diff --git a/src/radeon.h b/src/radeon.h
index 66b2330..633385f 100644
--- a/src/radeon.h
+++ b/src/radeon.h
@@ -893,6 +893,7 @@ extern void RADEONEngineFlush(ScrnInfoPtr pScrn);
 extern void RADEONEngineInit(ScrnInfoPtr pScrn);
 extern void RADEONEngineReset(ScrnInfoPtr pScrn);
 extern void RADEONEngineRestore(ScrnInfoPtr pScrn);
+extern Bool RADEONWaitforIdlePoll(ScrnInfoPtr pScrn);
 extern uint8_t *RADEONHostDataBlit(ScrnInfoPtr pScrn, unsigned int cpp,
 				 unsigned int w, uint32_t dstPitchOff,
 				 uint32_t *bufPitch, int x, int *y,
diff --git a/src/radeon_accel.c b/src/radeon_accel.c
index 96570e8..d2ae2e6 100644
--- a/src/radeon_accel.c
+++ b/src/radeon_accel.c
@@ -472,6 +472,23 @@ void RADEONEngineInit(ScrnInfoPtr pScrn)
     RADEONEngineRestore(pScrn);
 }
 
+/* really would be better to wait on a timestamp shadowed in memory,
+ * but this will do for now.
+ */
+Bool
+RADEONWaitforIdlePoll(ScrnInfoPtr pScrn)
+{
+    RADEONInfoPtr info = RADEONPTR(pScrn);
+    unsigned char *RADEONMMIO = info->MMIO;
+    uint32_t i;
+
+    for (i = 0; i < 1000000; i++) {
+	if ((INREG(RADEON_RBBM_STATUS) & RADEON_RBBM_ACTIVE) == 0)
+	    return TRUE;
+    }
+    return FALSE;
+}
+
 
 #define ACCEL_MMIO
 #define ACCEL_PREAMBLE()        unsigned char *RADEONMMIO = info->MMIO
@@ -872,6 +889,8 @@ RADEONHostDataBlitCopyPass(
     /* RADEONHostDataBlitCopy can return NULL ! */
     if( (dst==NULL) || (src==NULL)) return;
 
+    RADEONWaitforIdlePoll(pScrn);
+
     if ( dstPitch == srcPitch )
     {
 #if X_BYTE_ORDER == X_BIG_ENDIAN
diff --git a/src/radeon_exa.c b/src/radeon_exa.c
index c4bc1bb..ebe65d1 100644
--- a/src/radeon_exa.c
+++ b/src/radeon_exa.c
@@ -220,17 +220,22 @@ int RADEONBiggerCrtcArea(PixmapPtr pPix)
 }
 
 #if X_BYTE_ORDER == X_BIG_ENDIAN
-
 static unsigned long swapper_surfaces[3];
+#endif /* X_BYTE_ORDER == X_BIG_ENDIAN */
 
 static Bool RADEONPrepareAccess(PixmapPtr pPix, int index)
 {
     RINFO_FROM_SCREEN(pPix->drawable.pScreen);
+#if X_BYTE_ORDER == X_BIG_ENDIAN
     unsigned char *RADEONMMIO = info->MMIO;
     uint32_t offset = exaGetPixmapOffset(pPix);
     int bpp, soff;
     uint32_t size, flags;
+#endif /* X_BYTE_ORDER == X_BIG_ENDIAN */
+
+    RADEONWaitforIdlePoll(pScrn);
 
+#if X_BYTE_ORDER == X_BIG_ENDIAN
     /* Front buffer is always set with proper swappers */
     if (offset == 0)
         return TRUE;
@@ -286,11 +291,13 @@ static Bool RADEONPrepareAccess(PixmapPtr pPix, int index)
     OUTREG(RADEON_SURFACE0_LOWER_BOUND + soff, offset);
     OUTREG(RADEON_SURFACE0_UPPER_BOUND + soff, offset + size - 1);
     swapper_surfaces[index] = offset;
+#endif /* X_BYTE_ORDER == X_BIG_ENDIAN */
     return TRUE;
 }
 
 static void RADEONFinishAccess(PixmapPtr pPix, int index)
 {
+#if X_BYTE_ORDER == X_BIG_ENDIAN
     RINFO_FROM_SCREEN(pPix->drawable.pScreen);
     unsigned char *RADEONMMIO = info->MMIO;
     uint32_t offset = exaGetPixmapOffset(pPix);
@@ -318,9 +325,9 @@ static void RADEONFinishAccess(PixmapPtr pPix, int index)
     OUTREG(RADEON_SURFACE0_LOWER_BOUND + soff, 0);
     OUTREG(RADEON_SURFACE0_UPPER_BOUND + soff, 0);
     swapper_surfaces[index] = 0;
+#endif /* X_BYTE_ORDER == X_BIG_ENDIAN */
 }
 
-#endif /* X_BYTE_ORDER == X_BIG_ENDIAN */
 
 #define ENTER_DRAW(x) TRACE
 #define LEAVE_DRAW(x) TRACE
diff --git a/src/radeon_exa_funcs.c b/src/radeon_exa_funcs.c
index cd97cc6..eb714ab 100644
--- a/src/radeon_exa_funcs.c
+++ b/src/radeon_exa_funcs.c
@@ -408,6 +408,8 @@ RADEONDownloadFromScreenCP(PixmapPtr pSrc, int x, int y, int w, int h,
 	    while ((drmCommandNone(info->dri->drmFD, DRM_RADEON_CP_IDLE) == -EBUSY)
 		   && (i++ < RADEON_TIMEOUT))
 		;
+	    /* make sure the engine is idle */
+	    RADEONWaitforIdlePoll(pScrn);
 
 	    /* Kick next blit */
 	    if (hpass)
@@ -470,10 +472,8 @@ Bool FUNC_NAME(RADEONDrawInit)(ScreenPtr pScreen)
 	info->accel_state->exa->DownloadFromScreen = RADEONDownloadFromScreenCP;
 #endif
 
-#if X_BYTE_ORDER == X_BIG_ENDIAN
     info->accel_state->exa->PrepareAccess = RADEONPrepareAccess;
     info->accel_state->exa->FinishAccess = RADEONFinishAccess;
-#endif /* X_BYTE_ORDER == X_BIG_ENDIAN */
 
     info->accel_state->exa->flags = EXA_OFFSCREEN_PIXMAPS;
     info->accel_state->exa->pixmapOffsetAlign = RADEON_BUFFER_ALIGN + 1;