Re: [PATCH 1/1] Fix qcow2 corruption on discard

[Kevin Wolf <kwolf@redhat.com>]

Am 23.11.2020 um 16:49 hat Maxim Levitsky geschrieben:
> Commit 205fa50750 ("qcow2: Add subcluster support to zero_in_l2_slice()")
> introduced a subtle change to code in zero_in_l2_slice:
> 
> It swapped the order of
> 
> 1. qcow2_cache_entry_mark_dirty(s->l2_table_cache, l2_slice);
> 2. set_l2_entry(s, l2_slice, l2_index + i, QCOW_OFLAG_ZERO);
> 3. qcow2_free_any_clusters(bs, old_offset, 1, QCOW2_DISCARD_REQUEST);
> 
> To
> 
> 1. qcow2_cache_entry_mark_dirty(s->l2_table_cache, l2_slice);
> 2. qcow2_free_any_clusters(bs, old_offset, 1, QCOW2_DISCARD_REQUEST);
> 3. set_l2_entry(s, l2_slice, l2_index + i, QCOW_OFLAG_ZERO);
> 
> It seems harmless, however the call to qcow2_free_any_clusters
> can trigger a cache flush which can mark the L2 table as clean,
> and assuming that this was the last write to it,
> a stale version of it will remain on the disk.
> 
> Now we have a valid L2 entry pointing to a freed cluster. Oops.
> 
> Fixes: 205fa50750 ("qcow2: Add subcluster support to zero_in_l2_slice()")
> Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
> ---
>  block/qcow2-cluster.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c
> index 485b4cb92e..267b46a4ca 100644
> --- a/block/qcow2-cluster.c
> +++ b/block/qcow2-cluster.c
> @@ -2010,11 +2010,11 @@ static int zero_in_l2_slice(BlockDriverState *bs, uint64_t offset,
>              continue;
>          }
>  
> -        qcow2_cache_entry_mark_dirty(s->l2_table_cache, l2_slice);
>          if (unmap) {
>              qcow2_free_any_cluster(bs, old_l2_entry, QCOW2_DISCARD_REQUEST);
>          }
>          set_l2_entry(s, l2_slice, l2_index + i, new_l2_entry);
> +        qcow2_cache_entry_mark_dirty(s->l2_table_cache, l2_slice);

Good catch, but I think your order is wrong, too. We need the original
order from before 205fa50750:

1. qcow2_cache_entry_mark_dirty()
   set_l2_entry() + set_l2_bitmap()

2. qcow2_free_any_cluster()

The order between qcow2_cache_entry_mark_dirty() and set_l2_entry()
shouldn't matter, but it's important that we update the refcount table
only after the L2 table has been updated to not reference the cluster
any more.

Otherwise a crash could lead to a situation where the cluster is
allocated (because it has refcount 0), but it was still in use in an L2
table. This is a classic corruption scenario.

Kevin