Re: [Qemu-devel] [PATCH v2 3/5] block: Change BlockDriverInfo.cluster_size to 64 bits

[Stefan Hajnoczi <stefanha@gmail.com>]

On Tue, Feb 11, 2014 at 11:28:37AM +0800, Fam Zheng wrote:
> VMDK could have big cluster_size for monolithicFlat. It implements
> .bdrv_get_info now, a 32 bit field is likely to overflow.
> 
> Signed-off-by: Fam Zheng <famz@redhat.com>
> ---
>  include/block/block.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/include/block/block.h b/include/block/block.h
> index 826b5da..3eb4e54 100644
> --- a/include/block/block.h
> +++ b/include/block/block.h
> @@ -14,7 +14,7 @@ typedef struct BlockJob BlockJob;
>  
>  typedef struct BlockDriverInfo {
>      /* in bytes, 0 if irrelevant */
> -    int cluster_size;
> +    int64_t cluster_size;

Not sure if this commit is enough:

If the cluster size is really larger than 2 GB and a compressed image is
in use, then we will attempt 2 GB I/O.  There are also other codepaths
that use cluster_size and would similarly allocate 2 GB bounce buffers!

We need to think through what happens in those cases.  Should we refuse
after a certain size?  Will we need to enhance some code paths to
actually split up the bounce buffers (if possible)?

Stefan