From: Christian König christian.koenig@amd.com
Large amounts of VRAM are usually not CPU accessible, so they are not mapped into the processes address space. But since the device drivers usually support swapping buffers from VRAM to system memory we can still run into an out of memory situation when userspace starts to allocate to much.
This patch gives the OOM and lower memory killer another hint which process is holding how many resources.
Signed-off-by: Christian König christian.koenig@amd.com Reviewed-by: Alex Deucher alexander.deucher@amd.com --- drivers/gpu/drm/drm_gem.c | 9 +++++++++ 1 file changed, 9 insertions(+)
diff --git a/drivers/gpu/drm/drm_gem.c b/drivers/gpu/drm/drm_gem.c index 16a1647..17d14c1 100644 --- a/drivers/gpu/drm/drm_gem.c +++ b/drivers/gpu/drm/drm_gem.c @@ -282,6 +282,9 @@ drm_gem_handle_delete(struct drm_file *filp, u32 handle)
if (dev->driver->gem_close_object) dev->driver->gem_close_object(obj, filp); + + atomic_long_sub(obj->size >> PAGE_SHIFT, &filp->filp->f_oom_badness); + drm_gem_object_handle_unreference_unlocked(obj);
return 0; @@ -358,6 +361,9 @@ drm_gem_handle_create_tail(struct drm_file *file_priv, } }
+ atomic_long_add(obj->size >> PAGE_SHIFT, + &file_priv->filp->f_oom_badness); + return 0; }
@@ -717,6 +723,9 @@ drm_gem_object_release_handle(int id, void *ptr, void *data) if (dev->driver->gem_close_object) dev->driver->gem_close_object(obj, file_priv);
+ atomic_long_sub(obj->size >> PAGE_SHIFT, + &file_priv->filp->f_oom_badness); + drm_gem_object_handle_unreference_unlocked(obj);
return 0;