Kernel
Configuration
Enable the following options in the kernel configuration (choose
client and/or server support as appropriate) and recompile the
kernel if necessary:
File systems --->
[*] Network File Systems ---> [CONFIG_NETWORK_FILESYSTEMS]
<*/M> NFS client support [CONFIG_NFS_FS]
<*/M> NFS server support [CONFIG_NFSD]
Select the appropriate sub-options that appear when the above
options are selected.
Note
In BLFS we assume that nfs v3 will be used. If the server offers nfs v4 (for linux,
CONFIG_NFSD_V4) then auto-negotiation for v3 will fail and you
will need to add nfsver=3
to the
mount options. This also applies if that option is enabled in the
client's kernel, for
example in a distro trying to mount from a BLFS v3 server.
Installation
of NFS Utilities
Before you compile the program, ensure that the nobody
user and nogroup
group have been created as done in the
current LFS book. You can add them by running the following
commands as the root
user:
groupadd -g 99 nogroup &&
useradd -c "Unprivileged Nobody" -d /dev/null -g nogroup \
-s /bin/false -u 99 nobody
Note
The classic uid and gid values are 65534 which is also -2 when
interpreted as a signed 16-bit number. These values impact other
files on some filesystems that do not have support for sparse
files. The nobody
and
nogroup
values are relatively
arbitrary. The impact on a server is nil if the exports
file is configured correctly. If it is
misconfigured, an ls
-l or ps listing will show a uid or
gid number of 65534 instead of a name. The client uses
nobody
only as the user running
rpc.statd.
Install NFS Utilities by running
the following commands:
./configure --prefix=/usr \
--sysconfdir=/etc \
--without-tcp-wrappers \
--disable-nfsv4 \
--disable-gss &&
make
If your /usr
directory is NFS
mounted, you should install the executables in /sbin
by passing an additional parameter
--sbindir=/sbin
to the above
./configure command.
Now, as the root
user:
make install &&
chmod u+w,go+r /sbin/mount.nfs
The tests for this package require that the package be installed
and configured as specified below. In addition, the the daemon from
rpcbind-0.2.3 needs to be running and the tests
need to be run as the root
user.
To test the results, issue, as root: make check.
Command Explanations
--without-tcp-wrappers
:
This option is needed because TCP Wrappers is not in BLFS.
--disable-nfsv4
: This
allows the package to be built when libnfsidmap has not been
installed.
--disable-gss
: Disables
support for RPCSEC GSS (RPC Security).
Configuring NFS Utilities
Server Configuration
/etc/exports
contains the exported
directories on NFS servers. Refer to the exports.5
manual page for the syntax of this
file. Also refer to the "NFS HowTo" available at http://nfs.sourceforge.net/nfs-howto/
for information on how to configure the servers and clients in a
secure manner. For example, for sharing the /home
directory over the local network, the
following line may be added:
/home 192.168.0.0/24
(rw,subtree_check,anonuid=99,anongid=99)
Note
Be sure to replace the directory, network address. and prefix
above to match your network. The only space in the line above
should be between the directory and the network address.
Install the NFSv4 server units included in the
blfs-systemd-units-20160602
package to start the server at boot.
make install-nfsv4-server
If you have disabled NFSv4 support, run the following command
as the root
user to omit the
NFSv4 specific systemd units:
make install-nfs-server
You can edit the /etc/default/nfs-utils
file to change the
startup options for NFS daemons. Defaults should be fine for
most use cases.
Client Configuration
/etc/fstab
contains the directories
that are to be mounted on the client. Alternately the partitions
can be mounted by using the mount command with the proper
options. To mount the /home
and
/usr
partitions, add the following
to the /etc/fstab
:
<server-name>
:/home /home nfs rw,_netdev 0 0
<server-name>
:/usr /usr nfs ro,_netdev 0 0
The options which can be used are specified in man 5 nfs. If both the client
and server are running recent versions of linux, most of the
options will be negotiated. You can specify either rw
or ro
,
_netdev
if the filesystem is to be
automatically mounted at boot, or noauto
(and perhaps user
) for other filesystems.
If the fileserver is not running a recent version of linux, you
may need to specifiy other options.
If you are using systemd, you may need to enable autofs v4 in
your kernel, and add the option comment=systemd.automount
. Some machines need
this, because systemd tries to mount the external fs's before the
network is up, others do not need it. An alternative is for
root
to run mount -a.
Note
The following systemd units are not required if
the nfs-server units are installed.
Install the units included in the blfs-systemd-units-20160602 package to
start the client services at boot.
make install-nfs-client