This reverts commit 
d323bb44e4d23802eb25d13de1f93f2335bd60d0.
Fixes a double-free regression:
[    4.357928]  drm_gem_shmem_free_object+0xb4/0x100
[    4.358983]  virtio_gpu_dequeue_ctrl_func+0xd9/0x290
[    4.360343]  process_one_work+0x1d2/0x3a0
[    4.361581]  worker_thread+0x45/0x3c0
[    4.362645]  kthread+0xf6/0x130
[    4.363543]  ? process_one_work+0x3a0/0x3a0
[    4.364770]  ? kthread_park+0x80/0x80
[    4.365799]  ret_from_fork+0x35/0x40
[    4.367103] Modules linked in:
[    4.367958] CR2: 
0000000000000018
[    4.368857] ---[ end trace 
db84f7a2974d5c79 ]---
[    4.370118] RIP: 0010:dma_direct_unmap_sg+0x1f/0x60
In addition, virtio has it's own set of dma-ops so there's not an
obviously clean way to transition to shmem helpers.
Fixes: d323bb44e4d2 ("drm/virtio: Call the right shmem helpers")
Signed-off-by: Gurchetan Singh <gurchetansingh@chromium.org>
Link: http://patchwork.freedesktop.org/patch/msgid/20200615230500.551-1-gurchetansingh@chromium.org
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
        if (ret < 0)
                return -EINVAL;
 
-       shmem->pages = drm_gem_shmem_get_pages_sgt(&bo->base.base);
+       /*
+        * virtio_gpu uses drm_gem_shmem_get_sg_table instead of
+        * drm_gem_shmem_get_pages_sgt because virtio has it's own set of
+        * dma-ops. This is discouraged for other drivers, but should be fine
+        * since virtio_gpu doesn't support dma-buf import from other devices.
+        */
+       shmem->pages = drm_gem_shmem_get_sg_table(&bo->base.base);
        if (!shmem->pages) {
                drm_gem_shmem_unpin(&bo->base.base);
                return -EINVAL;