>>>>> "Kevin" == "Kevin P Neal" <kpneal@pobox.com> writes:

    Kevin> It would actually break FFS for a partition to grow
    Kevin> underneath it?

Not necessarily. For those who are unfortunate to have LVM, there is a
utility called extendfs which allows you to expand the filesystem -
here it's just ufs, but I suppose would work for vxfs too - into new
logical volumes.

    Kevin> Would newfs_ffs (or whatever) know which devices could
    Kevin> support the filesystem? How would that work?

I guess it works by having some kernel struct which has a number of
block counts and disk devices and offsets which comprise each logical
volume.

    Kevin> Lemmie see if I have this straight: 1) User program decides
    Kevin> to grow/shrink a partition.  2) User program tells
    Kevin> partition to grow/shrink 3) Virtual partition asks for the
    Kevin> permission of the filesystem 4) Assuming they agree, they
    Kevin> work together somehow to do whatever -- Filesystem must
    Kevin> make sure there is enough space available to shrink down
    Kevin> into. Non-resizable filesystems would just veto the whole
    Kevin> shebang.

Not exactly. In the LVM paradigm, physical volumes - disks or bits of
disks can be added and removed from a logical volume more or less at
will. [Obviously there are sanity checks that the PV isn't used.] Once
you've added a new PV to the logical volume, extendfs is used to
diddle the superblocks so the filesystem can use the new space.

    Kevin> Ok. Doesn't this assume that whatever is sitting on top of
    Kevin> the partition is a filesystem? What if it isn't? What if a
    Kevin> database or something or other is sitting on top? Would
    Kevin> resizing be possible? Wouldn't the mechanism of
    Kevin> communicating the resize be different because of the user
    Kevin> program hitting the disk instead of going through a
    Kevin> filesystem?

It depends on what it's being used for.

    Kevin> Third case: What do you do about resizing a swap partition?

This is hardly different from adding an extra swap device. Just add a
vnode which points at the extra space and use swapon to make the
kernel use it.