Jens Axboe wrote:

I updated that patch, and converted IDE and SCSI to use it. See the
results here:

http://brick.kernel.dk/git/?p=linux-2.6-block.git;a=shortlog;h=blk-softirq

The main change from the version posted last october is killing the
'slightly' overdesigned completion queue hashing.

Nifty, I like.  Comments:

* use of spin_lock_irq() in all completion paths now makes me nervous.

* certainly it's what SCSI does now, but is a softirq really necessary? 
  Using a tasklet would kill all that per-cpu code, and notifier.