From Gentoo Wiki
(Redirected from NFSv4)
Jump to:navigation Jump to:search
This page contains changes which are not marked for translation.
Other languages:
Not to be confused with NTFS.

This article has some todo items:
  • Review systemd support in article. man nfs.systemd, /etc/nfs.conf, etc.

Network File System (NFS) is a file system protocol that allows client machines to access network attached filesystems (called exports) from a host system. NFS is supported by the Linux kernel and userspace daemons and utilities are found in the net-fs/nfs-utils package.



NFS server support is not required for NFS clients. Conversely NFS client support is not required for NFS servers. Inotify support is only required for NFSv4. NFSv3 is only required for compatibility with legacy clients e.g. the BusyBox mount command does not support NFSv4.

Client support

Client kernel support must be enabled on each system connecting to the host running the NFS exports.

KERNEL Enabling NFS client support
File systems --->
  [*] Inotify support for userspace
  [*] Network File Systems --->
        <*>   NFS client support
        < >     NFS client support for NFS version 2
        <*>     NFS client support for NFS version 3
        [ ]       NFS client support for the NFSv3 ACL protocol extension (NEW)
        <*>     NFS client support for NFS version 4
        [ ]     Provide swap over NFS support
        [ ]   NFS client support for NFSv4.1
        [ ]   Use the legacy NFS DNS resolver
        [ ]   NFS: Disable NFS UDP protocol support

Server support

Server kernel support is only necessary on the system hosting the NFS exports. For local testing purposes, it can be helpful to also enable client support as defined in the previous section on the server as well.

KERNEL Enabling NFS server support
File systems --->
  [*] Inotify support for userspace
  [*] Network File Systems --->
        <*>   NFS server support
        -*-     NFS server support for NFS version 3
        [ ]       NFS server support for the NFSv3 ACL protocol extension (NEW)
        [*]     NFS server support for NFS version 4
        [ ]   NFSv4.1 server support for pNFS block layouts (NEW)
        [ ]   NFSv4.1 server support for pNFS SCSI layouts (NEW)
        [ ]   NFSv4.1 server support for pNFS Flex File layouts (NEW)
        [ ]   Provide Security Label support for NFSv4 server (NEW)

USE flags

USE flags for net-fs/nfs-utils NFS client and server daemons

caps Use Linux capabilities library to control privilege
junction Enable NFS junction support in nfsref
kerberos Add kerberos support
ldap Add ldap support
libmount Link mount.nfs with libmount
nfsv3 Enable support for NFSv2 and NFSv3
nfsv4 Enable support for NFSv4 (includes NFSv4.1 and NFSv4.2)
sasl Add support for the Simple Authentication and Security Layer
selinux !!internal use only!! Security Enhanced Linux support, this must be set by the selinux profile or breakage will occur
tcpd Add support for TCP wrappers
uuid Support UUID lookups in rpc.mountd


Install net-fs/nfs-utils:

root #emerge --ask net-fs/nfs-utils



The following table describes the filesystems that will be exported by the server:

Device Mount directory Description
/dev/sdb1 /home Filesystem containing user home directories.
/dev/sdc1 /data Filesystem containing user data.

Virtual root

While this article demonstrates a best-practice NFSv4 deployment using a virtual root, it is possible to directly export the required directories without using one. If that is desired this section can be skipped and the exports file populated as follows, instead:
FILE /etc/exports

The filesystems to be exported can be made available under a single directory. This directory is known as the virtual root directory:

root #mkdir /export
The /export directory is used throughout this article as the virtual root directory, although any directory can be used e.g. /nfs or /srv/nfs

Create directories in the virtual root directory for the filesystems (e.g. /home and /data) that are to be exported:

root #mkdir /export/home
root #mkdir /export/data

The filesystems to be exported need to be made available under their respective directories in the virtual root directory. This is accomplished with the --bind option of the mount command (if you need also mount something that is mounted inside, use --rbind instead:

root #mount --bind /home /export/home
root #mount --bind /data /export/data

To make the above mounts persistent, add the following to /etc/fstab:

FILE /etc/fstab
/home    /export/home    none    bind    0    0
/data    /export/data    none    bind    0    0


The filesystems to be made accessible for clients are specified in /etc/exports. This file consists of the directories to be exported, the clients allowed to access those directories, and a list options for each client. Refer to man exports for more information about the NFS export configuration options.

The following table briefly describes the server options used in the configuration below:

Option Description
insecure The server will require that client requests originate on unprivileged ports (those above 1024). This option is required when mounting exported directories from OS X or by the nfs:/ kioslave in KDE. The default is to use privileged ports.
rw The client will have read and write access to the exported directory. The default is to allow read-only access.
sync The server must wait until filesystem changes are committed to storage before responding to further client requests. This is the default.
no_subtree_check The server will not verify that a file requested by a client is in the appropriate filesystem and exported tree. This is the default.
crossmnt The server will reveal filesystems that are mounted under the virtual root directory that would otherwise be hidden when a client mounts the virtual root directory.
fsid=0 This option is required to uniquely identify the virtual root directory.

If changes are made to /etc/exports after the NFS server has started, issue the following command to propagate the changes to clients:

root #exportfs -rv

Configuration grants access to the exported local shares, access is granted to the clients in the192.0.2.0/24 IP network. Client access can also be specified as a single host (IP address or fully qualified domain name), NIS netgroup, or with a single * character which grants all clients access.

FILE /etc/exports

IPv6 only configuration. Allowed IPv6 prefixes are put after the already configured IPv4 networks. The above configuration grants access to the exported directories by IP network, in this case 2001:db8:1::/64. These IP networks are allowed to access the exported shares on the NFS server:

FILE /etc/exports
/export         2001:db8:1::/64(insecure,rw,sync,no_subtree_check,crossmnt,fsid=0)
/export/home    2001:db8:1::/64(insecure,rw,sync,no_subtree_check)
/export/data    2001:db8:1::/64(insecure,rw,sync,no_subtree_check)
Dual stack configuration

IPv4 and IPv6 networks which are allowed to access the exported shares on the NFS server, here and 2001:db8:1::/64.

FILE /etc/exports
/export,rw,sync,no_subtree_check,crossmnt,fsid=0) 2001:db8:1::/64(insecure,rw,sync,no_subtree_check,crossmnt,fsid=0)
/export/home,rw,sync,no_subtree_check) 2001:db8:1::/64(insecure,rw,sync,no_subtree_check)
/export/data,rw,sync,no_subtree_check) 2001:db8:1::/64(insecure,rw,sync,no_subtree_check)



The NFS daemon on OpenRC is configured via the OPTS_RPC_NFSD variable:

FILE /etc/conf.d/nfs
OPTS_RPC_NFSD="8 -V 3 -V 4 -V 4.1"

The NFS daemon on systemd is configured via the RPCNFSDARGS variable:

FILE /etc/conf.d/nfs
RPCNFSDARGS="8 -V 3 -V 4 -V 4.1"

The option 8 is the number of NFS server threads to start. Since only one thread is started by default, the thread count should be increased for optimal performance. The options -V 3, -V 4 and -V 4.1 enable NFS versions 3, 4, and 4.1. Refer to man nfsd for more information about the NFS daemon configuration options. Technical differences between major NFS versions explained in the wikipedia article.



To start the NFS server:

root #rc-service nfs start
 * Starting rpcbind ...                                                   [ ok ]
 * Starting NFS statd ...                                                 [ ok ]
 * Starting idmapd ...                                                    [ ok ]
 * Exporting NFS directories ...                                          [ ok ]
 * Starting NFS mountd ...                                                [ ok ]
 * Starting NFS daemon ...                                                [ ok ]
 * Starting NFS smnotify ...                                              [ ok ]

The above output shows that many other services are also started along with the nfs service. To stop all NFS services, stop the rpcbind service:

root #rc-service rpcbind stop

To start the NFS server at boot:

root #rc-update add nfs default

To start the NFS server:

root #systemctl start rpcbind nfs-server

To start the NFS server at boot:

root #systemctl enable rpcbind nfs-server




To be able to mount exported directories, start the NFS client:

root #rc-service nfsclient start
 * Starting rpcbind                                                       [ ok ]
 * Starting NFS statd                                                     [ ok ]
 * Starting NFS sm-notify                                                 [ ok ]

To start the NFS client at boot:

root #rc-update add nfsclient default


The nfs-client service will be started automatically when systemd detects that exported directories are being mounted.

Mounting exports

The commands and configuration files below use the IPv4 address and IPv6 address2001:db8:1::1 to represent the NFS server.

Mount the exported directories:

root #mount /home
root #mount /data

To make the above mounts persistent, add the following to /etc/fstab:

FILE /etc/fstab    /home    nfs    rw,_netdev    0    0    /data    nfs    rw,_netdev    0    0

root #mount -t nfs4 -o _netdev,rsize=1048576,wsize=1048576,vers=4

The virtual root directory can be mounted instead of each individual exported directory. This will make all exported directories available to the client:

root #mount /mnt

To make the above mount persistent, add the following to /etc/fstab:

FILE /etc/fstab        /mnt     nfs    rw,_netdev    0    0

When using /etc/fstab to mount the exported directories, add the netmount service to the default runlevel:

root #rc-update add netmount default
It will probably be necessary to specify the network management dependencies in /etc/conf.d/netmount.

If the NFS server or client support NFSv3 only, the full path to the exported directory (e.g. /export/home or /export/data) needs to be specified when mounting:

root #mount /home
root #mount /data

The same applies when mounting the virtual root directory:

root #mount /mnt

When mounting exported directories on an IPv6 network, enclose the IPv6 NFS server address in square brackets:

root #mount [2001:db8:1::1]:/home /home
root #mount [2001:db8:1::1]:/data /data

When mounting a link-local IPv6 address, the outgoing local network interface must also be specified:

root #mount [fe80::215:c5ff:fb3e:e2b1%eth0]:/home /home
root #mount [fe80::215:c5ff:fb3e:e2b1%eth0]:/data /data

With NFSv4, the virtual root directory can be rather 'invisible' depending on server configuration; you may need to use relative path:

root #mount -t nfs4 /home
root #mount -t nfs4 /data
I/O on large files over NFSv4 can be *strongly* improved by the following, which increases the maximum read and write size to 1024^2 bytes, or 1MB.
root #mount /home -o rsize=1048576,wsize=1048576,vers=4

For persistence:

FILE /etc/fstab   /data nfs4 _netdev,rw,rsize=1048576,wsize=1048576,vers=4


It is possible to identify NFS client using Kerberos GSS. This will require a few modifications. In the following instruction, it is supposed that Kerberos is already installed on the same server as NFS (which hostname is server.domain.tld) and that the client (client.domain.tld) is able to kinit to it. The Kerberos default realm it DOMAIN_REALM.TLD.

First, enable the following kernel option (CONFIG_RPCSEC_GSS_KRB5) for both server and client. Note that this option may not appear if all cryptographic dependencies are not selected. See kernel option dependencies for more information:

KERNEL Enabling Kerberos for RPC
File systems --->
  [*] Network File Systems --->
        <*>   Secure RPC: Kerberos V mechanism

Then, create principals for the NFS service for both the server and the client. On the server, execute:

root #kadmin.local add_principal -randkey nfs/server.domain.tld
root #kadmin.local add_principal -randkey nfs/client.domain.tld

Each computer must have its password saved in a local keytab. The easiest way to do it is (on the server):

root #kadmin.local ktadd nfs/server.domain.tld
root #kadmin.local ktadd -k /root/krb5.keytab nfs/client.domain.tld

and then transfer the /root/krb5.keytab to the client, with the name /etc/krb5.keytab. Note that the file should be owned by root with 0600 mode.

The service rpc.gssd must run at client side. The following line must appear in /etc/conf.d/nfsclient of the client:

FILE /etc/conf.d/nfsclient
rc_need="!rpc.statd rpc.gssd"

The services rpc.idmapd and rpc.svcgssd must run at server side. The following line must appear in /etc/conf.d/nfs of the server:

FILE /etc/conf.d/nfs
NFS_NEEDED_SERVICES="rpc.idmapd rpc.svcgssd"

The rpc.idmapd service must be correctly configured (on the server):

FILE /etc/idmapd.conf
Domain = domain.tld

Add sec=krb5 to the export options.

FILE /etc/exports

It is also possible to increase security with sec=krb5i (user authentication and integrity checking) or even sec=krb5p (user authentication, integrity checking and NFS traffic encryption). The more security, the more resources are needed.

The same option must be added to the mount command at client side.


  • Verify that the NFS server is running and listening for connections:
root #ss -tulpn | grep rpc
udp   UNCONN 0      0  *    users:(("rpcbind",pid=4020,fd=6))
udp   UNCONN 0      0  *    users:(("rpc.mountd",pid=4149,fd=4))
udp   UNCONN 0      0  *    users:(("rpc.mountd",pid=4149,fd=12))
udp   UNCONN 0      0*    users:(("rpc.statd",pid=4050,fd=5))
udp   UNCONN 0      0  *    users:(("rpc.statd",pid=4050,fd=8))
udp   UNCONN 0      0  *    users:(("rpc.mountd",pid=4149,fd=8))
udp   UNCONN 0      0                  *:111              *:*    users:(("rpcbind",pid=4020,fd=8))
udp   UNCONN 0      0                  *:49463            *:*    users:(("rpc.mountd",pid=4149,fd=14))
udp   UNCONN 0      0                  *:43316            *:*    users:(("rpc.mountd",pid=4149,fd=10))
udp   UNCONN 0      0                  *:44048            *:*    users:(("rpc.mountd",pid=4149,fd=6))
udp   UNCONN 0      0                  *:44332            *:*    users:(("rpc.statd",pid=4050,fd=10))
tcp   LISTEN 0      0  *    users:(("rpc.mountd",pid=4149,fd=5))
tcp   LISTEN 0      0  *    users:(("rpc.mountd",pid=4149,fd=9))
tcp   LISTEN 0      0  *    users:(("rpcbind",pid=4020,fd=7))
tcp   LISTEN 0      0  *    users:(("rpc.mountd",pid=4149,fd=13))
tcp   LISTEN 0      0  *    users:(("rpc.statd",pid=4050,fd=9))
tcp   LISTEN 0      0                  *:52293            *:*    users:(("rpc.mountd",pid=4149,fd=7))
tcp   LISTEN 0      0                  *:43983            *:*    users:(("rpc.mountd",pid=4149,fd=15))
tcp   LISTEN 0      0                  *:111              *:*    users:(("rpcbind",pid=4020,fd=9))
tcp   LISTEN 0      0                  *:40105            *:*    users:(("rpc.statd",pid=4050,fd=11))
tcp   LISTEN 0      0                  *:38481            *:*    users:(("rpc.mountd",pid=4149,fd=11))
  • Verify which NFS daemons are running:
root #rpcinfo -p
   program vers proto   port  service
    100000    4   tcp    111  portmapper
    100000    3   tcp    111  portmapper
    100000    2   tcp    111  portmapper
    100000    4   udp    111  portmapper
    100000    3   udp    111  portmapper
    100000    2   udp    111  portmapper
    100024    1   udp  57655  status
    100024    1   tcp  34950  status
    100003    2   tcp   2049  nfs
    100003    3   tcp   2049  nfs
    100003    4   tcp   2049  nfs
    100003    2   udp   2049  nfs
    100003    3   udp   2049  nfs
    100003    4   udp   2049  nfs
    100021    1   udp  44208  nlockmgr
    100021    3   udp  44208  nlockmgr
    100021    4   udp  44208  nlockmgr
    100021    1   tcp  44043  nlockmgr
    100021    3   tcp  44043  nlockmgr
    100021    4   tcp  44043  nlockmgr
  • List the exported directories from the NFS server:
root #exportfs -v
  • List the current open connections to the NFS server:
user $ss -tun|grep -E 'Sta|2049'
Netid State Recv-Q Send-Q       Local Address:Port     Peer Address:Port   Process
tcp   ESTAB 0      0      
  • Verify that the exported directories are mounted by the NFS client:
user $ss -tun|grep -E 'Sta|2049'
Netid State Recv-Q Send-Q       Local Address:Port     Peer Address:Port   Process
tcp   ESTAB 0      0            

Unresponsiveness of the system

The system may become unresponsive during shutdown when the NFS client attempts to unmount exported directories after udev has stopped. To prevent this a local.d script can be used to forcibly unmount the exported directories during shutdown.

Create the file nfs.stop:

FILE /etc/local.d/nfs.stop
/bin/umount -a -f -t nfs,nfs4

Set the according file bits:

root #chmod a+x /etc/local.d/nfs.stop

See also

  • Samba — a re-implementation of the SMB/CIFS networking protocol, a Microsoft Windows alternative to Network File System (NFS).

External resources