Re: [PATCH v4 2/2] quorum: Implement bdrv_co_pwrite_zeroes()

[Max Reitz <mreitz@redhat.com>]

On 13.11.20 17:52, Alberto Garcia wrote:
> This simply calls bdrv_co_pwrite_zeroes() in all children.
> 
> bs->supported_zero_flags is also set to the flags that are supported
> by all children.
> 
> Signed-off-by: Alberto Garcia <berto@igalia.com>
> ---
>   block/quorum.c             | 36 ++++++++++++++++++++++++++++++++++--
>   tests/qemu-iotests/312     | 11 +++++++++++
>   tests/qemu-iotests/312.out |  8 ++++++++
>   3 files changed, 53 insertions(+), 2 deletions(-)

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

> @@ -897,6 +910,21 @@ static QemuOptsList quorum_runtime_opts = {
>       },
>   };
>   
> +static void quorum_refresh_flags(BlockDriverState *bs)
> +{
> +    BDRVQuorumState *s = bs->opaque;
> +    int i;
> +
> +    bs->supported_zero_flags =
> +        BDRV_REQ_FUA | BDRV_REQ_MAY_UNMAP | BDRV_REQ_NO_FALLBACK;
> +
> +    for (i = 0; i < s->num_children; i++) {
> +        bs->supported_zero_flags &= s->children[i]->bs->supported_zero_flags;
> +    }
> +
> +    bs->supported_zero_flags |= BDRV_REQ_WRITE_UNCHANGED;

This made me wonder whether it’s true when rewrite_corrupted is set.  I 
think it is, because that only does something when reading from the 
children (i.e. not for write requests from parents, where this flag 
might be set).

Looking into quorum_child_perm(), quorum doesn’t take the WRITE 
permission on its children even if rewrite_corrupted is true.  Hm... 
Something to look into.

Max

> +}
> +
>   static int quorum_open(BlockDriverState *bs, QDict *options, int flags,
>                          Error **errp)
>   {