On Wed, Sep 04, 2002 at 01:37:28PM -0700, Jason R Thorpe wrote:
> On Wed, Sep 04, 2002 at 10:12:19PM +0200, Manuel Bouyer wrote:
> 
>  > Ho, I didn't say it shouldn't be done, I just said it was not the best
>  > fix for the raidframe problem :)
> 
> It could actually turn out to be quite a good thing for the "raidframe
> problem".
> 
> Consider the case of writing two full stripes worth of data... as the
> second stripe comes around, you're going to end up with adjacent requests
> for the underlying components, no?  (This would certainly be the case
> with RAID-0.)

But in this case wouldn't they already be adjacent at the upper level ?
What we say here is that MAXPHYS for a raid-0 raid is the sum of the MAXPHYSs
of the underlying devices.

Hum, I didn't consider the case when the stripe unit size is smaller
than the underlying MAXPHYS. Really raidframe's MAXPHYS is n*stripe unit size.
So sure, merging requests at the request queue would help there too.

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
--