SMP VFS Last call

The SMP FFS/VFS patch has undergone several iterations and lots of serious
testing over the past few weeks.  Many people, especially Peter Holm, have
sent me good bug reports.  It's currently running on the port build
cluster and I have done extended load testing in small memory
configurations.  What this means is, after I get back from vacation, it's
going to go into the tree.  If you don't test it now, you will be in a few
weeks. :-)

http://www.chesapeake.net/~jroberson/smpffs.diff

For those of you that missed the last few emails on the topic;  This patch
dynamically acquires Giant in high-level VFS code only if the underlying
filesystem needs it.  It also adds locking to ffs and softupdates so that
FFS no longer needs Giant, unless you have QUOTA in your kernel config.
This patch also fixes a half dozen or so race conditions that were
discovered in the earlier VFS locking work that had not been tested until
now.

Cheers,
Jeff