Re: [PATCH] cpumask_t

["David S. Miller" <davem@redhat.com>]

   From: Rusty Russell <rusty@rustcorp.com.au>
   Date: Thu, 08 Aug 2002 17:39:18 +1000

   I've tested this now with making cpumask_t a struct, and it works fine
   (at the moment it's unsigned long for every arch, no change).

It worked because you cast the thing to (unsigned long *) in every
bitops.  We either:

1) shouldn't need to do that, meaning cpumask_t must be a long
   or array or longs

2) you need to abstract away bitops on cpumask_t so that one
   _really_ could make cpumask_t a struct with things other
   than the mask itself inside, so cpumask_test, cpumask_add,
   cpumask_remove or however you'd like to name them

Didn't we go through a lot of effort to sanitize bitops types
and kill the ugly casts? :-))))