yet more cluster checks

give up on rotational layout if there is only one rotational position

doasyncfree and prtrealloc move FFS sysctl variables

daddr_t => ufs_daddr_t

more debugging for clustering

convert quad => int64_t (from mycroft)

more consistent with DEBUG (from mycroft)

add more debugging code for ffs_reallocblks

use and maintain fs_maxcluster array

debugging sysctl needs to be ifdef DEBUG (from mycroft)

conditionally avoid synchronous writes (and indicate real fix)

If the offset within the indirect block of the "end block" of our
current cluster is greater than the length of the cluster, then we know
that the start block of the cluster must also be in that indi

If the offset within the indirect block of the "end block" of our
current cluster is greater than the length of the cluster, then we know
that the start block of the cluster must also be in that indirect block
(since the cluster is contiguous) hence the entire cluster is mapped by
a single indirect block (or completely in the direct blocks if end_lvl == 0).
From: mike@cs.utah.edu (Mike Hibler)

show more ...

first cut at improving cluster allocation

lint

lint

changes for 4.4BSD-Lite requested by USL

changes for 4.4BSD-Lite requested by USL

4.4BSD snapshot (revision 8.1); add 1993 to copyright

take next available block on drives of unknown geometry

Detect contiguous allocation at the beginning of files and
indirect blocks so that you can allocate "maxcontig" blocks before paying
a rot-delay.

delete USES; add parameter declarations; streamline unmarshalling

registerized vnodeop ops after vnode interface conversion

vnodeop argument macro expansion

ffs_vfree now returns a (0) error

vnode interface conversion