diff --git a/linux-core/i810_dma.c b/linux-core/i810_dma.c
index e9b0e807..87145941 100644
--- a/linux-core/i810_dma.c
+++ b/linux-core/i810_dma.c
@@ -26,7 +26,7 @@
  *
  * Authors: Rickard E. (Rik) Faith <faith@valinux.com>
  *	    Jeff Hartmann <jhartmann@valinux.com>
- *          Keith Whitwell <keithw@valinux.com>
+ *          Keith Whitwell <keith_whitwell@yahoo.com>
  *
  */
 
@@ -36,11 +36,6 @@
 #include "i810_drv.h"
 #include <linux/interrupt.h>	/* For task queue support */
 
-/* in case we don't have a 2.3.99-pre6 kernel or later: */
-#ifndef VM_DONTCOPY
-#define VM_DONTCOPY 0
-#endif
-
 #define I810_BUF_FREE		2
 #define I810_BUF_CLIENT		1
 #define I810_BUF_HARDWARE      	0
@@ -50,29 +45,27 @@
 
 #define RING_LOCALS	unsigned int outring, ringmask; volatile char *virt;
 
-#define BEGIN_LP_RING(n) do {				\
-	if (I810_VERBOSE)				\
-		DRM_DEBUG("BEGIN_LP_RING(%d) in %s\n",	\
-			  n, __FUNCTION__);		\
-	if (dev_priv->ring.space < n*4) 		\
-		i810_wait_ring(dev, n*4);		\
-	dev_priv->ring.space -= n*4;			\
-	outring = dev_priv->ring.tail;			\
-	ringmask = dev_priv->ring.tail_mask;		\
-	virt = dev_priv->ring.virtual_start;		\
+#define BEGIN_LP_RING(n) do {						\
+	if (0) DRM_DEBUG("BEGIN_LP_RING(%d) in %s\n", n, __FUNCTION__);	\
+	if (dev_priv->ring.space < n*4)					\
+		i810_wait_ring(dev, n*4);				\
+	dev_priv->ring.space -= n*4;					\
+	outring = dev_priv->ring.tail;					\
+	ringmask = dev_priv->ring.tail_mask;				\
+	virt = dev_priv->ring.virtual_start;				\
 } while (0)
 
-#define ADVANCE_LP_RING() do {					\
-	if (I810_VERBOSE) DRM_DEBUG("ADVANCE_LP_RING\n");	\
-	dev_priv->ring.tail = outring;				\
-	I810_WRITE(LP_RING + RING_TAIL, outring);		\
+#define ADVANCE_LP_RING() do {				\
+	if (0) DRM_DEBUG("ADVANCE_LP_RING\n");			\
+	dev_priv->ring.tail = outring;			\
+	I810_WRITE(LP_RING + RING_TAIL, outring);	\
 } while(0)
 
-#define OUT_RING(n) do {						\
-	if (I810_VERBOSE) DRM_DEBUG("   OUT_RING %x\n", (int)(n));	\
-	*(volatile unsigned int *)(virt + outring) = n;			\
-	outring += 4;							\
-	outring &= ringmask;						\
+#define OUT_RING(n) do {				\
+	if (0) DRM_DEBUG("   OUT_RING %x\n", (int)(n));	\
+	*(volatile unsigned int *)(virt + outring) = n;	\
+	outring += 4;					\
+	outring &= ringmask;				\
 } while (0);
 
 static inline void i810_print_status_page(drm_device_t *dev)
@@ -181,36 +174,31 @@ static int i810_map_buffer(drm_buf_t *buf, struct file *filp)
 
 	if(buf_priv->currently_mapped == I810_BUF_MAPPED) return -EINVAL;
 
-	if(VM_DONTCOPY != 0) {
 #if LINUX_VERSION_CODE <= 0x020402
-		down( &current->mm->mmap_sem );
+	down( &current->mm->mmap_sem );
 #else
-		down_write( &current->mm->mmap_sem );
+	down_write( &current->mm->mmap_sem );
 #endif
-		old_fops = filp->f_op;
-		filp->f_op = &i810_buffer_fops;
-		dev_priv->mmap_buffer = buf;
-		buf_priv->virtual = (void *)do_mmap(filp, 0, buf->total,
-						    PROT_READ|PROT_WRITE,
-						    MAP_SHARED,
-						    buf->bus_address);
-		dev_priv->mmap_buffer = NULL;
-   		filp->f_op = old_fops;
-		if ((unsigned long)buf_priv->virtual > -1024UL) {
-			/* Real error */
-			DRM_DEBUG("mmap error\n");
-			retcode = (signed int)buf_priv->virtual;
-			buf_priv->virtual = 0;
-		}
-#if LINUX_VERSION_CODE <= 0x020402
-		up( &current->mm->mmap_sem );
-#else
-		up_write( &current->mm->mmap_sem );
-#endif
-	} else {
-		buf_priv->virtual = buf_priv->kernel_virtual;
-   		buf_priv->currently_mapped = I810_BUF_MAPPED;
+	old_fops = filp->f_op;
+	filp->f_op = &i810_buffer_fops;
+	dev_priv->mmap_buffer = buf;
+	buf_priv->virtual = (void *)do_mmap(filp, 0, buf->total,
+					    PROT_READ|PROT_WRITE,
+					    MAP_SHARED,
+					    buf->bus_address);
+	dev_priv->mmap_buffer = NULL;
+	filp->f_op = old_fops;
+	if ((unsigned long)buf_priv->virtual > -1024UL) {
+		/* Real error */
+		DRM_DEBUG("mmap error\n");
+		retcode = (signed int)buf_priv->virtual;
+		buf_priv->virtual = 0;
 	}
+#if LINUX_VERSION_CODE <= 0x020402
+	up( &current->mm->mmap_sem );
+#else
+	up_write( &current->mm->mmap_sem );
+#endif
 	return retcode;
 }
 
@@ -219,23 +207,21 @@ static int i810_unmap_buffer(drm_buf_t *buf)
 	drm_i810_buf_priv_t *buf_priv = buf->dev_private;
 	int retcode = 0;
 
-	if(VM_DONTCOPY != 0) {
-		if(buf_priv->currently_mapped != I810_BUF_MAPPED)
-			return -EINVAL;
+	if(buf_priv->currently_mapped != I810_BUF_MAPPED)
+		return -EINVAL;
 #if LINUX_VERSION_CODE <= 0x020402
-		down( &current->mm->mmap_sem );
+	down( &current->mm->mmap_sem );
 #else
-		down_write( &current->mm->mmap_sem );
+	down_write( &current->mm->mmap_sem );
 #endif
-        	retcode = do_munmap(current->mm,
-				    (unsigned long)buf_priv->virtual,
-				    (size_t) buf->total);
+	retcode = do_munmap(current->mm,
+			    (unsigned long)buf_priv->virtual,
+			    (size_t) buf->total);
 #if LINUX_VERSION_CODE <= 0x020402
-		up( &current->mm->mmap_sem );
+	up( &current->mm->mmap_sem );
 #else
-		up_write( &current->mm->mmap_sem );
+	up_write( &current->mm->mmap_sem );
 #endif
-	}
    	buf_priv->currently_mapped = I810_BUF_UNMAPPED;
    	buf_priv->virtual = 0;
 
@@ -444,9 +430,6 @@ static int i810_dma_initialize(drm_device_t *dev,
 		((u8 *)dev_priv->sarea_map->handle +
 		 init->sarea_priv_offset);
 
-   	atomic_set(&dev_priv->flush_done, 0);
-	init_waitqueue_head(&dev_priv->flush_queue);
-
    	dev_priv->ring.Start = init->ring_start;
    	dev_priv->ring.End = init->ring_end;
    	dev_priv->ring.Size = init->ring_size;
@@ -536,16 +519,12 @@ int i810_dma_init(struct inode *inode, struct file *filp,
 
 /* Most efficient way to verify state for the i810 is as it is
  * emitted.  Non-conformant state is silently dropped.
- *
- * Use 'volatile' & local var tmp to force the emitted values to be
- * identical to the verified ones.
  */
 static void i810EmitContextVerified( drm_device_t *dev,
-				     volatile unsigned int *code )
+				     unsigned int *code )
 {
    	drm_i810_private_t *dev_priv = dev->dev_private;
 	int i, j = 0;
-	unsigned int tmp;
 	RING_LOCALS;
 
 	BEGIN_LP_RING( I810_CTX_SETUP_SIZE );
@@ -557,12 +536,10 @@ static void i810EmitContextVerified( drm_device_t *dev,
 	OUT_RING( code[I810_CTXREG_ST1] );
 
 	for ( i = 4 ; i < I810_CTX_SETUP_SIZE ; i++ ) {
-		tmp = code[i];
-
-		if ((tmp & (7<<29)) == (3<<29) &&
-		    (tmp & (0x1f<<24)) < (0x1d<<24))
+		if ((code[i] & (7<<29)) == (3<<29) &&
+		    (code[i] & (0x1f<<24)) < (0x1d<<24))
 		{
-			OUT_RING( tmp );
+			OUT_RING( code[i] );
 			j++;
 		}
 	}
@@ -578,7 +555,6 @@ static void i810EmitTexVerified( drm_device_t *dev,
 {
    	drm_i810_private_t *dev_priv = dev->dev_private;
 	int i, j = 0;
-	unsigned int tmp;
 	RING_LOCALS;
 
 	BEGIN_LP_RING( I810_TEX_SETUP_SIZE );
@@ -589,12 +565,11 @@ static void i810EmitTexVerified( drm_device_t *dev,
 	OUT_RING( code[I810_TEXREG_MI3] );
 
 	for ( i = 4 ; i < I810_TEX_SETUP_SIZE ; i++ ) {
-		tmp = code[i];
 
-		if ((tmp & (7<<29)) == (3<<29) &&
-		    (tmp & (0x1f<<24)) < (0x1d<<24))
+		if ((code[i] & (7<<29)) == (3<<29) &&
+		    (code[i] & (0x1f<<24)) < (0x1d<<24))
 		{
-			OUT_RING( tmp );
+			OUT_RING( code[i] );
 			j++;
 		}
 	}
@@ -621,9 +596,7 @@ static void i810EmitDestVerified( drm_device_t *dev,
 	if (tmp == dev_priv->front_di1 || tmp == dev_priv->back_di1) {
 		OUT_RING( CMD_OP_DESTBUFFER_INFO );
 		OUT_RING( tmp );
-	} else
-	   DRM_DEBUG("bad di1 %x (allow %x or %x)\n",
-		     tmp, dev_priv->front_di1, dev_priv->back_di1);
+	} 
 
 	/* invarient:
 	 */
@@ -645,30 +618,24 @@ static void i810EmitDestVerified( drm_device_t *dev,
 
 
 
-static void i810EmitState( drm_device_t *dev )
+static void i810EmitState( drm_device_t *dev,
+			   drm_i810_state_t *state,
+			   unsigned int dirty )
 {
-   	drm_i810_private_t *dev_priv = dev->dev_private;
-      	drm_i810_sarea_t *sarea_priv = dev_priv->sarea_priv;
-	unsigned int dirty = sarea_priv->dirty;
-
 	if (dirty & I810_UPLOAD_BUFFERS) {
-		i810EmitDestVerified( dev, sarea_priv->BufferState );
-		sarea_priv->dirty &= ~I810_UPLOAD_BUFFERS;
+		i810EmitDestVerified( dev, state->BufferSetup );
 	}
 
 	if (dirty & I810_UPLOAD_CTX) {
-		i810EmitContextVerified( dev, sarea_priv->ContextState );
-		sarea_priv->dirty &= ~I810_UPLOAD_CTX;
+		i810EmitContextVerified( dev, state->Setup );
 	}
 
 	if (dirty & I810_UPLOAD_TEX0) {
-		i810EmitTexVerified( dev, sarea_priv->TexState[0] );
-		sarea_priv->dirty &= ~I810_UPLOAD_TEX0;
+		i810EmitTexVerified( dev, state->TexSetup[0] );
 	}
 
 	if (dirty & I810_UPLOAD_TEX1) {
-		i810EmitTexVerified( dev, sarea_priv->TexState[1] );
-		sarea_priv->dirty &= ~I810_UPLOAD_TEX1;
+		i810EmitTexVerified( dev, state->TexSetup[1] );
 	}
 }
 
@@ -708,7 +675,6 @@ static void i810_dma_dispatch_clear( drm_device_t *dev, int flags,
 			continue;
 
 	   	if ( flags & I810_FRONT ) {
-		   	DRM_DEBUG("clear front\n");
 			BEGIN_LP_RING( 6 );
 			OUT_RING( BR00_BITBLT_CLIENT |
 				  BR00_OP_COLOR_BLT | 0x3 );
@@ -721,7 +687,6 @@ static void i810_dma_dispatch_clear( drm_device_t *dev, int flags,
 		}
 
 		if ( flags & I810_BACK ) {
-			DRM_DEBUG("clear back\n");
 			BEGIN_LP_RING( 6 );
 			OUT_RING( BR00_BITBLT_CLIENT |
 				  BR00_OP_COLOR_BLT | 0x3 );
@@ -734,7 +699,6 @@ static void i810_dma_dispatch_clear( drm_device_t *dev, int flags,
 		}
 
 		if ( flags & I810_DEPTH ) {
-			DRM_DEBUG("clear depth\n");
 			BEGIN_LP_RING( 6 );
 			OUT_RING( BR00_BITBLT_CLIENT |
 				  BR00_OP_COLOR_BLT | 0x3 );
@@ -760,8 +724,6 @@ static void i810_dma_dispatch_swap( drm_device_t *dev )
 	int i;
 	RING_LOCALS;
 
-	DRM_DEBUG("swapbuffers\n");
-
   	i810_kernel_lost_context(dev);
 
       	if (nbox > I810_NR_SAREA_CLIPRECTS)
@@ -780,10 +742,6 @@ static void i810_dma_dispatch_swap( drm_device_t *dev )
 		    pbox->y2 > dev_priv->h)
 			continue;
 
-	   	DRM_DEBUG("dispatch swap %d,%d-%d,%d!\n",
-			  pbox[i].x1, pbox[i].y1,
-			  pbox[i].x2, pbox[i].y2);
-
 		BEGIN_LP_RING( 6 );
 		OUT_RING( BR00_BITBLT_CLIENT | BR00_OP_SRC_COPY_BLT | 0x4 );
 		OUT_RING( pitch | (0xCC << 16));
@@ -799,7 +757,10 @@ static void i810_dma_dispatch_swap( drm_device_t *dev )
 static void i810_dma_dispatch_vertex(drm_device_t *dev,
 				     drm_buf_t *buf,
 				     int discard,
-				     int used)
+				     drm_i810_prim_t *prim,
+				     unsigned int nrprim,
+				     drm_i810_state_t *state,
+				     unsigned int nrstate )
 {
    	drm_i810_private_t *dev_priv = dev->dev_private;
 	drm_i810_buf_priv_t *buf_priv = buf->dev_private;
@@ -807,139 +768,92 @@ static void i810_dma_dispatch_vertex(drm_device_t *dev,
    	drm_clip_rect_t *box = sarea_priv->boxes;
    	int nbox = sarea_priv->nbox;
 	unsigned long address = (unsigned long)buf->bus_address;
-	unsigned long start = address - dev->agp->base;
-	int i = 0, u;
+	unsigned long bufstart = address - dev->agp->base;
+	char *buf_virtual = (char *)buf_priv->virtual;
+	unsigned int i;
    	RING_LOCALS;
 
+   	if (nbox > I810_NR_SAREA_CLIPRECTS)
+		nrprim = 0;
+
    	i810_kernel_lost_context(dev);
 
-   	if (nbox > I810_NR_SAREA_CLIPRECTS)
-		nbox = I810_NR_SAREA_CLIPRECTS;
+	for (i = 0 ; i < nrprim ; i++, prim++) {
+		int j = 0;
 
-	if (discard) {
-		u = cmpxchg(buf_priv->in_use, I810_BUF_CLIENT,
-			    I810_BUF_HARDWARE);
-		if(u != I810_BUF_CLIENT) {
-			DRM_DEBUG("xxxx 2\n");
-		}
-	}
+/*  		printk("prim %d start %x fin %x\n", i, prim->start, */
+/*  		       prim->finish); */
 
-	if (used > 4*1024)
-		used = 0;
-
-	if (sarea_priv->dirty)
-	   i810EmitState( dev );
-
-  	DRM_DEBUG("dispatch vertex addr 0x%lx, used 0x%x nbox %d\n",
-		  address, used, nbox);
-
-   	dev_priv->counter++;
-   	DRM_DEBUG(  "dispatch counter : %ld\n", dev_priv->counter);
-   	DRM_DEBUG(  "i810_dma_dispatch\n");
-   	DRM_DEBUG(  "start : %lx\n", start);
-	DRM_DEBUG(  "used : %d\n", used);
-   	DRM_DEBUG(  "start + used - 4 : %ld\n", start + used - 4);
-
-	if (buf_priv->currently_mapped == I810_BUF_MAPPED) {
-		*(u32 *)buf_priv->virtual = (GFX_OP_PRIMITIVE |
-					     sarea_priv->vertex_prim |
-					     ((used/4)-2));
-
-		if (used & 4) {
-			*(u32 *)((u32)buf_priv->virtual + used) = 0;
-			used += 4;
+		if ((unsigned)prim->start >= I810_DMA_BUF_SZ ||
+		    prim->start >= prim->finish ||
+		    (prim->start & 0x7) != 0x4) {
+/*  			printk("alignment/other\n"); */
+			continue;
 		}
 
-		i810_unmap_buffer(buf);
-	}
+		if (prim->dirty && prim->stateidx < nrstate) {
+			i810EmitState( dev,
+				       &state[(int)prim->stateidx],
+				       prim->dirty );
+		}
+
+		*(int *)&buf_virtual[prim->start-4] =
+			(GFX_OP_PRIMITIVE | (((int)prim->prim) << 18) |
+			 ((prim->finish - prim->start)/4-1));
+
+
+		if (prim->finish & 0x4) {
+			*(int *)&buf_virtual[prim->finish] = 0;
+			prim->finish += 4;
+		}
+
+
 
-	if (used) {
 		do {
-			if (i < nbox) {
+			if (j < nbox) {
 				BEGIN_LP_RING(4);
-				OUT_RING( GFX_OP_SCISSOR | SC_UPDATE_SCISSOR |
+				OUT_RING( GFX_OP_SCISSOR |
+					  SC_UPDATE_SCISSOR |
 					  SC_ENABLE );
 				OUT_RING( GFX_OP_SCISSOR_INFO );
-				OUT_RING( box[i].x1 | (box[i].y1<<16) );
-				OUT_RING( (box[i].x2-1) | ((box[i].y2-1)<<16) );
+				OUT_RING( box[j].x1 | (box[j].y1<<16) );
+				OUT_RING( (box[j].x2-1) | ((box[j].y2-1)<<16) );
 				ADVANCE_LP_RING();
+				if (nbox == 1) nbox = 0;
 			}
 
 			BEGIN_LP_RING(4);
 			OUT_RING( CMD_OP_BATCH_BUFFER );
-			OUT_RING( start | BB1_PROTECTED );
-			OUT_RING( start + used - 4 );
+			OUT_RING( (bufstart + prim->start - 4) |
+				  BB1_PROTECTED );
+			OUT_RING( (bufstart + prim->finish - 4) );
 			OUT_RING( 0 );
 			ADVANCE_LP_RING();
-
-		} while (++i < nbox);
+		} while (++j < nbox);
 	}
 
-	BEGIN_LP_RING(10);
-	OUT_RING( CMD_STORE_DWORD_IDX );
-	OUT_RING( 20 );
-	OUT_RING( dev_priv->counter );
-	OUT_RING( 0 );
 
 	if (discard) {
+		dev_priv->counter++;
+
+		(void) cmpxchg(buf_priv->in_use, I810_BUF_CLIENT,
+			       I810_BUF_HARDWARE);
+
+		BEGIN_LP_RING( 8 );
+		OUT_RING( CMD_STORE_DWORD_IDX );
+		OUT_RING( 20 );
+		OUT_RING( dev_priv->counter );
 		OUT_RING( CMD_STORE_DWORD_IDX );
 		OUT_RING( buf_priv->my_use_idx );
 		OUT_RING( I810_BUF_FREE );
+		OUT_RING( CMD_REPORT_HEAD );
 		OUT_RING( 0 );
+		ADVANCE_LP_RING();
 	}
-
-      	OUT_RING( CMD_REPORT_HEAD );
-	OUT_RING( 0 );
-   	ADVANCE_LP_RING();
 }
 
 
-/* Interrupts are only for flushing */
-void i810_dma_service(int irq, void *device, struct pt_regs *regs)
-{
-	drm_device_t	 *dev = (drm_device_t *)device;
-      	drm_i810_private_t *dev_priv = (drm_i810_private_t *)dev->dev_private;
-   	u16 temp;
-
-	atomic_inc(&dev->counts[_DRM_STAT_IRQ]);
-      	temp = I810_READ16(I810REG_INT_IDENTITY_R);
-   	temp = temp & ~(0x6000);
-   	if(temp != 0) I810_WRITE16(I810REG_INT_IDENTITY_R,
-				   temp); /* Clear all interrupts */
-	else
-	   return;
-
-   	queue_task(&dev->tq, &tq_immediate);
-   	mark_bh(IMMEDIATE_BH);
-}
-
-void i810_dma_immediate_bh(void *device)
-{
-	drm_device_t *dev = (drm_device_t *) device;
-      	drm_i810_private_t *dev_priv = (drm_i810_private_t *)dev->dev_private;
-
-   	atomic_set(&dev_priv->flush_done, 1);
-   	wake_up_interruptible(&dev_priv->flush_queue);
-}
-
-static inline void i810_dma_emit_flush(drm_device_t *dev)
-{
-   	drm_i810_private_t *dev_priv = dev->dev_private;
-   	RING_LOCALS;
-
-   	i810_kernel_lost_context(dev);
-
-   	BEGIN_LP_RING(2);
-      	OUT_RING( CMD_REPORT_HEAD );
-      	OUT_RING( GFX_OP_USER_INTERRUPT );
-      	ADVANCE_LP_RING();
-
-/*  	i810_wait_ring( dev, dev_priv->ring.Size - 8 ); */
-/*     	atomic_set(&dev_priv->flush_done, 1); */
-/*     	wake_up_interruptible(&dev_priv->flush_queue); */
-}
-
-static inline void i810_dma_quiescent_emit(drm_device_t *dev)
+void i810_dma_quiescent(drm_device_t *dev)
 {
       	drm_i810_private_t *dev_priv = dev->dev_private;
    	RING_LOCALS;
@@ -950,79 +864,27 @@ static inline void i810_dma_quiescent_emit(drm_device_t *dev)
    	OUT_RING( INST_PARSER_CLIENT | INST_OP_FLUSH | INST_FLUSH_MAP_CACHE );
    	OUT_RING( CMD_REPORT_HEAD );
       	OUT_RING( 0 );
-      	OUT_RING( GFX_OP_USER_INTERRUPT );
+      	OUT_RING( 0 );
    	ADVANCE_LP_RING();
 
-/*  	i810_wait_ring( dev, dev_priv->ring.Size - 8 ); */
-/*     	atomic_set(&dev_priv->flush_done, 1); */
-/*     	wake_up_interruptible(&dev_priv->flush_queue); */
-}
-
-void i810_dma_quiescent(drm_device_t *dev)
-{
-      	DECLARE_WAITQUEUE(entry, current);
-  	drm_i810_private_t *dev_priv = (drm_i810_private_t *)dev->dev_private;
-	unsigned long end;
-
-   	if(dev_priv == NULL) {
-	   	return;
-	}
-      	atomic_set(&dev_priv->flush_done, 0);
-   	add_wait_queue(&dev_priv->flush_queue, &entry);
-   	end = jiffies + (HZ*3);
-
-   	for (;;) {
-		current->state = TASK_INTERRUPTIBLE;
-	      	i810_dma_quiescent_emit(dev);
-	   	if (atomic_read(&dev_priv->flush_done) == 1) break;
-		if((signed)(end - jiffies) <= 0) {
-		   	DRM_ERROR("lockup\n");
-		   	break;
-		}
-	      	schedule_timeout(HZ*3);
-	      	if (signal_pending(current)) {
-		   	break;
-		}
-	}
-
-   	current->state = TASK_RUNNING;
-   	remove_wait_queue(&dev_priv->flush_queue, &entry);
-
-   	return;
+	i810_wait_ring( dev, dev_priv->ring.Size - 8 );
 }
 
 static int i810_flush_queue(drm_device_t *dev)
 {
-   	DECLARE_WAITQUEUE(entry, current);
-  	drm_i810_private_t *dev_priv = (drm_i810_private_t *)dev->dev_private;
+   	drm_i810_private_t *dev_priv = dev->dev_private;
 	drm_device_dma_t *dma = dev->dma;
-	unsigned long end;
    	int i, ret = 0;
+   	RING_LOCALS;
+	
+   	i810_kernel_lost_context(dev);
 
-   	if(dev_priv == NULL) {
-	   	return 0;
-	}
-      	atomic_set(&dev_priv->flush_done, 0);
-   	add_wait_queue(&dev_priv->flush_queue, &entry);
-   	end = jiffies + (HZ*3);
-   	for (;;) {
-		current->state = TASK_INTERRUPTIBLE;
-	      	i810_dma_emit_flush(dev);
-	   	if (atomic_read(&dev_priv->flush_done) == 1) break;
-		if((signed)(end - jiffies) <= 0) {
-		   	DRM_ERROR("lockup\n");
-		   	break;
-		}
-	      	schedule_timeout(HZ*3);
-	      	if (signal_pending(current)) {
-		   	ret = -EINTR; /* Can't restart */
-		   	break;
-		}
-	}
-
-   	current->state = TASK_RUNNING;
-   	remove_wait_queue(&dev_priv->flush_queue, &entry);
+   	BEGIN_LP_RING(2);
+      	OUT_RING( CMD_REPORT_HEAD );
+      	OUT_RING( 0 );
+      	ADVANCE_LP_RING();
 
+	i810_wait_ring( dev, dev_priv->ring.Size - 8 );
 
    	for (i = 0; i < dma->buf_count; i++) {
 	   	drm_buf_t *buf = dma->buflist[ i ];
@@ -1034,7 +896,7 @@ static int i810_flush_queue(drm_device_t *dev)
 		if (used == I810_BUF_HARDWARE)
 			DRM_DEBUG("reclaimed from HARDWARE\n");
 		if (used == I810_BUF_CLIENT)
-			DRM_DEBUG("still on client HARDWARE\n");
+			DRM_DEBUG("still on client\n");
 	}
 
    	return ret;
@@ -1074,7 +936,6 @@ int i810_flush_ioctl(struct inode *inode, struct file *filp,
    	drm_file_t	  *priv	  = filp->private_data;
    	drm_device_t	  *dev	  = priv->dev;
 
-   	DRM_DEBUG("i810_flush_ioctl\n");
    	if(!_DRM_LOCK_IS_HELD(dev->lock.hw_lock->lock)) {
 		DRM_ERROR("i810_flush_ioctl called without lock held\n");
 		return -EINVAL;
@@ -1085,8 +946,32 @@ int i810_flush_ioctl(struct inode *inode, struct file *filp,
 }
 
 
+/* Copy sarea data into temporary structs.
+ */
+void i810_copy_state( drm_i810_sarea_t *sarea, drm_i810_state_t *state )
+{
+	if (sarea->dirty & I810_UPLOAD_CTX)
+		memcpy(state->Setup, sarea->ContextState,
+		       sizeof(state->Setup));
+
+	if (sarea->dirty & I810_UPLOAD_BUFFERS)
+		memcpy(state->BufferSetup, sarea->BufferState,
+		       sizeof(state->BufferSetup));
+
+	if (sarea->dirty & I810_UPLOAD_TEX0)
+		memcpy(state->TexSetup[0], sarea->TexState[0],
+		       sizeof(state->TexSetup[0]));
+
+	if (sarea->dirty & I810_UPLOAD_TEX1)
+		memcpy(state->TexSetup[1], sarea->TexState[1],
+		       sizeof(state->TexSetup[1]));
+}
+
+
+/* Obsolete, backwards compatibility ioctl:
+ */
 int i810_dma_vertex(struct inode *inode, struct file *filp,
-	       unsigned int cmd, unsigned long arg)
+		    unsigned int cmd, unsigned long arg)
 {
 	drm_file_t *priv = filp->private_data;
 	drm_device_t *dev = priv->dev;
@@ -1096,6 +981,12 @@ int i810_dma_vertex(struct inode *inode, struct file *filp,
    	drm_i810_sarea_t *sarea_priv = (drm_i810_sarea_t *)
      					dev_priv->sarea_priv;
 	drm_i810_vertex_t vertex;
+	drm_buf_t *buf;
+	drm_i810_buf_priv_t *buf_priv;
+	drm_i810_state_t tmpstate;
+	drm_i810_prim_t tmpprim;
+
+/*  	printk("%s\n", __FUNCTION__); */
 
 	if (copy_from_user(&vertex, (drm_i810_vertex_t *)arg, sizeof(vertex)))
 		return -EFAULT;
@@ -1105,24 +996,116 @@ int i810_dma_vertex(struct inode *inode, struct file *filp,
 		return -EINVAL;
 	}
 
-	DRM_DEBUG("i810 dma vertex, idx %d used %d discard %d\n",
-		  vertex.idx, vertex.used, vertex.discard);
+	buf = dma->buflist[ vertex.idx ];
+   	buf_priv = buf->dev_private;
 
-	if(vertex.idx < 0 || vertex.idx > dma->buf_count) return -EINVAL;
+	if (vertex.idx < 0 || vertex.idx >= I810_DMA_BUF_NR) {
+		DRM_ERROR("i810_dma_vertex2 bad buffer idx\n");
+		return -EINVAL;
+	}
+		
+	if (sarea_priv->dirty)
+		i810_copy_state( sarea_priv, &tmpstate );
 
-	i810_dma_dispatch_vertex( dev,
-				  dma->buflist[ vertex.idx ],
-				  vertex.discard, vertex.used );
+	tmpprim.start = 0;
+	tmpprim.finish = vertex.used;
+	tmpprim.prim = (char) sarea_priv->vertex_prim;
+	tmpprim.dirty = (char) sarea_priv->dirty;
+	tmpprim.stateidx = 0;
+	
+	i810_dma_dispatch_vertex( dev, buf,
+				  vertex.discard,
+				  &tmpprim, 1,
+				  &tmpstate, 1 );
+
+	if (buf_priv->currently_mapped == I810_BUF_MAPPED) {
+		i810_unmap_buffer(buf);
+	}
 
-   	atomic_add(vertex.used, &dev->counts[_DRM_STAT_SECONDARY]);
 	atomic_inc(&dev->counts[_DRM_STAT_DMA]);
+
 	sarea_priv->last_enqueue = dev_priv->counter-1;
    	sarea_priv->last_dispatch = (int) hw_status[5];
-
-	return 0;
+   	return 0;
 }
 
 
+int i810_dma_vertex2(struct inode *inode, struct file *filp,
+		    unsigned int cmd, unsigned long arg)
+{
+	drm_file_t *priv = filp->private_data;
+	drm_device_t *dev = priv->dev;
+	drm_device_dma_t *dma = dev->dma;
+	drm_i810_private_t *dev_priv = (drm_i810_private_t *)dev->dev_private;
+	u32 *hw_status = (u32 *)dev_priv->hw_status_page;
+	drm_i810_sarea_t *sarea_priv = (drm_i810_sarea_t *)
+		dev_priv->sarea_priv;
+	drm_i810_vertex2_t vertex;
+	drm_buf_t *buf;
+	drm_i810_buf_priv_t *buf_priv;
+
+/*  	printk("%s\n", __FUNCTION__); */
+	
+	if (copy_from_user(&vertex, (drm_i810_vertex2_t *)arg, sizeof(vertex)))
+		return -EFAULT;
+
+	if(!_DRM_LOCK_IS_HELD(dev->lock.hw_lock->lock)) {
+		DRM_ERROR("i810_dma_vertex2 called without lock held\n");
+		return -EINVAL;
+	}
+
+	if (vertex.idx >= I810_DMA_BUF_NR) {
+		DRM_ERROR("i810_dma_vertex2 bad buffer idx\n");
+		return -EINVAL;
+	}
+		
+
+/*  	printk("%s buf %d nstates %d nprim %d\n",  */
+/*  	       __FUNCTION__,  */
+/*  	       vertex.idx, */
+/*  	       vertex.nr_states, */
+/*  	       vertex.nr_prims ); */
+
+	buf = dma->buflist[ vertex.idx ];
+	buf_priv = buf->dev_private;
+
+	if (vertex.nr_states) {
+		if (vertex.nr_states >= I810_MAX_STATES)
+			return -EINVAL;
+
+		if (copy_from_user(dev_priv->statetmp,
+				   vertex.state_address,
+				   vertex.nr_states * sizeof(drm_i810_state_t)))
+			return -EFAULT;
+	}
+
+	if (vertex.nr_prims) {
+		if (vertex.nr_prims >= I810_MAX_PRIMS)
+			return -EINVAL;
+
+		if (copy_from_user(dev_priv->primtmp,
+				   vertex.prim_address,
+				   vertex.nr_prims * sizeof(drm_i810_prim_t)))
+			return -EFAULT;
+	}
+
+	i810_dma_dispatch_vertex( dev, buf,
+				  vertex.discard,
+				  dev_priv->primtmp,
+				  vertex.nr_prims,
+				  dev_priv->statetmp,
+				  vertex.nr_states );
+
+	if (buf_priv->currently_mapped == I810_BUF_MAPPED) 
+		i810_unmap_buffer(buf);
+	
+	atomic_inc(&dev->counts[_DRM_STAT_DMA]);
+
+	sarea_priv->last_enqueue = dev_priv->counter-1;
+	sarea_priv->last_dispatch = (int) hw_status[5];
+	return 0;
+}
+
 
 int i810_clear_bufs(struct inode *inode, struct file *filp,
 		   unsigned int cmd, unsigned long arg)
@@ -1156,8 +1139,6 @@ int i810_swap_bufs(struct inode *inode, struct file *filp,
 	drm_file_t *priv = filp->private_data;
 	drm_device_t *dev = priv->dev;
 
-	DRM_DEBUG("i810_swap_bufs\n");
-
    	if(!_DRM_LOCK_IS_HELD(dev->lock.hw_lock->lock)) {
 		DRM_ERROR("i810_swap_buf called without lock held\n");
 		return -EINVAL;
@@ -1193,7 +1174,6 @@ int i810_getbuf(struct inode *inode, struct file *filp, unsigned int cmd,
    	drm_i810_sarea_t *sarea_priv = (drm_i810_sarea_t *)
      					dev_priv->sarea_priv;
 
-	DRM_DEBUG("getbuf\n");
    	if (copy_from_user(&d, (drm_i810_dma_t *)arg, sizeof(d)))
 		return -EFAULT;
 
@@ -1206,9 +1186,6 @@ int i810_getbuf(struct inode *inode, struct file *filp, unsigned int cmd,
 
 	retcode = i810_dma_get_buffer(dev, &d, filp);
 
-	DRM_DEBUG("i810_dma: %d returning %d, granted = %d\n",
-		  current->pid, retcode, d.granted);
-
 	if (copy_to_user((drm_dma_t *)arg, &d, sizeof(d)))
 		return -EFAULT;
    	sarea_priv->last_dispatch = (int) hw_status[5];
@@ -1216,46 +1193,18 @@ int i810_getbuf(struct inode *inode, struct file *filp, unsigned int cmd,
 	return retcode;
 }
 
-int i810_copybuf(struct inode *inode, struct file *filp, unsigned int cmd,
-		unsigned long arg)
+int i810_copybuf(struct inode *inode,
+		 struct file *filp, 
+		 unsigned int cmd,
+		 unsigned long arg)
 {
-	drm_file_t	  *priv	    = filp->private_data;
-	drm_device_t	  *dev	    = priv->dev;
-	drm_i810_copy_t	  d;
-   	drm_i810_private_t *dev_priv = (drm_i810_private_t *)dev->dev_private;
-   	u32 *hw_status = (u32 *)dev_priv->hw_status_page;
-   	drm_i810_sarea_t *sarea_priv = (drm_i810_sarea_t *)
-     					dev_priv->sarea_priv;
-	drm_buf_t *buf;
-	drm_i810_buf_priv_t *buf_priv;
-	drm_device_dma_t *dma = dev->dma;
-
-	if(!_DRM_LOCK_IS_HELD(dev->lock.hw_lock->lock)) {
-		DRM_ERROR("i810_dma called without lock held\n");
-		return -EINVAL;
-	}
-
-   	if (copy_from_user(&d, (drm_i810_copy_t *)arg, sizeof(d)))
-		return -EFAULT;
-
-        if(d.idx < 0 || d.idx > dma->buf_count) return -EINVAL;
-	buf = dma->buflist[ d.idx ];
-   	buf_priv = buf->dev_private;
-	if (buf_priv->currently_mapped != I810_BUF_MAPPED) return -EPERM;
-
-	if(d.used < 0 || d.used > buf->total) return -EINVAL;
-
-   	if (copy_from_user(buf_priv->virtual, d.address, d.used))
-		return -EFAULT;
-
-   	sarea_priv->last_dispatch = (int) hw_status[5];
-
+	/* Never copy - 2.4.x doesn't need it */
 	return 0;
 }
 
 int i810_docopy(struct inode *inode, struct file *filp, unsigned int cmd,
 		unsigned long arg)
 {
-	if(VM_DONTCOPY == 0) return 1;
+	/* Never copy - 2.4.x doesn't need it */
 	return 0;
 }
diff --git a/linux-core/i810_drm.h b/linux-core/i810_drm.h
index 5d47adda..3af02be9 100644
--- a/linux-core/i810_drm.h
+++ b/linux-core/i810_drm.h
@@ -88,6 +88,8 @@
 #define I810_TEXREG_MCS  7	/* GFX_OP_MAP_COORD_SETS ??? */
 #define I810_TEX_SETUP_SIZE 8
 
+/* Flags for clear ioctl
+ */
 #define I810_FRONT   0x1
 #define I810_BACK    0x2
 #define I810_DEPTH   0x4
@@ -189,6 +191,52 @@ typedef struct _drm_i810_copy_t {
 	void *address;		/* Address to copy from */
 } drm_i810_copy_t;
 
+
+typedef struct _drm_i810_state {
+	unsigned int Setup[I810_CTX_SETUP_SIZE];
+	unsigned int BufferSetup[I810_DEST_SETUP_SIZE];
+	unsigned int TexSetup[2][I810_TEX_SETUP_SIZE];
+} drm_i810_state_t;
+
+#define PR_TRIANGLES         (0x0)
+#define PR_TRISTRIP_0        (0x1)
+#define PR_TRISTRIP_1        (0x2)
+#define PR_TRIFAN            (0x3)
+#define PR_POLYGON           (0x4)
+#define PR_LINES             (0x5)
+#define PR_LINESTRIP         (0x6)
+#define PR_RECTS             (0x7)
+
+typedef struct _drm_i810_prim {
+	int start;	
+	int finish;	
+	char prim;
+	char dirty;
+	char stateidx;
+} drm_i810_prim_t;
+
+
+#define I810_MAX_PRIMS 100
+#define I810_MAX_STATES 20
+
+/* Still uses the sarea to pass cliprects (fix?).  State and primitive
+ * information are passed directly as pointers to userspace data.
+ * Copying of vertex data is performed by this ioctl as well.  
+ * 
+ * This obsoletes _drm_i810_vertex, _drm_i810_copy and the state and
+ * primitive fields in the sarea.
+ */
+typedef struct _drm_i810_vertex2 {
+   	unsigned int idx;	   /* buffer index */
+	unsigned int discard;	   /* client is finished with the buffer? */
+	void *vertex_address;  
+	unsigned int nr_states;
+	drm_i810_state_t *state_address;
+	unsigned int nr_prims;
+	drm_i810_prim_t *prim_address;
+} drm_i810_vertex2_t;
+
+
 typedef struct drm_i810_dma {
 	void *virtual;
 	int request_idx;
diff --git a/linux-core/i810_drv.c b/linux-core/i810_drv.c
index 73df313c..ebed0d66 100644
--- a/linux-core/i810_drv.c
+++ b/linux-core/i810_drv.c
@@ -54,7 +54,8 @@
 	[DRM_IOCTL_NR(DRM_IOCTL_I810_GETBUF)] = { i810_getbuf,      1, 0 }, \
    	[DRM_IOCTL_NR(DRM_IOCTL_I810_SWAP)]   = { i810_swap_bufs,   1, 0 }, \
    	[DRM_IOCTL_NR(DRM_IOCTL_I810_COPY)]   = { i810_copybuf,     1, 0 }, \
-   	[DRM_IOCTL_NR(DRM_IOCTL_I810_DOCOPY)] = { i810_docopy,      1, 0 },
+   	[DRM_IOCTL_NR(DRM_IOCTL_I810_DOCOPY)] = { i810_docopy,      1, 0 }, \
+   	[DRM_IOCTL_NR(DRM_IOCTL_I810_VERTEX2)] = { i810_dma_vertex2,  1, 0 },
 
 
 #define __HAVE_COUNTERS         4
diff --git a/linux-core/i810_drv.h b/linux-core/i810_drv.h
index e1b17148..533a6b18 100644
--- a/linux-core/i810_drv.h
+++ b/linux-core/i810_drv.h
@@ -64,8 +64,6 @@ typedef struct drm_i810_private {
       	unsigned long hw_status_page;
    	unsigned long counter;
 
-   	atomic_t flush_done;
-   	wait_queue_head_t flush_queue;	/* Processes waiting until flush    */
 	drm_buf_t *mmap_buffer;
 
 
@@ -75,6 +73,12 @@ typedef struct drm_i810_private {
 	int depth_offset;
 	int w, h;
 	int pitch;
+
+	/* Temporaries to hold userspace data before submitting to the
+	 * ring.
+	 */
+	drm_i810_state_t statetmp[I810_MAX_STATES];
+	drm_i810_prim_t primtmp[I810_MAX_STATES];
 } drm_i810_private_t;
 
 				/* i810_dma.c */
@@ -89,19 +93,29 @@ extern void i810_reclaim_buffers(drm_device_t *dev, pid_t pid);
 extern int  i810_getage(struct inode *inode, struct file *filp,
 			unsigned int cmd, unsigned long arg);
 extern int i810_mmap_buffers(struct file *filp, struct vm_area_struct *vma);
+
+/* Obsolete:
+ */
 extern int i810_copybuf(struct inode *inode, struct file *filp,
 			unsigned int cmd, unsigned long arg);
+/* Obsolete:
+ */
 extern int i810_docopy(struct inode *inode, struct file *filp,
 		       unsigned int cmd, unsigned long arg);
 
 extern void i810_dma_quiescent(drm_device_t *dev);
 
-#define I810_VERBOSE 0
-
-
+/* Obsolete
+ */
 int i810_dma_vertex(struct inode *inode, struct file *filp,
 		    unsigned int cmd, unsigned long arg);
 
+/* Handles multiple prims/ioctl, does any necessary copying in this
+ * ioctl: 
+ */
+int i810_dma_vertex2(struct inode *inode, struct file *filp,
+		    unsigned int cmd, unsigned long arg);
+
 int i810_swap_bufs(struct inode *inode, struct file *filp,
 		   unsigned int cmd, unsigned long arg);
 
diff --git a/linux/drm.h b/linux/drm.h
index ac9f407a..0d2d762d 100644
--- a/linux/drm.h
+++ b/linux/drm.h
@@ -104,6 +104,7 @@ typedef struct drm_tex_region {
 #include "i810_drm.h"
 #include "r128_drm.h"
 #include "radeon_drm.h"
+#include "gamma_drm.h"
 #ifdef CONFIG_DRM_SIS
 #include "sis_drm.h"
 #endif
@@ -449,6 +450,7 @@ typedef struct drm_scatter_gather {
 #define DRM_IOCTL_I810_SWAP		DRM_IO(  0x46)
 #define DRM_IOCTL_I810_COPY		DRM_IOW( 0x47, drm_i810_copy_t)
 #define DRM_IOCTL_I810_DOCOPY		DRM_IO(  0x48)
+#define DRM_IOCTL_I810_VERTEX2		DRM_IOW( 0x49, drm_i810_vertex2_t)
 
 /* Rage 128 specific ioctls */
 #define DRM_IOCTL_R128_INIT		DRM_IOW( 0x40, drm_r128_init_t)
@@ -466,6 +468,7 @@ typedef struct drm_scatter_gather {
 #define DRM_IOCTL_R128_STIPPLE		DRM_IOW( 0x4d, drm_r128_stipple_t)
 #define DRM_IOCTL_R128_INDIRECT		DRM_IOWR(0x4f, drm_r128_indirect_t)
 #define DRM_IOCTL_R128_FULLSCREEN	DRM_IOW( 0x50, drm_r128_fullscreen_t)
+#define DRM_IOCTL_R128_CLEAR2		DRM_IOW( 0x51, drm_r128_clear2_t)
 
 /* Radeon specific ioctls */
 #define DRM_IOCTL_RADEON_CP_INIT	DRM_IOW( 0x40, drm_radeon_init_t)
@@ -483,6 +486,10 @@ typedef struct drm_scatter_gather {
 #define DRM_IOCTL_RADEON_INDIRECT	DRM_IOWR(0x4d, drm_radeon_indirect_t)
 #define DRM_IOCTL_RADEON_TEXTURE	DRM_IOWR(0x4e, drm_radeon_texture_t)
 
+/* Gamma specific ioctls */
+#define DRM_IOCTL_GAMMA_INIT		DRM_IOW( 0x40, drm_gamma_init_t)
+#define DRM_IOCTL_GAMMA_COPY		DRM_IOW( 0x41, drm_gamma_copy_t)
+
 #ifdef CONFIG_DRM_SIS
 /* SiS specific ioctls */
 #define SIS_IOCTL_FB_ALLOC		DRM_IOWR(0x44, drm_sis_mem_t)
diff --git a/linux/i810.h b/linux/i810.h
index a5152bc0..4a7f5207 100644
--- a/linux/i810.h
+++ b/linux/i810.h
@@ -60,50 +60,9 @@
 	i810_dma_quiescent( dev );					\
 } while (0)
 
-#define __HAVE_DMA_IRQ		1
-#define __HAVE_DMA_IRQ_BH	1
-#define __HAVE_SHARED_IRQ       1
-#define DRIVER_PREINSTALL() do {					\
-	drm_i810_private_t *dev_priv =					\
-		(drm_i810_private_t *)dev->dev_private;			\
-	u16 tmp;							\
-   	tmp = I810_READ16( I810REG_HWSTAM );				\
-   	tmp = tmp & 0x6000;						\
-   	I810_WRITE16( I810REG_HWSTAM, tmp );				\
-									\
-      	tmp = I810_READ16( I810REG_INT_MASK_R );			\
-   	tmp = tmp & 0x6000;		/* Unmask interrupts */		\
-   	I810_WRITE16( I810REG_INT_MASK_R, tmp );			\
-   	tmp = I810_READ16( I810REG_INT_ENABLE_R );			\
-   	tmp = tmp & 0x6000;		/* Disable all interrupts */	\
-      	I810_WRITE16( I810REG_INT_ENABLE_R, tmp );			\
-} while (0)
-
-#define DRIVER_POSTINSTALL() do {					\
-	drm_i810_private_t *dev_priv =					\
-		(drm_i810_private_t *)dev->dev_private;			\
-	u16 tmp;							\
-   	tmp = I810_READ16( I810REG_INT_ENABLE_R );			\
-   	tmp = tmp & 0x6000;						\
-   	tmp = tmp | 0x0003;	/* Enable bp & user interrupts */	\
-   	I810_WRITE16( I810REG_INT_ENABLE_R, tmp );			\
-} while (0)
-
-#define DRIVER_UNINSTALL() do {						\
-	drm_i810_private_t *dev_priv =					\
-		(drm_i810_private_t *)dev->dev_private;			\
-	u16 tmp;							\
-	if ( dev_priv ) {						\
-		tmp = I810_READ16( I810REG_INT_IDENTITY_R );		\
-		tmp = tmp & ~(0x6000);	/* Clear all interrupts */	\
-		if ( tmp != 0 )						\
-			I810_WRITE16( I810REG_INT_IDENTITY_R, tmp );	\
-									\
-		tmp = I810_READ16( I810REG_INT_ENABLE_R );		\
-		tmp = tmp & 0x6000;	/* Disable all interrupts */	\
-		I810_WRITE16( I810REG_INT_ENABLE_R, tmp );		\
-	}								\
-} while (0)
+#define __HAVE_DMA_IRQ		0
+#define __HAVE_DMA_IRQ_BH	0
+#define __HAVE_SHARED_IRQ	1
 
 /* Buffer customization:
  */
diff --git a/linux/i810_dma.c b/linux/i810_dma.c
index e9b0e807..87145941 100644
--- a/linux/i810_dma.c
+++ b/linux/i810_dma.c
@@ -26,7 +26,7 @@
  *
  * Authors: Rickard E. (Rik) Faith <faith@valinux.com>
  *	    Jeff Hartmann <jhartmann@valinux.com>
- *          Keith Whitwell <keithw@valinux.com>
+ *          Keith Whitwell <keith_whitwell@yahoo.com>
  *
  */
 
@@ -36,11 +36,6 @@
 #include "i810_drv.h"
 #include <linux/interrupt.h>	/* For task queue support */
 
-/* in case we don't have a 2.3.99-pre6 kernel or later: */
-#ifndef VM_DONTCOPY
-#define VM_DONTCOPY 0
-#endif
-
 #define I810_BUF_FREE		2
 #define I810_BUF_CLIENT		1
 #define I810_BUF_HARDWARE      	0
@@ -50,29 +45,27 @@
 
 #define RING_LOCALS	unsigned int outring, ringmask; volatile char *virt;
 
-#define BEGIN_LP_RING(n) do {				\
-	if (I810_VERBOSE)				\
-		DRM_DEBUG("BEGIN_LP_RING(%d) in %s\n",	\
-			  n, __FUNCTION__);		\
-	if (dev_priv->ring.space < n*4) 		\
-		i810_wait_ring(dev, n*4);		\
-	dev_priv->ring.space -= n*4;			\
-	outring = dev_priv->ring.tail;			\
-	ringmask = dev_priv->ring.tail_mask;		\
-	virt = dev_priv->ring.virtual_start;		\
+#define BEGIN_LP_RING(n) do {						\
+	if (0) DRM_DEBUG("BEGIN_LP_RING(%d) in %s\n", n, __FUNCTION__);	\
+	if (dev_priv->ring.space < n*4)					\
+		i810_wait_ring(dev, n*4);				\
+	dev_priv->ring.space -= n*4;					\
+	outring = dev_priv->ring.tail;					\
+	ringmask = dev_priv->ring.tail_mask;				\
+	virt = dev_priv->ring.virtual_start;				\
 } while (0)
 
-#define ADVANCE_LP_RING() do {					\
-	if (I810_VERBOSE) DRM_DEBUG("ADVANCE_LP_RING\n");	\
-	dev_priv->ring.tail = outring;				\
-	I810_WRITE(LP_RING + RING_TAIL, outring);		\
+#define ADVANCE_LP_RING() do {				\
+	if (0) DRM_DEBUG("ADVANCE_LP_RING\n");			\
+	dev_priv->ring.tail = outring;			\
+	I810_WRITE(LP_RING + RING_TAIL, outring);	\
 } while(0)
 
-#define OUT_RING(n) do {						\
-	if (I810_VERBOSE) DRM_DEBUG("   OUT_RING %x\n", (int)(n));	\
-	*(volatile unsigned int *)(virt + outring) = n;			\
-	outring += 4;							\
-	outring &= ringmask;						\
+#define OUT_RING(n) do {				\
+	if (0) DRM_DEBUG("   OUT_RING %x\n", (int)(n));	\
+	*(volatile unsigned int *)(virt + outring) = n;	\
+	outring += 4;					\
+	outring &= ringmask;				\
 } while (0);
 
 static inline void i810_print_status_page(drm_device_t *dev)
@@ -181,36 +174,31 @@ static int i810_map_buffer(drm_buf_t *buf, struct file *filp)
 
 	if(buf_priv->currently_mapped == I810_BUF_MAPPED) return -EINVAL;
 
-	if(VM_DONTCOPY != 0) {
 #if LINUX_VERSION_CODE <= 0x020402
-		down( &current->mm->mmap_sem );
+	down( &current->mm->mmap_sem );
 #else
-		down_write( &current->mm->mmap_sem );
+	down_write( &current->mm->mmap_sem );
 #endif
-		old_fops = filp->f_op;
-		filp->f_op = &i810_buffer_fops;
-		dev_priv->mmap_buffer = buf;
-		buf_priv->virtual = (void *)do_mmap(filp, 0, buf->total,
-						    PROT_READ|PROT_WRITE,
-						    MAP_SHARED,
-						    buf->bus_address);
-		dev_priv->mmap_buffer = NULL;
-   		filp->f_op = old_fops;
-		if ((unsigned long)buf_priv->virtual > -1024UL) {
-			/* Real error */
-			DRM_DEBUG("mmap error\n");
-			retcode = (signed int)buf_priv->virtual;
-			buf_priv->virtual = 0;
-		}
-#if LINUX_VERSION_CODE <= 0x020402
-		up( &current->mm->mmap_sem );
-#else
-		up_write( &current->mm->mmap_sem );
-#endif
-	} else {
-		buf_priv->virtual = buf_priv->kernel_virtual;
-   		buf_priv->currently_mapped = I810_BUF_MAPPED;
+	old_fops = filp->f_op;
+	filp->f_op = &i810_buffer_fops;
+	dev_priv->mmap_buffer = buf;
+	buf_priv->virtual = (void *)do_mmap(filp, 0, buf->total,
+					    PROT_READ|PROT_WRITE,
+					    MAP_SHARED,
+					    buf->bus_address);
+	dev_priv->mmap_buffer = NULL;
+	filp->f_op = old_fops;
+	if ((unsigned long)buf_priv->virtual > -1024UL) {
+		/* Real error */
+		DRM_DEBUG("mmap error\n");
+		retcode = (signed int)buf_priv->virtual;
+		buf_priv->virtual = 0;
 	}
+#if LINUX_VERSION_CODE <= 0x020402
+	up( &current->mm->mmap_sem );
+#else
+	up_write( &current->mm->mmap_sem );
+#endif
 	return retcode;
 }
 
@@ -219,23 +207,21 @@ static int i810_unmap_buffer(drm_buf_t *buf)
 	drm_i810_buf_priv_t *buf_priv = buf->dev_private;
 	int retcode = 0;
 
-	if(VM_DONTCOPY != 0) {
-		if(buf_priv->currently_mapped != I810_BUF_MAPPED)
-			return -EINVAL;
+	if(buf_priv->currently_mapped != I810_BUF_MAPPED)
+		return -EINVAL;
 #if LINUX_VERSION_CODE <= 0x020402
-		down( &current->mm->mmap_sem );
+	down( &current->mm->mmap_sem );
 #else
-		down_write( &current->mm->mmap_sem );
+	down_write( &current->mm->mmap_sem );
 #endif
-        	retcode = do_munmap(current->mm,
-				    (unsigned long)buf_priv->virtual,
-				    (size_t) buf->total);
+	retcode = do_munmap(current->mm,
+			    (unsigned long)buf_priv->virtual,
+			    (size_t) buf->total);
 #if LINUX_VERSION_CODE <= 0x020402
-		up( &current->mm->mmap_sem );
+	up( &current->mm->mmap_sem );
 #else
-		up_write( &current->mm->mmap_sem );
+	up_write( &current->mm->mmap_sem );
 #endif
-	}
    	buf_priv->currently_mapped = I810_BUF_UNMAPPED;
    	buf_priv->virtual = 0;
 
@@ -444,9 +430,6 @@ static int i810_dma_initialize(drm_device_t *dev,
 		((u8 *)dev_priv->sarea_map->handle +
 		 init->sarea_priv_offset);
 
-   	atomic_set(&dev_priv->flush_done, 0);
-	init_waitqueue_head(&dev_priv->flush_queue);
-
    	dev_priv->ring.Start = init->ring_start;
    	dev_priv->ring.End = init->ring_end;
    	dev_priv->ring.Size = init->ring_size;
@@ -536,16 +519,12 @@ int i810_dma_init(struct inode *inode, struct file *filp,
 
 /* Most efficient way to verify state for the i810 is as it is
  * emitted.  Non-conformant state is silently dropped.
- *
- * Use 'volatile' & local var tmp to force the emitted values to be
- * identical to the verified ones.
  */
 static void i810EmitContextVerified( drm_device_t *dev,
-				     volatile unsigned int *code )
+				     unsigned int *code )
 {
    	drm_i810_private_t *dev_priv = dev->dev_private;
 	int i, j = 0;
-	unsigned int tmp;
 	RING_LOCALS;
 
 	BEGIN_LP_RING( I810_CTX_SETUP_SIZE );
@@ -557,12 +536,10 @@ static void i810EmitContextVerified( drm_device_t *dev,
 	OUT_RING( code[I810_CTXREG_ST1] );
 
 	for ( i = 4 ; i < I810_CTX_SETUP_SIZE ; i++ ) {
-		tmp = code[i];
-
-		if ((tmp & (7<<29)) == (3<<29) &&
-		    (tmp & (0x1f<<24)) < (0x1d<<24))
+		if ((code[i] & (7<<29)) == (3<<29) &&
+		    (code[i] & (0x1f<<24)) < (0x1d<<24))
 		{
-			OUT_RING( tmp );
+			OUT_RING( code[i] );
 			j++;
 		}
 	}
@@ -578,7 +555,6 @@ static void i810EmitTexVerified( drm_device_t *dev,
 {
    	drm_i810_private_t *dev_priv = dev->dev_private;
 	int i, j = 0;
-	unsigned int tmp;
 	RING_LOCALS;
 
 	BEGIN_LP_RING( I810_TEX_SETUP_SIZE );
@@ -589,12 +565,11 @@ static void i810EmitTexVerified( drm_device_t *dev,
 	OUT_RING( code[I810_TEXREG_MI3] );
 
 	for ( i = 4 ; i < I810_TEX_SETUP_SIZE ; i++ ) {
-		tmp = code[i];
 
-		if ((tmp & (7<<29)) == (3<<29) &&
-		    (tmp & (0x1f<<24)) < (0x1d<<24))
+		if ((code[i] & (7<<29)) == (3<<29) &&
+		    (code[i] & (0x1f<<24)) < (0x1d<<24))
 		{
-			OUT_RING( tmp );
+			OUT_RING( code[i] );
 			j++;
 		}
 	}
@@ -621,9 +596,7 @@ static void i810EmitDestVerified( drm_device_t *dev,
 	if (tmp == dev_priv->front_di1 || tmp == dev_priv->back_di1) {
 		OUT_RING( CMD_OP_DESTBUFFER_INFO );
 		OUT_RING( tmp );
-	} else
-	   DRM_DEBUG("bad di1 %x (allow %x or %x)\n",
-		     tmp, dev_priv->front_di1, dev_priv->back_di1);
+	} 
 
 	/* invarient:
 	 */
@@ -645,30 +618,24 @@ static void i810EmitDestVerified( drm_device_t *dev,
 
 
 
-static void i810EmitState( drm_device_t *dev )
+static void i810EmitState( drm_device_t *dev,
+			   drm_i810_state_t *state,
+			   unsigned int dirty )
 {
-   	drm_i810_private_t *dev_priv = dev->dev_private;
-      	drm_i810_sarea_t *sarea_priv = dev_priv->sarea_priv;
-	unsigned int dirty = sarea_priv->dirty;
-
 	if (dirty & I810_UPLOAD_BUFFERS) {
-		i810EmitDestVerified( dev, sarea_priv->BufferState );
-		sarea_priv->dirty &= ~I810_UPLOAD_BUFFERS;
+		i810EmitDestVerified( dev, state->BufferSetup );
 	}
 
 	if (dirty & I810_UPLOAD_CTX) {
-		i810EmitContextVerified( dev, sarea_priv->ContextState );
-		sarea_priv->dirty &= ~I810_UPLOAD_CTX;
+		i810EmitContextVerified( dev, state->Setup );
 	}
 
 	if (dirty & I810_UPLOAD_TEX0) {
-		i810EmitTexVerified( dev, sarea_priv->TexState[0] );
-		sarea_priv->dirty &= ~I810_UPLOAD_TEX0;
+		i810EmitTexVerified( dev, state->TexSetup[0] );
 	}
 
 	if (dirty & I810_UPLOAD_TEX1) {
-		i810EmitTexVerified( dev, sarea_priv->TexState[1] );
-		sarea_priv->dirty &= ~I810_UPLOAD_TEX1;
+		i810EmitTexVerified( dev, state->TexSetup[1] );
 	}
 }
 
@@ -708,7 +675,6 @@ static void i810_dma_dispatch_clear( drm_device_t *dev, int flags,
 			continue;
 
 	   	if ( flags & I810_FRONT ) {
-		   	DRM_DEBUG("clear front\n");
 			BEGIN_LP_RING( 6 );
 			OUT_RING( BR00_BITBLT_CLIENT |
 				  BR00_OP_COLOR_BLT | 0x3 );
@@ -721,7 +687,6 @@ static void i810_dma_dispatch_clear( drm_device_t *dev, int flags,
 		}
 
 		if ( flags & I810_BACK ) {
-			DRM_DEBUG("clear back\n");
 			BEGIN_LP_RING( 6 );
 			OUT_RING( BR00_BITBLT_CLIENT |
 				  BR00_OP_COLOR_BLT | 0x3 );
@@ -734,7 +699,6 @@ static void i810_dma_dispatch_clear( drm_device_t *dev, int flags,
 		}
 
 		if ( flags & I810_DEPTH ) {
-			DRM_DEBUG("clear depth\n");
 			BEGIN_LP_RING( 6 );
 			OUT_RING( BR00_BITBLT_CLIENT |
 				  BR00_OP_COLOR_BLT | 0x3 );
@@ -760,8 +724,6 @@ static void i810_dma_dispatch_swap( drm_device_t *dev )
 	int i;
 	RING_LOCALS;
 
-	DRM_DEBUG("swapbuffers\n");
-
   	i810_kernel_lost_context(dev);
 
       	if (nbox > I810_NR_SAREA_CLIPRECTS)
@@ -780,10 +742,6 @@ static void i810_dma_dispatch_swap( drm_device_t *dev )
 		    pbox->y2 > dev_priv->h)
 			continue;
 
-	   	DRM_DEBUG("dispatch swap %d,%d-%d,%d!\n",
-			  pbox[i].x1, pbox[i].y1,
-			  pbox[i].x2, pbox[i].y2);
-
 		BEGIN_LP_RING( 6 );
 		OUT_RING( BR00_BITBLT_CLIENT | BR00_OP_SRC_COPY_BLT | 0x4 );
 		OUT_RING( pitch | (0xCC << 16));
@@ -799,7 +757,10 @@ static void i810_dma_dispatch_swap( drm_device_t *dev )
 static void i810_dma_dispatch_vertex(drm_device_t *dev,
 				     drm_buf_t *buf,
 				     int discard,
-				     int used)
+				     drm_i810_prim_t *prim,
+				     unsigned int nrprim,
+				     drm_i810_state_t *state,
+				     unsigned int nrstate )
 {
    	drm_i810_private_t *dev_priv = dev->dev_private;
 	drm_i810_buf_priv_t *buf_priv = buf->dev_private;
@@ -807,139 +768,92 @@ static void i810_dma_dispatch_vertex(drm_device_t *dev,
    	drm_clip_rect_t *box = sarea_priv->boxes;
    	int nbox = sarea_priv->nbox;
 	unsigned long address = (unsigned long)buf->bus_address;
-	unsigned long start = address - dev->agp->base;
-	int i = 0, u;
+	unsigned long bufstart = address - dev->agp->base;
+	char *buf_virtual = (char *)buf_priv->virtual;
+	unsigned int i;
    	RING_LOCALS;
 
+   	if (nbox > I810_NR_SAREA_CLIPRECTS)
+		nrprim = 0;
+
    	i810_kernel_lost_context(dev);
 
-   	if (nbox > I810_NR_SAREA_CLIPRECTS)
-		nbox = I810_NR_SAREA_CLIPRECTS;
+	for (i = 0 ; i < nrprim ; i++, prim++) {
+		int j = 0;
 
-	if (discard) {
-		u = cmpxchg(buf_priv->in_use, I810_BUF_CLIENT,
-			    I810_BUF_HARDWARE);
-		if(u != I810_BUF_CLIENT) {
-			DRM_DEBUG("xxxx 2\n");
-		}
-	}
+/*  		printk("prim %d start %x fin %x\n", i, prim->start, */
+/*  		       prim->finish); */
 
-	if (used > 4*1024)
-		used = 0;
-
-	if (sarea_priv->dirty)
-	   i810EmitState( dev );
-
-  	DRM_DEBUG("dispatch vertex addr 0x%lx, used 0x%x nbox %d\n",
-		  address, used, nbox);
-
-   	dev_priv->counter++;
-   	DRM_DEBUG(  "dispatch counter : %ld\n", dev_priv->counter);
-   	DRM_DEBUG(  "i810_dma_dispatch\n");
-   	DRM_DEBUG(  "start : %lx\n", start);
-	DRM_DEBUG(  "used : %d\n", used);
-   	DRM_DEBUG(  "start + used - 4 : %ld\n", start + used - 4);
-
-	if (buf_priv->currently_mapped == I810_BUF_MAPPED) {
-		*(u32 *)buf_priv->virtual = (GFX_OP_PRIMITIVE |
-					     sarea_priv->vertex_prim |
-					     ((used/4)-2));
-
-		if (used & 4) {
-			*(u32 *)((u32)buf_priv->virtual + used) = 0;
-			used += 4;
+		if ((unsigned)prim->start >= I810_DMA_BUF_SZ ||
+		    prim->start >= prim->finish ||
+		    (prim->start & 0x7) != 0x4) {
+/*  			printk("alignment/other\n"); */
+			continue;
 		}
 
-		i810_unmap_buffer(buf);
-	}
+		if (prim->dirty && prim->stateidx < nrstate) {
+			i810EmitState( dev,
+				       &state[(int)prim->stateidx],
+				       prim->dirty );
+		}
+
+		*(int *)&buf_virtual[prim->start-4] =
+			(GFX_OP_PRIMITIVE | (((int)prim->prim) << 18) |
+			 ((prim->finish - prim->start)/4-1));
+
+
+		if (prim->finish & 0x4) {
+			*(int *)&buf_virtual[prim->finish] = 0;
+			prim->finish += 4;
+		}
+
+
 
-	if (used) {
 		do {
-			if (i < nbox) {
+			if (j < nbox) {
 				BEGIN_LP_RING(4);
-				OUT_RING( GFX_OP_SCISSOR | SC_UPDATE_SCISSOR |
+				OUT_RING( GFX_OP_SCISSOR |
+					  SC_UPDATE_SCISSOR |
 					  SC_ENABLE );
 				OUT_RING( GFX_OP_SCISSOR_INFO );
-				OUT_RING( box[i].x1 | (box[i].y1<<16) );
-				OUT_RING( (box[i].x2-1) | ((box[i].y2-1)<<16) );
+				OUT_RING( box[j].x1 | (box[j].y1<<16) );
+				OUT_RING( (box[j].x2-1) | ((box[j].y2-1)<<16) );
 				ADVANCE_LP_RING();
+				if (nbox == 1) nbox = 0;
 			}
 
 			BEGIN_LP_RING(4);
 			OUT_RING( CMD_OP_BATCH_BUFFER );
-			OUT_RING( start | BB1_PROTECTED );
-			OUT_RING( start + used - 4 );
+			OUT_RING( (bufstart + prim->start - 4) |
+				  BB1_PROTECTED );
+			OUT_RING( (bufstart + prim->finish - 4) );
 			OUT_RING( 0 );
 			ADVANCE_LP_RING();
-
-		} while (++i < nbox);
+		} while (++j < nbox);
 	}
 
-	BEGIN_LP_RING(10);
-	OUT_RING( CMD_STORE_DWORD_IDX );
-	OUT_RING( 20 );
-	OUT_RING( dev_priv->counter );
-	OUT_RING( 0 );
 
 	if (discard) {
+		dev_priv->counter++;
+
+		(void) cmpxchg(buf_priv->in_use, I810_BUF_CLIENT,
+			       I810_BUF_HARDWARE);
+
+		BEGIN_LP_RING( 8 );
+		OUT_RING( CMD_STORE_DWORD_IDX );
+		OUT_RING( 20 );
+		OUT_RING( dev_priv->counter );
 		OUT_RING( CMD_STORE_DWORD_IDX );
 		OUT_RING( buf_priv->my_use_idx );
 		OUT_RING( I810_BUF_FREE );
+		OUT_RING( CMD_REPORT_HEAD );
 		OUT_RING( 0 );
+		ADVANCE_LP_RING();
 	}
-
-      	OUT_RING( CMD_REPORT_HEAD );
-	OUT_RING( 0 );
-   	ADVANCE_LP_RING();
 }
 
 
-/* Interrupts are only for flushing */
-void i810_dma_service(int irq, void *device, struct pt_regs *regs)
-{
-	drm_device_t	 *dev = (drm_device_t *)device;
-      	drm_i810_private_t *dev_priv = (drm_i810_private_t *)dev->dev_private;
-   	u16 temp;
-
-	atomic_inc(&dev->counts[_DRM_STAT_IRQ]);
-      	temp = I810_READ16(I810REG_INT_IDENTITY_R);
-   	temp = temp & ~(0x6000);
-   	if(temp != 0) I810_WRITE16(I810REG_INT_IDENTITY_R,
-				   temp); /* Clear all interrupts */
-	else
-	   return;
-
-   	queue_task(&dev->tq, &tq_immediate);
-   	mark_bh(IMMEDIATE_BH);
-}
-
-void i810_dma_immediate_bh(void *device)
-{
-	drm_device_t *dev = (drm_device_t *) device;
-      	drm_i810_private_t *dev_priv = (drm_i810_private_t *)dev->dev_private;
-
-   	atomic_set(&dev_priv->flush_done, 1);
-   	wake_up_interruptible(&dev_priv->flush_queue);
-}
-
-static inline void i810_dma_emit_flush(drm_device_t *dev)
-{
-   	drm_i810_private_t *dev_priv = dev->dev_private;
-   	RING_LOCALS;
-
-   	i810_kernel_lost_context(dev);
-
-   	BEGIN_LP_RING(2);
-      	OUT_RING( CMD_REPORT_HEAD );
-      	OUT_RING( GFX_OP_USER_INTERRUPT );
-      	ADVANCE_LP_RING();
-
-/*  	i810_wait_ring( dev, dev_priv->ring.Size - 8 ); */
-/*     	atomic_set(&dev_priv->flush_done, 1); */
-/*     	wake_up_interruptible(&dev_priv->flush_queue); */
-}
-
-static inline void i810_dma_quiescent_emit(drm_device_t *dev)
+void i810_dma_quiescent(drm_device_t *dev)
 {
       	drm_i810_private_t *dev_priv = dev->dev_private;
    	RING_LOCALS;
@@ -950,79 +864,27 @@ static inline void i810_dma_quiescent_emit(drm_device_t *dev)
    	OUT_RING( INST_PARSER_CLIENT | INST_OP_FLUSH | INST_FLUSH_MAP_CACHE );
    	OUT_RING( CMD_REPORT_HEAD );
       	OUT_RING( 0 );
-      	OUT_RING( GFX_OP_USER_INTERRUPT );
+      	OUT_RING( 0 );
    	ADVANCE_LP_RING();
 
-/*  	i810_wait_ring( dev, dev_priv->ring.Size - 8 ); */
-/*     	atomic_set(&dev_priv->flush_done, 1); */
-/*     	wake_up_interruptible(&dev_priv->flush_queue); */
-}
-
-void i810_dma_quiescent(drm_device_t *dev)
-{
-      	DECLARE_WAITQUEUE(entry, current);
-  	drm_i810_private_t *dev_priv = (drm_i810_private_t *)dev->dev_private;
-	unsigned long end;
-
-   	if(dev_priv == NULL) {
-	   	return;
-	}
-      	atomic_set(&dev_priv->flush_done, 0);
-   	add_wait_queue(&dev_priv->flush_queue, &entry);
-   	end = jiffies + (HZ*3);
-
-   	for (;;) {
-		current->state = TASK_INTERRUPTIBLE;
-	      	i810_dma_quiescent_emit(dev);
-	   	if (atomic_read(&dev_priv->flush_done) == 1) break;
-		if((signed)(end - jiffies) <= 0) {
-		   	DRM_ERROR("lockup\n");
-		   	break;
-		}
-	      	schedule_timeout(HZ*3);
-	      	if (signal_pending(current)) {
-		   	break;
-		}
-	}
-
-   	current->state = TASK_RUNNING;
-   	remove_wait_queue(&dev_priv->flush_queue, &entry);
-
-   	return;
+	i810_wait_ring( dev, dev_priv->ring.Size - 8 );
 }
 
 static int i810_flush_queue(drm_device_t *dev)
 {
-   	DECLARE_WAITQUEUE(entry, current);
-  	drm_i810_private_t *dev_priv = (drm_i810_private_t *)dev->dev_private;
+   	drm_i810_private_t *dev_priv = dev->dev_private;
 	drm_device_dma_t *dma = dev->dma;
-	unsigned long end;
    	int i, ret = 0;
+   	RING_LOCALS;
+	
+   	i810_kernel_lost_context(dev);
 
-   	if(dev_priv == NULL) {
-	   	return 0;
-	}
-      	atomic_set(&dev_priv->flush_done, 0);
-   	add_wait_queue(&dev_priv->flush_queue, &entry);
-   	end = jiffies + (HZ*3);
-   	for (;;) {
-		current->state = TASK_INTERRUPTIBLE;
-	      	i810_dma_emit_flush(dev);
-	   	if (atomic_read(&dev_priv->flush_done) == 1) break;
-		if((signed)(end - jiffies) <= 0) {
-		   	DRM_ERROR("lockup\n");
-		   	break;
-		}
-	      	schedule_timeout(HZ*3);
-	      	if (signal_pending(current)) {
-		   	ret = -EINTR; /* Can't restart */
-		   	break;
-		}
-	}
-
-   	current->state = TASK_RUNNING;
-   	remove_wait_queue(&dev_priv->flush_queue, &entry);
+   	BEGIN_LP_RING(2);
+      	OUT_RING( CMD_REPORT_HEAD );
+      	OUT_RING( 0 );
+      	ADVANCE_LP_RING();
 
+	i810_wait_ring( dev, dev_priv->ring.Size - 8 );
 
    	for (i = 0; i < dma->buf_count; i++) {
 	   	drm_buf_t *buf = dma->buflist[ i ];
@@ -1034,7 +896,7 @@ static int i810_flush_queue(drm_device_t *dev)
 		if (used == I810_BUF_HARDWARE)
 			DRM_DEBUG("reclaimed from HARDWARE\n");
 		if (used == I810_BUF_CLIENT)
-			DRM_DEBUG("still on client HARDWARE\n");
+			DRM_DEBUG("still on client\n");
 	}
 
    	return ret;
@@ -1074,7 +936,6 @@ int i810_flush_ioctl(struct inode *inode, struct file *filp,
    	drm_file_t	  *priv	  = filp->private_data;
    	drm_device_t	  *dev	  = priv->dev;
 
-   	DRM_DEBUG("i810_flush_ioctl\n");
    	if(!_DRM_LOCK_IS_HELD(dev->lock.hw_lock->lock)) {
 		DRM_ERROR("i810_flush_ioctl called without lock held\n");
 		return -EINVAL;
@@ -1085,8 +946,32 @@ int i810_flush_ioctl(struct inode *inode, struct file *filp,
 }
 
 
+/* Copy sarea data into temporary structs.
+ */
+void i810_copy_state( drm_i810_sarea_t *sarea, drm_i810_state_t *state )
+{
+	if (sarea->dirty & I810_UPLOAD_CTX)
+		memcpy(state->Setup, sarea->ContextState,
+		       sizeof(state->Setup));
+
+	if (sarea->dirty & I810_UPLOAD_BUFFERS)
+		memcpy(state->BufferSetup, sarea->BufferState,
+		       sizeof(state->BufferSetup));
+
+	if (sarea->dirty & I810_UPLOAD_TEX0)
+		memcpy(state->TexSetup[0], sarea->TexState[0],
+		       sizeof(state->TexSetup[0]));
+
+	if (sarea->dirty & I810_UPLOAD_TEX1)
+		memcpy(state->TexSetup[1], sarea->TexState[1],
+		       sizeof(state->TexSetup[1]));
+}
+
+
+/* Obsolete, backwards compatibility ioctl:
+ */
 int i810_dma_vertex(struct inode *inode, struct file *filp,
-	       unsigned int cmd, unsigned long arg)
+		    unsigned int cmd, unsigned long arg)
 {
 	drm_file_t *priv = filp->private_data;
 	drm_device_t *dev = priv->dev;
@@ -1096,6 +981,12 @@ int i810_dma_vertex(struct inode *inode, struct file *filp,
    	drm_i810_sarea_t *sarea_priv = (drm_i810_sarea_t *)
      					dev_priv->sarea_priv;
 	drm_i810_vertex_t vertex;
+	drm_buf_t *buf;
+	drm_i810_buf_priv_t *buf_priv;
+	drm_i810_state_t tmpstate;
+	drm_i810_prim_t tmpprim;
+
+/*  	printk("%s\n", __FUNCTION__); */
 
 	if (copy_from_user(&vertex, (drm_i810_vertex_t *)arg, sizeof(vertex)))
 		return -EFAULT;
@@ -1105,24 +996,116 @@ int i810_dma_vertex(struct inode *inode, struct file *filp,
 		return -EINVAL;
 	}
 
-	DRM_DEBUG("i810 dma vertex, idx %d used %d discard %d\n",
-		  vertex.idx, vertex.used, vertex.discard);
+	buf = dma->buflist[ vertex.idx ];
+   	buf_priv = buf->dev_private;
 
-	if(vertex.idx < 0 || vertex.idx > dma->buf_count) return -EINVAL;
+	if (vertex.idx < 0 || vertex.idx >= I810_DMA_BUF_NR) {
+		DRM_ERROR("i810_dma_vertex2 bad buffer idx\n");
+		return -EINVAL;
+	}
+		
+	if (sarea_priv->dirty)
+		i810_copy_state( sarea_priv, &tmpstate );
 
-	i810_dma_dispatch_vertex( dev,
-				  dma->buflist[ vertex.idx ],
-				  vertex.discard, vertex.used );
+	tmpprim.start = 0;
+	tmpprim.finish = vertex.used;
+	tmpprim.prim = (char) sarea_priv->vertex_prim;
+	tmpprim.dirty = (char) sarea_priv->dirty;
+	tmpprim.stateidx = 0;
+	
+	i810_dma_dispatch_vertex( dev, buf,
+				  vertex.discard,
+				  &tmpprim, 1,
+				  &tmpstate, 1 );
+
+	if (buf_priv->currently_mapped == I810_BUF_MAPPED) {
+		i810_unmap_buffer(buf);
+	}
 
-   	atomic_add(vertex.used, &dev->counts[_DRM_STAT_SECONDARY]);
 	atomic_inc(&dev->counts[_DRM_STAT_DMA]);
+
 	sarea_priv->last_enqueue = dev_priv->counter-1;
    	sarea_priv->last_dispatch = (int) hw_status[5];
-
-	return 0;
+   	return 0;
 }
 
 
+int i810_dma_vertex2(struct inode *inode, struct file *filp,
+		    unsigned int cmd, unsigned long arg)
+{
+	drm_file_t *priv = filp->private_data;
+	drm_device_t *dev = priv->dev;
+	drm_device_dma_t *dma = dev->dma;
+	drm_i810_private_t *dev_priv = (drm_i810_private_t *)dev->dev_private;
+	u32 *hw_status = (u32 *)dev_priv->hw_status_page;
+	drm_i810_sarea_t *sarea_priv = (drm_i810_sarea_t *)
+		dev_priv->sarea_priv;
+	drm_i810_vertex2_t vertex;
+	drm_buf_t *buf;
+	drm_i810_buf_priv_t *buf_priv;
+
+/*  	printk("%s\n", __FUNCTION__); */
+	
+	if (copy_from_user(&vertex, (drm_i810_vertex2_t *)arg, sizeof(vertex)))
+		return -EFAULT;
+
+	if(!_DRM_LOCK_IS_HELD(dev->lock.hw_lock->lock)) {
+		DRM_ERROR("i810_dma_vertex2 called without lock held\n");
+		return -EINVAL;
+	}
+
+	if (vertex.idx >= I810_DMA_BUF_NR) {
+		DRM_ERROR("i810_dma_vertex2 bad buffer idx\n");
+		return -EINVAL;
+	}
+		
+
+/*  	printk("%s buf %d nstates %d nprim %d\n",  */
+/*  	       __FUNCTION__,  */
+/*  	       vertex.idx, */
+/*  	       vertex.nr_states, */
+/*  	       vertex.nr_prims ); */
+
+	buf = dma->buflist[ vertex.idx ];
+	buf_priv = buf->dev_private;
+
+	if (vertex.nr_states) {
+		if (vertex.nr_states >= I810_MAX_STATES)
+			return -EINVAL;
+
+		if (copy_from_user(dev_priv->statetmp,
+				   vertex.state_address,
+				   vertex.nr_states * sizeof(drm_i810_state_t)))
+			return -EFAULT;
+	}
+
+	if (vertex.nr_prims) {
+		if (vertex.nr_prims >= I810_MAX_PRIMS)
+			return -EINVAL;
+
+		if (copy_from_user(dev_priv->primtmp,
+				   vertex.prim_address,
+				   vertex.nr_prims * sizeof(drm_i810_prim_t)))
+			return -EFAULT;
+	}
+
+	i810_dma_dispatch_vertex( dev, buf,
+				  vertex.discard,
+				  dev_priv->primtmp,
+				  vertex.nr_prims,
+				  dev_priv->statetmp,
+				  vertex.nr_states );
+
+	if (buf_priv->currently_mapped == I810_BUF_MAPPED) 
+		i810_unmap_buffer(buf);
+	
+	atomic_inc(&dev->counts[_DRM_STAT_DMA]);
+
+	sarea_priv->last_enqueue = dev_priv->counter-1;
+	sarea_priv->last_dispatch = (int) hw_status[5];
+	return 0;
+}
+
 
 int i810_clear_bufs(struct inode *inode, struct file *filp,
 		   unsigned int cmd, unsigned long arg)
@@ -1156,8 +1139,6 @@ int i810_swap_bufs(struct inode *inode, struct file *filp,
 	drm_file_t *priv = filp->private_data;
 	drm_device_t *dev = priv->dev;
 
-	DRM_DEBUG("i810_swap_bufs\n");
-
    	if(!_DRM_LOCK_IS_HELD(dev->lock.hw_lock->lock)) {
 		DRM_ERROR("i810_swap_buf called without lock held\n");
 		return -EINVAL;
@@ -1193,7 +1174,6 @@ int i810_getbuf(struct inode *inode, struct file *filp, unsigned int cmd,
    	drm_i810_sarea_t *sarea_priv = (drm_i810_sarea_t *)
      					dev_priv->sarea_priv;
 
-	DRM_DEBUG("getbuf\n");
    	if (copy_from_user(&d, (drm_i810_dma_t *)arg, sizeof(d)))
 		return -EFAULT;
 
@@ -1206,9 +1186,6 @@ int i810_getbuf(struct inode *inode, struct file *filp, unsigned int cmd,
 
 	retcode = i810_dma_get_buffer(dev, &d, filp);
 
-	DRM_DEBUG("i810_dma: %d returning %d, granted = %d\n",
-		  current->pid, retcode, d.granted);
-
 	if (copy_to_user((drm_dma_t *)arg, &d, sizeof(d)))
 		return -EFAULT;
    	sarea_priv->last_dispatch = (int) hw_status[5];
@@ -1216,46 +1193,18 @@ int i810_getbuf(struct inode *inode, struct file *filp, unsigned int cmd,
 	return retcode;
 }
 
-int i810_copybuf(struct inode *inode, struct file *filp, unsigned int cmd,
-		unsigned long arg)
+int i810_copybuf(struct inode *inode,
+		 struct file *filp, 
+		 unsigned int cmd,
+		 unsigned long arg)
 {
-	drm_file_t	  *priv	    = filp->private_data;
-	drm_device_t	  *dev	    = priv->dev;
-	drm_i810_copy_t	  d;
-   	drm_i810_private_t *dev_priv = (drm_i810_private_t *)dev->dev_private;
-   	u32 *hw_status = (u32 *)dev_priv->hw_status_page;
-   	drm_i810_sarea_t *sarea_priv = (drm_i810_sarea_t *)
-     					dev_priv->sarea_priv;
-	drm_buf_t *buf;
-	drm_i810_buf_priv_t *buf_priv;
-	drm_device_dma_t *dma = dev->dma;
-
-	if(!_DRM_LOCK_IS_HELD(dev->lock.hw_lock->lock)) {
-		DRM_ERROR("i810_dma called without lock held\n");
-		return -EINVAL;
-	}
-
-   	if (copy_from_user(&d, (drm_i810_copy_t *)arg, sizeof(d)))
-		return -EFAULT;
-
-        if(d.idx < 0 || d.idx > dma->buf_count) return -EINVAL;
-	buf = dma->buflist[ d.idx ];
-   	buf_priv = buf->dev_private;
-	if (buf_priv->currently_mapped != I810_BUF_MAPPED) return -EPERM;
-
-	if(d.used < 0 || d.used > buf->total) return -EINVAL;
-
-   	if (copy_from_user(buf_priv->virtual, d.address, d.used))
-		return -EFAULT;
-
-   	sarea_priv->last_dispatch = (int) hw_status[5];
-
+	/* Never copy - 2.4.x doesn't need it */
 	return 0;
 }
 
 int i810_docopy(struct inode *inode, struct file *filp, unsigned int cmd,
 		unsigned long arg)
 {
-	if(VM_DONTCOPY == 0) return 1;
+	/* Never copy - 2.4.x doesn't need it */
 	return 0;
 }
diff --git a/linux/i810_drm.h b/linux/i810_drm.h
index 5d47adda..3af02be9 100644
--- a/linux/i810_drm.h
+++ b/linux/i810_drm.h
@@ -88,6 +88,8 @@
 #define I810_TEXREG_MCS  7	/* GFX_OP_MAP_COORD_SETS ??? */
 #define I810_TEX_SETUP_SIZE 8
 
+/* Flags for clear ioctl
+ */
 #define I810_FRONT   0x1
 #define I810_BACK    0x2
 #define I810_DEPTH   0x4
@@ -189,6 +191,52 @@ typedef struct _drm_i810_copy_t {
 	void *address;		/* Address to copy from */
 } drm_i810_copy_t;
 
+
+typedef struct _drm_i810_state {
+	unsigned int Setup[I810_CTX_SETUP_SIZE];
+	unsigned int BufferSetup[I810_DEST_SETUP_SIZE];
+	unsigned int TexSetup[2][I810_TEX_SETUP_SIZE];
+} drm_i810_state_t;
+
+#define PR_TRIANGLES         (0x0)
+#define PR_TRISTRIP_0        (0x1)
+#define PR_TRISTRIP_1        (0x2)
+#define PR_TRIFAN            (0x3)
+#define PR_POLYGON           (0x4)
+#define PR_LINES             (0x5)
+#define PR_LINESTRIP         (0x6)
+#define PR_RECTS             (0x7)
+
+typedef struct _drm_i810_prim {
+	int start;	
+	int finish;	
+	char prim;
+	char dirty;
+	char stateidx;
+} drm_i810_prim_t;
+
+
+#define I810_MAX_PRIMS 100
+#define I810_MAX_STATES 20
+
+/* Still uses the sarea to pass cliprects (fix?).  State and primitive
+ * information are passed directly as pointers to userspace data.
+ * Copying of vertex data is performed by this ioctl as well.  
+ * 
+ * This obsoletes _drm_i810_vertex, _drm_i810_copy and the state and
+ * primitive fields in the sarea.
+ */
+typedef struct _drm_i810_vertex2 {
+   	unsigned int idx;	   /* buffer index */
+	unsigned int discard;	   /* client is finished with the buffer? */
+	void *vertex_address;  
+	unsigned int nr_states;
+	drm_i810_state_t *state_address;
+	unsigned int nr_prims;
+	drm_i810_prim_t *prim_address;
+} drm_i810_vertex2_t;
+
+
 typedef struct drm_i810_dma {
 	void *virtual;
 	int request_idx;
diff --git a/linux/i810_drv.c b/linux/i810_drv.c
index 73df313c..ebed0d66 100644
--- a/linux/i810_drv.c
+++ b/linux/i810_drv.c
@@ -54,7 +54,8 @@
 	[DRM_IOCTL_NR(DRM_IOCTL_I810_GETBUF)] = { i810_getbuf,      1, 0 }, \
    	[DRM_IOCTL_NR(DRM_IOCTL_I810_SWAP)]   = { i810_swap_bufs,   1, 0 }, \
    	[DRM_IOCTL_NR(DRM_IOCTL_I810_COPY)]   = { i810_copybuf,     1, 0 }, \
-   	[DRM_IOCTL_NR(DRM_IOCTL_I810_DOCOPY)] = { i810_docopy,      1, 0 },
+   	[DRM_IOCTL_NR(DRM_IOCTL_I810_DOCOPY)] = { i810_docopy,      1, 0 }, \
+   	[DRM_IOCTL_NR(DRM_IOCTL_I810_VERTEX2)] = { i810_dma_vertex2,  1, 0 },
 
 
 #define __HAVE_COUNTERS         4
diff --git a/linux/i810_drv.h b/linux/i810_drv.h
index e1b17148..533a6b18 100644
--- a/linux/i810_drv.h
+++ b/linux/i810_drv.h
@@ -64,8 +64,6 @@ typedef struct drm_i810_private {
       	unsigned long hw_status_page;
    	unsigned long counter;
 
-   	atomic_t flush_done;
-   	wait_queue_head_t flush_queue;	/* Processes waiting until flush    */
 	drm_buf_t *mmap_buffer;
 
 
@@ -75,6 +73,12 @@ typedef struct drm_i810_private {
 	int depth_offset;
 	int w, h;
 	int pitch;
+
+	/* Temporaries to hold userspace data before submitting to the
+	 * ring.
+	 */
+	drm_i810_state_t statetmp[I810_MAX_STATES];
+	drm_i810_prim_t primtmp[I810_MAX_STATES];
 } drm_i810_private_t;
 
 				/* i810_dma.c */
@@ -89,19 +93,29 @@ extern void i810_reclaim_buffers(drm_device_t *dev, pid_t pid);
 extern int  i810_getage(struct inode *inode, struct file *filp,
 			unsigned int cmd, unsigned long arg);
 extern int i810_mmap_buffers(struct file *filp, struct vm_area_struct *vma);
+
+/* Obsolete:
+ */
 extern int i810_copybuf(struct inode *inode, struct file *filp,
 			unsigned int cmd, unsigned long arg);
+/* Obsolete:
+ */
 extern int i810_docopy(struct inode *inode, struct file *filp,
 		       unsigned int cmd, unsigned long arg);
 
 extern void i810_dma_quiescent(drm_device_t *dev);
 
-#define I810_VERBOSE 0
-
-
+/* Obsolete
+ */
 int i810_dma_vertex(struct inode *inode, struct file *filp,
 		    unsigned int cmd, unsigned long arg);
 
+/* Handles multiple prims/ioctl, does any necessary copying in this
+ * ioctl: 
+ */
+int i810_dma_vertex2(struct inode *inode, struct file *filp,
+		    unsigned int cmd, unsigned long arg);
+
 int i810_swap_bufs(struct inode *inode, struct file *filp,
 		   unsigned int cmd, unsigned long arg);
 
diff --git a/linux/radeon_drm.h b/linux/radeon_drm.h
index 4ae387ff..ab7b370f 100644
--- a/linux/radeon_drm.h
+++ b/linux/radeon_drm.h
@@ -26,6 +26,7 @@
  * Authors:
  *    Kevin E. Martin <martin@valinux.com>
  *    Gareth Hughes <gareth@valinux.com>
+ *    Keith Whitwell <keith_whitwell@yahoo.com>
  */
 
 #ifndef __RADEON_DRM_H__
@@ -46,7 +47,6 @@
 #define RADEON_UPLOAD_MASKS		0x00000010
 #define RADEON_UPLOAD_VIEWPORT		0x00000020
 #define RADEON_UPLOAD_SETUP		0x00000040
-#define RADEON_UPLOAD_TCL		0x00000080
 #define RADEON_UPLOAD_MISC		0x00000100
 #define RADEON_UPLOAD_TEX0		0x00000200
 #define RADEON_UPLOAD_TEX1		0x00000400
@@ -82,8 +82,6 @@
 
 #define RADEON_SCRATCH_REG_OFFSET	32
 
-/* Keep these small for testing
- */
 #define RADEON_NR_SAREA_CLIPRECTS	12
 
 /* There are 2 heaps (local/AGP).  Each region within a heap is a
@@ -202,8 +200,9 @@ typedef struct {
 } drm_radeon_tex_region_t;
 
 typedef struct {
-	/* The channel for communication of state information to the kernel
-	 * on firing a vertex buffer.
+	/* The channel for communication of state information to the
+	 * kernel on firing a vertex buffer with either of the
+	 * obsoleted vertex/index ioctls.
 	 */
 	drm_radeon_context_regs_t context_state;
 	drm_radeon_texture_regs_t tex_state[RADEON_MAX_TEXTURE_UNITS];
@@ -296,6 +295,15 @@ typedef struct drm_radeon_vertex {
 	int discard;			/* Client finished with buffer? */
 } drm_radeon_vertex_t;
 
+typedef struct drm_radeon_vertex {
+	int idx;			/* Index of vertex buffer */
+	int discard;			/* Client finished with buffer? */
+	int nr_states;
+	drm_radeon_state_t *state;
+	int nr_prims;
+	drm_radeon_prim_t *prim;
+} drm_radeon_vertex2_t;
+
 typedef struct drm_radeon_indices {
 	int prim;
 	int idx;
@@ -319,6 +327,16 @@ typedef struct drm_radeon_texture {
 	drm_radeon_tex_image_t *image;
 } drm_radeon_texture_t;
 
+typedef struct drm_radeon_texture2 {
+	int offset;
+	int pitch;
+	int format;
+	int width;			/* Texture image coordinates */
+	int height;
+	drm_radeon_tex_image_t blit;
+	const void *data;
+} drm_radeon_texture2_t;
+
 typedef struct drm_radeon_stipple {
 	unsigned int *mask;
 } drm_radeon_stipple_t;
diff --git a/shared-core/drm.h b/shared-core/drm.h
index ac9f407a..0d2d762d 100644
--- a/shared-core/drm.h
+++ b/shared-core/drm.h
@@ -104,6 +104,7 @@ typedef struct drm_tex_region {
 #include "i810_drm.h"
 #include "r128_drm.h"
 #include "radeon_drm.h"
+#include "gamma_drm.h"
 #ifdef CONFIG_DRM_SIS
 #include "sis_drm.h"
 #endif
@@ -449,6 +450,7 @@ typedef struct drm_scatter_gather {
 #define DRM_IOCTL_I810_SWAP		DRM_IO(  0x46)
 #define DRM_IOCTL_I810_COPY		DRM_IOW( 0x47, drm_i810_copy_t)
 #define DRM_IOCTL_I810_DOCOPY		DRM_IO(  0x48)
+#define DRM_IOCTL_I810_VERTEX2		DRM_IOW( 0x49, drm_i810_vertex2_t)
 
 /* Rage 128 specific ioctls */
 #define DRM_IOCTL_R128_INIT		DRM_IOW( 0x40, drm_r128_init_t)
@@ -466,6 +468,7 @@ typedef struct drm_scatter_gather {
 #define DRM_IOCTL_R128_STIPPLE		DRM_IOW( 0x4d, drm_r128_stipple_t)
 #define DRM_IOCTL_R128_INDIRECT		DRM_IOWR(0x4f, drm_r128_indirect_t)
 #define DRM_IOCTL_R128_FULLSCREEN	DRM_IOW( 0x50, drm_r128_fullscreen_t)
+#define DRM_IOCTL_R128_CLEAR2		DRM_IOW( 0x51, drm_r128_clear2_t)
 
 /* Radeon specific ioctls */
 #define DRM_IOCTL_RADEON_CP_INIT	DRM_IOW( 0x40, drm_radeon_init_t)
@@ -483,6 +486,10 @@ typedef struct drm_scatter_gather {
 #define DRM_IOCTL_RADEON_INDIRECT	DRM_IOWR(0x4d, drm_radeon_indirect_t)
 #define DRM_IOCTL_RADEON_TEXTURE	DRM_IOWR(0x4e, drm_radeon_texture_t)
 
+/* Gamma specific ioctls */
+#define DRM_IOCTL_GAMMA_INIT		DRM_IOW( 0x40, drm_gamma_init_t)
+#define DRM_IOCTL_GAMMA_COPY		DRM_IOW( 0x41, drm_gamma_copy_t)
+
 #ifdef CONFIG_DRM_SIS
 /* SiS specific ioctls */
 #define SIS_IOCTL_FB_ALLOC		DRM_IOWR(0x44, drm_sis_mem_t)
diff --git a/shared/drm.h b/shared/drm.h
index ac9f407a..0d2d762d 100644
--- a/shared/drm.h
+++ b/shared/drm.h
@@ -104,6 +104,7 @@ typedef struct drm_tex_region {
 #include "i810_drm.h"
 #include "r128_drm.h"
 #include "radeon_drm.h"
+#include "gamma_drm.h"
 #ifdef CONFIG_DRM_SIS
 #include "sis_drm.h"
 #endif
@@ -449,6 +450,7 @@ typedef struct drm_scatter_gather {
 #define DRM_IOCTL_I810_SWAP		DRM_IO(  0x46)
 #define DRM_IOCTL_I810_COPY		DRM_IOW( 0x47, drm_i810_copy_t)
 #define DRM_IOCTL_I810_DOCOPY		DRM_IO(  0x48)
+#define DRM_IOCTL_I810_VERTEX2		DRM_IOW( 0x49, drm_i810_vertex2_t)
 
 /* Rage 128 specific ioctls */
 #define DRM_IOCTL_R128_INIT		DRM_IOW( 0x40, drm_r128_init_t)
@@ -466,6 +468,7 @@ typedef struct drm_scatter_gather {
 #define DRM_IOCTL_R128_STIPPLE		DRM_IOW( 0x4d, drm_r128_stipple_t)
 #define DRM_IOCTL_R128_INDIRECT		DRM_IOWR(0x4f, drm_r128_indirect_t)
 #define DRM_IOCTL_R128_FULLSCREEN	DRM_IOW( 0x50, drm_r128_fullscreen_t)
+#define DRM_IOCTL_R128_CLEAR2		DRM_IOW( 0x51, drm_r128_clear2_t)
 
 /* Radeon specific ioctls */
 #define DRM_IOCTL_RADEON_CP_INIT	DRM_IOW( 0x40, drm_radeon_init_t)
@@ -483,6 +486,10 @@ typedef struct drm_scatter_gather {
 #define DRM_IOCTL_RADEON_INDIRECT	DRM_IOWR(0x4d, drm_radeon_indirect_t)
 #define DRM_IOCTL_RADEON_TEXTURE	DRM_IOWR(0x4e, drm_radeon_texture_t)
 
+/* Gamma specific ioctls */
+#define DRM_IOCTL_GAMMA_INIT		DRM_IOW( 0x40, drm_gamma_init_t)
+#define DRM_IOCTL_GAMMA_COPY		DRM_IOW( 0x41, drm_gamma_copy_t)
+
 #ifdef CONFIG_DRM_SIS
 /* SiS specific ioctls */
 #define SIS_IOCTL_FB_ALLOC		DRM_IOWR(0x44, drm_sis_mem_t)