The qcow2_truncate() code is mostly independent from whether
we're using L2 slices or full L2 tables, but in full and
falloc preallocation modes new L2 tables are allocated using
qcow2_alloc_cluster_link_l2().  Therefore the code needs to be
modified to ensure that all nb_clusters that are processed in each
call can be allocated with just one L2 slice.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
---
 block/qcow2.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/block/qcow2.c b/block/qcow2.c
index 78f067cae7..529becfa30 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -3261,8 +3261,9 @@ static int qcow2_truncate(BlockDriverState *bs, int64_t offset,
         guest_offset = old_length;
         while (nb_new_data_clusters) {
             int64_t guest_cluster = guest_offset >> s->cluster_bits;
-            int64_t nb_clusters = MIN(nb_new_data_clusters,
-                                      s->l2_size - guest_cluster % s->l2_size);
+            int64_t nb_clusters = MIN(
+                nb_new_data_clusters,
+                s->l2_slice_size - guest_cluster % s->l2_slice_size);
             QCowL2Meta allocation = {
                 .offset       = guest_offset,
                 .alloc_offset = host_offset,
-- 
2.11.0

On 2018-01-26 15:59, Alberto Garcia wrote:
> The qcow2_truncate() code is mostly independent from whether
> we're using L2 slices or full L2 tables, but in full and
> falloc preallocation modes new L2 tables are allocated using
> qcow2_alloc_cluster_link_l2().  Therefore the code needs to be
> modified to ensure that all nb_clusters that are processed in each
> call can be allocated with just one L2 slice.
> 
> Signed-off-by: Alberto Garcia <berto@igalia.com>
> Reviewed-by: Eric Blake <eblake@redhat.com>
> ---
>  block/qcow2.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/block/qcow2.c b/block/qcow2.c
> index 78f067cae7..529becfa30 100644
> --- a/block/qcow2.c
> +++ b/block/qcow2.c
> @@ -3261,8 +3261,9 @@ static int qcow2_truncate(BlockDriverState *bs, int64_t offset,
>          guest_offset = old_length;
>          while (nb_new_data_clusters) {
>              int64_t guest_cluster = guest_offset >> s->cluster_bits;
> -            int64_t nb_clusters = MIN(nb_new_data_clusters,
> -                                      s->l2_size - guest_cluster % s->l2_size);
> +            int64_t nb_clusters = MIN(
> +                nb_new_data_clusters,
> +                s->l2_slice_size - guest_cluster % s->l2_slice_size);

An alternative would be the
"s->l2_slice_size - offset_to_l2_slice_index(s, guest_offset)" we
basically have elsewhere, but that's longer and doesn't really matter:

Reviewed-by: Max Reitz <mreitz@redhat.com>

>              QCowL2Meta allocation = {
>                  .offset       = guest_offset,
>                  .alloc_offset = host_offset,
> 

On Thu 01 Feb 2018 08:46:46 PM CET, Max Reitz <mreitz@redhat.com> wrote:
>> @@ -3261,8 +3261,9 @@ static int qcow2_truncate(BlockDriverState *bs, int64_t offset,
>>          guest_offset = old_length;
>>          while (nb_new_data_clusters) {
>>              int64_t guest_cluster = guest_offset >> s->cluster_bits;
>> -            int64_t nb_clusters = MIN(nb_new_data_clusters,
>> -                                      s->l2_size - guest_cluster % s->l2_size);
>> +            int64_t nb_clusters = MIN(
>> +                nb_new_data_clusters,
>> +                s->l2_slice_size - guest_cluster % s->l2_slice_size);
>
> An alternative would be the
> "s->l2_slice_size - offset_to_l2_slice_index(s, guest_offset)" we
> basically have elsewhere, but that's longer and doesn't really matter:

It's a bit longer but it looks better and we can get rid of
guest_cluster, so I think I'll change it.

Berto