Fix NFS-based MFS support for ZFS_NOEXPORTS
With ZFS_NOEXPORTS we try to avoid using the ZFS sharenfs
attribute as it results in numerous HUPs of mountd
. However, we still use this with the NFS-based admin MFS on the m400 nodes. We need to integrate those exports into exports_setup and have nfsmfs_setup:
-
Not change the
sharenfs
attribute for every MFS we setup, which currently causes a HUP tomountd
on every call, -
Call exports_setup at the end to do one incremental update, rather than calling "zfs export -a" which causes another (in theory, redundant) HUP to
mountd
.