Looks like the NFS startup procedure is incorrect compared with other distributions:
WRL5:
start_nfsd "$NFS_SERVERS"
start_mountd
exportfs -a
Debian:
$PREFIX/sbin/exportfs -r
/sbin/rpc.nfsd ...
/sbin/rpc.mountd ...
nfs-utils upstream:
22 Type=oneshot
23 RemainAfterExit=yes
24 ExecStartPre=/usr/sbin/exportfs -r
25 ExecStart=/usr/sbin/rpc.nfsd $RPCNFSDARGS
26 ExecStop=/usr/sbin/rpc.nfsd 0
27 ExecStopPost=/usr/sbin/exportfs -au
28 ExecStopPost=/usr/sbin/exportfs -f
exportfs -r is supposed to setup a proper etab (at-least needed by
mountd). But in Yocto this will will be 0 bytes when mountd starts,
instead of proper data.
The upshot of this is that on some very fast machines the following may be observed on first boot:
2015-03-02T10:09:15+00:00 xxxx rpc.mountd[3749]: couldn't open
/var/lib/nfs/etab
