From: Tina Zhang <tina.zhang@intel.com>
Date: Mon, 15 Jan 2018 14:41:42 +0800
Subject: drm/i915/gvt: Keep obj->dma_buf link NULL during exporting
Git-commit: 412718a10926f77052d8107cb3c5d9dbc96cf8c9
Patch-mainline: v4.16-rc1
References: FATE#326289 FATE#326079 FATE#326049 FATE#322398 FATE#326166

According to commit (319c933c71f3dbdb2b3274d1634d3494c70efa06)
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Thu Aug 15 00:02:46 2013 +0200

    drm/prime: proper locking+refcounting for obj->dma_buf link

obj->dma_buf link should be reinstated at import time.

Gvt-g dma-buf buffer exposeing might be simpler, as there won't be much
racing during Gvt-g dma-buf exposing. In other words, Gvt-g dma-buf
exposing can guarantee exposing happens before gem close ioctl, and Gvt-g
is the only exporter of the guest framebuffer.

But following the drm prime scheme can give Gvt-g a chance to increase a
dma-buf reference count during importing. Otherwise, we have to increase
the reference during exposing, which will break the case that the only
reference userspace has held was through the dma-buf fd and the reference
count is one.

Signed-off-by: Tina Zhang <tina.zhang@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Zhenyu Wang <zhenyuw@linux.intel.com>
Cc: Zhi Wang <zhi.a.wang@intel.com>
Cc: Hang Yuan <hang.yuan@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Acked-by: Petr Tesarik <ptesarik@suse.com>
---
 drivers/gpu/drm/i915/gvt/dmabuf.c |    1 -
 1 file changed, 1 deletion(-)

--- a/drivers/gpu/drm/i915/gvt/dmabuf.c
+++ b/drivers/gpu/drm/i915/gvt/dmabuf.c
@@ -472,7 +472,6 @@ int intel_vgpu_get_dmabuf(struct intel_v
 		ret = PTR_ERR(dmabuf);
 		goto out_free_gem;
 	}
-	obj->base.dma_buf = dmabuf;
 
 	i915_gem_object_put(obj);