On 11/13/2017 10:20 AM, Vladimir Sementsov-Ogievskiy wrote:
> Like other setters here these functions should take a lock.
> 
> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
> ---
>  block/dirty-bitmap.c | 4 ++++
>  1 file changed, 4 insertions(+)

Should this patch be in 2.11?

> 
> diff --git a/block/dirty-bitmap.c b/block/dirty-bitmap.c
> index bd04e991b1..2a0bcd9e51 100644
> --- a/block/dirty-bitmap.c
> +++ b/block/dirty-bitmap.c
> @@ -397,15 +397,19 @@ void bdrv_remove_persistent_dirty_bitmap(BlockDriverState *bs,
>  /* Called with BQL taken.  */
>  void bdrv_disable_dirty_bitmap(BdrvDirtyBitmap *bitmap)
>  {
> +    bdrv_dirty_bitmap_lock(bitmap);
>      assert(!bdrv_dirty_bitmap_frozen(bitmap));
>      bitmap->disabled = true;
> +    bdrv_dirty_bitmap_unlock(bitmap);

Why do we need this lock in addition to BQL?

>  }
>  
>  /* Called with BQL taken.  */
>  void bdrv_enable_dirty_bitmap(BdrvDirtyBitmap *bitmap)
>  {
> +    bdrv_dirty_bitmap_lock(bitmap);
>      assert(!bdrv_dirty_bitmap_frozen(bitmap));
>      bitmap->disabled = false;
> +    bdrv_dirty_bitmap_unlock(bitmap);

Again, why do we need this in addition to BQL?

The commit message needs more details about a scenario where our
existing BQL lock is insufficient to prevent misuse of bitmap->disabled.

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3266
Virtualization:  qemu.org | libvirt.org