On 04.11.19 16:04, Max Reitz wrote:
> On 26.10.19 23:25, Alberto Garcia wrote:
>> Setting the QCOW_OFLAG_ZERO bit of the L2 entry is forbidden if an
>> image has subclusters. Instead, the individual 'all zeroes' bits must
>> be used.
>>
>> Signed-off-by: Alberto Garcia <berto@igalia.com>
>> ---
>>  block/qcow2-cluster.c | 14 ++++++++++----
>>  1 file changed, 10 insertions(+), 4 deletions(-)
>>
>> diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c
>> index e67559152f..3e4ba8d448 100644
>> --- a/block/qcow2-cluster.c
>> +++ b/block/qcow2-cluster.c
>> @@ -1852,7 +1852,7 @@ static int zero_in_l2_slice(BlockDriverState *bs, uint64_t offset,
>>      assert(nb_clusters <= INT_MAX);
>>  
>>      for (i = 0; i < nb_clusters; i++) {
>> -        uint64_t old_offset;
>> +        uint64_t old_offset, l2_entry = 0;
>>          QCow2ClusterType cluster_type;
>>  
>>          old_offset = get_l2_entry(s, l2_slice, l2_index + i);
>> @@ -1869,12 +1869,18 @@ static int zero_in_l2_slice(BlockDriverState *bs, uint64_t offset,
>>  
>>          qcow2_cache_entry_mark_dirty(s->l2_table_cache, l2_slice);
>>          if (cluster_type == QCOW2_CLUSTER_COMPRESSED || unmap) {
>> -            set_l2_entry(s, l2_slice, l2_index + i, QCOW_OFLAG_ZERO);
>>              qcow2_free_any_clusters(bs, old_offset, 1, QCOW2_DISCARD_REQUEST);
> 
> It feels wrong to me to free the cluster before updating the L2 entry.

(Although it’s pre-existing, as set_l2_entry() is just an in-cache
operation anyway :-/)

Max