reassign instead of invalidating buffers associated with other vnodes
when doing vinvalbuf (fix for itrunc from noemi@osf.org)

minor optimization

manage dirty and clean buffer chains;
track number of active I/O's;
expand notavail inline to avoid spurious spl's

blkflush is no longer used;
VREF => VHOLD and vrele => HOLDRELE as buffers use holdcnts not usecounts

have to save v_mountf before vput'ing the vnode

add missing splx's

major overhaul to generalize for a logical buffer pool

have to check both v_mount for logical buffer cache and v_mounton for
block devices using the physical buffer cache when doing bflush and binval

updated version of blkflush from Torek

minor fixes for NFS

get the splx's in the right place

generalize the buffer pool so that NFS can become a client

have to remember to lock buffers before trying to grab them

binval must wait for I/O to complete before marking a buffer INVAL;
use bawrite rather than setting ASYNC flag and calling bwrite

update reference counts on vnodes using VREF

bug fix from Rick Macklem

merge in vnodes

../machine => machine

write dirty buffers that are growing to avoid bread'ing stale
data from the disk over the newer data in the buffer.

working with pack labels

add ifdefs for dev_bsize (almost working)

4.3BSD release version

lint and cleanups from Sam

sanity checks in getblk() and geteblk()

I've wanted to make this comment untrue for a long time...