That sounds like a plan. I'm guessing Jens Axboe is probably the one who ultimately has to be convinced. On Wed, Jun 30, 2004 at 01:46:43PM -0500, Kevin Corry wrote: > > I don't think it would be impossible to get this change into fs/bio.c during > 2.6. It would mostly be a matter of setting up the infrastructure (for which > there would be no initial users). Then we'd want to create a global pool > (sometime during kernel initialization) that would mimic the existing global > bio pool. Then make it so calls to bio_clone() and bio_alloc() would always > go to this global pool, which would mean a minimum of existing kernel code > would need to change. Then we'd be able to switch users of > bio_alloc/bio_clone over to private pools as necessary. > > -- > Kevin Corry > kevcorry@xxxxxxxxxx > http://evms.sourceforge.net/ > -- > > dm-devel@xxxxxxxxxx > https://www.redhat.com/mailman/listinfo/dm-devel