[dm-devel] Re: [PATCH block#for-2.6.31] block: add request clone interface (v2)

[Kiyoshi Ueda]

Hi Boaz, Jeff, Jens,

Thank you for your ideas.
It's time to decide now?  Please see below.

On 2009/06/15 18:30 +0900, Boaz Harrosh wrote:
> > On 06/15/2009 06:31 AM, Kiyoshi Ueda wrote:
>> >> On 06/12/2009 11:33 PM +0900, Jeff Moyer wrote:
>>> >>> Jens Axboe <jens.axboe at oracle.com> writes:
>>>> >>>> On Thu, Jun 11 2009, Jeff Moyer wrote:
>>>>> >>>>> Jens Axboe <jens.axboe at oracle.com> writes:
>>>>> >>>>> Is blk_rq_unprep_clone really the best name?
>>>>> >>>>>           ^^^^^^
>>>> >>>> Probably not, but I'm not very good at coming up with elegant names.
>>>> >>>> Your email should have included a new suggestion :-)
>>> >>> Fair enough.  ;)
>>> >>>
>>>> >>>>  - blk_rq_unprep_clone(struct request *clone)
>>>> >>>>      * Frees cloned bios from the clone request.
>>> >>> Why not blk_rq_free_clone?
>> >> Because the 'clone' is not freed in this interface.
>> >> This interface frees only bios in the 'clone'.
>> >> Allocating/freeing the 'clone' are the caller's work, since
>> >> only the caller knows how to allocate/free it.
>> >>
>> >> 'prep' after 'alloc' and 'unprep' before 'free' is symmetric
>> >> and I feel a good candidate for my request-stacking driver,
>> >> so I chose it. 
> > 
> > I'm not a native English speaker as well, so I'm fine
> > with blk_rq_{prep,unprep}_clone. But maybe the English
> > people don't like it?
> > 
> > Perhaps
> > blk_rq_{clone,declone} or blk_rq_{clone,declone}_bios
> > 
> > (Both unclone and declone are found on the net but are not
> >  found in the free dictionary)

I had a feeling that blk_rq_{clone,declone} allocates/frees
the clone request inside the interfaces like bio_clone(), so
I didn't take such namings.
And, the clone setup interface may not only make bio clones
but also do something else (for other request members), so
I didn't add any 'bio' namings to the interfaces.

Jens, what do you prefer?

Thanks,
Kiyoshi Ueda