drm/nv10: Fix crash when allocating a BO larger than half the available VRAM.
authorFrancisco Jerez <currojerez@riseup.net>
Thu, 3 Feb 2011 00:49:33 +0000 (01:49 +0100)
committerBen Skeggs <bskeggs@redhat.com>
Wed, 16 Feb 2011 23:39:45 +0000 (09:39 +1000)
Reported-by: Alex Buell <alex.buell@munted.org.uk>
Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
drivers/gpu/drm/nouveau/nouveau_bo.c

index a7fae26f4654e6bf1b32902084b83cd332e8522f..98dd970cea243b60a5a7018c06d6d07c3661c01b 100644 (file)
@@ -128,6 +128,7 @@ nouveau_bo_new(struct drm_device *dev, struct nouveau_channel *chan,
                }
        }
 
+       nvbo->bo.mem.num_pages = size >> PAGE_SHIFT;
        nouveau_bo_placement_set(nvbo, flags, 0);
 
        nvbo->channel = chan;
@@ -166,17 +167,17 @@ static void
 set_placement_range(struct nouveau_bo *nvbo, uint32_t type)
 {
        struct drm_nouveau_private *dev_priv = nouveau_bdev(nvbo->bo.bdev);
+       int vram_pages = dev_priv->vram_size >> PAGE_SHIFT;
 
        if (dev_priv->card_type == NV_10 &&
-           nvbo->tile_mode && (type & TTM_PL_FLAG_VRAM)) {
+           nvbo->tile_mode && (type & TTM_PL_FLAG_VRAM) &&
+           nvbo->bo.mem.num_pages < vram_pages / 2) {
                /*
                 * Make sure that the color and depth buffers are handled
                 * by independent memory controller units. Up to a 9x
                 * speed up when alpha-blending and depth-test are enabled
                 * at the same time.
                 */
-               int vram_pages = dev_priv->vram_size >> PAGE_SHIFT;
-
                if (nvbo->tile_flags & NOUVEAU_GEM_TILE_ZETA) {
                        nvbo->placement.fpfn = vram_pages / 2;
                        nvbo->placement.lpfn = ~0;