[PATCHv3,20/20] block/raw: copy BlockLimits on raw_open

Signed-off-by: Peter Lieven <pl@kamp.de>
---
 block/raw_bsd.c |    1 +
 1 file changed, 1 insertion(+)

On 09/24/2013 07:35 AM, Peter Lieven wrote:
> Signed-off-by: Peter Lieven <pl@kamp.de>
> ---
>  block/raw_bsd.c |    1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/block/raw_bsd.c b/block/raw_bsd.c
> index 8dc7bba..7af26ad 100644
> --- a/block/raw_bsd.c
> +++ b/block/raw_bsd.c
> @@ -159,6 +159,7 @@ static int raw_open(BlockDriverState *bs, QDict *options, int flags,
>                      Error **errp)
>  {
>      bs->sg = bs->file->sg;
> +    memcpy(&bs->bl, &bs->file->bl, sizeof(struct BlockLimits));

Personally, I think that sizeof(var) is more robust, because if the
declaration of var is ever changed, you don't have to remember to also
touch up the memcpy.  As in:

memcpy(&bs->bl, &bs->file->bl, sizeof(bs->bl));

But there's plenty of examples of sizeof(type) in the codebase even when
a var is handy, so you are not alone, and that's not a reason for me to
request a respin.

On the other hand, why use memcpy() at all?

bs->bl = bs->file->bl;

should do the same trick, with less typing.

On Wed, Oct 02, 2013 at 11:11:05AM -0600, Eric Blake wrote:
> On 09/24/2013 07:35 AM, Peter Lieven wrote:
> > Signed-off-by: Peter Lieven <pl@kamp.de>
> > ---
> >  block/raw_bsd.c |    1 +
> >  1 file changed, 1 insertion(+)
> > 
> > diff --git a/block/raw_bsd.c b/block/raw_bsd.c
> > index 8dc7bba..7af26ad 100644
> > --- a/block/raw_bsd.c
> > +++ b/block/raw_bsd.c
> > @@ -159,6 +159,7 @@ static int raw_open(BlockDriverState *bs, QDict *options, int flags,
> >                      Error **errp)
> >  {
> >      bs->sg = bs->file->sg;
> > +    memcpy(&bs->bl, &bs->file->bl, sizeof(struct BlockLimits));
> 
> Personally, I think that sizeof(var) is more robust, because if the
> declaration of var is ever changed, you don't have to remember to also
> touch up the memcpy.  As in:
> 
> memcpy(&bs->bl, &bs->file->bl, sizeof(bs->bl));
> 
> But there's plenty of examples of sizeof(type) in the codebase even when
> a var is handy, so you are not alone, and that's not a reason for me to
> request a respin.
> 
> On the other hand, why use memcpy() at all?
> 
> bs->bl = bs->file->bl;
> 
> should do the same trick, with less typing.

Yes, please use struct assignment.

Stefan

Am 07.10.2013 um 10:38 schrieb Stefan Hajnoczi <stefanha@gmail.com>:

> On Wed, Oct 02, 2013 at 11:11:05AM -0600, Eric Blake wrote:
>> On 09/24/2013 07:35 AM, Peter Lieven wrote:
>>> Signed-off-by: Peter Lieven <pl@kamp.de>
>>> ---
>>> block/raw_bsd.c |    1 +
>>> 1 file changed, 1 insertion(+)
>>> 
>>> diff --git a/block/raw_bsd.c b/block/raw_bsd.c
>>> index 8dc7bba..7af26ad 100644
>>> --- a/block/raw_bsd.c
>>> +++ b/block/raw_bsd.c
>>> @@ -159,6 +159,7 @@ static int raw_open(BlockDriverState *bs, QDict *options, int flags,
>>>                     Error **errp)
>>> {
>>>     bs->sg = bs->file->sg;
>>> +    memcpy(&bs->bl, &bs->file->bl, sizeof(struct BlockLimits));
>> 
>> Personally, I think that sizeof(var) is more robust, because if the
>> declaration of var is ever changed, you don't have to remember to also
>> touch up the memcpy.  As in:
>> 
>> memcpy(&bs->bl, &bs->file->bl, sizeof(bs->bl));
>> 
>> But there's plenty of examples of sizeof(type) in the codebase even when
>> a var is handy, so you are not alone, and that's not a reason for me to
>> request a respin.
>> 
>> On the other hand, why use memcpy() at all?
>> 
>> bs->bl = bs->file->bl;
>> 
>> should do the same trick, with less typing.
> 
> Yes, please use struct assignment.

Okay, I was unsure because when looking at bdrv_move_feature_fields I found that there memcpy was used.

Peter