PATCH improve worst case module load times

Recently I ran in to an apparently well known ppc/ppc64 issue regarding slow 
module load times which I see was posted to lkml back on '05.

   http://lkml.org/lkml/2005/9/28/180

  In my environment I'm forced to live with a few very large kernel modules 
and frequent reboots so this was become a bit of an issue.  I took a look at 
the root cause and put together a patch to address the issue.  The proposed 
patch resolves the problem by removing the order(n^2) algorithm and replaces 
it with an order(n) chained hash.

  It works well for me but I'm not a powerpc guru so I'd appreciate a review 
of the patch.  I thought about posting it to lkml but after I learned there 
was a ppc dev list this seemed like the right place to post.
  
-- 
Thanks,
Brian Behlendorf
behlendorf1@llnl.gov