The first patch is unchanged. The 2nd and 3rd patches reflact Dan's
suggestions. (wrt Cole's concern about the pool object being unlocked
during the buildVol call - Dan pointed out that this is read-only
usage of the object, and it can't be deleted while in use, so it's
safe to pass it in here.)
Note that I've tested this as much as I can with my own rig plus some
DEBUG statements and small code changes (in particular, forcing
creation of volumes on local filesystems to use the code intended for
netfs), but testing on real NFS (as well as Samba to see if it solves
that problem too) would be very much appreciated.