i965: Enable resource streamer for the batchbuffer
[mesa.git] / src / mesa / drivers / dri / i965 / intel_screen.c
index 61972dfe4614236345663fab6ffdd6a8a8509717..1470b059d9b314b73fb5fe517210942dc8dcba38 100644 (file)
@@ -1,8 +1,8 @@
 /**************************************************************************
- * 
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ *
+ * Copyright 2003 VMware, Inc.
  * All Rights Reserved.
- * 
+ *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the
  * "Software"), to deal in the Software without restriction, including
  * distribute, sub license, and/or sell copies of the Software, and to
  * permit persons to whom the Software is furnished to do so, subject to
  * the following conditions:
- * 
+ *
  * The above copyright notice and this permission notice (including the
  * next paragraph) shall be included in all copies or substantial portions
  * of the Software.
- * 
+ *
  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
  * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
  * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- * 
+ *
  **************************************************************************/
 
 #include <errno.h>
 #include <time.h>
+#include <unistd.h>
 #include "main/glheader.h"
 #include "main/context.h"
 #include "main/framebuffer.h"
 #include "main/fbobject.h"
 #include "main/version.h"
 #include "swrast/s_renderbuffer.h"
+#include "util/ralloc.h"
+#include "brw_shader.h"
+#include "glsl/nir/nir.h"
 
 #include "utils.h"
 #include "xmlpool.h"
 
-PUBLIC const char __driConfigOptions[] =
-   DRI_CONF_BEGIN
+static const __DRIconfigOptionsExtension brw_config_options = {
+   .base = { __DRI_CONFIG_OPTIONS, 1 },
+   .xml =
+DRI_CONF_BEGIN
    DRI_CONF_SECTION_PERFORMANCE
       DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_ALWAYS_SYNC)
       /* Options correspond to DRI_CONF_BO_REUSE_DISABLED,
@@ -57,16 +63,17 @@ PUBLIC const char __driConfigOptions[] =
       DRI_CONF_OPT_BEGIN_B(hiz, "true")
         DRI_CONF_DESC(en, "Enable Hierarchical Z on gen6+")
       DRI_CONF_OPT_END
-
-      DRI_CONF_OPT_BEGIN_B(early_z, "false")
-        DRI_CONF_DESC(en, "Enable early Z in classic mode (unstable, 945-only).")
-      DRI_CONF_OPT_END
-
    DRI_CONF_SECTION_END
+
    DRI_CONF_SECTION_QUALITY
       DRI_CONF_FORCE_S3TC_ENABLE("false")
-      DRI_CONF_ALLOW_LARGE_TEXTURES(2)
+
+      DRI_CONF_OPT_BEGIN(clamp_max_samples, int, -1)
+              DRI_CONF_DESC(en, "Clamp the value of GL_MAX_SAMPLES to the "
+                            "given integer. If negative, then do not clamp.")
+      DRI_CONF_OPT_END
    DRI_CONF_SECTION_END
+
    DRI_CONF_SECTION_DEBUG
       DRI_CONF_NO_RAST("false")
       DRI_CONF_ALWAYS_FLUSH_BATCH("false")
@@ -75,33 +82,28 @@ PUBLIC const char __driConfigOptions[] =
       DRI_CONF_FORCE_GLSL_EXTENSIONS_WARN("false")
       DRI_CONF_DISABLE_GLSL_LINE_CONTINUATIONS("false")
       DRI_CONF_DISABLE_BLEND_FUNC_EXTENDED("false")
+      DRI_CONF_ALLOW_GLSL_EXTENSION_DIRECTIVE_MIDSHADER("false")
 
       DRI_CONF_OPT_BEGIN_B(shader_precompile, "true")
         DRI_CONF_DESC(en, "Perform code generation at shader link time.")
       DRI_CONF_OPT_END
    DRI_CONF_SECTION_END
-DRI_CONF_END;
-
-const GLuint __driNConfigOptions = 14;
+DRI_CONF_END
+};
 
 #include "intel_batchbuffer.h"
 #include "intel_buffers.h"
 #include "intel_bufmgr.h"
-#include "intel_chipset.h"
 #include "intel_fbo.h"
 #include "intel_mipmap_tree.h"
 #include "intel_screen.h"
 #include "intel_tex.h"
-#include "intel_regions.h"
+#include "intel_image.h"
 
 #include "brw_context.h"
 
 #include "i915_drm.h"
 
-#ifdef USE_NEW_INTERFACE
-static PFNGLXCREATECONTEXTMODES create_context_modes = NULL;
-#endif /*USE_NEW_INTERFACE */
-
 /**
  * For debugging purposes, this returns a time in seconds.
  */
@@ -128,29 +130,28 @@ aub_dump_bmp(struct gl_context *ctx)
         enum aub_dump_bmp_format format;
 
         switch (irb->Base.Base.Format) {
-        case MESA_FORMAT_ARGB8888:
-        case MESA_FORMAT_XRGB8888:
+        case MESA_FORMAT_B8G8R8A8_UNORM:
+        case MESA_FORMAT_B8G8R8X8_UNORM:
            format = AUB_DUMP_BMP_FORMAT_ARGB_8888;
            break;
         default:
            continue;
         }
 
-         assert(irb->mt->region->pitch % irb->mt->region->cpp == 0);
-        drm_intel_gem_bo_aub_dump_bmp(irb->mt->region->bo,
+         drm_intel_gem_bo_aub_dump_bmp(irb->mt->bo,
                                       irb->draw_x,
                                       irb->draw_y,
                                       irb->Base.Base.Width,
                                       irb->Base.Base.Height,
                                       format,
-                                      irb->mt->region->pitch,
+                                      irb->mt->pitch,
                                       0);
       }
    }
 }
 
 static const __DRItexBufferExtension intelTexBufferExtension = {
-   .base = { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION },
+   .base = { __DRI_TEX_BUFFER, 3 },
 
    .setTexBuffer        = intelSetTexBuffer,
    .setTexBuffer2       = intelSetTexBuffer2,
@@ -158,38 +159,76 @@ static const __DRItexBufferExtension intelTexBufferExtension = {
 };
 
 static void
-intelDRI2Flush(__DRIdrawable *drawable)
+intel_dri2_flush_with_flags(__DRIcontext *cPriv,
+                            __DRIdrawable *dPriv,
+                            unsigned flags,
+                            enum __DRI2throttleReason reason)
 {
-   GET_CURRENT_CONTEXT(ctx);
-   struct intel_context *intel = intel_context(ctx);
-   if (intel == NULL)
+   struct brw_context *brw = cPriv->driverPrivate;
+
+   if (!brw)
       return;
 
-   intel_resolve_for_dri2_flush(intel, drawable);
-   intel->need_throttle = true;
+   struct gl_context *ctx = &brw->ctx;
+
+   FLUSH_VERTICES(ctx, 0);
+
+   if (flags & __DRI2_FLUSH_DRAWABLE)
+      intel_resolve_for_dri2_flush(brw, dPriv);
+
+   if (reason == __DRI2_THROTTLE_SWAPBUFFER)
+      brw->need_swap_throttle = true;
+   if (reason == __DRI2_THROTTLE_FLUSHFRONT)
+      brw->need_flush_throttle = true;
 
-   if (intel->batch.used)
-      intel_batchbuffer_flush(intel);
+   intel_batchbuffer_flush(brw);
 
    if (INTEL_DEBUG & DEBUG_AUB) {
       aub_dump_bmp(ctx);
    }
 }
 
+/**
+ * Provides compatibility with loaders that only support the older (version
+ * 1-3) flush interface.
+ *
+ * That includes libGL up to Mesa 9.0, and the X Server at least up to 1.13.
+ */
+static void
+intel_dri2_flush(__DRIdrawable *drawable)
+{
+   intel_dri2_flush_with_flags(drawable->driContextPriv, drawable,
+                               __DRI2_FLUSH_DRAWABLE,
+                               __DRI2_THROTTLE_SWAPBUFFER);
+}
+
 static const struct __DRI2flushExtensionRec intelFlushExtension = {
-    .base = { __DRI2_FLUSH, 3 },
+    .base = { __DRI2_FLUSH, 4 },
 
-    .flush              = intelDRI2Flush,
+    .flush              = intel_dri2_flush,
     .invalidate         = dri2InvalidateDrawable,
+    .flush_with_flags   = intel_dri2_flush_with_flags,
 };
 
 static struct intel_image_format intel_image_formats[] = {
    { __DRI_IMAGE_FOURCC_ARGB8888, __DRI_IMAGE_COMPONENTS_RGBA, 1,
      { { 0, 0, 0, __DRI_IMAGE_FORMAT_ARGB8888, 4 } } },
 
+   { __DRI_IMAGE_FOURCC_ABGR8888, __DRI_IMAGE_COMPONENTS_RGBA, 1,
+     { { 0, 0, 0, __DRI_IMAGE_FORMAT_ABGR8888, 4 } } },
+
+   { __DRI_IMAGE_FOURCC_SARGB8888, __DRI_IMAGE_COMPONENTS_RGBA, 1,
+     { { 0, 0, 0, __DRI_IMAGE_FORMAT_SARGB8, 4 } } },
+
    { __DRI_IMAGE_FOURCC_XRGB8888, __DRI_IMAGE_COMPONENTS_RGB, 1,
      { { 0, 0, 0, __DRI_IMAGE_FORMAT_XRGB8888, 4 }, } },
 
+   { __DRI_IMAGE_FOURCC_XBGR8888, __DRI_IMAGE_COMPONENTS_RGB, 1,
+     { { 0, 0, 0, __DRI_IMAGE_FORMAT_XBGR8888, 4 }, } },
+
+   { __DRI_IMAGE_FOURCC_RGB565, __DRI_IMAGE_COMPONENTS_RGB, 1,
+     { { 0, 0, 0, __DRI_IMAGE_FORMAT_RGB565, 2 } } },
+
    { __DRI_IMAGE_FOURCC_YUV410, __DRI_IMAGE_COMPONENTS_Y_U_V, 3,
      { { 0, 0, 0, __DRI_IMAGE_FORMAT_R8, 1 },
        { 1, 2, 2, __DRI_IMAGE_FORMAT_R8, 1 },
@@ -236,6 +275,44 @@ static struct intel_image_format intel_image_formats[] = {
        { 0, 1, 0, __DRI_IMAGE_FORMAT_ARGB8888, 4 } } }
 };
 
+static void
+intel_image_warn_if_unaligned(__DRIimage *image, const char *func)
+{
+   uint32_t tiling, swizzle;
+   drm_intel_bo_get_tiling(image->bo, &tiling, &swizzle);
+
+   if (tiling != I915_TILING_NONE && (image->offset & 0xfff)) {
+      _mesa_warning(NULL, "%s: offset 0x%08x not on tile boundary",
+                    func, image->offset);
+   }
+}
+
+static struct intel_image_format *
+intel_image_format_lookup(int fourcc)
+{
+   struct intel_image_format *f = NULL;
+
+   for (unsigned i = 0; i < ARRAY_SIZE(intel_image_formats); i++) {
+      if (intel_image_formats[i].fourcc == fourcc) {
+        f = &intel_image_formats[i];
+        break;
+      }
+   }
+
+   return f;
+}
+
+static boolean intel_lookup_fourcc(int dri_format, int *fourcc)
+{
+   for (unsigned i = 0; i < ARRAY_SIZE(intel_image_formats); i++) {
+      if (intel_image_formats[i].planes[0].dri_format == dri_format) {
+         *fourcc = intel_image_formats[i].fourcc;
+         return true;
+      }
+   }
+   return false;
+}
+
 static __DRIimage *
 intel_allocate_image(int dri_format, void *loaderPrivate)
 {
@@ -248,32 +325,9 @@ intel_allocate_image(int dri_format, void *loaderPrivate)
     image->dri_format = dri_format;
     image->offset = 0;
 
-    switch (dri_format) {
-    case __DRI_IMAGE_FORMAT_RGB565:
-       image->format = MESA_FORMAT_RGB565;
-       break;
-    case __DRI_IMAGE_FORMAT_XRGB8888:
-       image->format = MESA_FORMAT_XRGB8888;
-       break;
-    case __DRI_IMAGE_FORMAT_ARGB8888:
-       image->format = MESA_FORMAT_ARGB8888;
-       break;
-    case __DRI_IMAGE_FORMAT_ABGR8888:
-       image->format = MESA_FORMAT_RGBA8888_REV;
-       break;
-    case __DRI_IMAGE_FORMAT_XBGR8888:
-       image->format = MESA_FORMAT_RGBX8888_REV;
-       break;
-    case __DRI_IMAGE_FORMAT_R8:
-       image->format = MESA_FORMAT_R8;
-       break;
-    case __DRI_IMAGE_FORMAT_GR88:
-       image->format = MESA_FORMAT_GR88;
-       break;
-    case __DRI_IMAGE_FORMAT_NONE:
-       image->format = MESA_FORMAT_NONE;
-       break;
-    default:
+    image->format = driImageFormatToGLFormat(dri_format);
+    if (dri_format != __DRI_IMAGE_FORMAT_NONE &&
+        image->format == MESA_FORMAT_NONE) {
        free(image);
        return NULL;
     }
@@ -285,65 +339,28 @@ intel_allocate_image(int dri_format, void *loaderPrivate)
 }
 
 /**
- * Sets up a DRIImage structure to point to our shared image in a region
+ * Sets up a DRIImage structure to point to a slice out of a miptree.
  */
 static void
-intel_setup_image_from_mipmap_tree(struct intel_context *intel, __DRIimage *image,
+intel_setup_image_from_mipmap_tree(struct brw_context *brw, __DRIimage *image,
                                    struct intel_mipmap_tree *mt, GLuint level,
                                    GLuint zoffset)
 {
-   unsigned int draw_x, draw_y;
-   uint32_t mask_x, mask_y;
-
-   intel_miptree_make_shareable(intel, mt);
+   intel_miptree_make_shareable(brw, mt);
 
    intel_miptree_check_level_layer(mt, level, zoffset);
 
-   intel_region_get_tile_masks(mt->region, &mask_x, &mask_y, false);
-   intel_miptree_get_image_offset(mt, level, zoffset, &draw_x, &draw_y);
-
-   image->width = mt->level[level].width;
-   image->height = mt->level[level].height;
-   image->tile_x = draw_x & mask_x;
-   image->tile_y = draw_y & mask_y;
-
-   image->offset = intel_region_get_aligned_offset(mt->region,
-                                                   draw_x & ~mask_x,
-                                                   draw_y & ~mask_y,
-                                                   false);
-
-   intel_region_reference(&image->region, mt->region);
-}
-
-static void
-intel_setup_image_from_dimensions(__DRIimage *image)
-{
-   image->width    = image->region->width;
-   image->height   = image->region->height;
-   image->tile_x = 0;
-   image->tile_y = 0;
-   image->has_depthstencil = false;
-}
+   image->width = minify(mt->physical_width0, level - mt->first_level);
+   image->height = minify(mt->physical_height0, level - mt->first_level);
+   image->pitch = mt->pitch;
 
-static inline uint32_t
-intel_dri_format(GLuint format)
-{
-   switch (format) {
-   case MESA_FORMAT_RGB565:
-      return __DRI_IMAGE_FORMAT_RGB565;
-   case MESA_FORMAT_XRGB8888:
-      return __DRI_IMAGE_FORMAT_XRGB8888;
-   case MESA_FORMAT_ARGB8888:
-      return __DRI_IMAGE_FORMAT_ARGB8888;
-   case MESA_FORMAT_RGBA8888_REV:
-      return __DRI_IMAGE_FORMAT_ABGR8888;
-   case MESA_FORMAT_R8:
-      return __DRI_IMAGE_FORMAT_R8;
-   case MESA_FORMAT_RG88:
-      return __DRI_IMAGE_FORMAT_GR88;
-   }
+   image->offset = intel_miptree_get_tile_offsets(mt, level, zoffset,
+                                                  &image->tile_x,
+                                                  &image->tile_y);
 
-   return MESA_FORMAT_NONE;
+   drm_intel_bo_unreference(image->bo);
+   image->bo = mt->bo;
+   drm_intel_bo_reference(mt->bo);
 }
 
 static __DRIimage *
@@ -363,16 +380,17 @@ intel_create_image_from_name(__DRIscreen *screen,
        cpp = 1;
     else
        cpp = _mesa_get_format_bytes(image->format);
-    image->region = intel_region_alloc_for_handle(intelScreen,
-                                                 cpp, width, height,
-                                                 pitch * cpp, name, "image");
-    if (image->region == NULL) {
+
+    image->width = width;
+    image->height = height;
+    image->pitch = pitch * cpp;
+    image->bo = drm_intel_bo_gem_create_from_name(intelScreen->bufmgr, "image",
+                                                  name);
+    if (!image->bo) {
        free(image);
        return NULL;
     }
 
-    intel_setup_image_from_dimensions(image);
-
     return image;      
 }
 
@@ -381,19 +399,19 @@ intel_create_image_from_renderbuffer(__DRIcontext *context,
                                     int renderbuffer, void *loaderPrivate)
 {
    __DRIimage *image;
-   struct intel_context *intel = context->driverPrivate;
+   struct brw_context *brw = context->driverPrivate;
+   struct gl_context *ctx = &brw->ctx;
    struct gl_renderbuffer *rb;
    struct intel_renderbuffer *irb;
 
-   rb = _mesa_lookup_renderbuffer(&intel->ctx, renderbuffer);
+   rb = _mesa_lookup_renderbuffer(ctx, renderbuffer);
    if (!rb) {
-      _mesa_error(&intel->ctx,
-                 GL_INVALID_OPERATION, "glRenderbufferExternalMESA");
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glRenderbufferExternalMESA");
       return NULL;
    }
 
    irb = intel_renderbuffer(rb);
-   intel_miptree_make_shareable(intel, irb->mt);
+   intel_miptree_make_shareable(brw, irb->mt);
    image = calloc(1, sizeof *image);
    if (image == NULL)
       return NULL;
@@ -402,9 +420,13 @@ intel_create_image_from_renderbuffer(__DRIcontext *context,
    image->format = rb->Format;
    image->offset = 0;
    image->data = loaderPrivate;
-   intel_region_reference(&image->region, irb->mt->region);
-   intel_setup_image_from_dimensions(image);
-   image->dri_format = intel_dri_format(image->format);
+   drm_intel_bo_unreference(image->bo);
+   image->bo = irb->mt->bo;
+   drm_intel_bo_reference(irb->mt->bo);
+   image->width = rb->Width;
+   image->height = rb->Height;
+   image->pitch = irb->mt->pitch;
+   image->dri_format = driGLFormatToImageFormat(image->format);
    image->has_depthstencil = irb->mt->stencil_mt? true : false;
 
    rb->NeedsFinishRenderTexture = true;
@@ -419,12 +441,12 @@ intel_create_image_from_texture(__DRIcontext *context, int target,
                                 void *loaderPrivate)
 {
    __DRIimage *image;
-   struct intel_context *intel = context->driverPrivate;
+   struct brw_context *brw = context->driverPrivate;
    struct gl_texture_object *obj;
    struct intel_texture_object *iobj;
    GLuint face = 0;
 
-   obj = _mesa_lookup_texture(&intel->ctx, texture);
+   obj = _mesa_lookup_texture(&brw->ctx, texture);
    if (!obj || obj->Target != target) {
       *error = __DRI_IMAGE_ERROR_BAD_PARAMETER;
       return NULL;
@@ -433,7 +455,7 @@ intel_create_image_from_texture(__DRIcontext *context, int target,
    if (target == GL_TEXTURE_CUBE_MAP)
       face = zoffset;
 
-   _mesa_test_texobj_completeness(&intel->ctx, obj);
+   _mesa_test_texobj_completeness(&brw->ctx, obj);
    iobj = intel_texture_object(obj);
    if (!obj->_BaseComplete || (level > 0 && !obj->_MipmapComplete)) {
       *error = __DRI_IMAGE_ERROR_BAD_PARAMETER;
@@ -458,8 +480,8 @@ intel_create_image_from_texture(__DRIcontext *context, int target,
    image->internal_format = obj->Image[face][level]->InternalFormat;
    image->format = obj->Image[face][level]->TexFormat;
    image->data = loaderPrivate;
-   intel_setup_image_from_mipmap_tree(intel, image, iobj->mt, level, zoffset);
-   image->dri_format = intel_dri_format(image->format);
+   intel_setup_image_from_mipmap_tree(brw, image, iobj->mt, level, zoffset);
+   image->dri_format = driGLFormatToImageFormat(image->format);
    image->has_depthstencil = iobj->mt->stencil_mt? true : false;
    if (image->dri_format == MESA_FORMAT_NONE) {
       *error = __DRI_IMAGE_ERROR_BAD_PARAMETER;
@@ -474,8 +496,8 @@ intel_create_image_from_texture(__DRIcontext *context, int target,
 static void
 intel_destroy_image(__DRIimage *image)
 {
-    intel_region_release(&image->region);
-    free(image);
+   drm_intel_bo_unreference(image->bo);
+   free(image);
 }
 
 static __DRIimage *
@@ -488,6 +510,7 @@ intel_create_image(__DRIscreen *screen,
    struct intel_screen *intelScreen = screen->driverPrivate;
    uint32_t tiling;
    int cpp;
+   unsigned long pitch;
 
    tiling = I915_TILING_X;
    if (use & __DRI_IMAGE_USE_CURSOR) {
@@ -496,19 +519,25 @@ intel_create_image(__DRIscreen *screen,
       tiling = I915_TILING_NONE;
    }
 
+   if (use & __DRI_IMAGE_USE_LINEAR)
+      tiling = I915_TILING_NONE;
+
    image = intel_allocate_image(format, loaderPrivate);
    if (image == NULL)
       return NULL;
 
+   
    cpp = _mesa_get_format_bytes(image->format);
-   image->region =
-      intel_region_alloc(intelScreen, tiling, cpp, width, height, true);
-   if (image->region == NULL) {
+   image->bo = drm_intel_bo_alloc_tiled(intelScreen->bufmgr, "image",
+                                        width, height, cpp, &tiling,
+                                        &pitch, 0);
+   if (image->bo == NULL) {
       free(image);
       return NULL;
    }
-   
-   intel_setup_image_from_dimensions(image);
+   image->width = width;
+   image->height = height;
+   image->pitch = pitch;
 
    return image;
 }
@@ -518,21 +547,21 @@ intel_query_image(__DRIimage *image, int attrib, int *value)
 {
    switch (attrib) {
    case __DRI_IMAGE_ATTRIB_STRIDE:
-      *value = image->region->pitch;
+      *value = image->pitch;
       return true;
    case __DRI_IMAGE_ATTRIB_HANDLE:
-      *value = image->region->bo->handle;
+      *value = image->bo->handle;
       return true;
    case __DRI_IMAGE_ATTRIB_NAME:
-      return intel_region_flink(image->region, (uint32_t *) value);
+      return !drm_intel_bo_flink(image->bo, (uint32_t *) value);
    case __DRI_IMAGE_ATTRIB_FORMAT:
       *value = image->dri_format;
       return true;
    case __DRI_IMAGE_ATTRIB_WIDTH:
-      *value = image->region->width;
+      *value = image->width;
       return true;
    case __DRI_IMAGE_ATTRIB_HEIGHT:
-      *value = image->region->height;
+      *value = image->height;
       return true;
    case __DRI_IMAGE_ATTRIB_COMPONENTS:
       if (image->planar_format == NULL)
@@ -540,9 +569,17 @@ intel_query_image(__DRIimage *image, int attrib, int *value)
       *value = image->planar_format->components;
       return true;
    case __DRI_IMAGE_ATTRIB_FD:
-      if (drm_intel_bo_gem_export_to_prime(image->region->bo, value) == 0)
+      if (drm_intel_bo_gem_export_to_prime(image->bo, value) == 0)
          return true;
       return false;
+   case __DRI_IMAGE_ATTRIB_FOURCC:
+      if (intel_lookup_fourcc(image->dri_format, value))
+         return true;
+      return false;
+   case __DRI_IMAGE_ATTRIB_NUM_PLANES:
+      *value = 1;
+      return true;
+
   default:
       return false;
    }
@@ -557,12 +594,8 @@ intel_dup_image(__DRIimage *orig_image, void *loaderPrivate)
    if (image == NULL)
       return NULL;
 
-   intel_region_reference(&image->region, orig_image->region);
-   if (image->region == NULL) {
-      free(image);
-      return NULL;
-   }
-
+   drm_intel_bo_reference(orig_image->bo);
+   image->bo              = orig_image->bo;
    image->internal_format = orig_image->internal_format;
    image->planar_format   = orig_image->planar_format;
    image->dri_format      = orig_image->dri_format;
@@ -570,6 +603,7 @@ intel_dup_image(__DRIimage *orig_image, void *loaderPrivate)
    image->offset          = orig_image->offset;
    image->width           = orig_image->width;
    image->height          = orig_image->height;
+   image->pitch           = orig_image->pitch;
    image->tile_x          = orig_image->tile_x;
    image->tile_y          = orig_image->tile_y;
    image->has_depthstencil = orig_image->has_depthstencil;
@@ -585,7 +619,7 @@ static GLboolean
 intel_validate_usage(__DRIimage *image, unsigned int use)
 {
    if (use & __DRI_IMAGE_USE_CURSOR) {
-      if (image->region->width != 64 || image->region->height != 64)
+      if (image->width != 64 || image->height != 64)
         return GL_FALSE;
    }
 
@@ -606,12 +640,7 @@ intel_create_image_from_names(__DRIscreen *screen,
     if (screen == NULL || names == NULL || num_names != 1)
         return NULL;
 
-    for (i = 0; i < ARRAY_SIZE(intel_image_formats); i++) {
-        if (intel_image_formats[i].fourcc == fourcc) {
-           f = &intel_image_formats[i];
-        }
-    }
-
+    f = intel_image_format_lookup(fourcc);
     if (f == NULL)
         return NULL;
 
@@ -640,33 +669,35 @@ intel_create_image_from_fds(__DRIscreen *screen,
                             void *loaderPrivate)
 {
    struct intel_screen *intelScreen = screen->driverPrivate;
-   struct intel_image_format *f = NULL;
+   struct intel_image_format *f;
    __DRIimage *image;
    int i, index;
 
    if (fds == NULL || num_fds != 1)
       return NULL;
 
-   for (i = 0; i < ARRAY_SIZE(intel_image_formats); i++) {
-      if (intel_image_formats[i].fourcc == fourcc) {
-         f = &intel_image_formats[i];
-      }
-   }
-
+   f = intel_image_format_lookup(fourcc);
    if (f == NULL)
       return NULL;
 
-   image = intel_allocate_image(__DRI_IMAGE_FORMAT_NONE, loaderPrivate);
+   if (f->nplanes == 1)
+      image = intel_allocate_image(f->planes[0].dri_format, loaderPrivate);
+   else
+      image = intel_allocate_image(__DRI_IMAGE_FORMAT_NONE, loaderPrivate);
+
    if (image == NULL)
       return NULL;
 
-   image->region = intel_region_alloc_for_fd(intelScreen,
-                                             1, width, height,
-                                             strides[0], fds[0], "image");
-   if (image->region == NULL) {
+   image->bo = drm_intel_bo_gem_create_from_prime(intelScreen->bufmgr,
+                                                  fds[0],
+                                                  height * strides[0]);
+   if (image->bo == NULL) {
       free(image);
       return NULL;
    }
+   image->width = width;
+   image->height = height;
+   image->pitch = strides[0];
 
    image->planar_format = f;
    for (i = 0; i < f->nplanes; i++) {
@@ -675,16 +706,64 @@ intel_create_image_from_fds(__DRIscreen *screen,
       image->strides[index] = strides[index];
    }
 
+   if (f->nplanes == 1) {
+      image->offset = image->offsets[0];
+      intel_image_warn_if_unaligned(image, __func__);
+   }
+
    return image;
 }
 
+static __DRIimage *
+intel_create_image_from_dma_bufs(__DRIscreen *screen,
+                                 int width, int height, int fourcc,
+                                 int *fds, int num_fds,
+                                 int *strides, int *offsets,
+                                 enum __DRIYUVColorSpace yuv_color_space,
+                                 enum __DRISampleRange sample_range,
+                                 enum __DRIChromaSiting horizontal_siting,
+                                 enum __DRIChromaSiting vertical_siting,
+                                 unsigned *error,
+                                 void *loaderPrivate)
+{
+   __DRIimage *image;
+   struct intel_image_format *f = intel_image_format_lookup(fourcc);
+
+   /* For now only packed formats that have native sampling are supported. */
+   if (!f || f->nplanes != 1) {
+      *error = __DRI_IMAGE_ERROR_BAD_MATCH;
+      return NULL;
+   }
+
+   image = intel_create_image_from_fds(screen, width, height, fourcc, fds,
+                                       num_fds, strides, offsets,
+                                       loaderPrivate);
+
+   /*
+    * Invalid parameters and any inconsistencies between are assumed to be
+    * checked by the caller. Therefore besides unsupported formats one can fail
+    * only in allocation.
+    */
+   if (!image) {
+      *error = __DRI_IMAGE_ERROR_BAD_ALLOC;
+      return NULL;
+   }
+
+   image->dma_buf_imported = true;
+   image->yuv_color_space = yuv_color_space;
+   image->sample_range = sample_range;
+   image->horizontal_siting = horizontal_siting;
+   image->vertical_siting = vertical_siting;
+
+   *error = __DRI_IMAGE_ERROR_SUCCESS;
+   return image;
+}
 
 static __DRIimage *
 intel_from_planar(__DRIimage *parent, int plane, void *loaderPrivate)
 {
     int width, height, offset, stride, dri_format, index;
     struct intel_image_format *f;
-    uint32_t mask_x, mask_y;
     __DRIimage *image;
 
     if (parent == NULL || parent->planar_format == NULL)
@@ -695,8 +774,8 @@ intel_from_planar(__DRIimage *parent, int plane, void *loaderPrivate)
     if (plane >= f->nplanes)
         return NULL;
 
-    width = parent->region->width >> f->planes[plane].width_shift;
-    height = parent->region->height >> f->planes[plane].height_shift;
+    width = parent->width >> f->planes[plane].width_shift;
+    height = parent->height >> f->planes[plane].height_shift;
     dri_format = f->planes[plane].dri_format;
     index = f->planes[plane].buffer_index;
     offset = parent->offsets[index];
@@ -706,39 +785,27 @@ intel_from_planar(__DRIimage *parent, int plane, void *loaderPrivate)
     if (image == NULL)
        return NULL;
 
-    if (offset + height * stride > parent->region->bo->size) {
+    if (offset + height * stride > parent->bo->size) {
        _mesa_warning(NULL, "intel_create_sub_image: subimage out of bounds");
        free(image);
        return NULL;
     }
 
-    image->region = calloc(sizeof(*image->region), 1);
-    if (image->region == NULL) {
-       free(image);
-       return NULL;
-    }
+    image->bo = parent->bo;
+    drm_intel_bo_reference(parent->bo);
 
-    image->region->cpp = _mesa_get_format_bytes(image->format);
-    image->region->width = width;
-    image->region->height = height;
-    image->region->pitch = stride;
-    image->region->refcount = 1;
-    image->region->bo = parent->region->bo;
-    drm_intel_bo_reference(image->region->bo);
-    image->region->tiling = parent->region->tiling;
+    image->width = width;
+    image->height = height;
+    image->pitch = stride;
     image->offset = offset;
-    intel_setup_image_from_dimensions(image);
 
-    intel_region_get_tile_masks(image->region, &mask_x, &mask_y, false);
-    if (offset & mask_x)
-       _mesa_warning(NULL,
-                     "intel_create_sub_image: offset not on tile boundary");
+    intel_image_warn_if_unaligned(image, __func__);
 
     return image;
 }
 
-static struct __DRIimageExtensionRec intelImageExtension = {
-    .base = { __DRI_IMAGE, 7 },
+static const __DRIimageExtension intelImageExtension = {
+    .base = { __DRI_IMAGE, 11 },
 
     .createImageFromName                = intel_create_image_from_name,
     .createImageFromRenderbuffer        = intel_create_image_from_renderbuffer,
@@ -750,17 +817,118 @@ static struct __DRIimageExtensionRec intelImageExtension = {
     .createImageFromNames               = intel_create_image_from_names,
     .fromPlanar                         = intel_from_planar,
     .createImageFromTexture             = intel_create_image_from_texture,
-    .createImageFromFds                 = intel_create_image_from_fds
+    .createImageFromFds                 = intel_create_image_from_fds,
+    .createImageFromDmaBufs             = intel_create_image_from_dma_bufs,
+    .blitImage                          = NULL,
+    .getCapabilities                    = NULL
+};
+
+static int
+brw_query_renderer_integer(__DRIscreen *psp, int param, unsigned int *value)
+{
+   const struct intel_screen *const intelScreen =
+      (struct intel_screen *) psp->driverPrivate;
+
+   switch (param) {
+   case __DRI2_RENDERER_VENDOR_ID:
+      value[0] = 0x8086;
+      return 0;
+   case __DRI2_RENDERER_DEVICE_ID:
+      value[0] = intelScreen->deviceID;
+      return 0;
+   case __DRI2_RENDERER_ACCELERATED:
+      value[0] = 1;
+      return 0;
+   case __DRI2_RENDERER_VIDEO_MEMORY: {
+      /* Once a batch uses more than 75% of the maximum mappable size, we
+       * assume that there's some fragmentation, and we start doing extra
+       * flushing, etc.  That's the big cliff apps will care about.
+       */
+      size_t aper_size;
+      size_t mappable_size;
+
+      drm_intel_get_aperture_sizes(psp->fd, &mappable_size, &aper_size);
+
+      const unsigned gpu_mappable_megabytes =
+         (aper_size / (1024 * 1024)) * 3 / 4;
+
+      const long system_memory_pages = sysconf(_SC_PHYS_PAGES);
+      const long system_page_size = sysconf(_SC_PAGE_SIZE);
+
+      if (system_memory_pages <= 0 || system_page_size <= 0)
+         return -1;
+
+      const uint64_t system_memory_bytes = (uint64_t) system_memory_pages
+         * (uint64_t) system_page_size;
+
+      const unsigned system_memory_megabytes =
+         (unsigned) (system_memory_bytes / (1024 * 1024));
+
+      value[0] = MIN2(system_memory_megabytes, gpu_mappable_megabytes);
+      return 0;
+   }
+   case __DRI2_RENDERER_UNIFIED_MEMORY_ARCHITECTURE:
+      value[0] = 1;
+      return 0;
+   default:
+      return driQueryRendererIntegerCommon(psp, param, value);
+   }
+
+   return -1;
+}
+
+static int
+brw_query_renderer_string(__DRIscreen *psp, int param, const char **value)
+{
+   const struct intel_screen *intelScreen =
+      (struct intel_screen *) psp->driverPrivate;
+
+   switch (param) {
+   case __DRI2_RENDERER_VENDOR_ID:
+      value[0] = brw_vendor_string;
+      return 0;
+   case __DRI2_RENDERER_DEVICE_ID:
+      value[0] = brw_get_renderer_string(intelScreen->deviceID);
+      return 0;
+   default:
+      break;
+   }
+
+   return -1;
+}
+
+static const __DRI2rendererQueryExtension intelRendererQueryExtension = {
+   .base = { __DRI2_RENDERER_QUERY, 1 },
+
+   .queryInteger = brw_query_renderer_integer,
+   .queryString = brw_query_renderer_string
+};
+
+static const __DRIrobustnessExtension dri2Robustness = {
+   .base = { __DRI2_ROBUSTNESS, 1 }
 };
 
 static const __DRIextension *intelScreenExtensions[] = {
     &intelTexBufferExtension.base,
+    &intelFenceExtension.base,
     &intelFlushExtension.base,
     &intelImageExtension.base,
+    &intelRendererQueryExtension.base,
     &dri2ConfigQueryExtension.base,
     NULL
 };
 
+static const __DRIextension *intelRobustScreenExtensions[] = {
+    &intelTexBufferExtension.base,
+    &intelFenceExtension.base,
+    &intelFlushExtension.base,
+    &intelImageExtension.base,
+    &intelRendererQueryExtension.base,
+    &dri2ConfigQueryExtension.base,
+    &dri2Robustness.base,
+    NULL
+};
+
 static bool
 intel_get_param(__DRIscreen *psp, int param, int *value)
 {
@@ -796,7 +964,7 @@ intelDestroyScreen(__DRIscreen * sPriv)
    dri_bufmgr_destroy(intelScreen->bufmgr);
    driDestroyOptionInfo(&intelScreen->optionCache);
 
-   free(intelScreen);
+   ralloc_free(intelScreen);
    sPriv->driverPrivate = NULL;
 }
 
@@ -811,7 +979,7 @@ intelCreateBuffer(__DRIscreen * driScrnPriv,
 {
    struct intel_renderbuffer *rb;
    struct intel_screen *screen = (struct intel_screen*) driScrnPriv->driverPrivate;
-   gl_format rgbFormat;
+   mesa_format rgbFormat;
    unsigned num_samples = intel_quantize_num_samples(screen, mesaVis->samples);
    struct gl_framebuffer *fb;
 
@@ -824,14 +992,19 @@ intelCreateBuffer(__DRIscreen * driScrnPriv,
 
    _mesa_initialize_window_framebuffer(fb, mesaVis);
 
+   if (screen->winsys_msaa_samples_override != -1) {
+      num_samples = screen->winsys_msaa_samples_override;
+      fb->Visual.samples = num_samples;
+   }
+
    if (mesaVis->redBits == 5)
-      rgbFormat = MESA_FORMAT_RGB565;
+      rgbFormat = MESA_FORMAT_B5G6R5_UNORM;
    else if (mesaVis->sRGBCapable)
-      rgbFormat = MESA_FORMAT_SARGB8;
+      rgbFormat = MESA_FORMAT_B8G8R8A8_SRGB;
    else if (mesaVis->alphaBits == 0)
-      rgbFormat = MESA_FORMAT_XRGB8888;
+      rgbFormat = MESA_FORMAT_B8G8R8X8_UNORM;
    else {
-      rgbFormat = MESA_FORMAT_SARGB8;
+      rgbFormat = MESA_FORMAT_B8G8R8A8_SRGB;
       fb->Visual.sRGBCapable = true;
    }
 
@@ -852,11 +1025,11 @@ intelCreateBuffer(__DRIscreen * driScrnPriv,
    if (mesaVis->depthBits == 24) {
       assert(mesaVis->stencilBits == 8);
 
-      if (screen->hw_has_separate_stencil) {
-         rb = intel_create_private_renderbuffer(MESA_FORMAT_X8_Z24,
+      if (screen->devinfo->has_hiz_and_separate_stencil) {
+         rb = intel_create_private_renderbuffer(MESA_FORMAT_Z24_UNORM_X8_UINT,
                                                 num_samples);
          _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &rb->Base.Base);
-         rb = intel_create_private_renderbuffer(MESA_FORMAT_S8,
+         rb = intel_create_private_renderbuffer(MESA_FORMAT_S_UINT8,
                                                 num_samples);
          _mesa_add_renderbuffer(fb, BUFFER_STENCIL, &rb->Base.Base);
       } else {
@@ -864,7 +1037,7 @@ intelCreateBuffer(__DRIscreen * driScrnPriv,
           * Use combined depth/stencil. Note that the renderbuffer is
           * attached to two attachment points.
           */
-         rb = intel_create_private_renderbuffer(MESA_FORMAT_S8_Z24,
+         rb = intel_create_private_renderbuffer(MESA_FORMAT_Z24_UNORM_S8_UINT,
                                                 num_samples);
          _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &rb->Base.Base);
          _mesa_add_renderbuffer(fb, BUFFER_STENCIL, &rb->Base.Base);
@@ -872,7 +1045,7 @@ intelCreateBuffer(__DRIscreen * driScrnPriv,
    }
    else if (mesaVis->depthBits == 16) {
       assert(mesaVis->stencilBits == 0);
-      rb = intel_create_private_renderbuffer(MESA_FORMAT_Z16,
+      rb = intel_create_private_renderbuffer(MESA_FORMAT_Z_UNORM16,
                                              num_samples);
       _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &rb->Base.Base);
    }
@@ -898,39 +1071,8 @@ static void
 intelDestroyBuffer(__DRIdrawable * driDrawPriv)
 {
     struct gl_framebuffer *fb = driDrawPriv->driverPrivate;
-  
-    _mesa_reference_framebuffer(&fb, NULL);
-}
-
-/* There are probably better ways to do this, such as an
- * init-designated function to register chipids and createcontext
- * functions.
- */
-
-static GLboolean
-intelCreateContext(gl_api api,
-                  const struct gl_config * mesaVis,
-                   __DRIcontext * driContextPriv,
-                  unsigned major_version,
-                  unsigned minor_version,
-                  uint32_t flags,
-                  unsigned *error,
-                   void *sharedContextPrivate)
-{
-   bool success = false;
-
-   success = brwCreateContext(api, mesaVis,
-                              driContextPriv,
-                              major_version, minor_version, flags,
-                              error, sharedContextPrivate);
-
-   if (success)
-      return true;
-
-   if (driContextPriv->driverPrivate != NULL)
-      intelDestroyContext(driContextPriv);
 
-   return false;
+    _mesa_reference_framebuffer(&fb, NULL);
 }
 
 static bool
@@ -957,31 +1099,6 @@ intel_init_bufmgr(struct intel_screen *intelScreen)
    return true;
 }
 
-/**
- * Override intel_screen.hw_has_separate_stencil with environment variable
- * INTEL_SEPARATE_STENCIL.
- *
- * Valid values for INTEL_SEPARATE_STENCIL are "0" and "1". If an invalid
- * valid value is encountered, a warning is emitted and INTEL_SEPARATE_STENCIL
- * is ignored.
- */
-static void
-intel_override_separate_stencil(struct intel_screen *screen)
-{
-   const char *s = getenv("INTEL_SEPARATE_STENCIL");
-   if (!s) {
-      return;
-   } else if (!strncmp("0", s, 2)) {
-      screen->hw_has_separate_stencil = false;
-   } else if (!strncmp("1", s, 2)) {
-      screen->hw_has_separate_stencil = true;
-   } else {
-      fprintf(stderr,
-             "warning: env variable INTEL_SEPARATE_STENCIL=\"%s\" has "
-             "invalid value and is ignored", s);
-   }
-}
-
 static bool
 intel_detect_swizzling(struct intel_screen *screen)
 {
@@ -1006,12 +1123,57 @@ intel_detect_swizzling(struct intel_screen *screen)
       return true;
 }
 
+static bool
+intel_detect_timestamp(struct intel_screen *screen)
+{
+   uint64_t dummy = 0;
+   int loop = 10;
+
+   /*
+    * On 32bit systems, some old kernels trigger a hw bug resulting in the
+    * TIMESTAMP register being shifted and the low 32bits always zero. Detect
+    * this by repeating the read a few times and check the register is
+    * incrementing every 80ns as expected and not stuck on zero (as would be
+    * the case with the buggy kernel/hw.).
+    */
+   do {
+      if (drm_intel_reg_read(screen->bufmgr, TIMESTAMP, &dummy))
+        return false;
+   } while ((dummy & 0xffffffff) == 0 && --loop);
+
+   return loop > 0;
+}
+
+/**
+ * Return array of MSAA modes supported by the hardware. The array is
+ * zero-terminated and sorted in decreasing order.
+ */
+const int*
+intel_supported_msaa_modes(const struct intel_screen  *screen)
+{
+   static const int gen8_modes[] = {8, 4, 2, 0, -1};
+   static const int gen7_modes[] = {8, 4, 0, -1};
+   static const int gen6_modes[] = {4, 0, -1};
+   static const int gen4_modes[] = {0, -1};
+
+   if (screen->devinfo->gen >= 8) {
+      return gen8_modes;
+   } else if (screen->devinfo->gen >= 7) {
+      return gen7_modes;
+   } else if (screen->devinfo->gen == 6) {
+      return gen6_modes;
+   } else {
+      return gen4_modes;
+   }
+}
+
 static __DRIconfig**
 intel_screen_make_configs(__DRIscreen *dri_screen)
 {
-   static const gl_format formats[] = {
-      MESA_FORMAT_RGB565,
-      MESA_FORMAT_ARGB8888
+   static const mesa_format formats[] = {
+      MESA_FORMAT_B5G6R5_UNORM,
+      MESA_FORMAT_B8G8R8A8_UNORM,
+      MESA_FORMAT_B8G8R8X8_UNORM
    };
 
    /* GLX_SWAP_COPY_OML is not supported due to page flipping. */
@@ -1023,6 +1185,7 @@ intel_screen_make_configs(__DRIscreen *dri_screen)
    static const uint8_t multisample_samples[2]  = {4, 8};
 
    struct intel_screen *screen = dri_screen->driverPrivate;
+   const struct brw_device_info *devinfo = screen->devinfo;
    uint8_t depth_bits[4], stencil_bits[4];
    __DRIconfig **configs = NULL;
 
@@ -1038,10 +1201,10 @@ intel_screen_make_configs(__DRIscreen *dri_screen)
       depth_bits[0] = 0;
       stencil_bits[0] = 0;
 
-      if (formats[i] == MESA_FORMAT_RGB565) {
+      if (formats[i] == MESA_FORMAT_B5G6R5_UNORM) {
          depth_bits[1] = 16;
          stencil_bits[1] = 0;
-         if (screen->gen >= 6) {
+         if (devinfo->gen >= 6) {
              depth_bits[2] = 24;
              stencil_bits[2] = 8;
              num_depth_stencil_bits = 3;
@@ -1067,7 +1230,7 @@ intel_screen_make_configs(__DRIscreen *dri_screen)
    for (int i = 0; i < ARRAY_SIZE(formats); i++) {
       __DRIconfig **new_configs;
 
-      if (formats[i] == MESA_FORMAT_RGB565) {
+      if (formats[i] == MESA_FORMAT_B5G6R5_UNORM) {
          depth_bits[0] = 16;
          stencil_bits[0] = 0;
       } else {
@@ -1097,7 +1260,7 @@ intel_screen_make_configs(__DRIscreen *dri_screen)
     * them.
     */
    for (int i = 0; i < ARRAY_SIZE(formats); i++) {
-      if (screen->gen < 6)
+      if (devinfo->gen < 6)
          break;
 
       __DRIconfig **new_configs;
@@ -1107,7 +1270,7 @@ intel_screen_make_configs(__DRIscreen *dri_screen)
       depth_bits[0] = 0;
       stencil_bits[0] = 0;
 
-      if (formats[i] == MESA_FORMAT_RGB565) {
+      if (formats[i] == MESA_FORMAT_B5G6R5_UNORM) {
          depth_bits[1] = 16;
          stencil_bits[1] = 0;
       } else {
@@ -1115,9 +1278,9 @@ intel_screen_make_configs(__DRIscreen *dri_screen)
          stencil_bits[1] = 8;
       }
 
-      if (screen->gen >= 7)
+      if (devinfo->gen >= 7)
          num_msaa_modes = 2;
-      else if (screen->gen == 6)
+      else if (devinfo->gen == 6)
          num_msaa_modes = 1;
 
       new_configs = driCreateConfigs(formats[i],
@@ -1143,64 +1306,53 @@ intel_screen_make_configs(__DRIscreen *dri_screen)
 static void
 set_max_gl_versions(struct intel_screen *screen)
 {
-   int gl_version_override = _mesa_get_gl_version_override();
+   __DRIscreen *psp = screen->driScrnPriv;
 
-   switch (screen->gen) {
+   switch (screen->devinfo->gen) {
+   case 9:
+   case 8:
    case 7:
-      screen->max_gl_core_version = 31;
-      screen->max_gl_compat_version = 30;
-      screen->max_gl_es1_version = 11;
-      screen->max_gl_es2_version = 30;
-      break;
    case 6:
-      screen->max_gl_core_version = 31;
-      screen->max_gl_compat_version = 30;
-      screen->max_gl_es1_version = 11;
-      screen->max_gl_es2_version = 30;
+      psp->max_gl_core_version = 33;
+      psp->max_gl_compat_version = 30;
+      psp->max_gl_es1_version = 11;
+      psp->max_gl_es2_version = 30;
       break;
    case 5:
    case 4:
-      screen->max_gl_core_version = 0;
-      screen->max_gl_compat_version = 21;
-      screen->max_gl_es1_version = 11;
-      screen->max_gl_es2_version = 20;
-      break;
-   case 3: {
-      screen->max_gl_core_version = 0;
-      screen->max_gl_es1_version = 11;
-      screen->max_gl_compat_version = 21;
-      screen->max_gl_es2_version = 20;
-
-      break;
-   }
-   case 2:
-      screen->max_gl_core_version = 0;
-      screen->max_gl_compat_version = 13;
-      screen->max_gl_es1_version = 11;
-      screen->max_gl_es2_version = 0;
+      psp->max_gl_core_version = 0;
+      psp->max_gl_compat_version = 21;
+      psp->max_gl_es1_version = 11;
+      psp->max_gl_es2_version = 20;
       break;
    default:
-      assert(!"unrecognized intel_screen::gen");
-      break;
+      unreachable("unrecognized intel_screen::gen");
    }
+}
 
-   if (gl_version_override >= 31) {
-      screen->max_gl_core_version = MAX2(screen->max_gl_core_version,
-                                         gl_version_override);
-   } else {
-      screen->max_gl_compat_version = MAX2(screen->max_gl_compat_version,
-                                           gl_version_override);
-   }
+static int
+brw_get_revision(int fd)
+{
+   struct drm_i915_getparam gp;
+   int revision;
+   int ret;
 
-#ifndef FEATURE_ES1
-   screen->max_gl_es1_version = 0;
-#endif
+   memset(&gp, 0, sizeof(gp));
+   gp.param = I915_PARAM_REVISION;
+   gp.value = &revision;
 
-#ifndef FEATURE_ES2
-   screen->max_gl_es2_version = 0;
-#endif
+   ret = drmCommandWriteRead(fd, DRM_I915_GETPARAM, &gp, sizeof(gp));
+   if (ret)
+      revision = -1;
+
+   return revision;
 }
 
+/* Drop when RS headers get pulled to libdrm */
+#ifndef I915_PARAM_HAS_RESOURCE_STREAMER
+#define I915_PARAM_HAS_RESOURCE_STREAMER 36
+#endif
+
 /**
  * This is the driver specific part of the createNewScreen entry point.
  * Called when using DRI2.
@@ -1212,7 +1364,8 @@ __DRIconfig **intelInitScreen2(__DRIscreen *psp)
 {
    struct intel_screen *intelScreen;
 
-   if (psp->dri2.loader->base.version <= 2 ||
+   if (psp->image.loader) {
+   } else if (psp->dri2.loader->base.version <= 2 ||
        psp->dri2.loader->getBuffersWithFormat == NULL) {
       fprintf(stderr,
              "\nERROR!  DRI2 loader with getBuffersWithFormat() "
@@ -1221,14 +1374,13 @@ __DRIconfig **intelInitScreen2(__DRIscreen *psp)
    }
 
    /* Allocate the private area */
-   intelScreen = calloc(1, sizeof *intelScreen);
+   intelScreen = rzalloc(NULL, struct intel_screen);
    if (!intelScreen) {
       fprintf(stderr, "\nERROR!  Allocating private area failed\n");
       return false;
    }
    /* parse information in __driConfigOptions */
-   driParseOptionInfo(&intelScreen->optionCache,
-                      __driConfigOptions, __driNConfigOptions);
+   driParseOptionInfo(&intelScreen->optionCache, brw_config_options.xml);
 
    intelScreen->driScrnPriv = psp;
    psp->driverPrivate = (void *) intelScreen;
@@ -1237,56 +1389,77 @@ __DRIconfig **intelInitScreen2(__DRIscreen *psp)
        return false;
 
    intelScreen->deviceID = drm_intel_bufmgr_gem_get_devid(intelScreen->bufmgr);
+   intelScreen->devinfo = brw_get_device_info(intelScreen->deviceID,
+                                              brw_get_revision(psp->fd));
+   if (!intelScreen->devinfo)
+      return false;
+
+   brw_process_intel_debug_variable(intelScreen);
+
+   intelScreen->hw_must_use_separate_stencil = intelScreen->devinfo->gen >= 7;
 
-   if (IS_GEN7(intelScreen->deviceID)) {
-      intelScreen->gen = 7;
-   } else if (IS_GEN6(intelScreen->deviceID)) {
-      intelScreen->gen = 6;
-   } else if (IS_GEN5(intelScreen->deviceID)) {
-      intelScreen->gen = 5;
-   } else if (IS_965(intelScreen->deviceID)) {
-      intelScreen->gen = 4;
-   } else if (IS_9XX(intelScreen->deviceID)) {
-      intelScreen->gen = 3;
+   intelScreen->hw_has_swizzling = intel_detect_swizzling(intelScreen);
+   intelScreen->hw_has_timestamp = intel_detect_timestamp(intelScreen);
+
+   const char *force_msaa = getenv("INTEL_FORCE_MSAA");
+   if (force_msaa) {
+      intelScreen->winsys_msaa_samples_override =
+         intel_quantize_num_samples(intelScreen, atoi(force_msaa));
+      printf("Forcing winsys sample count to %d\n",
+             intelScreen->winsys_msaa_samples_override);
    } else {
-      intelScreen->gen = 2;
+      intelScreen->winsys_msaa_samples_override = -1;
    }
 
-   intelScreen->hw_has_separate_stencil = intelScreen->gen >= 6;
-   intelScreen->hw_must_use_separate_stencil = intelScreen->gen >= 7;
+   set_max_gl_versions(intelScreen);
 
-   int has_llc = 0;
-   bool success = intel_get_param(intelScreen->driScrnPriv, I915_PARAM_HAS_LLC,
-                                 &has_llc);
-   if (success && has_llc)
-      intelScreen->hw_has_llc = true;
-   else if (!success && intelScreen->gen >= 6)
-      intelScreen->hw_has_llc = true;
+   /* Notification of GPU resets requires hardware contexts and a kernel new
+    * enough to support DRM_IOCTL_I915_GET_RESET_STATS.  If the ioctl is
+    * supported, calling it with a context of 0 will either generate EPERM or
+    * no error.  If the ioctl is not supported, it always generate EINVAL.
+    * Use this to determine whether to advertise the __DRI2_ROBUSTNESS
+    * extension to the loader.
+    *
+    * Don't even try on pre-Gen6, since we don't attempt to use contexts there.
+    */
+   if (intelScreen->devinfo->gen >= 6) {
+      struct drm_i915_reset_stats stats;
+      memset(&stats, 0, sizeof(stats));
 
-   intel_override_separate_stencil(intelScreen);
+      const int ret = drmIoctl(psp->fd, DRM_IOCTL_I915_GET_RESET_STATS, &stats);
 
-   intelScreen->hw_has_swizzling = intel_detect_swizzling(intelScreen);
+      intelScreen->has_context_reset_notification =
+         (ret != -1 || errno != EINVAL);
+   }
 
-   set_max_gl_versions(intelScreen);
+   struct drm_i915_getparam getparam;
+   getparam.param = I915_PARAM_CMD_PARSER_VERSION;
+   getparam.value = &intelScreen->cmd_parser_version;
+   const int ret = drmIoctl(psp->fd, DRM_IOCTL_I915_GETPARAM, &getparam);
+   if (ret == -1)
+      intelScreen->cmd_parser_version = 0;
 
-   psp->api_mask = (1 << __DRI_API_OPENGL);
-   if (intelScreen->max_gl_core_version > 0)
-      psp->api_mask |= (1 << __DRI_API_OPENGL_CORE);
-   if (intelScreen->max_gl_es1_version > 0)
-      psp->api_mask |= (1 << __DRI_API_GLES);
-   if (intelScreen->max_gl_es2_version > 0)
-      psp->api_mask |= (1 << __DRI_API_GLES2);
-   if (intelScreen->max_gl_es2_version >= 30)
-      psp->api_mask |= (1 << __DRI_API_GLES3);
+   psp->extensions = !intelScreen->has_context_reset_notification
+      ? intelScreenExtensions : intelRobustScreenExtensions;
 
-   psp->extensions = intelScreenExtensions;
+   intelScreen->compiler = brw_compiler_create(intelScreen,
+                                               intelScreen->devinfo);
+
+   if (intelScreen->devinfo->has_resource_streamer) {
+      int val = -1;
+      getparam.param = I915_PARAM_HAS_RESOURCE_STREAMER;
+      getparam.value = &val;
+
+      drmIoctl(psp->fd, DRM_IOCTL_I915_GETPARAM, &getparam);
+      intelScreen->has_resource_streamer = val > 0;
+   }
 
    return (const __DRIconfig**) intel_screen_make_configs(psp);
 }
 
 struct intel_buffer {
    __DRIbuffer base;
-   struct intel_region *region;
+   drm_intel_bo *bo;
 };
 
 static __DRIbuffer *
@@ -1305,23 +1478,27 @@ intelAllocateBuffer(__DRIscreen *screen,
       return NULL;
 
    /* The front and back buffers are color buffers, which are X tiled. */
-   intelBuffer->region = intel_region_alloc(intelScreen,
-                                            I915_TILING_X,
-                                            format / 8,
-                                            width,
-                                            height,
-                                            true);
-   
-   if (intelBuffer->region == NULL) {
+   uint32_t tiling = I915_TILING_X;
+   unsigned long pitch;
+   int cpp = format / 8;
+   intelBuffer->bo = drm_intel_bo_alloc_tiled(intelScreen->bufmgr,
+                                              "intelAllocateBuffer",
+                                              width,
+                                              height,
+                                              cpp,
+                                              &tiling, &pitch,
+                                              BO_ALLOC_FOR_RENDER);
+
+   if (intelBuffer->bo == NULL) {
           free(intelBuffer);
           return NULL;
    }
-   
-   intel_region_flink(intelBuffer->region, &intelBuffer->base.name);
+
+   drm_intel_bo_flink(intelBuffer->bo, &intelBuffer->base.name);
 
    intelBuffer->base.attachment = attachment;
-   intelBuffer->base.cpp = intelBuffer->region->cpp;
-   intelBuffer->base.pitch = intelBuffer->region->pitch;
+   intelBuffer->base.cpp = cpp;
+   intelBuffer->base.pitch = pitch;
 
    return &intelBuffer->base;
 }
@@ -1331,15 +1508,14 @@ intelReleaseBuffer(__DRIscreen *screen, __DRIbuffer *buffer)
 {
    struct intel_buffer *intelBuffer = (struct intel_buffer *) buffer;
 
-   intel_region_release(&intelBuffer->region);
+   drm_intel_bo_unreference(intelBuffer->bo);
    free(intelBuffer);
 }
 
-
-const struct __DriverAPIRec driDriverAPI = {
+static const struct __DriverAPIRec brw_driver_api = {
    .InitScreen          = intelInitScreen2,
    .DestroyScreen       = intelDestroyScreen,
-   .CreateContext       = intelCreateContext,
+   .CreateContext       = brwCreateContext,
    .DestroyContext      = intelDestroyContext,
    .CreateBuffer        = intelCreateBuffer,
    .DestroyBuffer       = intelDestroyBuffer,
@@ -1349,9 +1525,23 @@ const struct __DriverAPIRec driDriverAPI = {
    .ReleaseBuffer        = intelReleaseBuffer
 };
 
-/* This is the table of extensions that the loader will dlsym() for. */
-PUBLIC const __DRIextension *__driDriverExtensions[] = {
+static const struct __DRIDriverVtableExtensionRec brw_vtable = {
+   .base = { __DRI_DRIVER_VTABLE, 1 },
+   .vtable = &brw_driver_api,
+};
+
+static const __DRIextension *brw_driver_extensions[] = {
     &driCoreExtension.base,
+    &driImageDriverExtension.base,
     &driDRI2Extension.base,
+    &brw_vtable.base,
+    &brw_config_options.base,
     NULL
 };
+
+PUBLIC const __DRIextension **__driDriverGetExtensions_i965(void)
+{
+   globalDriverAPI = &brw_driver_api;
+
+   return brw_driver_extensions;
+}