In function virtio_gpufb_create, a virtio_gpu_object is allocated for
framebuffer using virtio_gpu_alloc_object.
In virtio_gpu_fbdev_destroy, instead of freeing the object, pointer to
it is set to NULL. This leads to memory leak during framebuffer
destruction, which is reported to kmesg with a message like this:
Memory manager not clean during takedown.
With DRM_DEBUG_MM enabled, following additional information is printed
about the leak:
node [00100000 + 000001d5]: inserted at
save_stack.isra.9+0x67/0xc0
drm_mm_insert_node_in_range+0x325/0x4f0
drm_vma_offset_add+0x46/0x60
ttm_bo_init_reserved+0x2c9/0x400
ttm_bo_init+0x2a/0x80
virtio_gpu_object_create+0x139/0x180
virtio_gpu_alloc_object+0x2f/0x60
virtio_gpufb_create+0xac/0x2a0
Correctly freeing virtio_gpu_object during framebuffer destruction
fixes the issue.
Signed-off-by: Damir Shaikhutdinov <[email protected]>
Signed-off-by: Kiran Pawar <[email protected]>
---
drivers/gpu/drm/virtio/virtgpu_fb.c | 7 +++++--
1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/virtio/virtgpu_fb.c
b/drivers/gpu/drm/virtio/virtgpu_fb.c
index 15d18fd0c64b..10a66a387bfb 100644
--- a/drivers/gpu/drm/virtio/virtgpu_fb.c
+++ b/drivers/gpu/drm/virtio/virtgpu_fb.c
@@ -301,11 +301,14 @@ static int virtio_gpu_fbdev_destroy(struct drm_device
*dev,
drm_fb_helper_unregister_fbi(&vgfbdev->helper);
- if (vgfb->obj)
- vgfb->obj = NULL;
drm_fb_helper_fini(&vgfbdev->helper);
drm_framebuffer_cleanup(&vgfb->base);
+ if (vgfb->obj) {
+ virtio_gpu_gem_free_object(vgfb->obj);
+ vgfb->obj = NULL;
+ }
+
return 0;
}
static const struct drm_fb_helper_funcs virtio_gpu_fb_helper_funcs = {
--
2.17.1
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
