Re: [Qemu-devel] [PATCH v20 17/26] rbd.c: replace QEMUOptionParameter with QemuOpts

[Chunyan Liu]

2014-02-18 2:57 GMT+08:00 Eric Blake <address@hidden>:

On 02/11/2014 11:33 PM, Chunyan Liu wrote:
> rbd.c: replace QEMUOptionParameter with QemuOpts
>
> Signed-off-by: Dong Xu Wang <address@hidden>
> Signed-off-by: Chunyan Liu <address@hidden>
> ---
>  block/rbd.c |   63 +++++++++++++++++++++++++++++------------------------------
>  1 files changed, 31 insertions(+), 32 deletions(-)
>

> +    objsize = qemu_opt_get_size_del(opts, BLOCK_OPT_CLUSTER_SIZE, 0);
> +    if (objsize) {
> +        if ((objsize - 1) & objsize) {    /* not a power of 2? */
> +            error_report("obj size needs to be power of 2");
> +            return -EINVAL;
> +        }
> +        if (objsize < 4096) {
> +            error_report("obj size too small");
> +            return -EINVAL;
>          }

> +        {
> +            .name = BLOCK_OPT_CLUSTER_SIZE,
> +            .type = QEMU_OPT_SIZE,
> +            .help = "RBD object size",
> +            .def_value_str = stringify(0),

Do we really want to list a default of 0, given that the code behaves
the same as if 0 had been specified or if the argument had been omitted?
 Passing 0 doesn't really mean using a cluster size of 0, which makes
listing it as an explicit default would look odd.

Inherited from old patches. Not necessary in fact. Will remove that.

 

> -    .create_options     = qemu_rbd_create_options,
> +    .create_opts        = qemu_rbd_create_opts,

Hmm, I've noticed that your series is inconsistent on whether it uses '=
foo_opts' or '= &foo_opts' in these initializers (for example, 15/26
used & even though nothing else did, 16/26 uses & but all other
initializers were also using &, and in this 17/26 nothing is using &).
Not that it matters to the compiler, but there's a lot to be said for
consistency.

Sorry, will correct that.

 

--
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org