_PITCH_SHIFT;
+
+    /* Set the primary texture state in r128ctx->regs */
+    r128ctx->regs.prim_tex_cntl_c             = t->regs.tex_cntl;
+    r128ctx->regs.prim_texture_combine_cntl_c = tex_combine_cntl;
+    r128ctx->regs.tex_size_pitch_c            = tex_size_pitch;
+    r128ctx->regs.prim_texture_border_color_c = t->regs.border_color;
+
+    /* Set texture offsets */
+    if (t->regs.tex_cntl & R128_MIP_MAP_DISABLE) {
+	for (i = 0; i < R128_TEX_MAXLEVELS; i++)
+	    r128ctx->regs.prim_tex_offset[i] = (CARD32)t->bufAddr;
+    } else {
+	int maxLevel = ((t->regs.size_pitch & R128_TEX_SIZE_MASK) >>
+			R128_TEX_SIZE_SHIFT);
+	int minLevel = ((t->regs.size_pitch & R128_TEX_MIN_SIZE_MASK) >>
+			R128_TEX_MIN_SIZE_SHIFT);
+	for (i = maxLevel; i >= minLevel; i--)
+	    r128ctx->regs.prim_tex_offset[i] = 
+		t->image[maxLevel-i].offset + (CARD32)t->bufAddr;
+    }
+    /* Fix AGP texture offsets */
+    if (t->heap == R128_AGP_TEX_HEAP)
+	for (i = 0; i < R128_TEX_MAXLEVELS; i++)
+	    r128ctx->regs.prim_tex_offset[i] +=
+		R128_AGP_TEX_OFFSET + r128ctx->r128Screen->agpTexOffset;
+
+    /* Force loading the new state into the hardware */
+    r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+    r128ctx->dirty_context |= R128_CTX_TEX0STATE | R128_CTX_ENGINESTATE;
+}
+
+/* Update the hardware state for texture unit 1 */
+/* NOTE: This function is only called while holding the hardware lock */
+static void r128UpdateTex1State(r128ContextPtr r128ctx)
+{
+    GLcontext                *ctx = r128ctx->glCtx;
+    r128TexObjPtr             t;
+    struct gl_texture_object *tObj;
+    int                       i;
+    CARD32                    tex_size_pitch, tex_combine_cntl, tex_cntl;
+
+    /* Only update the hardware texture state if the texture is current,
+       complete and enabled. */
+    if (!(tObj = ctx->Texture.Unit[1].Current))       return;
+    if ((tObj != ctx->Texture.Unit[1].CurrentD[2]) &&
+	(tObj != ctx->Texture.Unit[1].CurrentD[1]))   return;
+    if (!tObj->Complete)                              return;
+
+    /* If tex1 is not enabled, then disable it */
+    if (!(ctx->Texture.Enabled & ENABLE_TEX1)) {
+	r128ctx->regs.tex_cntl_c &= ~R128_SEC_TEXMAP_ENABLE;
+	return;
+    }
+
+    /* If this is the first time the texture has been used, then create
+       a new texture object for it. */
+    t = tObj->DriverData;
+    if (!t) t = r128CreateTexObj(r128ctx, tObj);
+    if (!t) return;
+
+    if (R128_DEBUG_FLAGS & DEBUG_VERBOSE_API)
+	fprintf(stderr, "r128UpdateTex1State(%p, 0x%08x)\n",
+		tObj, (int)t->dirty_images);
+
+    /* Force any texture images to be loaded into the hardware */
+    if (t->dirty_images) r128ctx->dirty |= R128_UPDATE_TEX1IMAGES;
+
+    /* Bind texture to texture 1 unit */
+    r128ctx->CurrentTexObj[1] = t;
+    t->bound = 2;
+
+    if (t->memBlock) r128UpdateTexLRU(r128ctx, t);
+
+    /* Set the texture environment state */
+    switch (ctx->Texture.Unit[1].EnvMode) {
+    case GL_REPLACE:
+	switch (tObj->Image[0]->Format) {
+	case GL_RGBA:
+	case GL_LUMINANCE_ALPHA:
+	case GL_INTENSITY:
+	    tex_combine_cntl = (R128_COMB_DIS                 |
+				R128_COLOR_FACTOR_TEX         |
+				R128_INPUT_FACTOR_PREV_COLOR  |
+				R128_COMB_ALPHA_DIS           |
+				R128_ALPHA_FACTOR_TEX_ALPHA   |
+				R128_INP_FACTOR_A_PREV_ALPHA);
+	    break;
+	case GL_RGB:
+	case GL_LUMINANCE:
+	    tex_combine_cntl = (R128_COMB_DIS                 |
+				R128_COLOR_FACTOR_TEX         |
+				R128_INPUT_FACTOR_PREV_COLOR  |
+				R128_COMB_ALPHA_COPY_INP      |
+				R128_ALPHA_FACTOR_TEX_ALPHA   |
+				R128_INP_FACTOR_A_PREV_ALPHA);
+	    break;
+	case GL_ALPHA:
+	    tex_combine_cntl = (R128_COMB_COPY_INP            |
+				R128_COLOR_FACTOR_TEX         |
+				R128_INPUT_FACTOR_PREV_COLOR  |
+				R128_COMB_ALPHA_DIS           |
+				R128_ALPHA_FACTOR_TEX_ALPHA   |
+				R128_INP_FACTOR_A_PREV_ALPHA);
+	    break;
+	case GL_COLOR_INDEX:
+	default:
+	    return;
+	}
+	break;
+
+    case GL_BLEND:
+	switch (tObj->Image[0]->Format) {
+	case GL_RGBA:
+	case GL_LUMINANCE_ALPHA:
+	    switch ( r128ctx->regs.constant_color_c & R128_CONSTANT_COLOR_MASK ) {
+	    case R128_CONSTANT_COLOR_ZERO:
+		tex_combine_cntl = (R128_COMB_MODULATE            |
+				    R128_COLOR_FACTOR_NTEX        |
+				    R128_INPUT_FACTOR_PREV_COLOR  |
+				    R128_COMB_ALPHA_MODULATE      |
+				    R128_ALPHA_FACTOR_TEX_ALPHA   |
+				    R128_INP_FACTOR_A_PREV_ALPHA);
+		break;
+	    case R128_CONSTANT_COLOR_ONE:
+	    default:
+		tex_combine_cntl = (R128_COMB_MODULATE            |
+				    R128_COLOR_FACTOR_TEX         |
+				    R128_INPUT_FACTOR_PREV_COLOR  |
+				    R128_COMB_ALPHA_MODULATE      |
+				    R128_ALPHA_FACTOR_TEX_ALPHA   |
+				    R128_INP_FACTOR_A_PREV_ALPHA);
+		r128ctx->Fallback |= R128_FALLBACK_TEXTURE;
+		break;
+	    }
+	    break;
+	case GL_RGB:
+	case GL_LUMINANCE:
+	    switch (r128ctx->regs.constant_color_c &
+		    R128_CONSTANT_COLOR_MASK) {
+	    case R128_CONSTANT_COLOR_ZERO:
+		tex_combine_cntl = (R128_COMB_MODULATE            |
+				    R128_COLOR_FACTOR_NTEX        |
+				    R128_INPUT_FACTOR_PREV_COLOR  |
+				    R128_COMB_ALPHA_COPY_INP      |
+				    R128_ALPHA_FACTOR_TEX_ALPHA   |
+				    R128_INP_FACTOR_A_PREV_ALPHA );
+		break;
+	    case R128_CONSTANT_COLOR_ONE:
+	    default:
+		tex_combine_cntl = (R128_COMB_MODULATE            |
+				    R128_COLOR_FACTOR_TEX         |
+				    R128_INPUT_FACTOR_PREV_COLOR  |
+				    R128_COMB_ALPHA_COPY_INP      |
+				    R128_ALPHA_FACTOR_TEX_ALPHA   |
+				    R128_INP_FACTOR_A_PREV_ALPHA );
+		r128ctx->Fallback |= R128_FALLBACK_TEXTURE;
+		break;
+	    }
+	    break;
+	case GL_ALPHA:
+	    tex_combine_cntl = (R128_COMB_COPY_INP            |
+				R128_COLOR_FACTOR_TEX         |
+				R128_INPUT_FACTOR_PREV_COLOR  |
+				R128_COMB_ALPHA_MODULATE      |
+				R128_ALPHA_FACTOR_TEX_ALPHA   |
+				R128_INP_FACTOR_A_PREV_ALPHA);
+	    break;
+	case GL_INTENSITY:
+	    switch (r128ctx->regs.constant_color_c &
+		    R128_CONSTANT_COLOR_MASK) {
+	    case R128_CONSTANT_COLOR_ZERO:
+		tex_combine_cntl = (R128_COMB_MODULATE            |
+				    R128_COLOR_FACTOR_NTEX        |
+				    R128_INPUT_FACTOR_PREV_COLOR  |
+				    R128_COMB_ALPHA_MODULATE      |
+				    R128_ALPHA_FACTOR_NTEX_ALPHA  |
+				    R128_INP_FACTOR_A_PREV_ALPHA);
+		break;
+	    case R128_CONSTANT_COLOR_ONE:
+	    default:
+		tex_combine_cntl = (R128_COMB_MODULATE            |
+				    R128_COLOR_FACTOR_TEX         |
+				    R128_INPUT_FACTOR_PREV_COLOR  |
+				    R128_COMB_ALPHA_MODULATE      |
+				    R128_ALPHA_FACTOR_TEX_ALPHA   |
+				    R128_INP_FACTOR_A_PREV_ALPHA);
+		r128ctx->Fallback |= R128_FALLBACK_TEXTURE;
+		break;
+	    }
+	    break;
+	case GL_COLOR_INDEX:
+	default:
+	    return;
+	}
+	break;
+
+    case GL_MODULATE:
+	switch (tObj->Image[0]->Format) {
+	case GL_RGBA:
+	case GL_LUMINANCE_ALPHA:
+	case GL_INTENSITY:
+	    tex_combine_cntl = (R128_COMB_MODULATE            |
+				R128_COLOR_FACTOR_TEX         |
+				R128_INPUT_FACTOR_PREV_COLOR  |
+				R128_COMB_ALPHA_MODULATE      |
+				R128_ALPHA_FACTOR_TEX_ALPHA   |
+				R128_INP_FACTOR_A_PREV_ALPHA);
+	    break;
+	case GL_RGB:
+	case GL_LUMINANCE:
+	    tex_combine_cntl = (R128_COMB_MODULATE            |
+				R128_COLOR_FACTOR_TEX         |
+				R128_INPUT_FACTOR_PREV_COLOR  |
+				R128_COMB_ALPHA_COPY_INP      |
+				R128_ALPHA_FACTOR_TEX_ALPHA   |
+				R128_INP_FACTOR_A_PREV_ALPHA);
+	    break;
+	case GL_ALPHA:
+	    tex_combine_cntl = (R128_COMB_COPY_INP            |
+				R128_COLOR_FACTOR_TEX         |
+				R128_INPUT_FACTOR_PREV_COLOR  |
+				R128_COMB_ALPHA_MODULATE      |
+				 R128_ALPHA_FACTOR_TEX_ALPHA   |
+				R128_INP_FACTOR_A_PREV_ALPHA);
+	    break;
+	case GL_COLOR_INDEX:
+	default:
+	    return;
+	}
+	break;
+
+    case GL_DECAL:
+	switch (tObj->Image[0]->Format) {
+	case GL_RGBA:
+	    tex_combine_cntl = (R128_COMB_BLEND_TEXTURE       |
+				R128_COLOR_FACTOR_TEX         |
+				R128_INPUT_FACTOR_PREV_COLOR  |
+				R128_COMB_ALPHA_COPY_INP      |
+				R128_ALPHA_FACTOR_TEX_ALPHA   |
+				R128_INP_FACTOR_A_PREV_ALPHA);
+	    break;
+	case GL_RGB:
+	    tex_combine_cntl = (R128_COMB_DIS                 |
+				R128_COLOR_FACTOR_TEX         |
+				R128_INPUT_FACTOR_PREV_COLOR  |
+				R128_COMB_ALPHA_COPY_INP      |
+				R128_ALPHA_FACTOR_TEX_ALPHA   |
+				R128_INP_FACTOR_A_PREV_ALPHA);
+	    break;
+	case GL_ALPHA:
+	case GL_LUMINANCE:
+	case GL_LUMINANCE_ALPHA:
+	case GL_INTENSITY:
+	    /* Undefined behaviour - just copy the input */
+	    tex_combine_cntl = (R128_COMB_COPY_INP            |
+				R128_COLOR_FACTOR_TEX         |
+				R128_INPUT_FACTOR_PREV_COLOR  |
+				R128_COMB_ALPHA_COPY_INP      |
+				R128_ALPHA_FACTOR_TEX_ALPHA   |
+				R128_INP_FACTOR_A_PREV_ALPHA);
+	    break;
+	case GL_COLOR_INDEX:
+	default:
+	    return;
+	}
+	break;
+
+    case GL_ADD:
+	switch (tObj->Image[0]->Format) {
+	case GL_RGBA:
+	case GL_LUMINANCE_ALPHA:
+	case GL_INTENSITY:
+	    tex_combine_cntl = (R128_COMB_ADD                 |
+				R128_COLOR_FACTOR_TEX         |
+				R128_INPUT_FACTOR_PREV_COLOR  |
+				R128_COMB_ALPHA_MODULATE      |
+				R128_ALPHA_FACTOR_TEX_ALPHA   |
+				R128_INP_FACTOR_A_PREV_ALPHA);
+	    break;
+	case GL_RGB:
+	case GL_LUMINANCE:
+	    tex_combine_cntl = (R128_COMB_ADD                 |
+				R128_COLOR_FACTOR_TEX         |
+				R128_INPUT_FACTOR_PREV_COLOR  |
+				R128_COMB_ALPHA_COPY_INP      |
+				R128_ALPHA_FACTOR_TEX_ALPHA   |
+				R128_INP_FACTOR_A_PREV_ALPHA);
+	    break;
+	case GL_ALPHA:
+	    tex_combine_cntl = (R128_COMB_COPY_INP            |
+				R128_COLOR_FACTOR_TEX         |
+				R128_INPUT_FACTOR_PREV_COLOR  |
+				R128_COMB_ALPHA_MODULATE      |
+				R128_ALPHA_FACTOR_TEX_ALPHA   |
+				R128_INP_FACTOR_A_PREV_ALPHA);
+	    break;
+	case GL_COLOR_INDEX:
+	default:
+	    return;
+	}
+	break;
+
+    default:
+	return;
+    }
+
+    /* Enable tex1 */
+    r128ctx->regs.tex_cntl_c |= R128_SEC_TEXMAP_ENABLE;
+
+    tex_size_pitch  = r128ctx->regs.tex_size_pitch_c;
+    tex_size_pitch &= ~R128_SEC_TEX_SIZE_PITCH_MASK;
+    tex_size_pitch |= t->regs.size_pitch << R128_SEC_TEX_SIZE_PITCH_SHIFT;
+
+    tex_cntl        = t->regs.tex_cntl | R128_SEC_SELECT_SEC_ST;
+
+    /* Set the secondary texture state in r128ctx->regs */
+    r128ctx->regs.sec_tex_cntl_c              = tex_cntl;
+    r128ctx->regs.sec_tex_combine_cntl_c      = tex_combine_cntl;
+    r128ctx->regs.tex_size_pitch_c            = tex_size_pitch;
+    r128ctx->regs.sec_texture_border_color_c  = t->regs.border_color;
+
+    /* Set texture offsets */
+    if (t->regs.tex_cntl & R128_MIP_MAP_DISABLE) {
+	for (i = 0; i < R128_TEX_MAXLEVELS; i++)
+	    r128ctx->regs.sec_tex_offset[i] = (CARD32)t->bufAddr;
+    } else {
+	int maxLevel = ((t->regs.size_pitch & R128_TEX_SIZE_MASK) >>
+			R128_TEX_SIZE_SHIFT);
+	int minLevel = ((t->regs.size_pitch & R128_TEX_MIN_SIZE_MASK) >>
+			R128_TEX_MIN_SIZE_SHIFT);
+	for (i = maxLevel; i >= minLevel; i--)
+	    r128ctx->regs.sec_tex_offset[i] = 
+		t->image[maxLevel-i].offset + (CARD32)t->bufAddr;
+    }
+    /* Fix AGP texture offsets */
+    if (t->heap == R128_AGP_TEX_HEAP)
+	for (i = 0; i < R128_TEX_MAXLEVELS; i++)
+	    r128ctx->regs.sec_tex_offset[i] +=
+		R128_AGP_TEX_OFFSET + r128ctx->r128Screen->agpTexOffset;
+
+    /* Force loading the new state into the hardware */
+    r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+    r128ctx->dirty_context |= R128_CTX_TEX1STATE | R128_CTX_ENGINESTATE;
+}
+
+/* Update the hardware texture state */
+/* NOTE: This function is only called while holding the hardware lock */
+void r128UpdateTextureState(r128ContextPtr r128ctx)
+{
+    /* Clear the GL_BLEND texturing fallback */
+    r128ctx->Fallback &= ~R128_FALLBACK_TEXTURE;
+
+    /* Unbind any currently bound textures */
+    if (r128ctx->CurrentTexObj[0]) r128ctx->CurrentTexObj[0]->bound = 0;
+    if (r128ctx->CurrentTexObj[1]) r128ctx->CurrentTexObj[1]->bound = 0;
+    r128ctx->CurrentTexObj[0] = NULL;
+    r128ctx->CurrentTexObj[1] = NULL;
+
+    /* Update the texture unit 0/1 state */
+    r128UpdateTex0State(r128ctx);
+    r128UpdateTex1State(r128ctx);
+}
+
+/* Set the texture wrap mode */
+static void r128SetTexWrap(r128TexObjPtr t, GLenum swrap, GLenum twrap)
+{
+    t->regs.tex_cntl &= ~(R128_TEX_CLAMP_S_MASK | R128_TEX_CLAMP_T_MASK);
+
+    switch (swrap) {
+    case GL_CLAMP:  t->regs.tex_cntl |= R128_TEX_CLAMP_S_CLAMP; break;
+    case GL_REPEAT: t->regs.tex_cntl |= R128_TEX_CLAMP_S_WRAP;  break;
+    default: /* ERROR!! */ return;
+    }
+
+    switch (twrap) {
+    case GL_CLAMP:  t->regs.tex_cntl |= R128_TEX_CLAMP_T_CLAMP; break;
+    case GL_REPEAT: t->regs.tex_cntl |= R128_TEX_CLAMP_T_WRAP;  break;
+    default: /* ERROR!! */ return;
+    }
+}
+
+/* Set the texture filter mode */
+static void r128SetTexFilter(r128TexObjPtr t, GLenum minf, GLenum magf)
+{
+    t->regs.tex_cntl &= ~(R128_MIN_BLEND_MASK | R128_MAG_BLEND_MASK);
+
+    switch (minf) {
+    case GL_NEAREST:
+	t->regs.tex_cntl |= R128_MIN_BLEND_NEAREST;
+	break;
+    case GL_LINEAR:
+	t->regs.tex_cntl |= R128_MIN_BLEND_LINEAR;
+	break;
+    case GL_NEAREST_MIPMAP_NEAREST:
+	t->regs.tex_cntl |= R128_MIN_BLEND_MIPNEAREST;
+	break;
+    case GL_LINEAR_MIPMAP_NEAREST:
+	t->regs.tex_cntl |= R128_MIN_BLEND_LINEARMIPNEAREST;
+	break;
+    case GL_NEAREST_MIPMAP_LINEAR:
+	t->regs.tex_cntl |= R128_MIN_BLEND_MIPLINEAR;
+	break;
+    case GL_LINEAR_MIPMAP_LINEAR:
+	t->regs.tex_cntl |= R128_MIN_BLEND_LINEARMIPLINEAR;
+	break;
+    default: /* ERROR!! */ return;
+    }
+
+    switch (magf) {
+    case GL_NEAREST:
+	t->regs.tex_cntl |= R128_MAG_BLEND_NEAREST;
+	break;
+    case GL_LINEAR:
+	t->regs.tex_cntl |= R128_MAG_BLEND_LINEAR;
+	break;
+    }
+}
+
+/* Set the texture border color */
+static void r128SetTexBorderColor(r128TexObjPtr t, GLubyte c[4])
+{
+    t->regs.border_color = r128PackColor(32, c[0], c[1], c[2], c[3]);
+}
+
+/* Set the texture environment state */
+static void r128DDTexEnv(GLcontext *ctx, GLenum target, GLenum pname,
+			 const GLfloat *param)
+{
+    r128ContextPtr          r128ctx = R128_CONTEXT(ctx);
+    struct gl_texture_unit *texUnit;
+    GLubyte                 c[4];
+    CARD32                  col;
+    
+    /* FIXME: Add texture LOD bias extension */
+
+    switch (pname) {
+    case GL_TEXTURE_ENV_MODE:
+	/* TexEnv modes are handled in UpdateTextureState */
+	R128CCE_FLUSH_VB_LOCK(r128ctx);
+	r128ctx->dirty |= R128_UPDATE_TEXSTATE;
+	break;
+    case GL_TEXTURE_ENV_COLOR:
+	texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit];
+	FLOAT_RGBA_TO_UBYTE_RGBA(texUnit->EnvColor, c);
+	col = r128PackColor(32, c[0], c[1], c[2], c[3]);
+	if (r128ctx->regs.constant_color_c != col) {
+	    R128CCE_FLUSH_VB_LOCK(r128ctx);
+	    r128ctx->regs.constant_color_c = col;
+
+	    /* FIXME: Load into hardware now??? */
+	    r128ctx->dirty         |= R128_UPDATE_CONTEXT;
+	    r128ctx->dirty_context |= R128_CTX_TEXENVSTATE;
+	}
+	break;
+    default:
+	return;
+    }
+}
+
+/* Upload a new texture image */
+static void r128DDTexImage(GLcontext *ctx, GLenum target,
+			   struct gl_texture_object *tObj, GLint level,
+			   GLint internalFormat,
+			   const struct gl_texture_image *image)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+    r128TexObjPtr  t;
+
+    if (R128_DEBUG_FLAGS & DEBUG_VERBOSE_API)
+	fprintf(stderr, "r128DDTexImage(%p, level %d)\n", tObj, level);
+
+    if ((target != GL_TEXTURE_2D) && (target != GL_TEXTURE_1D)) return;
+    if (level >= R128_TEX_MAXLEVELS) return;
+
+    t = (r128TexObjPtr)tObj->DriverData;
+    if (t) {
+	if (t->bound) R128CCE_FLUSH_VB_LOCK(r128ctx);
+
+	/* Destroy the old texture, and upload a new one.  The actual
+           uploading of the texture image occurs in the UploadSubImage
+           function. */
+	r128DestroyTexObj(r128ctx, t);
+	r128ctx->dirty |= R128_UPDATE_TEXSTATE;
+    }
+}
+
+/* Upload a new texture sub-image */
+static void r128DDTexSubImage(GLcontext *ctx, GLenum target,
+			      struct gl_texture_object *tObj, GLint level,
+			      GLint xoffset, GLint yoffset,
+			      GLsizei width, GLsizei height,
+			      GLint internalFormat,
+			      const struct gl_texture_image *image)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+    r128TexObjPtr  t;
+
+    if (R128_DEBUG_FLAGS & DEBUG_VERBOSE_API)
+	fprintf(stderr, "r128DDTexSubImage(%p, level %d) "
+		"size: %d,%d of %d,%d\n",
+		tObj, level, width, height, image->Width, image->Height);
+
+    if ((target != GL_TEXTURE_2D) && (target != GL_TEXTURE_1D)) return;
+    if (level >= R128_TEX_MAXLEVELS) return;
+
+    t = (r128TexObjPtr)tObj->DriverData;
+    if (t) {
+	if (t->bound) R128CCE_FLUSH_VB_LOCK(r128ctx);
+
+	LOCK_HARDWARE(r128ctx);
+	r128UploadSubImage(r128ctx, t, level,
+			   xoffset, yoffset, width, height);
+	UNLOCK_HARDWARE(r128ctx);
+
+	/* Update the context state */
+	r128ctx->regs.tex_cntl_c |= R128_TEX_CACHE_FLUSH;
+
+	r128ctx->dirty           |= (R128_UPDATE_CONTEXT |
+				     R128_UPDATE_TEXSTATE);
+	r128ctx->dirty_context   |= (R128_CTX_ENGINESTATE |
+				     R128_CTX_TEX0STATE   |
+				     R128_CTX_TEX1STATE);
+    }
+}
+
+/* Set the texture parameter state */
+static void r128DDTexParameter(GLcontext *ctx, GLenum target,
+			       struct gl_texture_object *tObj,
+			       GLenum pname, const GLfloat *params)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+    r128TexObjPtr  t       = (r128TexObjPtr)tObj->DriverData;
+
+    if (!t) return;
+    if ((target != GL_TEXTURE_2D) && (target != GL_TEXTURE_1D)) return;
+
+    switch (pname) {
+    case GL_TEXTURE_MIN_FILTER:
+    case GL_TEXTURE_MAG_FILTER:
+	if (t->bound) R128CCE_FLUSH_VB_LOCK(r128ctx);
+	r128SetTexFilter(t, tObj->MinFilter, tObj->MagFilter);
+	break;
+
+    case GL_TEXTURE_WRAP_S:
+    case GL_TEXTURE_WRAP_T:
+	if (t->bound) R128CCE_FLUSH_VB_LOCK(r128ctx);
+	r128SetTexWrap(t, tObj->WrapS, tObj->WrapT);
+	break;
+  
+    case GL_TEXTURE_BORDER_COLOR:
+	if (t->bound) R128CCE_FLUSH_VB_LOCK(r128ctx);
+	r128SetTexBorderColor(t, tObj->BorderColor);
+	break;
+
+    default:
+	return;
+    }
+
+    r128ctx->dirty |= R128_UPDATE_TEXSTATE;
+}
+
+/* Bind a texture to the currently active texture unit */
+static void r128DDBindTexture(GLcontext *ctx, GLenum target,
+			      struct gl_texture_object *tObj)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+
+    R128CCE_FLUSH_VB_LOCK(r128ctx);
+
+    /* Unbind the old texture */
+    if (r128ctx->CurrentTexObj[ctx->Texture.CurrentUnit]) {
+	r128ctx->CurrentTexObj[ctx->Texture.CurrentUnit]->bound = 0;
+	r128ctx->CurrentTexObj[ctx->Texture.CurrentUnit] = NULL;
+    }
+
+    /* The actualy binding occurs in the Tex[01]UpdateState function */
+    r128ctx->dirty |= R128_UPDATE_TEXSTATE;
+}
+
+/* Remove texture from AGP/local texture memory */
+static void r128DDDeleteTexture(GLcontext *ctx,
+				struct gl_texture_object *tObj)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+    r128TexObjPtr  t       = (r128TexObjPtr)tObj->DriverData;
+
+    if (t) {
+	if (t->bound) {
+	    R128CCE_FLUSH_VB_LOCK(r128ctx);
+
+	    r128ctx->CurrentTexObj[t->bound-1] = 0;
+	    r128ctx->dirty |= R128_UPDATE_TEXSTATE;
+	}
+
+	r128DestroyTexObj(r128ctx, t);
+	tObj->DriverData = NULL;
+    }
+}
+
+/* Determine if a texture is currently residing in either AGP/local
+   texture memory */
+static GLboolean r128DDIsTextureResident(GLcontext *ctx, 
+					 struct gl_texture_object *tObj)
+{
+    r128TexObjPtr t = (r128TexObjPtr)tObj->DriverData;
+
+    return t && t->memBlock;
+}
+
+/* Initialize the driver's texture functions */
+void r128DDInitTextureFuncs(GLcontext *ctx)
+{
+    ctx->Driver.TexEnv                  = r128DDTexEnv;
+    ctx->Driver.TexImage                = r128DDTexImage;
+    ctx->Driver.TexSubImage             = r128DDTexSubImage;
+    ctx->Driver.TexParameter            = r128DDTexParameter;
+    ctx->Driver.BindTexture             = r128DDBindTexture;
+    ctx->Driver.DeleteTexture           = r128DDDeleteTexture;
+    ctx->Driver.UpdateTexturePalette    = NULL;
+    ctx->Driver.ActiveTexture           = NULL;
+    ctx->Driver.IsTextureResident       = r128DDIsTextureResident;
+    ctx->Driver.PrioritizeTexture       = NULL;
+}
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_tex.h b/xc/lib/GL/mesa/src/drv/r128/r128_tex.h
new file mode 100644
index 000000000..20893110c
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_tex.h
@@ -0,0 +1,82 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *   Gareth Hughes <gareth@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_TEX_H_
+#define _R128_TEX_H_
+
+#ifdef GLX_DIRECT_RENDERING
+
+extern void r128AgeTextures(r128ContextPtr r128ctx, int heap);
+extern int  r128UploadTexImages(r128ContextPtr r128ctx, r128TexObjPtr t);
+extern void r128UpdateTextureState(r128ContextPtr r128ctx);
+extern void r128DestroyTexObj(r128ContextPtr r128ctx, r128TexObjPtr t);
+
+extern void r128DDInitTextureFuncs(GLcontext *ctx);
+
+#define R128PACKCOLOR332(r, g, b)                                             \
+    (((r) & 0xe0) | (((g) & 0xe0) >> 3) | (((b) & 0xc0) >> 6))
+
+#define R128PACKCOLOR1555(r, g, b, a)                                         \
+    ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) |        \
+     ((a) ? 0x8000 : 0))
+
+#define R128PACKCOLOR565(r, g, b)                                             \
+    ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3))
+
+#define R128PACKCOLOR888(r, g, b)                                             \
+    (((r) << 16) | ((g) << 8) | (b))
+
+#define R128PACKCOLOR8888(r, g, b, a)                                         \
+    (((a) << 24) | ((r) << 16) | ((g) << 8) | (b))
+
+#define R128PACKCOLOR4444(r, g, b, a)                                         \
+    ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4))
+
+static __inline__ CARD32 r128PackColor(GLuint depth,
+				       GLubyte r, GLubyte g,
+				       GLubyte b, GLubyte a)
+{
+    switch (depth) {
+    case  8: return R128PACKCOLOR332(r, g, b);
+    case 15: return R128PACKCOLOR1555(r, g, b, a);
+    case 16: return R128PACKCOLOR565(r, g, b);
+    case 24: return R128PACKCOLOR888(r, g, b);
+    case 32: return R128PACKCOLOR8888(r, g, b, a);
+    default: return 0;
+    }
+}
+
+#endif
+#endif /* _R128_TEX_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_texobj.h b/xc/lib/GL/mesa/src/drv/r128/r128_texobj.h
new file mode 100644
index 000000000..588b768ed
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_texobj.h
@@ -0,0 +1,87 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *   Gareth Hughes <gareth@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_TEXOBJ_H_
+#define _R128_TEXOBJ_H_
+
+#include "mm.h"
+
+#define R128_TEX_MAXLEVELS 11
+
+/* Setup registers for each texture */
+typedef struct {
+    CARD32  tex_cntl;
+    CARD32  size_pitch;
+    CARD32  border_color;
+} r128TextureRegs;
+
+/* Individual texture image information */
+typedef struct {
+    int offset;		/* Offset into locally shared texture space (i.e.,
+			   relative to bufAddr (below) */
+    int width;		/* Width of texture image */
+    int height;		/* Height of texture image */
+} r128TexImage;
+
+typedef struct r128_tex_obj r128TexObj, *r128TexObjPtr;
+
+/* Texture object in locally shared texture space */
+struct r128_tex_obj {
+    r128TexObjPtr    next, prev;
+
+    struct gl_texture_object *tObj;	/* Mesa texture object */
+
+    PMemBlock        memBlock;		/* Memory block containing texture */
+    unsigned char   *bufAddr;		/* Offset to start of locally
+					   shared texture block */
+
+    CARD32           dirty_images;	/* Flags for whether or not
+					   images need to be uploaded to
+					   local or AGP texture space */
+    int              bound;		/* Texture unit currently bound to */
+    int              heap;              /* Texture heap currently stored in */
+    r128TexImage     image[R128_TEX_MAXLEVELS]; /* Image data for all
+						   mipmap levels */
+    int              totalSize;		/* Total size of the texture
+					   including all mipmap levels */
+    int              internFormat;	/* Internal GL format used to store
+					   texture on card */
+    int              textureFormat;	/* Actual hardware format */
+    int              texelBytes;	/* Number of bytes per texel */
+
+    r128TextureRegs  regs;		/* Setup regs for texture */
+};
+
+#endif /* _R128_TEXOBJ_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_tris.c b/xc/lib/GL/mesa/src/drv/r128/r128_tris.c
new file mode 100644
index 000000000..1c1b45daf
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_tris.c
@@ -0,0 +1,548 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#include "r128_init.h"
+#include "r128_mesa.h"
+#include "r128_xmesa.h"
+#include "r128_context.h"
+#include "r128_lock.h"
+#include "r128_reg.h"
+#include "r128_cce.h"
+#include "r128_vb.h"
+#include "r128_tris.h"
+#include "r128_state.h"
+
+static triangle_func tri_tab[0x40];    /* only 0x20 actually used */
+static quad_func     quad_tab[0x40];   /* only 0x20 actually used */
+static line_func     line_tab[0x40];   /* less than 0x20 used */
+static points_func   points_tab[0x40]; /* less than 0x20 used */
+
+/* Draw a triangle from the vertices in the vertex buffer */
+void r128DrawTriangleVB(r128ContextPtr r128ctx,
+			r128_vertex *v0,
+			r128_vertex *v1,
+			r128_vertex *v2)
+{
+    r128_vertex *vbptr;
+
+    if (r128ctx->disableVB) {
+	r128DrawTriangle(r128ctx, v0, v1, v2);
+	return;
+    }
+
+    R128CCE_ALLOC_VB_SPACE(r128ctx, vbptr, 3);
+
+    *vbptr++ = *v0;
+    *vbptr++ = *v1;
+    *vbptr++ = *v2;
+}
+
+/* Draw a triangle */
+void r128DrawTriangle(r128ContextPtr r128ctx,
+		      r128_vertex *v0,
+		      r128_vertex *v1,
+		      r128_vertex *v2)
+{
+    LOCK_HARDWARE(r128ctx);
+    BEGIN_CLIP_LOOP(r128ctx);
+
+#if USE_RHW2
+    R128CCE3(R128_CCE_PACKET3_3D_RNDR_GEN_PRIM, 34);
+#else
+    R128CCE3(R128_CCE_PACKET3_3D_RNDR_GEN_PRIM, 31);
+#endif
+    R128CCE(R128_FULL_VERTEX_FORMAT);
+    R128CCE(R128_CCE_VC_CNTL_PRIM_TYPE_TRI_LIST |
+	    R128_CCE_VC_CNTL_PRIM_WALK_RING     |
+	    (3 << R128_CCE_VC_CNTL_NUM_SHIFT));
+
+    R128CCE_SEND_VERTEX(v0);
+    R128CCE_SEND_VERTEX(v1);
+    R128CCE_SEND_VERTEX(v2);
+
+    R128CCE_SUBMIT_PACKETS();
+
+    END_CLIP_LOOP(r128ctx);
+    UNLOCK_HARDWARE(r128ctx);
+}
+
+/* Draw a line from the vertices in the vertex buffer */
+void r128DrawLineVB(r128ContextPtr r128ctx,
+		    r128_vertex *v0, r128_vertex *v1,
+		    float width)
+{
+    float        dx, dy, ix, iy;
+    r128_vertex *vbptr;
+
+    /* FIXME: Use r128's line primitive for width 1 lines */
+    if (r128ctx->disableVB) {
+	r128DrawLine(r128ctx, v0, v1, width);
+	return;
+    }
+
+    R128CCE_ALLOC_VB_SPACE(r128ctx, vbptr, 3);
+
+    dx = v0->x - v1->x;
+    dy = v0->y - v1->y;
+
+    ix = width * 0.5; iy = 0;
+    if (dx*dx > dy*dy) {
+	iy = ix; ix = 0;
+    }
+
+    *vbptr   = *v0; vbptr->x = v0->x - ix; vbptr->y = v0->y - iy;
+    *++vbptr = *v1; vbptr->x = v1->x + ix; vbptr->y = v1->y + iy;
+    *++vbptr = *v0; vbptr->x = v0->x + ix; vbptr->y = v0->y + iy;
+    *++vbptr = *v0; vbptr->x = v0->x - ix; vbptr->y = v0->y - iy;
+    *++vbptr = *v1; vbptr->x = v1->x - ix; vbptr->y = v1->y - iy;
+    *++vbptr = *v1; vbptr->x = v1->x + ix; vbptr->y = v1->y + iy;
+}
+
+/* Draw a line */
+void r128DrawLine(r128ContextPtr r128ctx,
+		  r128_vertex *v0, r128_vertex *v1,
+		  float width)
+{
+    LOCK_HARDWARE(r128ctx);
+    if (width != 1) {
+	float dx, dy, ix, iy;
+
+	dx = v0->x - v1->x;
+	dy = v0->y - v1->y;
+
+	ix = width * 0.5; iy = 0;
+	if (dx*dx > dy*dy) {
+	    iy = ix; ix = 0;
+	}
+
+	BEGIN_CLIP_LOOP(r128ctx);
+
+#if USE_RHW2
+	R128CCE3(R128_CCE_PACKET3_3D_RNDR_GEN_PRIM, 67);
+#else
+	R128CCE3(R128_CCE_PACKET3_3D_RNDR_GEN_PRIM, 61);
+#endif
+	R128CCE(R128_FULL_VERTEX_FORMAT);
+	R128CCE(R128_CCE_VC_CNTL_PRIM_TYPE_TRI_LIST |
+		R128_CCE_VC_CNTL_PRIM_WALK_RING     |
+		(6 << R128_CCE_VC_CNTL_NUM_SHIFT));
+
+	R128CCEF(v0->x - ix);
+	R128CCEF(v0->y - iy);
+	R128CCEF(v0->z);
+	R128CCEF(v0->rhw);
+	R128CCE(*(int *)&v0->dif_argb);
+	R128CCE(*(int *)&v0->spec_frgb);
+	R128CCEF(v0->tu0);
+	R128CCEF(v0->tv0);
+	R128CCEF(v0->tu1);
+	R128CCEF(v0->tv1);
+#if USE_RHW2
+	R128CCEF(v0->rhw2);
+#endif
+
+	R128CCEF(v1->x + ix);
+	R128CCEF(v1->y + iy);
+	R128CCEF(v1->z);
+	R128CCEF(v1->rhw);
+	R128CCE(*(int *)&v1->dif_argb);
+	R128CCE(*(int *)&v1->spec_frgb);
+	R128CCEF(v1->tu0);
+	R128CCEF(v1->tv0);
+	R128CCEF(v1->tu1);
+	R128CCEF(v1->tv1);
+#if USE_RHW2
+	R128CCEF(v1->rhw2);
+#endif
+
+	R128CCEF(v0->x + ix);
+	R128CCEF(v0->y + iy);
+	R128CCEF(v0->z);
+	R128CCEF(v0->rhw);
+	R128CCE(*(int *)&v0->dif_argb);
+	R128CCE(*(int *)&v0->spec_frgb);
+	R128CCEF(v0->tu0);
+	R128CCEF(v0->tv0);
+	R128CCEF(v0->tu1);
+	R128CCEF(v0->tv1);
+#if USE_RHW2
+	R128CCEF(v0->rhw2);
+#endif
+
+	R128CCEF(v0->x - ix);
+	R128CCEF(v0->y - iy);
+	R128CCEF(v0->z);
+	R128CCEF(v0->rhw);
+	R128CCE(*(int *)&v0->dif_argb);
+	R128CCE(*(int *)&v0->spec_frgb);
+	R128CCEF(v0->tu0);
+	R128CCEF(v0->tv0);
+	R128CCEF(v0->tu1);
+	R128CCEF(v0->tv1);
+#if USE_RHW2
+	R128CCEF(v0->rhw2);
+#endif
+
+	R128CCEF(v1->x - ix);
+	R128CCEF(v1->y - iy);
+	R128CCEF(v1->z);
+	R128CCEF(v1->rhw);
+	R128CCE(*(int *)&v1->dif_argb);
+	R128CCE(*(int *)&v1->spec_frgb);
+	R128CCEF(v1->tu0);
+	R128CCEF(v1->tv0);
+	R128CCEF(v1->tu1);
+	R128CCEF(v1->tv1);
+#if USE_RHW2
+	R128CCEF(v1->rhw2);
+#endif
+
+	R128CCEF(v1->x + ix);
+	R128CCEF(v1->y + iy);
+	R128CCEF(v1->z);
+	R128CCEF(v1->rhw);
+	R128CCE(*(int *)&v1->dif_argb);
+	R128CCE(*(int *)&v1->spec_frgb);
+	R128CCEF(v1->tu0);
+	R128CCEF(v1->tv0);
+	R128CCEF(v1->tu1);
+	R128CCEF(v1->tv1);
+#if USE_RHW2
+	R128CCEF(v1->rhw2);
+#endif
+
+	R128CCE_SUBMIT_PACKETS();
+
+	END_CLIP_LOOP(r128ctx);
+    } else {
+	BEGIN_CLIP_LOOP(r128ctx);
+
+#if USE_RHW2
+	R128CCE3(R128_CCE_PACKET3_3D_RNDR_GEN_PRIM, 23);
+#else
+	R128CCE3(R128_CCE_PACKET3_3D_RNDR_GEN_PRIM, 21);
+#endif
+	R128CCE(R128_FULL_VERTEX_FORMAT);
+	R128CCE(R128_CCE_VC_CNTL_PRIM_TYPE_LINE     |
+		R128_CCE_VC_CNTL_PRIM_WALK_RING     |
+		(2 << R128_CCE_VC_CNTL_NUM_SHIFT));
+
+	R128CCE_SEND_VERTEX(v0);
+	R128CCE_SEND_VERTEX(v1);
+
+	R128CCE_SUBMIT_PACKETS();
+
+	END_CLIP_LOOP(r128ctx);
+    }
+    UNLOCK_HARDWARE(r128ctx);
+}
+
+/* Draw a point from the vertices in the vertex buffer */
+void r128DrawPointVB(r128ContextPtr r128ctx,
+		     r128_vertex *v0, float size)
+{
+    r128_vertex *vbptr;
+
+    /* FIXME: Use r128's point primitive for diameter 1 points */
+    if (r128ctx->disableVB) {
+	r128DrawPoint(r128ctx, v0, size);
+	return;
+    }
+
+    R128CCE_ALLOC_VB_SPACE(r128ctx, vbptr, 3);
+
+    *vbptr   = *v0; vbptr->x = v0->x - size; vbptr->y = v0->y - size;
+    *++vbptr = *v0; vbptr->x = v0->x + size; vbptr->y = v0->y - size;
+    *++vbptr = *v0; vbptr->x = v0->x + size; vbptr->y = v0->y + size;
+    *++vbptr = *v0; vbptr->x = v0->x + size; vbptr->y = v0->y + size;
+    *++vbptr = *v0; vbptr->x = v0->x - size; vbptr->y = v0->y + size;
+    *++vbptr = *v0; vbptr->x = v0->x - size; vbptr->y = v0->y - size;
+}
+
+/* Draw a point */
+void r128DrawPoint(r128ContextPtr r128ctx,
+		   r128_vertex *v0, float size)
+{
+    LOCK_HARDWARE(r128ctx);
+    if (size != 0.5) {
+	BEGIN_CLIP_LOOP(r128ctx);
+
+#if USE_RHW2
+	R128CCE3(R128_CCE_PACKET3_3D_RNDR_GEN_PRIM, 67);
+#else
+	R128CCE3(R128_CCE_PACKET3_3D_RNDR_GEN_PRIM, 61);
+#endif
+	R128CCE(R128_FULL_VERTEX_FORMAT);
+	R128CCE(R128_CCE_VC_CNTL_PRIM_TYPE_TRI_LIST |
+		R128_CCE_VC_CNTL_PRIM_WALK_RING     |
+		(6 << R128_CCE_VC_CNTL_NUM_SHIFT));
+
+	R128CCEF(v0->x - size);
+	R128CCEF(v0->y - size);
+	R128CCEF(v0->z);
+	R128CCEF(v0->rhw);
+	R128CCE(*(int *)&v0->dif_argb);
+	R128CCE(*(int *)&v0->spec_frgb);
+	R128CCEF(v0->tu0);
+	R128CCEF(v0->tv0);
+	R128CCEF(v0->tu1);
+	R128CCEF(v0->tv1);
+#if USE_RHW2
+	R128CCEF(v0->rhw2);
+#endif
+
+	R128CCEF(v0->x + size);
+	R128CCEF(v0->y - size);
+	R128CCEF(v0->z);
+	R128CCEF(v0->rhw);
+	R128CCE(*(int *)&v0->dif_argb);
+	R128CCE(*(int *)&v0->spec_frgb);
+	R128CCEF(v0->tu0);
+	R128CCEF(v0->tv0);
+	R128CCEF(v0->tu1);
+	R128CCEF(v0->tv1);
+#if USE_RHW2
+	R128CCEF(v0->rhw2);
+#endif
+
+	R128CCEF(v0->x + size);
+	R128CCEF(v0->y + size);
+	R128CCEF(v0->z);
+	R128CCEF(v0->rhw);
+	R128CCE(*(int *)&v0->dif_argb);
+	R128CCE(*(int *)&v0->spec_frgb);
+	R128CCEF(v0->tu0);
+	R128CCEF(v0->tv0);
+	R128CCEF(v0->tu1);
+	R128CCEF(v0->tv1);
+#if USE_RHW2
+	R128CCEF(v0->rhw2);
+#endif
+
+	R128CCEF(v0->x + size);
+	R128CCEF(v0->y + size);
+	R128CCEF(v0->z);
+	R128CCEF(v0->rhw);
+	R128CCE(*(int *)&v0->dif_argb);
+	R128CCE(*(int *)&v0->spec_frgb);
+	R128CCEF(v0->tu0);
+	R128CCEF(v0->tv0);
+	R128CCEF(v0->tu1);
+	R128CCEF(v0->tv1);
+#if USE_RHW2
+	R128CCEF(v0->rhw2);
+#endif
+
+	R128CCEF(v0->x - size);
+	R128CCEF(v0->y + size);
+	R128CCEF(v0->z);
+	R128CCEF(v0->rhw);
+	R128CCE(*(int *)&v0->dif_argb);
+	R128CCE(*(int *)&v0->spec_frgb);
+	R128CCEF(v0->tu0);
+	R128CCEF(v0->tv0);
+	R128CCEF(v0->tu1);
+	R128CCEF(v0->tv1);
+#if USE_RHW2
+	R128CCEF(v0->rhw2);
+#endif
+
+	R128CCEF(v0->x - size);
+	R128CCEF(v0->y - size);
+	R128CCEF(v0->z);
+	R128CCEF(v0->rhw);
+	R128CCE(*(int *)&v0->dif_argb);
+	R128CCE(*(int *)&v0->spec_frgb);
+	R128CCEF(v0->tu0);
+	R128CCEF(v0->tv0);
+	R128CCEF(v0->tu1);
+	R128CCEF(v0->tv1);
+#if USE_RHW2
+	R128CCEF(v0->rhw2);
+#endif
+
+	R128CCE_SUBMIT_PACKETS();
+
+	END_CLIP_LOOP(r128ctx);
+    } else {
+	BEGIN_CLIP_LOOP(r128ctx);
+
+#if USE_RHW2
+	R128CCE3(R128_CCE_PACKET3_3D_RNDR_GEN_PRIM, 12);
+#else
+	R128CCE3(R128_CCE_PACKET3_3D_RNDR_GEN_PRIM, 11);
+#endif
+	R128CCE(R128_FULL_VERTEX_FORMAT);
+	R128CCE(R128_CCE_VC_CNTL_PRIM_TYPE_POINT    |
+		R128_CCE_VC_CNTL_PRIM_WALK_RING     |
+		(1 << R128_CCE_VC_CNTL_NUM_SHIFT));
+
+	R128CCE_SEND_VERTEX(v0);
+
+	R128CCE_SUBMIT_PACKETS();
+
+	END_CLIP_LOOP(r128ctx);
+    }
+    UNLOCK_HARDWARE(r128ctx);
+}
+
+#define R128_COLOR(to, from)                                                \
+do {                                                                        \
+    (to)[0] = (from)[2];                                                    \
+    (to)[1] = (from)[1];                                                    \
+    (to)[2] = (from)[0];                                                    \
+    (to)[3] = (from)[3];                                                    \
+} while (0)
+
+#define IND (0)
+#define TAG(x) x
+#include "r128_tritmp.h"
+
+#define IND (R128_FLAT_BIT)
+#define TAG(x) x##_flat
+#include "r128_tritmp.h"
+
+#define IND (R128_OFFSET_BIT)
+#define TAG(x) x##_offset
+#include "r128_tritmp.h"
+
+#define IND (R128_OFFSET_BIT | R128_FLAT_BIT)
+#define TAG(x) x##_offset_flat
+#include "r128_tritmp.h"
+
+#define IND (R128_TWOSIDE_BIT)
+#define TAG(x) x##_twoside
+#include "r128_tritmp.h"
+
+#define IND (R128_TWOSIDE_BIT | R128_FLAT_BIT)
+#define TAG(x) x##_twoside_flat
+#include "r128_tritmp.h"
+
+#define IND (R128_TWOSIDE_BIT | R128_OFFSET_BIT)
+#define TAG(x) x##_twoside_offset
+#include "r128_tritmp.h"
+
+#define IND (R128_TWOSIDE_BIT | R128_OFFSET_BIT | R128_FLAT_BIT)
+#define TAG(x) x##_twoside_offset_flat
+#include "r128_tritmp.h"
+
+/* Initialize the table of points, line and triangle drawing functions */
+void r128TriangleFuncsInit(void)
+{
+    init();
+    init_flat();
+    init_offset();
+    init_offset_flat();
+    init_twoside();
+    init_twoside_flat();
+    init_twoside_offset();
+    init_twoside_offset_flat();
+}
+
+/* Setup the Point, Line, Triangle and Quad functions based on the
+   current rendering state.  Wherever possible, use the hardware to
+   render the primitive.  Otherwise, fallback to software rendering. */
+void r128ChooseRenderState(GLcontext *ctx)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+    GLuint         flags   = ctx->TriangleCaps;
+
+    if (r128ctx->Fallback) return;
+
+    if (r128ctx->SWonly) {
+	r128ctx->IndirectTriangles = DD_SW_RASTERIZE;
+	r128ctx->PointsFunc   = NULL;
+	r128ctx->LineFunc     = NULL;
+	r128ctx->TriangleFunc = NULL;
+	r128ctx->QuadFunc     = NULL;
+	return;
+    } else {
+	r128ctx->IndirectTriangles = 0;
+    }
+
+    if (flags) {
+	CARD32 index    = 0;
+	CARD32 shared   = 0;
+	CARD32 fallback = R128_FALLBACK_BIT;
+
+	if (r128ctx->SWfallbackDisable) fallback = 0; /* No fallbacks */
+	if (r128ctx->SWonly) shared = fallback; /* Everything's a fallback */
+
+	if (flags & DD_FLATSHADE)               shared |= R128_FLAT_BIT;
+	if (flags & DD_MULTIDRAW)               shared |= fallback;
+	if (flags & DD_SELECT)                  shared |= R128_FALLBACK_BIT;
+	if (flags & DD_FEEDBACK)                shared |= R128_FALLBACK_BIT;
+
+	/* Setup PointFunc */
+	index = shared;
+	if (flags & DD_POINT_SMOOTH)            index  |= fallback;
+	if (flags & DD_POINT_ATTEN)             index  |= fallback;
+
+	r128ctx->RenderIndex = index;
+	r128ctx->PointsFunc  = points_tab[index];
+	if (index & R128_FALLBACK_BIT)
+	    r128ctx->IndirectTriangles |= DD_POINT_SW_RASTERIZE;
+
+	/* Setup LineFunc */
+	index = shared;
+	if (flags & DD_LINE_SMOOTH)             index  |= fallback;
+	if (flags & DD_LINE_STIPPLE)            index  |= fallback;
+
+	r128ctx->RenderIndex |= index;
+	r128ctx->LineFunc     = line_tab[index];
+	if (index & R128_FALLBACK_BIT)
+	    r128ctx->IndirectTriangles |= DD_LINE_SW_RASTERIZE;
+
+	/* Setup TriangleFunc and QuadFunc */
+	index = shared;
+	if (flags & DD_TRI_SMOOTH)              index  |= fallback;
+	if (flags & DD_TRI_OFFSET)              index  |= R128_OFFSET_BIT;
+	if (flags & DD_TRI_LIGHT_TWOSIDE)       index  |= R128_TWOSIDE_BIT;
+	if (flags & DD_TRI_UNFILLED)            index  |= fallback;
+	if (flags & DD_TRI_STIPPLE)             index  |= fallback;
+
+	r128ctx->RenderIndex  |= index;
+	r128ctx->TriangleFunc  = tri_tab[index];
+	r128ctx->QuadFunc      = quad_tab[index];
+	if (index & R128_FALLBACK_BIT)
+	    r128ctx->IndirectTriangles |= (DD_TRI_SW_RASTERIZE |
+					   DD_QUAD_SW_RASTERIZE);
+    } else if (r128ctx->RenderIndex) {
+	r128ctx->RenderIndex  = 0;
+	r128ctx->PointsFunc   = points_tab[0];
+	r128ctx->LineFunc     = line_tab[0];
+	r128ctx->TriangleFunc = tri_tab[0];
+	r128ctx->QuadFunc     = quad_tab[0];
+    }
+}
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_tris.h b/xc/lib/GL/mesa/src/drv/r128/r128_tris.h
new file mode 100644
index 000000000..0980f2791
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_tris.h
@@ -0,0 +1,76 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_TRIS_H_
+#define _R128_TRIS_H_
+
+#ifdef GLX_DIRECT_RENDERING
+
+#include "r128_vb.h"
+
+#define R128_ANTIALIAS_BIT   0x00   /* ignored for now, no fallback */
+#define R128_FLAT_BIT        0x01
+#define R128_OFFSET_BIT      0x02
+#define R128_TWOSIDE_BIT     0x04
+#define R128_NODRAW_BIT      0x08
+#define R128_FALLBACK_BIT    0x10
+#define R128_FEEDBACK_BIT    0x20
+#define R128_SELECT_BIT      0x40
+#define R128_POINT_PARAM_BIT 0x80   /* not needed? */
+
+extern void r128DrawTriangleVB(r128ContextPtr r128ctx,
+			       r128_vertex *v0,
+			       r128_vertex *v1,
+			       r128_vertex *v2);
+extern void r128DrawLineVB(r128ContextPtr r128ctx,
+			   r128_vertex *tmp0, r128_vertex *tmp1,
+			   float width);
+extern void r128DrawPointVB(r128ContextPtr r128ctx,
+			    r128_vertex *tmp, float size);
+
+extern void r128DrawTriangle(r128ContextPtr r128ctx,
+			     r128_vertex *v0,
+			     r128_vertex *v1,
+			     r128_vertex *v2);
+extern void r128DrawLine(r128ContextPtr r128ctx,
+			 r128_vertex *tmp0, r128_vertex *tmp1,
+			 float width);
+extern void r128DrawPoint(r128ContextPtr r128ctx,
+			  r128_vertex *tmp, float size);
+
+extern void r128ChooseRenderState(GLcontext *ctx);
+extern void r128TriangleFuncsInit(void);
+
+#endif
+#endif /* _R128_TRIS_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_tritmp.h b/xc/lib/GL/mesa/src/drv/r128/r128_tritmp.h
new file mode 100644
index 000000000..833bd832a
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_tritmp.h
@@ -0,0 +1,327 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#if !defined(TAG) || !defined(IND)
+    this is an error
+#endif
+
+/* Draw a single triangle.  Note that the device-dependent vertex data
+   might need to be changed based on the render state. */
+static void TAG(triangle)(GLcontext *ctx,
+			  GLuint e0, GLuint e1, GLuint e2,
+			  GLuint pv)
+{
+    r128ContextPtr        r128ctx   = R128_CONTEXT(ctx);
+    struct vertex_buffer *VB        = ctx->VB;
+    r128VertexPtr         r128verts = R128_DRIVER_DATA(VB)->verts;
+    const r128_vertex    *v0        = &r128verts[e0].v;  
+    const r128_vertex    *v1        = &r128verts[e1].v;  
+    const r128_vertex    *v2        = &r128verts[e2].v;  
+
+#if (IND & R128_OFFSET_BIT)
+    GLfloat offset;
+#endif
+
+#if (IND & (R128_FLAT_BIT | R128_TWOSIDE_BIT))
+    int c0 = *(int *)&r128verts[pv].v.dif_argb; 
+    int c1 = c0;
+    int c2 = c0;
+#endif
+
+#if (IND & R128_OFFSET_BIT)
+    switch (ctx->Visual->DepthBits) {
+    case 16: offset = ctx->Polygon.OffsetUnits / 65536.0;      break;
+    case 24: offset = ctx->Polygon.OffsetUnits / 16777216.0;   break;
+    case 32: offset = ctx->Polygon.OffsetUnits / 4294967296.0; break;
+    default: offset = ctx->Polygon.OffsetUnits / 65536.0;      break;
+    }
+#endif
+
+#if (IND & (R128_TWOSIDE_BIT | R128_OFFSET_BIT)) 
+    {
+	GLfloat ex = v0->x - v2->x;
+	GLfloat ey = v0->y - v2->y;
+	GLfloat fx = v1->x - v2->x;
+	GLfloat fy = v1->y - v2->y;
+	GLfloat c  = ex*fy - ey*fx;
+         
+#if (IND & R128_TWOSIDE_BIT) 
+	{
+	    GLuint  facing        = (c > 0.0) ^ ctx->Polygon.FrontBit;
+	    GLubyte (*vbcolor)[4] = VB->Color[facing]->data;
+	    if (IND & R128_FLAT_BIT) {
+		R128_COLOR((char *)&c0, vbcolor[pv]);
+		c2 = c1 = c0;
+	    } else {
+		R128_COLOR((char *)&c0, vbcolor[e0]);
+		R128_COLOR((char *)&c1, vbcolor[e1]);
+		R128_COLOR((char *)&c2, vbcolor[e2]);
+	    }
+	}
+#endif
+      
+#if (IND & R128_OFFSET_BIT) 
+	{
+	    if (c * c > 1e-16) {
+		GLfloat factor = ctx->Polygon.OffsetFactor;
+		GLfloat ez     = v0->z - v2->z;
+		GLfloat fz     = v1->z - v2->z;
+		GLfloat a      = ey*fz - ez*fy;
+		GLfloat b      = ez*fx - ex*fz;
+		GLfloat ic     = 1.0 / c;
+		GLfloat ac     = a * ic;
+		GLfloat bc     = b * ic;
+		if (ac < 0.0f) ac = -ac;
+		if (bc < 0.0f) bc = -bc;
+		offset += MAX2(ac, bc) * factor;
+	    }
+	}
+#endif
+    }
+#endif  
+
+    LOCK_HARDWARE(r128ctx);
+    if (r128ctx->regs.tex_cntl_c & (R128_TEXMAP_ENABLE |
+				    R128_SEC_TEXMAP_ENABLE)) {
+	BEGIN_CLIP_LOOP(r128ctx);
+
+#if USE_RHW2
+	R128CCE3(R128_CCE_PACKET3_3D_RNDR_GEN_PRIM, 34);
+#else
+	R128CCE3(R128_CCE_PACKET3_3D_RNDR_GEN_PRIM, 31);
+#endif
+	R128CCE(R128_FULL_VERTEX_FORMAT);
+	R128CCE(R128_CCE_VC_CNTL_PRIM_TYPE_TRI_LIST |
+		R128_CCE_VC_CNTL_PRIM_WALK_RING     |
+		(3 << R128_CCE_VC_CNTL_NUM_SHIFT));
+
+	R128CCEF(v0->x);
+	R128CCEF(v0->y);
+#if (IND & R128_OFFSET_BIT)
+	R128CCEF(v0->z + offset);
+#else
+	R128CCEF(v0->z);
+#endif
+	R128CCEF(v0->rhw);
+
+#if (IND & (R128_FLAT_BIT | R128_TWOSIDE_BIT))
+	R128CCE(c0);
+#else
+	R128CCE(*(int *)&v0->dif_argb);
+#endif
+	R128CCE(*(int *)&v0->spec_frgb);
+
+	R128CCEF(v0->tu0);
+	R128CCEF(v0->tv0);
+	R128CCEF(v0->tu1);
+	R128CCEF(v0->tv1);
+#if USE_RHW2
+	R128CCEF(v0->rhw2);
+#endif
+
+	R128CCEF(v1->x);
+	R128CCEF(v1->y);
+#if (IND & R128_OFFSET_BIT)
+	R128CCEF(v1->z + offset);
+#else
+	R128CCEF(v1->z);
+#endif
+	R128CCEF(v1->rhw);
+
+#if (IND & (R128_FLAT_BIT | R128_TWOSIDE_BIT))
+	R128CCE(c1);
+#else
+	R128CCE(*(int *)&v1->dif_argb);
+#endif
+	R128CCE(*(int *)&v1->spec_frgb);
+
+	R128CCEF(v1->tu0);
+	R128CCEF(v1->tv0);
+	R128CCEF(v1->tu1);
+	R128CCEF(v1->tv1);
+#if USE_RHW2
+	R128CCEF(v1->rhw2);
+#endif
+
+	R128CCEF(v2->x);
+	R128CCEF(v2->y);
+#if (IND & R128_OFFSET_BIT)
+	R128CCEF(v2->z + offset);
+#else
+	R128CCEF(v2->z);
+#endif
+	R128CCEF(v2->rhw);
+
+#if (IND & (R128_FLAT_BIT | R128_TWOSIDE_BIT))
+	R128CCE(c2);
+#else
+	R128CCE(*(int *)&v2->dif_argb);
+#endif
+	R128CCE(*(int *)&v2->spec_frgb);
+
+	R128CCEF(v2->tu0);
+	R128CCEF(v2->tv0);
+	R128CCEF(v2->tu1);
+	R128CCEF(v2->tv1);
+#if USE_RHW2
+	R128CCEF(v2->rhw2);
+#endif
+
+	R128CCE_SUBMIT_PACKETS();
+
+	END_CLIP_LOOP(r128ctx);
+    } else {
+	BEGIN_CLIP_LOOP(r128ctx);
+
+	R128CCE3(R128_CCE_PACKET3_3D_RNDR_GEN_PRIM, 13);
+	R128CCE(R128_CCE_VC_FRMT_DIFFUSE_ARGB);
+	R128CCE(R128_CCE_VC_CNTL_PRIM_TYPE_TRI_LIST |
+		R128_CCE_VC_CNTL_PRIM_WALK_RING     |
+		(3 << R128_CCE_VC_CNTL_NUM_SHIFT));
+
+	R128CCEF(v0->x);
+	R128CCEF(v0->y);
+#if (IND & R128_OFFSET_BIT)
+	R128CCEF(v0->z + offset);
+#else
+	R128CCEF(v0->z);
+#endif
+
+#if (IND & (R128_FLAT_BIT | R128_TWOSIDE_BIT))
+	R128CCE(c0);
+#else
+	R128CCE(*(int *)&v0->dif_argb);
+#endif
+
+	R128CCEF(v1->x);
+	R128CCEF(v1->y);
+#if (IND & R128_OFFSET_BIT)
+	R128CCEF(v1->z + offset);
+#else
+	R128CCEF(v1->z);
+#endif
+
+#if (IND & (R128_FLAT_BIT | R128_TWOSIDE_BIT))
+	R128CCE(c1);
+#else
+	R128CCE(*(int *)&v1->dif_argb);
+#endif
+
+	R128CCEF(v2->x);
+	R128CCEF(v2->y);
+#if (IND & R128_OFFSET_BIT)
+	R128CCEF(v2->z + offset);
+#else
+	R128CCEF(v2->z);
+#endif
+
+#if (IND & (R128_FLAT_BIT | R128_TWOSIDE_BIT))
+	R128CCE(c2);
+#else
+	R128CCE(*(int *)&v2->dif_argb);
+#endif
+
+	R128CCE_SUBMIT_PACKETS();
+
+	END_CLIP_LOOP(r128ctx);
+    }
+    UNLOCK_HARDWARE(r128ctx);
+}
+
+static void TAG(quad)(GLcontext *ctx,
+		      GLuint v0, GLuint v1, GLuint v2, GLuint v3,
+		      GLuint pv)
+{
+    TAG(triangle)(ctx, v0, v1, v3, pv);
+    TAG(triangle)(ctx, v1, v2, v3, pv);
+}
+
+#if ((IND & ~R128_FLAT_BIT) == 0)
+
+/* Draw a single line.  Note that the device-dependent vertex data might
+   need to be changed based on the render state. */
+static void TAG(line)(GLcontext *ctx,
+		      GLuint v0, GLuint v1,
+		      GLuint pv)
+{
+    r128ContextPtr  r128ctx   = R128_CONTEXT(ctx);
+    r128VertexPtr   r128verts = R128_DRIVER_DATA(ctx->VB)->verts;
+    r128_vertex     tmp0      = r128verts[v0].v;
+    r128_vertex     tmp1      = r128verts[v1].v;
+    float           width     = ctx->Line.Width;
+
+    if (IND & R128_FLAT_BIT) {
+	*(int *)&tmp1.dif_argb =
+	    *(int *)&tmp0.dif_argb = 
+	    *(int *)&r128verts[pv].v.dif_argb;
+    }
+
+    r128DrawLine(r128ctx, &tmp0, &tmp1, width);
+}
+
+/* Draw a set of points.  Note that the device-dependent vertex data
+   might need to be changed based on the render state. */
+static void TAG(points)(GLcontext *ctx,
+			GLuint first, GLuint last)
+{
+    r128ContextPtr        r128ctx   = R128_CONTEXT(ctx);
+    struct vertex_buffer *VB        = ctx->VB;
+    r128VertexPtr         r128verts = R128_DRIVER_DATA(VB)->verts;
+    GLfloat               size      = ctx->Point.Size * 0.5;
+    int                   i;
+
+    for(i = first; i <= last; i++) {
+	if(VB->ClipMask[i] == 0) {
+            r128_vertex *tmp = &r128verts[i].v;
+            r128DrawPoint(r128ctx, tmp, size);
+	}
+    }
+}
+
+#endif
+
+/* Initialize the table of primitives to render. */
+static void TAG(init)(void)
+{
+    tri_tab[IND]    = TAG(triangle);
+    quad_tab[IND]   = TAG(quad);
+
+#if ((IND & ~R128_FLAT_BIT) == 0)
+    line_tab[IND]   = TAG(line);
+    points_tab[IND] = TAG(points);
+#endif
+}
+
+#undef IND
+#undef TAG
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_vb.c b/xc/lib/GL/mesa/src/drv/r128/r128_vb.c
new file mode 100644
index 000000000..f14ec21bd
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_vb.c
@@ -0,0 +1,470 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#include "r128_init.h"
+#include "r128_mesa.h"
+#include "r128_xmesa.h"
+#include "r128_context.h"
+#include "r128_lock.h"
+#include "r128_reg.h"
+#include "r128_cce.h"
+#include "r128_state.h"
+#include "r128_vb.h"
+
+#include "stages.h"
+
+#define TEX0                                                            \
+do {                                                                    \
+    v->v.tu0 = tc0[i][0];                                               \
+    v->v.tv0 = tc0[i][1];                                               \
+} while (0)
+
+#define TEX1                                                            \
+do {                                                                    \
+    v->v.tu1 = tc1[i][0];                                               \
+    v->v.tv1 = tc1[i][1];                                               \
+} while (0)
+
+#define SPC                                                             \
+do {                                                                    \
+    GLubyte *spec = &(VB->Spec[0][i][0]);                               \
+    v->v.spec_frgb.r = spec[0];                                         \
+    v->v.spec_frgb.g = spec[1];                                         \
+    v->v.spec_frgb.b = spec[2];                                         \
+} while (0)
+
+#define FOG                                                             \
+do {                                                                    \
+    GLubyte *spec = &(VB->Spec[0][i][0]);                               \
+    v->v.spec_frgb.a = spec[3];                                         \
+} while (0)
+
+#define COL                                                             \
+do {                                                                    \
+    GLubyte *col = &(VB->Color[0]->data[i][0]);                         \
+    v->v.dif_argb.a = col[3];                                           \
+    v->v.dif_argb.r = col[0];                                           \
+    v->v.dif_argb.g = col[1];                                           \
+    v->v.dif_argb.b = col[2];                                           \
+} while (0)
+
+#if 1
+/* FIXME: These are handled by the Rage 128 */
+#define TEX0_4
+#define TEX1_4
+#else
+#define TEX0_4                                                          \
+do {                                                                    \
+    if (VB->TexCoordPtr[0]->size == 4) {                                \
+	GLfloat (*tc)[4] = VB->TexCoordPtr[0]->data;                    \
+	v = &(R128_DRIVER_DATA(VB)->verts[start]);                      \
+	for (i = start; i < end; i++, v++)   {                          \
+	    float oow = 1.0 / tc[i][3];                                 \
+	    v->v.rhw *= tc[i][3];                                       \
+	    v->v.tu0 *= oow;                                            \
+	    v->v.tv0 *= oow;                                            \
+	}                                                               \
+    }                                                                   \
+} while (0)
+
+#if USE_RHW2
+#define TEX1_4                                                          \
+do {                                                                    \
+    if (VB->TexCoordPtr[1]->size == 4) {                                \
+	GLfloat (*tc)[4] = VB->TexCoordPtr[1]->data;                    \
+	v = &(R128_DRIVER_DATA(VB)->verts[start]);                      \
+	for (i = start; i < end; i++, v++)   {                          \
+	    float oow = 1.0 / tc[i][3];                                 \
+	    v->v.rhw2 *= tc[i][3];                                      \
+	    v->v.tu1  *= oow;                                           \
+	    v->v.tv1  *= oow;                                           \
+	}                                                               \
+    }                                                                   \
+} while (0)
+#else
+#define TEX1_4
+#endif
+#endif
+
+#if USE_RHW2
+#define COORD                                                           \
+do {                                                                    \
+    GLfloat *win = VB->Win.data[i];                                     \
+    v->v.x =                win[0];                                     \
+    v->v.y =   r128height - win[1];                                     \
+    v->v.z =        scale * win[2];                                     \
+    v->v.rhw = v->v.rhw2 =  win[3];                                     \
+} while (0)
+#else
+#define COORD                                                           \
+do {                                                                    \
+    GLfloat *win = VB->Win.data[i];                                     \
+    v->v.x =                win[0];                                     \
+    v->v.y =   r128height - win[1];                                     \
+    v->v.z =        scale * win[2];                                     \
+    v->v.rhw =              win[3];                                     \
+} while (0)
+#endif
+
+#define NOP
+
+/* Setup the r128 vertex buffer entries */
+#define SETUPFUNC(name,win,col,tex0,tex1,tex0_4,tex1_4,spec,fog)        \
+static void name(struct vertex_buffer *VB, GLuint start, GLuint end)    \
+{                                                                       \
+    r128ContextPtr        r128ctx    = R128_CONTEXT(VB->ctx);           \
+    __DRIdrawablePrivate *dPriv      = r128ctx->driDrawable;            \
+    r128VertexPtr         v;                                            \
+    GLfloat               (*tc0)[4];                                    \
+    GLfloat               (*tc1)[4];                                    \
+    GLfloat               r128height = dPriv->h;                        \
+    GLfloat               scale;                                        \
+    int                   i;                                            \
+                                                                        \
+    (void) r128height; (void) r128ctx;                                  \
+                                                                        \
+    gl_import_client_data(VB, VB->ctx->RenderFlags,                     \
+			  (VB->ClipOrMask                               \
+			   ? VEC_WRITABLE | VEC_GOOD_STRIDE             \
+			   : VEC_GOOD_STRIDE));                         \
+                                                                        \
+    switch (VB->ctx->Visual->DepthBits) {                               \
+    case 16: scale = 1.0 / 65536.0;      break;                         \
+    case 24: scale = 1.0 / 16777216.0;   break;                         \
+    case 32: scale = 1.0 / 4294967296.0; break;                         \
+    default: scale = 1.0 / 65536.0;      break;                         \
+    }                                                                   \
+                                                                        \
+    tc0 = VB->TexCoordPtr[0]->data;                                     \
+    tc1 = VB->TexCoordPtr[1]->data;                                     \
+                                                                        \
+    v = &(R128_DRIVER_DATA(VB)->verts[start]);                          \
+                                                                        \
+    if (VB->ClipOrMask == 0)                                            \
+	for (i = start; i < end; i++, v++) {                            \
+	    win;                                                        \
+	    col;                                                        \
+	    spec;                                                       \
+	    fog;                                                        \
+	    tex0;                                                       \
+	    tex1;                                                       \
+	}                                                               \
+    else                                                                \
+	for (i = start; i < end; i++, v++) {                            \
+	    if (VB->ClipMask[i] == 0) {                                 \
+		win;                                                    \
+		spec;                                                   \
+		fog;                                                    \
+		tex0;                                                   \
+		tex1;                                                   \
+	    }                                                           \
+	    col;                                                        \
+	}                                                               \
+    tex0_4;                                                             \
+    tex1_4;                                                             \
+}
+
+
+SETUPFUNC(rs_wt0,       COORD, NOP, TEX0, NOP,  TEX0_4, NOP,    NOP, NOP)
+SETUPFUNC(rs_wt1,       COORD, NOP, NOP,  TEX1, NOP,    TEX1_4, NOP, NOP)
+SETUPFUNC(rs_wt0t1,     COORD, NOP, TEX0, TEX1, TEX0_4, TEX1_4, NOP, NOP)
+SETUPFUNC(rs_wft0,      COORD, NOP, TEX0, NOP,  TEX0_4, NOP,    NOP, FOG)
+SETUPFUNC(rs_wft1,      COORD, NOP, NOP,  TEX1, NOP,    TEX1_4, NOP, FOG)
+SETUPFUNC(rs_wft0t1,    COORD, NOP, TEX0, TEX1, TEX0_4, TEX1_4, NOP, FOG)
+SETUPFUNC(rs_wg,        COORD, COL, NOP,  NOP,  NOP,    NOP,    NOP, NOP)
+SETUPFUNC(rs_wgs,       COORD, COL, NOP,  NOP,  NOP,    NOP,    SPC, NOP)
+SETUPFUNC(rs_wgt0,      COORD, COL, TEX0, NOP,  TEX0_4, NOP,    NOP, NOP)
+SETUPFUNC(rs_wgt1,      COORD, COL, NOP,  TEX1, NOP,    TEX1_4, NOP, NOP)
+SETUPFUNC(rs_wgt0t1,    COORD, COL, TEX0, TEX1, TEX0_4, TEX1_4, NOP, NOP)
+SETUPFUNC(rs_wgst0,     COORD, COL, TEX0, NOP,  TEX0_4, NOP,    SPC, NOP)
+SETUPFUNC(rs_wgst1,     COORD, COL, NOP,  TEX1, NOP,    TEX1_4, SPC, NOP)
+SETUPFUNC(rs_wgst0t1,   COORD, COL, TEX0, TEX1, TEX0_4, TEX1_4, SPC, NOP)
+SETUPFUNC(rs_wgf,       COORD, COL, NOP,  NOP,  NOP,    NOP,    NOP, FOG)
+SETUPFUNC(rs_wgfs,      COORD, COL, NOP,  NOP,  NOP,    NOP,    SPC, FOG)
+SETUPFUNC(rs_wgft0,     COORD, COL, TEX0, NOP,  TEX0_4, NOP,    NOP, FOG)
+SETUPFUNC(rs_wgft1,     COORD, COL, NOP,  TEX1, NOP,    TEX1_4, NOP, FOG)
+SETUPFUNC(rs_wgft0t1,   COORD, COL, TEX0, TEX1, TEX0_4, TEX1_4, NOP, FOG)
+SETUPFUNC(rs_wgfst0,    COORD, COL, TEX0, NOP,  TEX0_4, NOP,    SPC, FOG)
+SETUPFUNC(rs_wgfst1  ,  COORD, COL, NOP,  TEX1, NOP,    TEX1_4, SPC, FOG)
+SETUPFUNC(rs_wgfst0t1,  COORD, COL, TEX0, TEX1, TEX0_4, TEX1_4, SPC, FOG)
+
+SETUPFUNC(rs_t0,        NOP,   NOP, TEX0, NOP,  TEX0_4, NOP,    NOP, NOP)
+SETUPFUNC(rs_t1,        NOP,   NOP, NOP,  TEX1, NOP,    TEX1_4, NOP, NOP)
+SETUPFUNC(rs_t0t1,      NOP,   NOP, TEX0, TEX1, TEX0_4, TEX1_4, NOP, NOP)
+SETUPFUNC(rs_f,         NOP,   NOP, NOP,  NOP,  NOP,    NOP,    NOP, FOG)
+SETUPFUNC(rs_ft0,       NOP,   NOP, TEX0, NOP,  TEX0_4, NOP,    NOP, FOG)
+SETUPFUNC(rs_ft1,       NOP,   NOP, NOP,  TEX1, NOP,    TEX1_4, NOP, FOG)
+SETUPFUNC(rs_ft0t1,     NOP,   NOP, TEX0, TEX1, TEX0_4, TEX1_4, NOP, FOG)
+SETUPFUNC(rs_g,         NOP,   COL, NOP,  NOP,  NOP,    NOP,    NOP, NOP)
+SETUPFUNC(rs_gs,        NOP,   COL, NOP,  NOP,  NOP,    NOP,    SPC, NOP)
+SETUPFUNC(rs_gt0,       NOP,   COL, TEX0, NOP,  TEX0_4, NOP,    NOP, NOP)
+SETUPFUNC(rs_gt1,       NOP,   COL, NOP,  TEX1, NOP,    TEX1_4, NOP, NOP)
+SETUPFUNC(rs_gt0t1,     NOP,   COL, TEX0, TEX1, TEX0_4, TEX1_4, NOP, NOP)
+SETUPFUNC(rs_gst0,      NOP,   COL, TEX0, NOP,  TEX0_4, NOP,    SPC, NOP)
+SETUPFUNC(rs_gst1,      NOP,   COL, NOP,  TEX1, NOP,    TEX1_4, SPC, NOP)
+SETUPFUNC(rs_gst0t1,    NOP,   COL, TEX0, TEX1, TEX0_4, TEX1_4, SPC, NOP)
+SETUPFUNC(rs_gf,        NOP,   COL, NOP,  NOP,  NOP,    NOP,    NOP, FOG)
+SETUPFUNC(rs_gfs,       NOP,   COL, NOP,  NOP,  NOP,    NOP,    SPC, FOG)
+SETUPFUNC(rs_gft0,      NOP,   COL, TEX0, NOP,  TEX0_4, NOP,    NOP, FOG)
+SETUPFUNC(rs_gft1,      NOP,   COL, NOP,  TEX1, NOP,    TEX1_4, NOP, FOG)
+SETUPFUNC(rs_gft0t1,    NOP,   COL, TEX0, TEX1, TEX0_4, TEX1_4, NOP, FOG)
+SETUPFUNC(rs_gfst0,     NOP,   COL, TEX0, NOP,  TEX0_4, NOP,    SPC, FOG)
+SETUPFUNC(rs_gfst1,     NOP,   COL, NOP,  TEX1, NOP,    TEX1_4, SPC, FOG)
+SETUPFUNC(rs_gfst0t1,   NOP,   COL, TEX0, TEX1, TEX0_4, TEX1_4, SPC, FOG)
+
+static void rs_invalid(struct vertex_buffer *VB, GLuint start, GLuint end)
+{
+    fprintf(stderr, "r128RasterSetup(): invalid setup function\n");
+}
+
+typedef void (*setupFunc)(struct vertex_buffer *, GLuint, GLuint);
+static setupFunc setup_func[0x80];
+
+/* Initialize the table of vertex buffer setup functions */
+void r128SetupInit(void)
+{
+    int i;
+
+    for (i = 0; i < 0x80; i++) setup_func[i] = rs_invalid;
+
+    /* Funcs to build vertices from scratch */
+    setup_func[R128_WIN_BIT|R128_TEX0_BIT] = rs_wt0;
+    setup_func[R128_WIN_BIT|R128_TEX1_BIT] = rs_wt1;
+    setup_func[R128_WIN_BIT|R128_TEX0_BIT|R128_TEX1_BIT] = rs_wt0t1;
+    setup_func[R128_WIN_BIT|R128_FOG_BIT|R128_TEX0_BIT] = rs_wft0;
+    setup_func[R128_WIN_BIT|R128_FOG_BIT|R128_TEX1_BIT] = rs_wft1;
+    setup_func[R128_WIN_BIT|R128_FOG_BIT|R128_TEX0_BIT|R128_TEX1_BIT] = rs_wft0t1;
+    setup_func[R128_WIN_BIT|R128_RGBA_BIT] = rs_wg;
+    setup_func[R128_WIN_BIT|R128_RGBA_BIT|R128_SPEC_BIT] = rs_wgs;
+    setup_func[R128_WIN_BIT|R128_RGBA_BIT|R128_TEX0_BIT] = rs_wgt0;
+    setup_func[R128_WIN_BIT|R128_RGBA_BIT|R128_TEX1_BIT] = rs_wgt1;
+    setup_func[R128_WIN_BIT|R128_RGBA_BIT|R128_TEX0_BIT|R128_TEX1_BIT] = rs_wgt0t1;
+    setup_func[R128_WIN_BIT|R128_RGBA_BIT|R128_SPEC_BIT|R128_TEX0_BIT] = rs_wgst0;
+    setup_func[R128_WIN_BIT|R128_RGBA_BIT|R128_SPEC_BIT|R128_TEX1_BIT] = rs_wgst1;
+    setup_func[R128_WIN_BIT|R128_RGBA_BIT|R128_SPEC_BIT|R128_TEX0_BIT|R128_TEX1_BIT] = rs_wgst0t1;
+    setup_func[R128_WIN_BIT|R128_RGBA_BIT|R128_FOG_BIT] = rs_wgf;
+    setup_func[R128_WIN_BIT|R128_RGBA_BIT|R128_FOG_BIT|R128_SPEC_BIT] = rs_wgfs;
+    setup_func[R128_WIN_BIT|R128_RGBA_BIT|R128_FOG_BIT|R128_TEX0_BIT] = rs_wgft0;
+    setup_func[R128_WIN_BIT|R128_RGBA_BIT|R128_FOG_BIT|R128_TEX1_BIT] = rs_wgft1;
+    setup_func[R128_WIN_BIT|R128_RGBA_BIT|R128_FOG_BIT|R128_TEX0_BIT|R128_TEX1_BIT] = rs_wgft0t1;
+    setup_func[R128_WIN_BIT|R128_RGBA_BIT|R128_FOG_BIT|R128_SPEC_BIT|R128_TEX0_BIT] = rs_wgfst0;
+    setup_func[R128_WIN_BIT|R128_RGBA_BIT|R128_FOG_BIT|R128_SPEC_BIT|R128_TEX1_BIT] = rs_wgfst1;
+    setup_func[R128_WIN_BIT|R128_RGBA_BIT|R128_FOG_BIT|R128_SPEC_BIT|R128_TEX0_BIT|R128_TEX1_BIT] = rs_wgfst0t1;
+
+    /* Funcs to repair vertices */
+    setup_func[R128_TEX0_BIT] = rs_t0;
+    setup_func[R128_TEX1_BIT] = rs_t1;
+    setup_func[R128_TEX0_BIT|R128_TEX1_BIT] = rs_t0t1;
+    setup_func[R128_FOG_BIT] = rs_f;
+    setup_func[R128_FOG_BIT|R128_TEX0_BIT] = rs_ft0;
+    setup_func[R128_FOG_BIT|R128_TEX1_BIT] = rs_ft1;
+    setup_func[R128_FOG_BIT|R128_TEX0_BIT|R128_TEX1_BIT] = rs_ft0t1;
+    setup_func[R128_RGBA_BIT] = rs_g;
+    setup_func[R128_RGBA_BIT|R128_SPEC_BIT] = rs_gs;
+    setup_func[R128_RGBA_BIT|R128_TEX0_BIT] = rs_gt0;
+    setup_func[R128_RGBA_BIT|R128_TEX1_BIT] = rs_gt1;
+    setup_func[R128_RGBA_BIT|R128_TEX0_BIT|R128_TEX1_BIT] = rs_gt0t1;
+    setup_func[R128_RGBA_BIT|R128_SPEC_BIT|R128_TEX0_BIT] = rs_gst0;
+    setup_func[R128_RGBA_BIT|R128_SPEC_BIT|R128_TEX1_BIT] = rs_gst1;
+    setup_func[R128_RGBA_BIT|R128_SPEC_BIT|R128_TEX0_BIT|R128_TEX1_BIT] = rs_gst0t1;
+    setup_func[R128_RGBA_BIT|R128_FOG_BIT] = rs_gf;
+    setup_func[R128_RGBA_BIT|R128_FOG_BIT|R128_SPEC_BIT] = rs_gfs;
+    setup_func[R128_RGBA_BIT|R128_FOG_BIT|R128_TEX0_BIT] = rs_gft0;
+    setup_func[R128_RGBA_BIT|R128_FOG_BIT|R128_TEX1_BIT] = rs_gft1;
+    setup_func[R128_RGBA_BIT|R128_FOG_BIT|R128_TEX0_BIT|R128_TEX1_BIT] = rs_gft0t1;
+    setup_func[R128_RGBA_BIT|R128_FOG_BIT|R128_SPEC_BIT|R128_TEX0_BIT] = rs_gfst0;
+    setup_func[R128_RGBA_BIT|R128_FOG_BIT|R128_SPEC_BIT|R128_TEX1_BIT] = rs_gfst1;
+    setup_func[R128_RGBA_BIT|R128_FOG_BIT|R128_SPEC_BIT|R128_TEX0_BIT|R128_TEX1_BIT] = rs_gfst0t1;
+}
+
+/* Initialize the vertex buffer setup functions based on the current
+   rendering state */
+void r128ChooseRasterSetupFunc(GLcontext *ctx)
+{
+    r128ContextPtr r128ctx   = R128_CONTEXT(ctx);
+    int            funcIndex = R128_WIN_BIT | R128_RGBA_BIT;
+
+    if (ctx->Texture.Enabled & 0xf) {
+	if (ctx->Texture.Unit[0].EnvMode == GL_REPLACE)
+	    funcIndex &= ~R128_RGBA_BIT;
+	funcIndex |= R128_TEX0_BIT;
+    }
+
+    if (ctx->Texture.Enabled & 0xf0)
+	funcIndex |= R128_TEX1_BIT;
+
+    /* FIXME: Verify this works properly */
+    if (ctx->Light.Model.ColorControl == GL_SEPARATE_SPECULAR_COLOR)
+	funcIndex |= R128_SPEC_BIT;
+
+    if (ctx->FogMode == FOG_FRAGMENT)
+	funcIndex |= R128_FOG_BIT;
+
+    r128ctx->SetupIndex = funcIndex;
+    ctx->Driver.RasterSetup = setup_func[funcIndex];
+}
+
+/* Check to see if any updates of the vertex buffer entries are needed */
+void r128CheckPartialRasterSetup(GLcontext *ctx,
+				 struct gl_pipeline_stage *s)
+{
+    r128ContextPtr r128ctx = R128_CONTEXT(ctx);
+    int            tmp     = r128ctx->SetupDone;
+
+    s->type = 0;
+    r128ctx->SetupDone = GL_FALSE;
+
+    if ((ctx->Array.Summary & VERT_OBJ_ANY) == 0) return;
+    if (ctx->IndirectTriangles) return;
+
+    r128ctx->SetupDone = tmp;
+}
+
+/* Update the vertex buffer entries, if necessary */
+void r128PartialRasterSetup(struct vertex_buffer *VB)
+{
+    r128ContextPtr r128ctx   = R128_CONTEXT(VB->ctx);
+    int            new       = VB->pipeline->new_outputs;
+    int            available = VB->pipeline->outputs;
+    int            index     = 0;
+
+    if (new & VERT_WIN) {
+	new = available;
+	index |= R128_WIN_BIT | R128_FOG_BIT;
+    }
+
+    if (new & VERT_RGBA)      index |= R128_RGBA_BIT | R128_SPEC_BIT;
+    if (new & VERT_TEX0_ANY)  index |= R128_TEX0_BIT;
+    if (new & VERT_TEX1_ANY)  index |= R128_TEX1_BIT;
+#if 0
+    /* FIXME */
+    if (new & VERT_FOG_COORD) index |= R128_FOG_BIT;
+#endif
+
+    r128ctx->SetupDone &= ~index;
+    index              &= r128ctx->SetupIndex;
+    r128ctx->SetupDone |= index;
+
+    if (index) setup_func[index & ~R128_ALPHA_BIT](VB, VB->Start, VB->Count);
+}
+
+/* Perform the raster setup for the fast path, if using CVA */
+void r128DoRasterSetup(struct vertex_buffer *VB)
+{
+    GLcontext *ctx = VB->ctx;
+
+    if (VB->Type == VB_CVA_PRECALC)   r128PartialRasterSetup(VB);
+    else if (ctx->Driver.RasterSetup) ctx->Driver.RasterSetup(VB,
+							      VB->CopyStart,
+							      VB->Count);
+}
+
+/* Resize an existing vertex buffer */
+void r128ResizeVB(struct vertex_buffer *VB, GLuint size)
+{
+    r128VertexBufferPtr r128vb = R128_DRIVER_DATA(VB);
+
+    while (r128vb->size < size)
+	r128vb->size *= 2;
+
+    free(r128vb->vert_store);
+    r128vb->vert_store = malloc(sizeof(r128Vertex) * r128vb->size + 31);
+    if (!r128vb->vert_store) {
+	fprintf(stderr, "Cannot allocate vertex store!  Exiting...\n");
+	exit(1);
+    }
+
+    r128vb->verts = (r128VertexPtr)(((CARD32)r128vb->vert_store + 31) & ~31);
+
+    gl_vector1ui_free(&r128vb->clipped_elements);
+    gl_vector1ui_alloc(&r128vb->clipped_elements,
+		       VEC_WRITABLE, r128vb->size, 32);
+    if (!r128vb->clipped_elements.start) {
+	fprintf(stderr, "Cannot allocate clipped elements!  Exiting...\n");
+	exit(1);
+    }
+
+    free(VB->ClipMask);
+    VB->ClipMask = (GLubyte *)malloc(sizeof(GLubyte) * r128vb->size);
+    if (!VB->ClipMask) {
+	fprintf(stderr, "Cannot allocate clipmask!  Exiting...\n");
+	exit(1);
+    }
+}
+
+/* Create a new device-dependent vertex buffer */
+void r128DDRegisterVB(struct vertex_buffer *VB)
+{
+    r128VertexBufferPtr r128vb;
+
+    r128vb = (r128VertexBufferPtr)calloc(1, sizeof(*r128vb));
+
+    r128vb->size = VB->Size * 2;
+    r128vb->vert_store = malloc(sizeof(r128Vertex) * r128vb->size + 31);
+    if (!r128vb->vert_store) {
+	fprintf(stderr, "Cannot allocate vertex store!  Exiting...\n");
+	exit(1);
+    }
+
+    r128vb->verts = (r128VertexPtr)(((CARD32)r128vb->vert_store + 31) & ~31);
+
+    gl_vector1ui_alloc(&r128vb->clipped_elements,
+		       VEC_WRITABLE, r128vb->size, 32);
+    if (!r128vb->clipped_elements.start) {
+	fprintf(stderr, "Cannot allocate clipped elements!  Exiting...\n");
+	exit(1);
+    }
+
+    free(VB->ClipMask);
+    VB->ClipMask = (GLubyte *)malloc(sizeof(GLubyte) * r128vb->size);
+    if (!VB->ClipMask) {
+	fprintf(stderr, "Cannot allocate clipmask!  Exiting...\n");
+	exit(1);
+    }
+
+    VB->driver_data = r128vb;
+}
+
+/* Destroy a device-dependent vertex buffer */
+void r128DDUnregisterVB(struct vertex_buffer *VB)
+{
+    r128VertexBufferPtr r128vb = R128_DRIVER_DATA(VB);
+
+    if (r128vb) {
+	if (r128vb->vert_store)     free(r128vb->vert_store);
+	gl_vector1ui_free(&r128vb->clipped_elements);
+	free(r128vb);
+	VB->driver_data = 0;
+    }
+}
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_vb.h b/xc/lib/GL/mesa/src/drv/r128/r128_vb.h
new file mode 100644
index 000000000..4f7c11d7f
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_vb.h
@@ -0,0 +1,150 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_VB_H_
+#define _R128_VB_H_
+
+#ifdef GLX_DIRECT_RENDERING
+
+#define USE_RHW2 0
+
+/* FIXME: This is endian-specific */
+typedef struct {
+    GLubyte b;
+    GLubyte g;
+    GLubyte r;
+    GLubyte a;
+} r128Color;
+
+typedef struct {
+    GLfloat   x, y, z;		/* Coordinates in screen space */
+    GLfloat   rhw;		/* Reciprocal homogeneous w */
+    r128Color dif_argb;		/* Diffuse color */
+    r128Color spec_frgb;	/* Specular color (alpha is fog) */
+    GLfloat   tu0, tv0;		/* Texture 0 coordinates */
+    GLfloat   tu1, tv1;		/* Texture 1 coordinates */
+#if USE_RHW2
+    GLfloat   rhw2;		/* Reciprocal homogeneous w */
+#endif
+} r128_vertex;
+
+#if USE_RHW2
+/* Format of vertices in r128_vertex struct */
+#define R128_FULL_VERTEX_FORMAT                                            \
+    R128_CCE_VC_FRMT_RHW          |                                        \
+    R128_CCE_VC_FRMT_DIFFUSE_ARGB |                                        \
+    R128_CCE_VC_FRMT_SPEC_FRGB    |                                        \
+    R128_CCE_VC_FRMT_S_T          |                                        \
+    R128_CCE_VC_FRMT_S2_T2        |                                        \
+    R128_CCE_VC_FRMT_RHW2
+
+/* Send a single vertex to the ring buffer */
+#define R128CCE_SEND_VERTEX(v)                                              \
+    do {                                                                    \
+	R128CCEF((v)->x);                                                   \
+	R128CCEF((v)->y);                                                   \
+	R128CCEF((v)->z);                                                   \
+	R128CCEF((v)->rhw);                                                 \
+	R128CCE(*(int *)&(v)->dif_argb);                                    \
+	R128CCE(*(int *)&(v)->spec_frgb);                                   \
+	R128CCEF((v)->tu0);                                                 \
+	R128CCEF((v)->tv0);                                                 \
+	R128CCEF((v)->tu1);                                                 \
+	R128CCEF((v)->tv1);                                                 \
+	R128CCEF((v)->rhw2);                                                \
+    } while (0)
+
+#else /* !USE_RHW2 */
+
+/* Format of vertices in r128_vertex struct */
+#define R128_FULL_VERTEX_FORMAT                                            \
+    R128_CCE_VC_FRMT_RHW          |                                        \
+    R128_CCE_VC_FRMT_DIFFUSE_ARGB |                                        \
+    R128_CCE_VC_FRMT_SPEC_FRGB    |                                        \
+    R128_CCE_VC_FRMT_S_T          |                                        \
+    R128_CCE_VC_FRMT_S2_T2
+
+/* Send a single vertex to the ring buffer */
+#define R128CCE_SEND_VERTEX(v)                                              \
+    do {                                                                    \
+	R128CCEF((v)->x);                                                   \
+	R128CCEF((v)->y);                                                   \
+	R128CCEF((v)->z);                                                   \
+	R128CCEF((v)->rhw);                                                 \
+	R128CCE(*(int *)&(v)->dif_argb);                                    \
+	R128CCE(*(int *)&(v)->spec_frgb);                                   \
+	R128CCEF((v)->tu0);                                                 \
+	R128CCEF((v)->tv0);                                                 \
+	R128CCEF((v)->tu1);                                                 \
+	R128CCEF((v)->tv1);                                                 \
+    } while (0)
+#endif
+
+/* FIXME: We currently only have assembly for 16-stride vertices */
+typedef union {
+    r128_vertex v;
+    float       f[16];
+} r128Vertex, *r128VertexPtr;
+
+/* Vertex buffer for use when on the fast path */
+typedef struct {
+    GLuint         size;		/* Number of vertices in store */
+    void          *vert_store;		/* Storage for vertex buffer */
+    r128VertexPtr  verts;		/* Aligned start of verts in storage */
+    int            last_vert;		/* Index of last vertex used */
+    GLvector1ui    clipped_elements;	/* List of clipped elements */
+} *r128VertexBufferPtr;
+
+#define R128_DRIVER_DATA(vb) ((r128VertexBufferPtr)((vb)->driver_data))
+
+#define R128_SPEC_BIT       0x01
+#define R128_FOG_BIT        0x02
+#define R128_ALPHA_BIT      0x04 /* GL_BLEND, not used */
+#define R128_TEX1_BIT       0x08
+#define R128_TEX0_BIT       0x10        
+#define R128_RGBA_BIT       0x20
+#define R128_WIN_BIT        0x40
+
+extern void         r128SetupInit(void);
+extern void         r128ChooseRasterSetupFunc(GLcontext *ctx);
+extern void         r128CheckPartialRasterSetup(GLcontext *ctx,
+						struct gl_pipeline_stage *s);
+extern void         r128PartialRasterSetup(struct vertex_buffer *VB);
+extern void         r128DoRasterSetup(struct vertex_buffer *VB);
+extern void         r128ResizeVB(struct vertex_buffer *VB, GLuint size);
+extern void         r128DDRegisterVB(struct vertex_buffer *VB);
+extern void         r128DDUnregisterVB(struct vertex_buffer *VB);
+
+#endif
+#endif /* _R128_VB_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_xmesa.c b/xc/lib/GL/mesa/src/drv/r128/r128_xmesa.c
new file mode 100644
index 000000000..60d528813
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_xmesa.c
@@ -0,0 +1,257 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#ifdef GLX_DIRECT_RENDERING
+
+/* r128 Mesa driver includes */
+#include "r128_init.h"
+#include "r128_context.h"
+#include "r128_xmesa.h"
+#include "r128_state.h"
+#include "r128_tex.h"
+#include "r128_swap.h"
+
+/* Mesa src includes */
+#include "context.h"
+#include "simple_list.h"
+#include "mmath.h"
+
+#ifndef R128_DEBUG_FLAGS
+int R128_DEBUG_FLAGS = (0
+#if 0
+			| DEBUG_ALWAYS_SYNC
+			| DEBUG_VERBOSE_CCE
+			| DEBUG_VERBOSE_OUTREG
+			| DEBUG_VERBOSE_MSG
+			| DEBUG_NO_OUTRING
+			| DEBUG_NO_OUTREG
+			| DEBUG_VERBOSE_API
+			| DEBUG_VERBOSE_2D
+			| DEBUG_VERBOSE_DRI
+			| DEBUG_VALIDATE_RING
+			| DEBUG_VERBOSE_IOCTL
+#endif
+    );
+#endif
+
+#if DEBUG_LOCKING
+char *prevLockFile = NULL;
+int   prevLockLine = 0;
+#endif
+
+static r128ContextPtr r128Context = NULL;
+
+
+/* Initialize the driver specific screen private data */
+GLboolean XMesaInitDriver(__DRIscreenPrivate *sPriv)
+{
+    sPriv->private = (void *)r128CreateScreen(sPriv);
+    if (!sPriv->private) {
+	r128DestroyScreen(sPriv);
+	return GL_FALSE;
+    }
+
+    return GL_TRUE;
+}
+
+/* Reset the driver specific screen private data */
+void XMesaResetDriver(__DRIscreenPrivate *sPriv)
+{
+    r128DestroyScreen(sPriv);
+}
+
+/* Create and initialize the Mesa and driver specific visual data */
+GLvisual *XMesaCreateVisual(Display *dpy,
+			    __DRIscreenPrivate *driScrnPriv,
+			    const XVisualInfo *visinfo,
+			    const __GLXvisualConfig *config)
+{
+    /* Drivers may change the args to _mesa_create_visual() in order to
+     * setup special visuals.
+     */
+    return _mesa_create_visual(config->rgba,
+			       config->doubleBuffer,
+			       config->stereo,
+			       _mesa_bitcount(visinfo->red_mask),
+			       _mesa_bitcount(visinfo->green_mask),
+			       _mesa_bitcount(visinfo->blue_mask),
+			       config->alphaSize,
+			       0, /* index bits */
+			       config->depthSize,
+			       config->stencilSize,
+			       config->accumRedSize,
+			       config->accumGreenSize,
+			       config->accumBlueSize,
+			       config->accumAlphaSize,
+			       0 /* num samples */);
+}
+
+/* Create and initialize the Mesa and driver specific context data */
+GLboolean XMesaCreateContext(Display *dpy, GLvisual *mesaVis,
+			     __DRIcontextPrivate *driContextPriv)
+{
+    return r128CreateContext(dpy, mesaVis, driContextPriv);
+}
+
+/* Destroy the Mesa and driver specific context data */
+void XMesaDestroyContext(__DRIcontextPrivate *driContextPriv)
+{
+    r128ContextPtr r128ctx = (r128ContextPtr)driContextPriv->driverPrivate;
+
+    if (r128ctx == (void *)r128Context) r128Context = NULL;
+    r128DestroyContext(r128ctx);
+}
+
+/* Create and initialize the Mesa and driver specific pixmap buffer data */
+GLframebuffer *XMesaCreateWindowBuffer(Display *dpy,
+				       __DRIscreenPrivate *driScrnPriv,
+				       __DRIdrawablePrivate *driDrawPriv,
+				       GLvisual *mesaVis)
+{
+   return gl_create_framebuffer(mesaVis,
+                                GL_FALSE,  /* software depth buffer? */
+                                mesaVis->StencilBits > 0,
+                                mesaVis->AccumRedBits > 0,
+                                mesaVis->AlphaBits > 0
+                                );
+}
+
+/* Create and initialize the Mesa and driver specific pixmap buffer data */
+GLframebuffer *XMesaCreatePixmapBuffer( Display *dpy,
+                                        __DRIscreenPrivate *driScrnPriv,
+                                        __DRIdrawablePrivate *driDrawPriv,
+                                        GLvisual *mesaVis)
+{
+#if 0
+    /* Different drivers may have different combinations of hardware and
+     * software ancillary buffers.
+     */
+    return gl_create_framebuffer(mesaVis,
+				 GL_FALSE,  /* software depth buffer? */
+				 mesaVis->StencilBits > 0,
+				 mesaVis->AccumRedBits > 0,
+				 mesaVis->AlphaBits > 0);
+#else
+    return NULL;  /* not implemented yet */
+#endif
+}
+
+/* Copy the back color buffer to the front color buffer */
+void XMesaSwapBuffers(__DRIdrawablePrivate *driDrawPriv)
+{
+    /* FIXME: This assumes buffer is currently bound to a context.  This
+       needs to be able to swap buffers when not currently bound.  Also,
+       this needs to swap according to buffer, and NOT according to
+       context! */
+    if (r128Context == NULL) return;
+
+    /* Only swap buffers when a back buffer exists */
+    if (R128_MESACTX(r128Context)->Visual->DBflag) {
+	FLUSH_VB(R128_MESACTX(r128Context), "swap buffers");
+	r128SwapBuffers(r128Context);
+    }
+}
+
+/* Force the context `c' to be the current context and associate with it
+   buffer `b' */
+GLboolean XMesaMakeCurrent(__DRIcontextPrivate *driContextPriv,
+			   __DRIdrawablePrivate *driDrawPriv,
+			   __DRIdrawablePrivate *driReadPriv)
+{
+    if (driContextPriv) {
+	r128ContextPtr r128ctx = (r128ContextPtr)driContextPriv->driverPrivate;
+
+	if (r128Context &&
+	    r128ctx == (void *)r128Context &&
+	    driDrawPriv == R128_DRIDRAWABLE(r128Context))
+	    return GL_TRUE;
+
+	r128Context = r128MakeCurrent(r128Context, r128ctx, driDrawPriv);
+
+	gl_make_current2(R128_MESACTX(r128Context),
+			 driDrawPriv->mesaBuffer, driReadPriv->mesaBuffer);
+
+	if (!R128_MESACTX(r128Context)->Viewport.Width) {
+	    gl_Viewport(R128_MESACTX(r128Context), 0, 0,
+			driDrawPriv->w, driDrawPriv->h);
+	}
+    } else {
+	gl_make_current(0,0);
+	r128Context = NULL;
+    }
+
+    return GL_TRUE;
+}
+
+/* Force the context `c' to be unbound from its buffer */
+GLboolean XMesaUnbindContext(__DRIcontextPrivate *driContextPriv)
+{
+    return GL_TRUE;
+}
+
+/* Update the hardware state.  This is called if another context has
+   grabbed the hardware lock, which includes the X server.  This
+   function also updates the driver's window state after the X server
+   moves, resizes or restacks a window -- the change will be reflected
+   in the drawable position and clip rects.  Since the X server grabs
+   the hardware lock when it changes the window state, this routine will
+   automatically be called after such a change. */
+/* NOTE: This routine is only called while holding the hardware lock. */
+void XMesaUpdateState(__DRIcontextPrivate *driContextPriv)
+{
+    r128ContextPtr        r128ctx = driContextPriv->driverPrivate;
+    __DRIscreenPrivate   *sPriv   = R128_DRISCREEN(r128ctx);
+    __DRIdrawablePrivate *dPriv   = R128_DRIDRAWABLE(r128ctx);
+    int                   stamp   = dPriv->lastStamp;
+
+    /* The window might have moved, so we might need to get new clip
+       rects.
+
+       NOTE: This releases and regrabs the hw lock to allow the X server
+       to respond to the DRI protocol request for new drawable info.
+       Since the hardware state depends on having the latest drawable
+       clip rects, all state checking must be done _after_ this call. */
+    XMESA_VALIDATE_DRAWABLE_INFO(r128ctx->display, sPriv, dPriv);
+
+    r128UpdateState(r128ctx, (stamp != dPriv->lastStamp));
+}
+
+/* This function is called by libGL.so as soon as libGL.so is loaded.
+ * This is where we'd register new extension functions with the dispatcher.
+ */
+void __driRegisterExtensions(void)
+{
+}
+
+#endif
diff --git a/xc/lib/GL/mesa/src/drv/r128/r128_xmesa.h b/xc/lib/GL/mesa/src/drv/r128/r128_xmesa.h
new file mode 100644
index 000000000..ca6fc127b
--- /dev/null
+++ b/xc/lib/GL/mesa/src/drv/r128/r128_xmesa.h
@@ -0,0 +1,44 @@
+/* $XFree86$ */
+/**************************************************************************
+
+Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, 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
+on 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
+ATI, PRECISION INSIGHT AND/OR THEIR 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.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <kevin@precisioninsight.com>
+ *
+ */
+
+#ifndef _R128_XMESA_H_
+#define _R128_XMESA_H_
+
+#ifdef GLX_DIRECT_RENDERING
+
+extern void XMesaUpdateState(__DRIcontextPrivate *driContextPriv);
+extern void __driRegisterExtensions(void);
+
+#endif
+#endif /* _R128_XMESA_H_ */
diff --git a/xc/lib/GL/mesa/src/drv/tdfx/Imakefile b/xc/lib/GL/mesa/src/drv/tdfx/Imakefile
index c159a8e76..4a6a8e0ff 100644
--- a/xc/lib/GL/mesa/src/drv/tdfx/Imakefile
+++ b/xc/lib/GL/mesa/src/drv/tdfx/Imakefile
@@ -79,7 +79,8 @@ LinkSourceFile(fxvs_tmp.h, ../../../../../../extras/Mesa/src/FX)
 		fxglidew.o fxpipeline.o fxrender.o fxsanity.o fxsetup.o \
 		fxtexman.o fxtrifuncs.o fxvsetup.o
 
-     MESASRCS = ../../accum.c \
+     MESASRCS = ../../aatriangle.c \
+		../../accum.c \
 		../../alpha.c \
 		../../alphabuf.c \
 		../../attrib.c \
@@ -109,7 +110,7 @@ LinkSourceFile(fxvs_tmp.h, ../../../../../../extras/Mesa/src/FX)
 		../../glthread.c \
 		../../hash.c \
 		../../image.c \
-		../../imaging.o \
+		../../imaging.c \
 		../../light.c \
 		../../lines.c \
 		../../logic.c \
@@ -120,6 +121,7 @@ LinkSourceFile(fxvs_tmp.h, ../../../../../../extras/Mesa/src/FX)
 		../../pb.c \
 		../../pipeline.c \
 		../../pixel.c \
+		../../pixeltex.c \
 		../../points.c \
 		../../polygon.c \
 		../../quads.c \
@@ -136,6 +138,7 @@ LinkSourceFile(fxvs_tmp.h, ../../../../../../extras/Mesa/src/FX)
 		../../texobj.c \
 		../../texstate.c \
 		../../texture.c \
+		../../texutil.c \
 		../../translate.c \
 		../../triangle.c \
 		../../varray.c \
@@ -152,7 +155,8 @@ LinkSourceFile(fxvs_tmp.h, ../../../../../../extras/Mesa/src/FX)
 		../../zoom.c \
 		../../X86/common_x86.c
 
-     MESAOBJS = ../../accum.o \
+     MESAOBJS = ../../aatriangle.o \
+		../../accum.o \
 		../../alpha.o \
 		../../alphabuf.o \
 		../../attrib.o \
@@ -191,6 +195,7 @@ LinkSourceFile(fxvs_tmp.h, ../../../../../../extras/Mesa/src/FX)
 		../../pb.o \
 		../../pipeline.o \
 		../../pixel.o \
+		../../pixeltex.o \
 		../../points.o \
 		../../polygon.o \
 		../../quads.o \
@@ -207,6 +212,7 @@ LinkSourceFile(fxvs_tmp.h, ../../../../../../extras/Mesa/src/FX)
 		../../texobj.o \
 		../../texstate.o \
 		../../texture.o \
+		../../texutil.o \
 		../../translate.o \
 		../../triangle.o \
 		../../varray.o \
@@ -239,8 +245,7 @@ LinkSourceFile(fxvs_tmp.h, ../../../../../../extras/Mesa/src/FX)
 
      MMX_OBJS = ../../X86/mmx_blend.o
 
-XCOMM Disabling 3Dnow code for the time being.
-#if 0
+#if MesaUse3DNow
    3DNOW_SRCS = ../../X86/3dnow.c \
 		../../X86/3dnow_norm_raw.S \
 		../../X86/3dnow_xform_masked1.S \
@@ -271,10 +276,14 @@ XCOMM Disabling 3Dnow code for the time being.
       ASMSRCS = $(X86_SRCS) $(MMX_SRCS) $(3DNOW_SRCS)
       ASMOBJS = $(X86_OBJS) $(MMX_OBJS) $(3DNOW_OBJS)
 
-	 SRCS = $(DRISRCS) $(DRMSRCS) $(TDFXSRCS) $(MESASRCS) $(ASMSRCS)
-	 OBJS = $(DRIOBJS) $(DRMOBJS) $(TDFXOBJS) $(MESAOBJS) $(ASMOBJS)
+         SRCS = $(DRISRCS) $(DRMSRCS) $(TDFXSRCS) $(MESASRCS) $(ASMSRCS)
+         OBJS = $(DRIOBJS) $(DRMOBJS) $(TDFXOBJS) $(MESAOBJS) $(ASMOBJS)
+
+REQUIREDLIBS = -lglide3x -lm
+#if !GlxBuiltInTdfx
+REQUIREDLIBS += -L../../../.. -lGL
+#endif
 
-REQUIREDLIBS += -lglide3x
 
 #if !GlxUseBuiltInDRIDriver
 #undef DoNormalLib NormalLibGlx
diff --git a/xc/lib/GL/mesa/src/drv/tdfx/tdfx_init.h b/xc/lib/GL/mesa/src/drv/tdfx/tdfx_init.h
index 3759512ee..251992d00 100644
--- a/xc/lib/GL/mesa/src/drv/tdfx/tdfx_init.h
+++ b/xc/lib/GL/mesa/src/drv/tdfx/tdfx_init.h
@@ -41,9 +41,8 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "dri_tmm.h"
 #include "dri_mesaint.h"
 #include "dri_mesa.h"
-#include "types.h"
-#include "xmesaP.h"
-/*  #include "fxdrv.h" */
+#include "dri_xmesaapi.h"
+
 
 typedef struct {
   drmHandle handle;
@@ -98,7 +97,8 @@ typedef struct tfxMesaContext tdfxContextPrivate;
 
 extern GLboolean tdfxMapAllRegions(__DRIscreenPrivate *driScrnPriv);
 extern void tdfxUnmapAllRegions(__DRIscreenPrivate *driScrnPriv);
-extern GLboolean tdfxInitHW(XMesaContext c);
+extern GLboolean tdfxInitHW(__DRIdrawablePrivate *driDrawPrivate,
+                            tdfxContextPrivate *cPriv);
 
 extern void XMesaWindowMoved(void);
 extern void XMesaUpdateState(int windowMoved);
@@ -119,8 +119,8 @@ extern void grDRIImportFifo(int fifoPtr, int fifoRead);
 extern void grDRIInvalidateAll(void);
 extern void grDRIResetSAREA(void);
 
-extern XMesaContext gCC;
-extern tdfxContextPrivate *gCCPriv;
+extern __DRIcontextPrivate *gCC;
+/*000extern tdfxContextPrivate *gCCPriv;*/
 
 /* You can turn this on to find locking conflicts.
 #define DEBUG_LOCKING 
@@ -167,7 +167,7 @@ extern int prevLockLine;
 #define LOCK_HARDWARE() \
   do { \
     char __ret=0; \
-    __DRIdrawablePrivate *dPriv = gCC->driContextPriv->driDrawablePriv; \
+    __DRIdrawablePrivate *dPriv = gCC->driDrawablePriv; \
     __DRIscreenPrivate *sPriv = dPriv->driScreenPriv; \
     DEBUG_CHECK_LOCK(); \
     DRM_CAS(&sPriv->pSAREA->lock, dPriv->driContextPriv->hHWContext, \
@@ -186,7 +186,7 @@ extern int prevLockLine;
 /* Unlock the hardware using the global current context */
 #define UNLOCK_HARDWARE() \
   do { \
-    __DRIdrawablePrivate *dPriv = gCC->driContextPriv->driDrawablePriv; \
+    __DRIdrawablePrivate *dPriv = gCC->driDrawablePriv; \
     __DRIscreenPrivate *sPriv = dPriv->driScreenPriv; \
     XMesaSetSAREA(); \
     DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock,  \
@@ -202,24 +202,26 @@ extern int prevLockLine;
   so it is not self contained and doesn't have the nice single 
   statement semantics of most macros
 */
-#define BEGIN_CLIP_LOOP()	\
-  do {				\
-    __DRIdrawablePrivate *dPriv = gCC->driContextPriv->driDrawablePriv; \
-    int _nc; \
-    LOCK_HARDWARE(); \
-    _nc = dPriv->numClipRects; \
-    while (_nc--) { \
-      if (gCCPriv->needClip) { \
-        gCCPriv->clipMinX=dPriv->pClipRects[_nc].x1; \
-        gCCPriv->clipMaxX=dPriv->pClipRects[_nc].x2; \
-        gCCPriv->clipMinY=dPriv->pClipRects[_nc].y1; \
-        gCCPriv->clipMaxY=dPriv->pClipRects[_nc].y2; \
-        fxSetScissorValues(gCCPriv->glCtx); \
+#define BEGIN_CLIP_LOOP()					\
+  do {								\
+    __DRIdrawablePrivate *dPriv = gCC->driDrawablePriv;		\
+    int _nc;							\
+    LOCK_HARDWARE();						\
+    _nc = dPriv->numClipRects;					\
+    while (_nc--) {						\
+      tdfxContextPrivate *gCCPriv =				\
+                    (tdfxContextPrivate *) gCC->driverPrivate;	\
+      if (gCCPriv->needClip) {					\
+        gCCPriv->clipMinX=dPriv->pClipRects[_nc].x1;		\
+        gCCPriv->clipMaxX=dPriv->pClipRects[_nc].x2;		\
+        gCCPriv->clipMinY=dPriv->pClipRects[_nc].y1;		\
+        gCCPriv->clipMaxY=dPriv->pClipRects[_nc].y2;		\
+        fxSetScissorValues(gCCPriv->glCtx);			\
       }
 
-#define END_CLIP_LOOP() \
-    } \
-    UNLOCK_HARDWARE(); \
+#define END_CLIP_LOOP()		\
+    }				\
+    UNLOCK_HARDWARE();		\
   } while (0)
 
 #endif
diff --git a/xc/lib/GL/mesa/src/drv/tdfx/tdfx_inithw.c b/xc/lib/GL/mesa/src/drv/tdfx/tdfx_inithw.c
index 044f30a67..0bafa3190 100644
--- a/xc/lib/GL/mesa/src/drv/tdfx/tdfx_inithw.c
+++ b/xc/lib/GL/mesa/src/drv/tdfx/tdfx_inithw.c
@@ -36,13 +36,11 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "tdfx_init.h"
 #include <glide.h>
 
-GLboolean tdfxInitHW(XMesaContext c)
+GLboolean tdfxInitHW(__DRIdrawablePrivate *driDrawPriv,
+                     tdfxContextPrivate *fxMesa)
 {
   /* KW: Would be nice to make one of these a member of the other.
    */
-  tdfxContextPrivate *cPriv = (tdfxContextPrivate*)c->private;
-  tdfxContextPrivate *fxMesa = cPriv;
-  __DRIdrawablePrivate *driDrawPriv = c->driContextPriv->driDrawablePriv;
   __DRIscreenPrivate *driScrnPriv = driDrawPriv->driScreenPriv;
   tdfxScreenPrivate *sPriv = (tdfxScreenPrivate*)driScrnPriv->private;
 
@@ -50,10 +48,10 @@ GLboolean tdfxInitHW(XMesaContext c)
   fprintf(stderr, "Debug locking enabled\n");
 #endif
     
-  if (cPriv->initDone) return GL_TRUE;
+  if (fxMesa->initDone) return GL_TRUE;
 
-  cPriv->width=driDrawPriv->w;
-  cPriv->height=driDrawPriv->h;
+  fxMesa->width=driDrawPriv->w;
+  fxMesa->height=driDrawPriv->h;
 
   /* We have to use a light lock here, because we can't do any glide
      operations yet. No use of FX_* functions in this function. */
@@ -89,7 +87,7 @@ GLboolean tdfxInitHW(XMesaContext c)
 
   fxInitPixelTables(fxMesa, GL_FALSE); /* Load tables of pixel colors */
 
-  cPriv->initDone=GL_TRUE;
+  fxMesa->initDone=GL_TRUE;
   return GL_TRUE;
 
 }
diff --git a/xc/lib/GL/mesa/src/drv/tdfx/tdfx_xmesa.c b/xc/lib/GL/mesa/src/drv/tdfx/tdfx_xmesa.c
index a3973a8ab..505f4bb99 100644
--- a/xc/lib/GL/mesa/src/drv/tdfx/tdfx_xmesa.c
+++ b/xc/lib/GL/mesa/src/drv/tdfx/tdfx_xmesa.c
@@ -28,7 +28,7 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 /*
  * Authors:
  *   Daryll Strauss <daryll@precisioninsight.com>
- *
+ *   Brian E. Paul <brian@precisioninsight.com>
  */
 
 #ifdef GLX_DIRECT_RENDERING
@@ -37,23 +37,13 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include <glide.h>
 #include "tdfx_init.h"
 #include "context.h"
-#include "vbxform.h"
 #include "matrix.h"
+#include "mmath.h"
+#include "vbxform.h"
 
-XMesaContext        nullCC  = 0;
-XMesaContext        gCC     = 0;
-tdfxContextPrivate *gCCPriv = 0;
 
-static int count_bits(unsigned int n)
-{
-  int bits = 0;
+__DRIcontextPrivate *gCC = 0;
 
-  while (n > 0) {
-    if (n & 1) bits++;
-    n >>= 1;
-  }
-  return bits;
-}
 
 GLboolean XMesaInitDriver(__DRIscreenPrivate *driScrnPriv)
 {
@@ -81,99 +71,44 @@ void XMesaResetDriver(__DRIscreenPrivate *driScrnPriv)
   Xfree(driScrnPriv->private);
 }
 
-XMesaVisual XMesaCreateVisual(XMesaDisplay *display,
-			      XMesaVisualInfo visinfo,
-			      GLboolean rgb_flag,
-			      GLboolean alpha_flag,
-			      GLboolean db_flag,
-			      GLboolean stereo_flag,
-			      GLboolean ximage_flag,
-			      GLint depth_size,
-			      GLint stencil_size,
-                              GLint accum_red_size,
-                              GLint accum_green_size,
-                              GLint accum_blue_size,
-                              GLint accum_alpha_size,
-                              GLint num_samples,
-                              GLint level,
-                              GLint visualCaveat )
+GLvisual *XMesaCreateVisual(Display *dpy,
+                            __DRIscreenPrivate *driScrnPriv,
+                            const XVisualInfo *visinfo,
+                            const __GLXvisualConfig *config)
 {
-   XMesaVisual v;
-   GLint redBits, greenBits, blueBits, alphaBits;
-
-   /* Only RGB visuals are supported on the TDFX boards */
-   if (!rgb_flag) return 0;
-
-   v = (XMesaVisual)Xmalloc(sizeof(struct xmesa_visual));
-   if (!v) return 0;
-
-   v->visinfo = (XVisualInfo *)Xmalloc(sizeof(*visinfo));
-   if(!v->visinfo) {
-      Xfree(v);
-      return 0;
-   }
-   memcpy(v->visinfo, visinfo, sizeof(*visinfo));
-
-   redBits = count_bits(visinfo->red_mask);
-   greenBits = count_bits(visinfo->green_mask);
-   blueBits = count_bits(visinfo->blue_mask);
-   alphaBits = (redBits == 8 && greenBits == 8 && blueBits == 8 && alpha_flag) ? 8 : 0;
-
-   v->display = display;
-   v->level = level;
-   v->VisualCaveat = visualCaveat;
-   v->gl_visual = _mesa_create_visual(rgb_flag,
-                                      GL_FALSE,  /* alpha flag */
-                                      db_flag,
-                                      stereo_flag,
-                                      redBits, greenBits, blueBits, alphaBits,
-                                      0, /* index bits */
-                                      depth_size,
-                                      stencil_size,
-                                      accum_red_size,
-                                      accum_green_size,
-                                      accum_blue_size,
-                                      accum_alpha_size,
-                                      0 /* num samples */ );
-   if (!v->gl_visual) {
-      Xfree(v->visinfo);
-      Xfree(v);
-      return NULL;
-   }
-   else
-      return v;
+   /* Drivers may change the args to _mesa_create_visual() in order to
+    * setup special visuals.
+    */
+   return _mesa_create_visual( config->rgba,
+                               config->doubleBuffer,
+                               config->stereo,
+                               _mesa_bitcount(visinfo->red_mask),
+                               _mesa_bitcount(visinfo->green_mask),
+                               _mesa_bitcount(visinfo->blue_mask),
+                               config->alphaSize,
+                               0, /* index bits */
+                               config->depthSize,
+                               config->stencilSize,
+                               config->accumRedSize,
+                               config->accumGreenSize,
+                               config->accumBlueSize,
+                               config->accumAlphaSize,
+                               0 /* num samples */ );
 }
 
-void XMesaDestroyVisual(XMesaVisual v)
-{
-  _mesa_destroy_visual(v->gl_visual);
-  Xfree(v->visinfo);
-  Xfree(v);
-}
 
-XMesaContext XMesaCreateContext(XMesaVisual v, XMesaContext share_list,
-					__DRIcontextPrivate *driContextPriv)
+GLboolean XMesaCreateContext(Display *dpy, GLvisual *mesaVis,
+                             __DRIcontextPrivate *driContextPriv)
 {
-  XMesaContext c;
   tdfxContextPrivate *cPriv;
   __DRIscreenPrivate *driScrnPriv = driContextPriv->driScreenPriv;
   tdfxScreenPrivate *sPriv = (tdfxScreenPrivate *)driScrnPriv->private;
   TDFXSAREAPriv *saPriv;
-  GLcontext *shareCtx;
   /*int **fifoPtr;*/
 
-  c = (XMesaContext)Xmalloc(sizeof(struct xmesa_context));
-  if (!c) return 0;
-
-  c->driContextPriv = driContextPriv;
-  c->xm_visual = v;
-  c->xm_buffer = 0; /* Set by MakeCurrent */
-  c->display = v->display;
-
   cPriv = (tdfxContextPrivate *)Xmalloc(sizeof(tdfxContextPrivate));
   if (!cPriv) {
-    Xfree(c);
-    return NULL;
+    return GL_FALSE;
   }
 
   cPriv->hHWContext = driContextPriv->hHWContext;
@@ -181,26 +116,12 @@ XMesaContext XMesaCreateContext(XMesaVisual v, XMesaContext share_list,
   /* deviceID = 0x05 = Voodoo3 */
   /* deviceID = 0x09 = Voodoo5 (and Voodoo5?) */
   cPriv->haveHwStencil = sPriv->deviceID == 0x9 && sPriv->cpp == 4;
-  c->private = (void *)cPriv;
-
-  cPriv->glVis=v->gl_visual;
-  cPriv->glBuffer=gl_create_framebuffer(v->gl_visual,
-                        GL_FALSE,  /* software depth buffer? */
-                        v->gl_visual->StencilBits > 0 && !cPriv->haveHwStencil,
-                        v->gl_visual->AccumRedBits > 0,
-                        GL_FALSE   /* software alpha channel */
-                      );
 
   cPriv->screen_width=sPriv->width;
   cPriv->screen_height=sPriv->height;
-
   cPriv->new_state = ~0;
 
-  if (share_list)
-    shareCtx=((tdfxContextPrivate*)(share_list->private))->glCtx;
-  else
-    shareCtx=0;
-  cPriv->glCtx=gl_create_context(v->gl_visual, shareCtx, (void*)cPriv, GL_TRUE);
+  cPriv->glCtx = driContextPriv->mesaContext;
   cPriv->initDone=GL_FALSE;
 
   saPriv=(TDFXSAREAPriv*)((char*)driScrnPriv->pSAREA+sizeof(XF86DRISAREARec));
@@ -210,41 +131,63 @@ XMesaContext XMesaCreateContext(XMesaVisual v, XMesaContext share_list,
 	    sPriv->backOffset, sPriv->depthOffset, sPriv->textureOffset, 
 	    sPriv->textureSize, &saPriv->fifoPtr, &saPriv->fifoRead);
 
-  return c;
+  driContextPriv->driverPrivate = (void *) cPriv;
+
+  return GL_TRUE;
 }
 
-void XMesaDestroyContext(XMesaContext c)
+void XMesaDestroyContext(__DRIcontextPrivate *driContextPriv)
 {
   tdfxContextPrivate *cPriv;
 
-  cPriv=(tdfxContextPrivate*)c->private;
+  cPriv = (tdfxContextPrivate *) driContextPriv->driverPrivate;
   if (cPriv) {
-    gl_destroy_context(cPriv->glCtx);
-    gl_destroy_framebuffer(cPriv->glBuffer);
+     /* XXX XFree the tdfxContextPrivate struct? */
+    driContextPriv->driverPrivate = NULL;
   }
-  if (c==gCC) {
-    gCC=0;
-    gCCPriv=0;
+
+  if (driContextPriv == gCC) {
+    gCC = 0;
   }
 }
 
-XMesaBuffer XMesaCreateWindowBuffer(XMesaVisual v, XMesaWindow w,
-				    __DRIdrawablePrivate *driDrawPriv)
-{
-  return (XMesaBuffer)1;
-}
 
-XMesaBuffer XMesaCreatePixmapBuffer(XMesaVisual v, XMesaPixmap p,
-				    XMesaColormap c, __DRIdrawablePrivate *driDrawPriv)
+GLframebuffer *XMesaCreateWindowBuffer( Display *dpy,
+                                        __DRIscreenPrivate *driScrnPriv,
+                                        __DRIdrawablePrivate *driDrawPriv,
+                                        GLvisual *mesaVis)
 {
-  return (XMesaBuffer)1;
+   return gl_create_framebuffer(mesaVis,
+                                GL_FALSE,  /* software depth buffer? */
+                                mesaVis->StencilBits > 0,
+                                mesaVis->AccumRedBits > 0,
+                                mesaVis->AlphaBits > 0
+                                );
 }
 
-void XMesaDestroyBuffer(XMesaBuffer b)
+
+GLframebuffer *XMesaCreatePixmapBuffer( Display *dpy,
+                                        __DRIscreenPrivate *driScrnPriv,
+                                        __DRIdrawablePrivate *driDrawPriv,
+                                        GLvisual *mesaVis)
 {
+#if 0
+   /* Different drivers may have different combinations of hardware and
+    * software ancillary buffers.
+    */
+   return gl_create_framebuffer(mesaVis,
+                                GL_FALSE,  /* software depth buffer? */
+                                mesaVis->StencilBits > 0,
+                                mesaVis->AccumRedBits > 0,
+                                mesaVis->AlphaBits > 0
+                                );
+#else
+   return NULL;  /* not implemented yet */
+#endif
 }
 
-void XMesaSwapBuffers(XMesaBuffer b)
+
+void XMesaSwapBuffers(__DRIdrawablePrivate *driDrawPriv)
 {
   FxI32 result;
 #ifdef STATS
@@ -252,15 +195,20 @@ void XMesaSwapBuffers(XMesaBuffer b)
   extern int texSwaps;
   static int prevStalls=0;
 #endif
+  tdfxContextPrivate *gCCPriv;
+
   /*
   ** NOT_DONE: This assumes buffer is currently bound to a context.
   ** This needs to be able to swap buffers when not currently bound.
   */
-  if (gCC == NULL || gCCPriv == NULL) return;
+  if (gCC == NULL)
+     return;
+
+  gCCPriv = (tdfxContextPrivate *) gCC->driverPrivate;
 
   FLUSH_VB( gCCPriv->glCtx, "swap buffers" );
 
-  if (gCCPriv->glVis->DBflag) {
+  if (gCC->mesaContext->Visual->DBflag) {
 #ifdef STATS
     stalls=grFifoGetStalls();
     if (stalls!=prevStalls) {
@@ -277,56 +225,62 @@ void XMesaSwapBuffers(XMesaBuffer b)
       result=FX_grGetInteger(FX_PENDING_BUFFERSWAPS);
     } while (result>gCCPriv->maxPendingSwapBuffers);
     gCCPriv->stats.swapBuffer++;
-  } else {
-    fprintf(stderr, "No double buffer\n");
   }
 }
 
-GLboolean XMesaUnbindContext(XMesaContext c)
+GLboolean XMesaUnbindContext(__DRIcontextPrivate *driContextPriv)
 {
-  if (c && c==gCC && gCCPriv) FX_grGlideGetState((GrState*)gCCPriv->state);
-  return GL_TRUE;
+   if (driContextPriv && driContextPriv == gCC) {
+      tdfxContextPrivate *gCCPriv;
+      gCCPriv = (tdfxContextPrivate *) gCC->driverPrivate;
+      FX_grGlideGetState((GrState*)gCCPriv->state);
+   }
+   return GL_TRUE;
 }
 
-GLboolean XMesaMakeCurrent(XMesaContext c, XMesaBuffer b)
+GLboolean XMesaMakeCurrent(__DRIcontextPrivate *driContextPriv,
+                           __DRIdrawablePrivate *driDrawPriv,
+                           __DRIdrawablePrivate *driReadPriv)
 {
-  __DRIdrawablePrivate *driDrawPriv;
+  if (driContextPriv) {
+    tdfxContextPrivate *gCCPriv;
 
-  if (c) {
-    if (c==gCC) return GL_TRUE;
-    gCC     = c;
-    gCCPriv = (tdfxContextPrivate *)c->private;
+    gCC = driContextPriv;
+    gCCPriv = (tdfxContextPrivate *)driContextPriv->driverPrivate;
 
-    driDrawPriv = gCC->driContextPriv->driDrawablePriv;
     if (!gCCPriv->initDone) {
-      if (!tdfxInitHW(c)) return GL_FALSE;
+      if (!tdfxInitHW(driDrawPriv, gCCPriv))
+        return GL_FALSE;
       gCCPriv->width=0;
       XMesaWindowMoved();
       FX_grGlideGetState((GrState*)gCCPriv->state);
-    } else {
+    }
+    else {
       FX_grSstSelect(gCCPriv->board);
       FX_grGlideSetState((GrState*)gCCPriv->state);
       XMesaWindowMoved();
     }
 
-    gl_make_current(gCCPriv->glCtx, gCCPriv->glBuffer);
+    gl_make_current2(gCCPriv->glCtx, driDrawPriv->mesaBuffer, driReadPriv->mesaBuffer);
+
     fxSetupDDPointers(gCCPriv->glCtx);
     if (!gCCPriv->glCtx->Viewport.Width)
       gl_Viewport(gCCPriv->glCtx, 0, 0, driDrawPriv->w, driDrawPriv->h);
-  } else {
+  }
+  else {
     gl_make_current(0,0);
-    gCC     = NULL;
-    gCCPriv = NULL;
+    gCC = NULL;
   }
   return GL_TRUE;
 }
 
-void
-XMesaWindowMoved() {
-  __DRIdrawablePrivate *dPriv = gCC->driContextPriv->driDrawablePriv;
-  GLcontext *ctx;
-  
-  ctx=gCCPriv->glCtx;
+
+void XMesaWindowMoved(void)
+{
+  __DRIdrawablePrivate *dPriv = gCC->driDrawablePriv;
+  tdfxContextPrivate *gCCPriv = (tdfxContextPrivate *) gCC->driverPrivate;
+  GLcontext *ctx = gCCPriv->glCtx;
+
   grDRIPosition(dPriv->x, dPriv->y, dPriv->w, dPriv->h,
 		dPriv->numClipRects, dPriv->pClipRects);
   gCCPriv->numClipRects=dPriv->numClipRects;
@@ -362,9 +316,11 @@ XMesaWindowMoved() {
 }
 
 /* This is called from within the LOCK_HARDWARE routine */
-void XMesaUpdateState(int windowMoved) {
-  __DRIdrawablePrivate *dPriv = gCC->driContextPriv->driDrawablePriv;
+void XMesaUpdateState(int windowMoved)
+{
+  __DRIdrawablePrivate *dPriv = gCC->driDrawablePriv;
   __DRIscreenPrivate *sPriv = dPriv->driScreenPriv;
+  tdfxContextPrivate *gCCPriv = (tdfxContextPrivate *) gCC->driverPrivate;
   TDFXSAREAPriv *saPriv=(TDFXSAREAPriv*)(((char*)sPriv->pSAREA)+sizeof(XF86DRISAREARec));
 
   /* fprintf(stderr, "In FifoPtr=%d FifoRead=%d\n", saPriv->fifoPtr, saPriv->fifoRead); */
@@ -389,8 +345,9 @@ void XMesaUpdateState(int windowMoved) {
     XMesaWindowMoved();
 }
 
-void XMesaSetSAREA() {
-  __DRIdrawablePrivate *dPriv = gCC->driContextPriv->driDrawablePriv;
+void XMesaSetSAREA(void)
+{
+  __DRIdrawablePrivate *dPriv = gCC->driDrawablePriv;
   __DRIscreenPrivate *sPriv = dPriv->driScreenPriv;
   TDFXSAREAPriv *saPriv=(TDFXSAREAPriv*)(((char*)sPriv->pSAREA)+sizeof(XF86DRISAREARec));