diff --git a/src/mesa/main/readpix.c b/src/mesa/main/readpix.c
index 4918549..3dd3384 100644
--- a/src/mesa/main/readpix.c
+++ b/src/mesa/main/readpix.c
@@ -220,9 +220,16 @@ fast_read_rgba_pixels_memcpy( struct gl_context *ctx,
    struct gl_renderbuffer *rb = ctx->ReadBuffer->_ColorReadBuffer;
    GLubyte *dst, *map;
    int dstStride, stride, j, texelBytes;
+   GLboolean swizzle_rb = GL_FALSE;
 
-   if (!_mesa_format_matches_format_and_type(rb->Format, format, type,
-                                             ctx->Pack.SwapBytes))
+   if (rb->Format == MESA_FORMAT_RGBA8888_REV &&
+       format == GL_BGRA &&
+       type == GL_UNSIGNED_INT_8_8_8_8_REV &&
+       !ctx->Pack.SwapBytes) {
+      swizzle_rb = GL_TRUE;
+   }
+   else if (!_mesa_format_matches_format_and_type(rb->Format, format, type,
+                                                  ctx->Pack.SwapBytes))
       return GL_FALSE;
 
    /* If the format is unsigned normalized then we can ignore clamping
@@ -247,10 +254,28 @@ fast_read_rgba_pixels_memcpy( struct gl_context *ctx,
    }
 
    texelBytes = _mesa_get_format_bytes(rb->Format);
-   for (j = 0; j < height; j++) {
-      memcpy(dst, map, width * texelBytes);
-      dst += dstStride;
-      map += stride;
+
+   if (swizzle_rb) {
+      /* swap R/B */
+      for (j = 0; j < height; j++) {
+         int i;
+         for (i = 0; i < width; i++) {
+            dst[i*4+0] = map[i*4+2];
+            dst[i*4+1] = map[i*4+1];
+            dst[i*4+2] = map[i*4+0];
+            dst[i*4+3] = map[i*4+3];
+         }
+         dst += dstStride;
+         map += stride;
+      }
+
+   } else {
+      /* just memcpy */
+      for (j = 0; j < height; j++) {
+         memcpy(dst, map, width * texelBytes);
+         dst += dstStride;
+         map += stride;
+      }
    }
 
    ctx->Driver.UnmapRenderbuffer(ctx, rb);