Hi,
On Wed, Apr 04, 2001 at 11:14:46PM +0200, Andi Kleen wrote:
> Hopefully there will some agreement on that final direction soon (so far
> none is in sight as far as I can see)
> Looks like design by comittee is not working very well here.
Final direction is looking much closer now. I spoke with Linus and
the "kiobuf controversy" turned out to be a trivial miscommunication
which we can easily fix. I'll start on that once I get back home some
time next week. Jens Axboe is already working on separating out the
buffer_head's buffer-description and cache-description fields and
pushing that knowledge into the block device layer, and we've got a
ton of other blockdev layer changes to work on once that is complete.
kiobufs are not hard. The hard thing here is really integration,
making sure that everybody is talking the same data structures. After
the kernel workshop I'm much happier about the prospect of getting
agreement on the data structures needed to get this to work.
Fortunately, the massive improvements in 2.4's elevator code have
bought us enough time to do this properly.
Cheers,
Stephen
|