[dm-devel] [PATCH 1/2] blkdev: fix merge_bvec_fn return value checks

[Alasdair G Kergon]

On Sat, Mar 06, 2010 at 10:52:37AM +1100, Neil Brown wrote:
> But my first suggestion, that splitting could be made easier, still stands.

In full agreement we should try to make it easier.

> And do you honour merge_bvec_fn's of underlying devices?  A quick grep
> suggests you do only for dm-linear and dm-crypt.  

We implemented it to address some specific performance problems reported by
people using those targets.  So far there hasn't been any pressing need to
implement it for other targets.

> This suggests to me that it
> is actually a hard interface to support completely in a stacked device, so we
> might be better off without it.

All this code is hard - but very useful.  The direction I want to explore when
I finally get some time to work on this is one which will probably make more
use of, and extend, merge_bvec.

> that it makes sense to only require that the lowest level does the
> splitting, as that is where the specifics on what might be required exists.

We have a hybrid model at the moment - some stuff dealt with top-down, other
stuff bottom-up.  This thread describes just one of several problems related to
this.  In the next incarnation of device-mapper I hope we'll find a way to
eliminate all of them together.  Maybe I'll finally get time later this year to
start working on these ideas properly.

Alasdair