XFree86 4.3.0 merge

7 files changed, 175 insertions, 178 deletions

diff --git a/shared-core/mga_drv.h b/shared-core/mga_drv.h
index 0e650b42..a5085b06 100644
--- a/shared-core/mga_drv.h
+++ b/shared-core/mga_drv.h
@@ -90,14 +90,14 @@ typedef struct drm_mga_private {
 	unsigned int texture_offset;
 	unsigned int texture_size;
 
-	drm_local_map_t *sarea;
-	drm_local_map_t *fb;
-	drm_local_map_t *mmio;
-	drm_local_map_t *status;
-	drm_local_map_t *warp;
-	drm_local_map_t *primary;
-	drm_local_map_t *buffers;
-	drm_local_map_t *agp_textures;
+	drm_map_t *sarea;
+	drm_map_t *fb;
+	drm_map_t *mmio;
+	drm_map_t *status;
+	drm_map_t *warp;
+	drm_map_t *primary;
+	drm_map_t *buffers;
+	drm_map_t *agp_textures;
 } drm_mga_private_t;
 
 				/* mga_dma.c */
@@ -131,30 +131,32 @@ extern int  mga_getparam( DRM_IOCTL_ARGS );
 extern int mga_warp_install_microcode( drm_mga_private_t *dev_priv );
 extern int mga_warp_init( drm_mga_private_t *dev_priv );
 
-#define mga_flush_write_combine()	DRM_WRITEMEMORYBARRIER(dev_priv->primary)
+#define mga_flush_write_combine()	DRM_WRITEMEMORYBARRIER()
+
 
-#if defined(__linux__) && defined(__alpha__)
 #define MGA_BASE( reg )		((unsigned long)(dev_priv->mmio->handle))
 #define MGA_ADDR( reg )		(MGA_BASE(reg) + reg)
 
 #define MGA_DEREF( reg )	*(volatile u32 *)MGA_ADDR( reg )
 #define MGA_DEREF8( reg )	*(volatile u8 *)MGA_ADDR( reg )
 
+#ifdef __alpha__
 #define MGA_READ( reg )		(_MGA_READ((u32 *)MGA_ADDR(reg)))
 #define MGA_READ8( reg )	(_MGA_READ((u8 *)MGA_ADDR(reg)))
-#define MGA_WRITE( reg, val )	do { DRM_WRITEMEMORYBARRIER(dev_priv->mmio); MGA_DEREF( reg ) = val; } while (0)
-#define MGA_WRITE8( reg, val )  do { DRM_WRITEMEMORYBARRIER(dev_priv->mmio); MGA_DEREF8( reg ) = val; } while (0)
+#define MGA_WRITE( reg, val )	do { DRM_WRITEMEMORYBARRIER(); MGA_DEREF( reg ) = val; } while (0)
+#define MGA_WRITE8( reg, val )  do { DRM_WRITEMEMORYBARRIER(); MGA_DEREF8( reg ) = val; } while (0)
 
 static inline u32 _MGA_READ(u32 *addr)
 {
-	DRM_READMEMORYBARRIER(dev_priv->mmio);
+	DRM_READMEMORYBARRIER();
 	return *(volatile u32 *)addr;
 }
+
 #else
-#define MGA_READ8( reg )	DRM_READ8(dev_priv->mmio, (reg))
-#define MGA_READ( reg )		DRM_READ32(dev_priv->mmio, (reg))
-#define MGA_WRITE8( reg, val )  DRM_WRITE8(dev_priv->mmio, (reg), (val))
-#define MGA_WRITE( reg, val )	DRM_WRITE32(dev_priv->mmio, (reg), (val))
+#define MGA_READ( reg )		MGA_DEREF( reg )
+#define MGA_READ8( reg )	MGA_DEREF8( reg )
+#define MGA_WRITE( reg, val )	do { MGA_DEREF( reg ) = val; } while (0)
+#define MGA_WRITE8( reg, val )  do { MGA_DEREF8( reg ) = val; } while (0)
 #endif
 
 #define DWGREG0 	0x1c00
diff --git a/shared-core/r128_cce.c b/shared-core/r128_cce.c
index 8d305b75..5175885e 100644
--- a/shared-core/r128_cce.c
+++ b/shared-core/r128_cce.c
@@ -579,7 +579,6 @@ static int r128_do_init_cce( drm_device_t *dev, drm_r128_init_t *init )
 		(dev_priv->ring.size / sizeof(u32)) - 1;
 
 	dev_priv->ring.high_mark = 128;
-	dev_priv->ring.ring_rptr = dev_priv->ring_rptr;
 
 	dev_priv->sarea_priv->last_frame = 0;
 	R128_WRITE( R128_LAST_FRAME_REG, dev_priv->sarea_priv->last_frame );
diff --git a/shared-core/r128_drv.h b/shared-core/r128_drv.h
index 5e6f1215..763fcb3a 100644
--- a/shared-core/r128_drv.h
+++ b/shared-core/r128_drv.h
@@ -34,8 +34,8 @@
 #ifndef __R128_DRV_H__
 #define __R128_DRV_H__
 
-#define GET_RING_HEAD(ring)		DRM_READ32(  (ring)->ring_rptr, 0 ) /* (ring)->head */
-#define SET_RING_HEAD(ring,val)		DRM_WRITE32( (ring)->ring_rptr, 0, (val) ) /* (ring)->head */
+#define GET_RING_HEAD(ring)		DRM_READ32(  (volatile u32 *) (ring)->head )
+#define SET_RING_HEAD(ring,val)		DRM_WRITE32( (volatile u32 *) (ring)->head, (val) )
 
 typedef struct drm_r128_freelist {
    	unsigned int age;
@@ -56,7 +56,6 @@ typedef struct drm_r128_ring_buffer {
 	int space;
 
 	int high_mark;
-	drm_local_map_t *ring_rptr;
 } drm_r128_ring_buffer_t;
 
 typedef struct drm_r128_private {
@@ -99,13 +98,13 @@ typedef struct drm_r128_private {
 	u32 depth_pitch_offset_c;
 	u32 span_pitch_offset_c;
 
-	drm_local_map_t *sarea;
-	drm_local_map_t *fb;
-	drm_local_map_t *mmio;
-	drm_local_map_t *cce_ring;
-	drm_local_map_t *ring_rptr;
-	drm_local_map_t *buffers;
-	drm_local_map_t *agp_textures;
+	drm_map_t *sarea;
+	drm_map_t *fb;
+	drm_map_t *mmio;
+	drm_map_t *cce_ring;
+	drm_map_t *ring_rptr;
+	drm_map_t *buffers;
+	drm_map_t *agp_textures;
 } drm_r128_private_t;
 
 typedef struct drm_r128_buf_priv {
@@ -371,10 +370,15 @@ extern int r128_cce_indirect( DRM_IOCTL_ARGS );
 
 #define R128_PERFORMANCE_BOXES		0
 
-#define R128_READ(reg)		DRM_READ32(  dev_priv->mmio, (reg) )
-#define R128_WRITE(reg,val)	DRM_WRITE32( dev_priv->mmio, (reg), (val) )
-#define R128_READ8(reg)		DRM_READ8(   dev_priv->mmio, (reg) )
-#define R128_WRITE8(reg,val)	DRM_WRITE8(  dev_priv->mmio, (reg), (val) )
+
+#define R128_BASE(reg)		((unsigned long)(dev_priv->mmio->handle))
+#define R128_ADDR(reg)		(R128_BASE( reg ) + reg)
+
+#define R128_READ(reg)		DRM_READ32(  (volatile u32 *) R128_ADDR(reg) )
+#define R128_WRITE(reg,val)	DRM_WRITE32( (volatile u32 *) R128_ADDR(reg), (val) )
+
+#define R128_READ8(reg)		DRM_READ8(  (volatile u8 *) R128_ADDR(reg) )
+#define R128_WRITE8(reg,val)	DRM_WRITE8( (volatile u8 *) R128_ADDR(reg), (val) )
 
 #define R128_WRITE_PLL(addr,val)					\
 do {									\
@@ -449,7 +453,7 @@ do {									\
 #if defined(__powerpc__)
 #define r128_flush_write_combine()	(void) GET_RING_HEAD( &dev_priv->ring )
 #else
-#define r128_flush_write_combine()	DRM_WRITEMEMORYBARRIER(dev_priv->ring_rptr)
+#define r128_flush_write_combine()	DRM_WRITEMEMORYBARRIER()
 #endif
 
 
diff --git a/shared-core/r128_state.c b/shared-core/r128_state.c
index 20307c04..68f73061 100644
--- a/shared-core/r128_state.c
+++ b/shared-core/r128_state.c
@@ -896,7 +896,7 @@ static int r128_cce_dispatch_write_span( drm_device_t *dev,
 	int count, x, y;
 	u32 *buffer;
 	u8 *mask;
-	int i, buffer_size, mask_size;
+	int i;
 	RING_LOCALS;
 	DRM_DEBUG( "\n" );
 
@@ -908,25 +908,25 @@ static int r128_cce_dispatch_write_span( drm_device_t *dev,
 		return DRM_ERR(EFAULT);
 	}
 
-	buffer_size = depth->n * sizeof(u32);
-	buffer = DRM_MALLOC( buffer_size );
+	buffer = DRM_MALLOC( depth->n * sizeof(u32) );
 	if ( buffer == NULL )
 		return DRM_ERR(ENOMEM);
-	if ( DRM_COPY_FROM_USER( buffer, depth->buffer, buffer_size ) ) {
-		DRM_FREE( buffer, buffer_size);
+	if ( DRM_COPY_FROM_USER( buffer, depth->buffer,
+			     depth->n * sizeof(u32) ) ) {
+		DRM_FREE( buffer );
 		return DRM_ERR(EFAULT);
 	}
 
-	mask_size = depth->n * sizeof(u8);
 	if ( depth->mask ) {
-		mask = DRM_MALLOC( mask_size );
+		mask = DRM_MALLOC( depth->n * sizeof(u8) );
 		if ( mask == NULL ) {
-			DRM_FREE( buffer, buffer_size );
+			DRM_FREE( buffer );
 			return DRM_ERR(ENOMEM);
 		}
-		if ( DRM_COPY_FROM_USER( mask, depth->mask, mask_size ) ) {
-			DRM_FREE( buffer, buffer_size );
-			DRM_FREE( mask, mask_size );
+		if ( DRM_COPY_FROM_USER( mask, depth->mask,
+				     depth->n * sizeof(u8) ) ) {
+			DRM_FREE( buffer );
+			DRM_FREE( mask );
 			return DRM_ERR(EFAULT);
 		}
 
@@ -953,7 +953,7 @@ static int r128_cce_dispatch_write_span( drm_device_t *dev,
 			}
 		}
 
-		DRM_FREE( mask, mask_size );
+		DRM_FREE( mask );
 	} else {
 		for ( i = 0 ; i < count ; i++, x++ ) {
 			BEGIN_RING( 6 );
@@ -977,7 +977,7 @@ static int r128_cce_dispatch_write_span( drm_device_t *dev,
 		}
 	}
 
-	DRM_FREE( buffer, buffer_size );
+	DRM_FREE( buffer );
 
 	return 0;
 }
@@ -989,62 +989,60 @@ static int r128_cce_dispatch_write_pixels( drm_device_t *dev,
 	int count, *x, *y;
 	u32 *buffer;
 	u8 *mask;
-	int i, xbuf_size, ybuf_size, buffer_size, mask_size;
+	int i;
 	RING_LOCALS;
 	DRM_DEBUG( "\n" );
 
 	count = depth->n;
 
-	xbuf_size = count * sizeof(*x);
-	ybuf_size = count * sizeof(*y);
-	x = DRM_MALLOC( xbuf_size );
+	x = DRM_MALLOC( count * sizeof(*x) );
 	if ( x == NULL ) {
 		return DRM_ERR(ENOMEM);
 	}
-	y = DRM_MALLOC( ybuf_size );
+	y = DRM_MALLOC( count * sizeof(*y) );
 	if ( y == NULL ) {
-		DRM_FREE( x, xbuf_size );
+		DRM_FREE( x );
 		return DRM_ERR(ENOMEM);
 	}
-	if ( DRM_COPY_FROM_USER( x, depth->x, xbuf_size ) ) {
-		DRM_FREE( x, xbuf_size );
-		DRM_FREE( y, ybuf_size );
+	if ( DRM_COPY_FROM_USER( x, depth->x, count * sizeof(int) ) ) {
+		DRM_FREE( x );
+		DRM_FREE( y );
 		return DRM_ERR(EFAULT);
 	}
-	if ( DRM_COPY_FROM_USER( y, depth->y, xbuf_size ) ) {
-		DRM_FREE( x, xbuf_size );
-		DRM_FREE( y, ybuf_size );
+	if ( DRM_COPY_FROM_USER( y, depth->y, count * sizeof(int) ) ) {
+		DRM_FREE( x );
+		DRM_FREE( y );
 		return DRM_ERR(EFAULT);
 	}
 
-	buffer_size = depth->n * sizeof(u32);
-	buffer = DRM_MALLOC( buffer_size );
+	buffer = DRM_MALLOC( depth->n * sizeof(u32) );
 	if ( buffer == NULL ) {
-		DRM_FREE( x, xbuf_size );
-		DRM_FREE( y, ybuf_size );
+		DRM_FREE( x );
+		DRM_FREE( y );
 		return DRM_ERR(ENOMEM);
 	}
-	if ( DRM_COPY_FROM_USER( buffer, depth->buffer, buffer_size ) ) {
-		DRM_FREE( x, xbuf_size );
-		DRM_FREE( y, ybuf_size );
-		DRM_FREE( buffer, buffer_size );
+	if ( DRM_COPY_FROM_USER( buffer, depth->buffer,
+			     depth->n * sizeof(u32) ) ) {
+		DRM_FREE( x );
+		DRM_FREE( y );
+		DRM_FREE( buffer );
 		return DRM_ERR(EFAULT);
 	}
 
 	if ( depth->mask ) {
-		mask_size = depth->n * sizeof(u8);
-		mask = DRM_MALLOC( mask_size );
+		mask = DRM_MALLOC( depth->n * sizeof(u8) );
 		if ( mask == NULL ) {
-			DRM_FREE( x, xbuf_size );
-			DRM_FREE( y, ybuf_size );
-			DRM_FREE( buffer, buffer_size );
+			DRM_FREE( x );
+			DRM_FREE( y );
+			DRM_FREE( buffer );
 			return DRM_ERR(ENOMEM);
 		}
-		if ( DRM_COPY_FROM_USER( mask, depth->mask, mask_size ) ) {
-			DRM_FREE( x, xbuf_size );
-			DRM_FREE( y, ybuf_size );
-			DRM_FREE( buffer, buffer_size );
-			DRM_FREE( mask, mask_size );
+		if ( DRM_COPY_FROM_USER( mask, depth->mask,
+				     depth->n * sizeof(u8) ) ) {
+			DRM_FREE( x );
+			DRM_FREE( y );
+			DRM_FREE( buffer );
+			DRM_FREE( mask );
 			return DRM_ERR(EFAULT);
 		}
 
@@ -1071,7 +1069,7 @@ static int r128_cce_dispatch_write_pixels( drm_device_t *dev,
 			}
 		}
 
-		DRM_FREE( mask, mask_size );
+		DRM_FREE( mask );
 	} else {
 		for ( i = 0 ; i < count ; i++ ) {
 			BEGIN_RING( 6 );
@@ -1095,9 +1093,9 @@ static int r128_cce_dispatch_write_pixels( drm_device_t *dev,
 		}
 	}
 
-	DRM_FREE( x, xbuf_size );
-	DRM_FREE( y, ybuf_size );
-	DRM_FREE( buffer, buffer_size );
+	DRM_FREE( x );
+	DRM_FREE( y );
+	DRM_FREE( buffer );
 
 	return 0;
 }
@@ -1148,7 +1146,7 @@ static int r128_cce_dispatch_read_pixels( drm_device_t *dev,
 {
 	drm_r128_private_t *dev_priv = dev->dev_private;
 	int count, *x, *y;
-	int i, xbuf_size, ybuf_size;
+	int i;
 	RING_LOCALS;
 	DRM_DEBUG( "%s\n", __FUNCTION__ );
 
@@ -1157,25 +1155,23 @@ static int r128_cce_dispatch_read_pixels( drm_device_t *dev,
 		count = dev_priv->depth_pitch;
 	}
 
-	xbuf_size = count * sizeof(*x);
-	ybuf_size = count * sizeof(*y);
-	x = DRM_MALLOC( xbuf_size );
+	x = DRM_MALLOC( count * sizeof(*x) );
 	if ( x == NULL ) {
 		return DRM_ERR(ENOMEM);
 	}
-	y = DRM_MALLOC( ybuf_size );
+	y = DRM_MALLOC( count * sizeof(*y) );
 	if ( y == NULL ) {
-		DRM_FREE( x, xbuf_size );
+		DRM_FREE( x );
 		return DRM_ERR(ENOMEM);
 	}
-	if ( DRM_COPY_FROM_USER( x, depth->x, xbuf_size ) ) {
-		DRM_FREE( x, xbuf_size );
-		DRM_FREE( y, ybuf_size );
+	if ( DRM_COPY_FROM_USER( x, depth->x, count * sizeof(int) ) ) {
+		DRM_FREE( x );
+		DRM_FREE( y );
 		return DRM_ERR(EFAULT);
 	}
-	if ( DRM_COPY_FROM_USER( y, depth->y, ybuf_size ) ) {
-		DRM_FREE( x, xbuf_size );
-		DRM_FREE( y, ybuf_size );
+	if ( DRM_COPY_FROM_USER( y, depth->y, count * sizeof(int) ) ) {
+		DRM_FREE( x );
+		DRM_FREE( y );
 		return DRM_ERR(EFAULT);
 	}
 
@@ -1203,8 +1199,8 @@ static int r128_cce_dispatch_read_pixels( drm_device_t *dev,
 		ADVANCE_RING();
 	}
 
-	DRM_FREE( x, xbuf_size );
-	DRM_FREE( y, ybuf_size );
+	DRM_FREE( x );
+	DRM_FREE( y );
 
 	return 0;
 }
diff --git a/shared-core/radeon_cp.c b/shared-core/radeon_cp.c
index 89c9eab1..b4d0e4b6 100644
--- a/shared-core/radeon_cp.c
+++ b/shared-core/radeon_cp.c
@@ -36,11 +36,6 @@
 
 #define RADEON_FIFO_DEBUG	0
 
-#if defined(__alpha__) || defined(__powerpc__)
-# define PCIGART_ENABLED
-#else
-# undef PCIGART_ENABLED
-#endif
 
 
 /* CP microcode (from ATI) */
@@ -885,6 +880,7 @@ static void radeon_cp_init_ring_buffer( drm_device_t *dev,
 
 	/* Set the write pointer delay */
 	RADEON_WRITE( RADEON_CP_RB_WPTR_DELAY, 0 );
+	RADEON_READ( RADEON_CP_RB_WPTR_DELAY ); /* read back to propagate */
 
 	/* Initialize the ring buffer's read and write pointers */
 	cur_read_ptr = RADEON_READ( RADEON_CP_RB_RPTR );
@@ -926,11 +922,11 @@ static void radeon_cp_init_ring_buffer( drm_device_t *dev,
 	RADEON_WRITE( RADEON_SCRATCH_UMSK, 0x7 );
 
 	/* Writeback doesn't seem to work everywhere, test it first */
-	DRM_WRITE32( dev_priv->ring_rptr, RADEON_SCRATCHOFF(1), 0 );
+	DRM_WRITE32( &dev_priv->scratch[1], 0 );
 	RADEON_WRITE( RADEON_SCRATCH_REG1, 0xdeadbeef );
 
 	for ( tmp = 0 ; tmp < dev_priv->usec_timeout ; tmp++ ) {
-		if ( DRM_READ32( dev_priv->ring_rptr, RADEON_SCRATCHOFF(1) ) == 0xdeadbeef )
+		if ( DRM_READ32( &dev_priv->scratch[1] ) == 0xdeadbeef )
 			break;
 		DRM_UDELAY( 1 );
 	}
@@ -990,17 +986,6 @@ static int radeon_do_init_cp( drm_device_t *dev, drm_radeon_init_t *init )
 
 	dev_priv->is_pci = init->is_pci;
 
-#if !defined(PCIGART_ENABLED)
-	/* PCI support is not 100% working, so we disable it here.
-	 */
-	if ( dev_priv->is_pci ) {
-		DRM_ERROR( "PCI GART not yet supported for Radeon!\n" );
-		dev->dev_private = (void *)dev_priv;
-		radeon_do_cleanup_cp(dev);
-		return DRM_ERR(EINVAL);
-	}
-#endif
-
 	if ( dev_priv->is_pci && !dev->sg ) {
 		DRM_ERROR( "PCI GART memory not allocated!\n" );
 		dev->dev_private = (void *)dev_priv;
@@ -1217,7 +1202,6 @@ static int radeon_do_init_cp( drm_device_t *dev, drm_radeon_init_t *init )
 		(dev_priv->ring.size / sizeof(u32)) - 1;
 
 	dev_priv->ring.high_mark = RADEON_RING_HIGH_MARK;
-	dev_priv->ring.ring_rptr = dev_priv->ring_rptr;
 
 #if __REALLY_HAVE_SG
 	if ( dev_priv->is_pci ) {
@@ -1355,9 +1339,6 @@ int radeon_cp_stop( DRM_IOCTL_ARGS )
 
 	DRM_COPY_FROM_USER_IOCTL( stop, (drm_radeon_cp_stop_t *)data, sizeof(stop) );
 
-	if (!dev_priv->cp_running)
-		return 0;
-
 	/* Flush any pending CP commands.  This ensures any outstanding
 	 * commands are exectuted by the engine before we turn it off.
 	 */
@@ -1385,39 +1366,6 @@ int radeon_cp_stop( DRM_IOCTL_ARGS )
 	return 0;
 }
 
-
-void radeon_do_release( drm_device_t *dev )
-{
-	drm_radeon_private_t *dev_priv = dev->dev_private;
-	int ret;
-
-	if (dev_priv) {
-		if (dev_priv->cp_running) {
-			/* Stop the cp */
-			while ((ret = radeon_do_cp_idle( dev_priv )) != 0) {
-				DRM_DEBUG("radeon_do_cp_idle %d\n", ret);
-#ifdef __linux__
-				schedule();
-#else
-				tsleep(&ret, PZERO, "rdnrel", 1);
-#endif
-			}
-			radeon_do_cp_stop( dev_priv );
-			radeon_do_engine_reset( dev );
-		}
-
-		/* Disable *all* interrupts */
-		RADEON_WRITE( RADEON_GEN_INT_CNTL, 0 );
-
-		/* Free memory heap structures */
-		radeon_mem_takedown( &(dev_priv->agp_heap) );
-		radeon_mem_takedown( &(dev_priv->fb_heap) );
-
-		/* deallocate kernel resources */
-		radeon_do_cleanup_cp( dev );
-	}
-}
-
 /* Just reset the CP ring.  Called as part of an X Server engine reset.
  */
 int radeon_cp_reset( DRM_IOCTL_ARGS )
@@ -1449,6 +1397,9 @@ int radeon_cp_idle( DRM_IOCTL_ARGS )
 
 	LOCK_TEST_WITH_RETURN( dev );
 
+/* 	if (dev->irq)  */
+/* 		radeon_emit_and_wait_irq( dev ); */
+
 	return radeon_do_cp_idle( dev_priv );
 }
 
@@ -1543,7 +1494,7 @@ drm_buf_t *radeon_freelist_get( drm_device_t *dev )
 	drm_buf_t *buf;
 	int i, t;
 	int start;
-	u32 done_age = DRM_READ32(dev_priv->ring_rptr, RADEON_SCRATCHOFF(1));
+	u32 done_age = DRM_READ32(&dev_priv->scratch[1]);
 
 	if ( ++dev_priv->last_buf >= dma->buf_count )
 		dev_priv->last_buf = 0;
diff --git a/shared-core/radeon_drv.h b/shared-core/radeon_drv.h
index 502ba89b..635ad14b 100644
--- a/shared-core/radeon_drv.h
+++ b/shared-core/radeon_drv.h
@@ -31,8 +31,8 @@
 #ifndef __RADEON_DRV_H__
 #define __RADEON_DRV_H__
 
-#define GET_RING_HEAD(ring)		DRM_READ32(  (ring)->ring_rptr, 0 ) /* (ring)->head */
-#define SET_RING_HEAD(ring,val)		DRM_WRITE32( (ring)->ring_rptr, 0, (val) ) /* (ring)->head */
+#define GET_RING_HEAD(ring)		DRM_READ32(  (volatile u32 *) (ring)->head )
+#define SET_RING_HEAD(ring,val)		DRM_WRITE32( (volatile u32 *) (ring)->head , (val))
 
 typedef struct drm_radeon_freelist {
    	unsigned int age;
@@ -53,7 +53,6 @@ typedef struct drm_radeon_ring_buffer {
 	int space;
 
 	int high_mark;
-	drm_local_map_t *ring_rptr;
 } drm_radeon_ring_buffer_t;
 
 typedef struct drm_radeon_depth_clear_t {
@@ -127,13 +126,13 @@ typedef struct drm_radeon_private {
 
 	drm_radeon_depth_clear_t depth_clear;
 
-	drm_local_map_t *sarea;
-	drm_local_map_t *fb;
-	drm_local_map_t *mmio;
-	drm_local_map_t *cp_ring;
-	drm_local_map_t *ring_rptr;
-	drm_local_map_t *buffers;
-	drm_local_map_t *agp_textures;
+	drm_map_t *sarea;
+	drm_map_t *fb;
+	drm_map_t *mmio;
+	drm_map_t *cp_ring;
+	drm_map_t *ring_rptr;
+	drm_map_t *buffers;
+	drm_map_t *agp_textures;
 
 	struct mem_block *agp_heap;
 	struct mem_block *fb_heap;
@@ -194,7 +193,6 @@ extern int radeon_emit_and_wait_irq(drm_device_t *dev);
 extern int radeon_wait_irq(drm_device_t *dev, int swi_nr);
 extern int radeon_emit_irq(drm_device_t *dev);
 
-extern void radeon_do_release(drm_device_t *dev);
 
 /* Flags for stats.boxes
  */
@@ -268,10 +266,8 @@ extern void radeon_do_release(drm_device_t *dev);
 #define RADEON_SCRATCH_UMSK		0x0770
 #define RADEON_SCRATCH_ADDR		0x0774
 
-#define RADEON_SCRATCHOFF( x )		(RADEON_SCRATCH_REG_OFFSET + 4*(x))
-
 #define GET_SCRATCH( x )	(dev_priv->writeback_works			\
-				? DRM_READ32( dev_priv->ring_rptr, RADEON_SCRATCHOFF(x) ) \
+				? DRM_READ32( &dev_priv->scratch[(x)] )		\
 				: RADEON_READ( RADEON_SCRATCH_REG0 + 4*(x) ) )
 
 
@@ -690,10 +686,15 @@ extern void radeon_do_release(drm_device_t *dev);
 
 #define RADEON_RING_HIGH_MARK		128
 
-#define RADEON_READ(reg)	DRM_READ32(  dev_priv->mmio, (reg) )
-#define RADEON_WRITE(reg,val)	DRM_WRITE32( dev_priv->mmio, (reg), (val) )
-#define RADEON_READ8(reg)	DRM_READ8(  dev_priv->mmio, (reg) )
-#define RADEON_WRITE8(reg,val)	DRM_WRITE8( dev_priv->mmio, (reg), (val) )
+
+#define RADEON_BASE(reg)	((unsigned long)(dev_priv->mmio->handle))
+#define RADEON_ADDR(reg)	(RADEON_BASE( reg ) + reg)
+
+#define RADEON_READ(reg)	DRM_READ32(  (volatile u32 *) RADEON_ADDR(reg) )
+#define RADEON_WRITE(reg,val)	DRM_WRITE32( (volatile u32 *) RADEON_ADDR(reg), (val) )
+
+#define RADEON_READ8(reg)	DRM_READ8(  (volatile u8 *) RADEON_ADDR(reg) )
+#define RADEON_WRITE8(reg,val)	DRM_WRITE8( (volatile u8 *) RADEON_ADDR(reg), (val) )
 
 #define RADEON_WRITE_PLL( addr, val )					\
 do {									\
@@ -826,10 +827,43 @@ do {									\
 
 #define RING_LOCALS	int write, _nr; unsigned int mask; u32 *ring;
 
-#define BEGIN_RING( n ) do {						\
+#if defined(__alpha__)
+#  define RADEON_PAD_RING 16 /* pad ring requests to 16 lw boundaries */
+#else
+#  define RADEON_PAD_RING 0
+#endif
+
+#if RADEON_PAD_RING
+# define radeon_pad_size(n) 	\
+	(((RADEON_PAD_RING) - ((n) % (RADEON_PAD_RING))) % (RADEON_PAD_RING))
+# define radeon_pad_ring() do {						\
+	if (RADEON_VERBOSE) {						\
+		DRM_INFO("Padding ring from %d (%x) with %d words\n", 	\
+			 write, write, radeon_pad_size(write));		\
+	}								\
+	switch (radeon_pad_size(write)) {			      	\
+	case 0:  /* aligned */					      	\
+		break;						      	\
+	case 1:  /* 1 word */					      	\
+		OUT_RING(CP_PACKET2());				      	\
+		break;						      	\
+	default: /* >= 2 words */				      	\
+		OUT_RING(CP_PACKET3(0x1000, radeon_pad_size(write) - 1));\
+		write = (write + radeon_pad_size(write)) & mask;	\
+		write &= mask;						\
+		break;						      	\
+	}							      	\
+} while(0)
+#else
+# define radeon_pad_size(n) 0
+# define radeon_pad_ring()
+#endif
+
+#define BEGIN_RING( req_n ) do {					\
+	int n = req_n + radeon_pad_size(req_n);				\
 	if ( RADEON_VERBOSE ) {						\
-		DRM_INFO( "BEGIN_RING( %d ) in %s\n",			\
-			   n, __FUNCTION__ );				\
+		DRM_INFO( "BEGIN_RING( %d (%d) ) in %s\n",		\
+			   n, req_n, __FUNCTION__ );			\
 	}								\
 	if ( dev_priv->ring.space <= (n) * sizeof(u32) ) {		\
                 COMMIT_RING();						\
@@ -846,6 +880,7 @@ do {									\
 		DRM_INFO( "ADVANCE_RING() wr=0x%06x tail=0x%06x\n",	\
 			  write, dev_priv->ring.tail );			\
 	}								\
+	radeon_pad_ring();						\
 	if (((dev_priv->ring.tail + _nr) & mask) != write) {		\
 		DRM_ERROR( 						\
 			"ADVANCE_RING(): mismatch: nr: %x write: %x line: %d\n",	\
@@ -857,7 +892,7 @@ do {									\
 
 #define COMMIT_RING() do {						\
 	/* Flush writes to ring */					\
-	DRM_READMEMORYBARRIER(dev_priv->mmio);					\
+	DRM_READMEMORYBARRIER();					\
 	GET_RING_HEAD( &dev_priv->ring );				\
 	RADEON_WRITE( RADEON_CP_RB_WPTR, dev_priv->ring.tail );		\
 	/* read from PCI bus to ensure correct posting */		\
diff --git a/shared-core/radeon_mem.c b/shared-core/radeon_mem.c
index 3a2c8969..5c07c1af 100644
--- a/shared-core/radeon_mem.c
+++ b/shared-core/radeon_mem.c
@@ -118,7 +118,7 @@ static void free_block( struct mem_block *p )
 		p->size += q->size;
 		p->next = q->next;
 		p->next->prev = p;
-		DRM_FREE(q, sizeof(*q));
+		DRM_FREE(q);
 	}
 
 	if (p->prev->pid == 0) {
@@ -126,10 +126,20 @@ static void free_block( struct mem_block *p )
 		q->size += p->size;
 		q->next = p->next;
 		q->next->prev = q;
-		DRM_FREE(p, sizeof(*q));
+		DRM_FREE(p);
 	}
 }
 
+static void print_heap( struct mem_block *heap )
+{
+	struct mem_block *p;
+
+	for (p = heap->next ; p != heap ; p = p->next) 
+		DRM_DEBUG("0x%x..0x%x (0x%x) -- owner %d\n",
+			  p->start, p->start + p->size,
+			  p->size, p->pid);
+}
+
 /* Initialize.  How to check for an uninitialized heap?
  */
 static int init_heap(struct mem_block **heap, int start, int size)
@@ -141,7 +151,7 @@ static int init_heap(struct mem_block **heap, int start, int size)
 	
 	*heap = DRM_MALLOC(sizeof(**heap));
 	if (!*heap) {
-		DRM_FREE( blocks, sizeof(*blocks) );
+		DRM_FREE( blocks );
 		return -ENOMEM;
 	}
 
@@ -181,7 +191,7 @@ void radeon_mem_release( struct mem_block *heap )
 			p->size += q->size;
 			p->next = q->next;
 			p->next->prev = p;
-			DRM_FREE(q, sizeof(*q));
+			DRM_FREE(q);
 		}
 	}
 }
@@ -198,10 +208,10 @@ void radeon_mem_takedown( struct mem_block **heap )
 	for (p = (*heap)->next ; p != *heap ; ) {
 		struct mem_block *q = p;
 		p = p->next;
-		DRM_FREE(q, sizeof(*q));
+		DRM_FREE(q);
 	}
 
-	DRM_FREE( *heap, sizeof(**heap) );
+	DRM_FREE( *heap );
 	*heap = 0;
 }