Brad Huntting and David Mertz
Welcome to "File Sharing Servers", the fifth of eight tutorials designed to prepare you for LPI exam 201. In this tutorial you will learn how to use a Linux system as a networked file server, using any of several protocols supported by Linux
The Linux Professional Institute (LPI) certifies Linux system administrators at junior and intermediate levels. There are two exams at each certification level. This series of eight tutorials helps you prepare for the first of the two LPI intermediate level system administrator exams--LPI exam 201. A companion series of tutorials is available for the other intermediate level exam--LPI exam 202. Both exam 201 and exam 202 are required for intermediate level certification. Intermediate level certification is also known as certification level 2.
Each exam covers several or topics and each topic has a weight. The weight indicate the relative importance of each topic. Very roughly, expect more questions on the exam for topics with higher weight. The topics and their weights for LPI exam 201 are:
Topic 201: Linux Kernel (5) Topic 202: System Startup (5) Topic 203: Filesystems (10) Topic 204: Hardware (8) Topic 209: File Sharing Servers (8) Topic 211: System Maintenance (4) Topic 213: System Customization and Automation (3) Topic 214: Troubleshooting (6)
Welcome to "File Sharing Servers", the fifth of eight tutorials designed to prepare you for LPI exam 201. In this tutorial you will learn how to use a Linux system as a networked file server, using any of several protocols supported by Linux.
The current LPI guidelines for the specific Topic 209 exam cover NFS and Samba. A system administrator designing a server configuration, however, should also keep in mind whether FTP, SCP/SSH, HTTP, or other protocols might, in fact, fit their specific requirements.
To get the most from this tutorial, you should already have a basic knowledge of Linux and a working Linux system on which you can practice the commands covered in this tutorial.
One of the most significant uses for Linux, particularly in a server context, is providing shared files to client systems. In fact, in a general way, serving files is probably most of what all networking is used for. This tutorial--and in fact, this series of tutorials--will not address peer-to-peer file sharing servers such as BitTorrent. Rather, we will look only at older client-server arrangements: a central server that provides disk stores for multiple clients. Even when clients upload files, those are always stored and served by the server, rather than in a decentralized fashion.
Protocols widely used for file serving include HTTP (the WWW), TFTP (trivial file transfer protocol), FTP (file transfer protocol), SCP (secure copy; a specialized use of SSH), RCP (remote copy; generally deprecated), NFS (network file system), and Samba (server message block). HTTP and SSH will be discussed in the tutorials for LPI exam 202, as will security issues around FTP. TFTP and RCP are special purpose or deprecated, and will not be addressed in these tutorials.
This tutorial looks at NFS and Samba in some detail, and briefly describes FTP. NFS and Samba are network file sharing protocols that allow mostly transparent access to remote filesystems. FTP might require a custom FTP client program, although many desktop environments or tools (on Linux or otherwise) hide the details of this negotiation, and effectively present the same user interface as an NFS or Samba mounted drive.
If the server is properly configured, and the client has appropriate
permissions, mounting a remote filesystem with NFS requires only the
mount -t nfs my.nfs.server.com:/path/on/server /path/on/client
or a suitable entry in
my.nfs.server.com:/path/on/server /path/on/client nfs rw,soft 0 0
soft option tells the kernel to send an IO error (EIO) to user
processes in the event of network difficulties. The default
option will cause processes to hang while the nfs server is
In addition, the helper programs
rpc.quotad may be run on client and/or server.
An NFS server requires 3 distinct programs, as well as 3 optional programs.
When an NFS client mounts an NFS file system, it contacts the
following server daemons, most of which must run stand alone (as
opposed to being started from
portmap (sometimes named
portmapper or rpc.bind),
In addition, there are three optional helper programs
rpc.quotad which respectively provide global
locking, accelerate the
lstat family of syscalls (used by
etc), and provide support for quotas.
All three NFS related servers use "TCPwrappers" (i.e.
access control, and hence may require entries in
portmap normally require any configuration beyond
The configuration file for
mountd is (indirectly)
says which filesystems can be mounted by which clients. Under the
Linux implementation of NFS,
/etc/exports is not directly parsed by
mountd. Instead, the
exportfs -a command parses
writes the result to
mountd can read it.
There are other flags to
exportfs which allow these two files to be
desynchronized. That is, you may temporarily add or remove exported
directories without modifying the semi-permanent records in
Administrators of other Unix-like servers should note that the syntax
of the Linux
/etc/exports file differs significantly from that of
SunOS or BSD.
The configuration file
/etc/hosts.allow describes hosts that are
allowed to connect to a Linux system. This configuration is not
specific to NFS, but a system needs to be permitted to connect in the
first place to use and NFS server. Similarly,
/etc/hosts.deny is a
list of hosts prohibited from connecting.
Slightly unintuitively, first allowed hosts are searched, then denied hosts, but anything left unmatched is granted access. This does not mean that the login mechanisms of individual servers are not still operative, but a cautious administrator might deny anything not explicitly permitted (a little paranoia is good) by using:
# /etc/hosts.deny ALL:ALL EXCEPT localhost:DENY
/etc/hosts.deny set to deny everything (except connections
from LOCALHOST), only those connections explicitly permitted will be
allowed. For example:
#/etc/hosts.allow # Allow localhost and intra-net domain to use all servers ALL : 127.0.0.1, 192.168. # Let everyone ssh here except 216.73.92.* and .microsoft.com sshd: ALL EXCEPT 216.73.92. .microsoft.com : ALLOW # Let users in the *.example.net domain ftp in ftpd: .example.net
Here's a sample /etc/export file:
# sample /etc/exports file / master(rw) trusty(rw,no_root_squash) /projects proj*.local.domain(rw) /usr *.local.domain(ro) @trusted(rw) /home/joe pc001(rw,all_squash,anonuid=150,anongid=100) /pub (ro,insecure,all_squash)
root (uid 0) on the client is treated as
65534) on the server; this is called
root squashing as it protects
files owned by root (and not group/other writable) from being altered
by NFS clients. The
no_root_squash tag disables this behavior, and
allows the root user on
trusty full access to the
This can be useful for installing and configuring software.
/usr partition will be read only for all hosts except those in
the "trusted" netgroup.
/home/joe is mounted by
pc001, all remote users (regardless
of uid/gid) will be treated as if they have uid=150, gid=100. This is
useful if the remote NFS client is a single user workstation or does
not support different users (e.g. DOS).
Normally, Linux (and other Unix-like operating systems) reserves the
use of TCP and UDP ports 1-1023 (so called
secure ports) for use by
processes running as root. To ensure that the root user has initiated
a remote NFS mount, the NFS server normally requires remote clients to
use "secure ports" when mounting NFS filesystems. This convention,
however, is not honored by some operating systems (notably Windows).
In such cases, the
insecure option allows the NFS client to use any
TCP/UDP port. This is usually required when serving Windows clients.
nfsstat displays a time series of NFS related statistics (client
and/or server) regarding the local machine similar to
showmount command queries
mountd and shows which clients are
currently mounting filesystems. As NFS is a stateless protocol, and
mountd daemon is queried infrequently, the output of
can become inaccurate. Unfortunately, there is not really any way to
showmount to become accurate. However, where it is
showmount almost always errs in showing stale mounts
rather than omitting active mounts (i.e. relatively harmlessly).
In this context, "stateless" means that the
nfsd daemons that serve
up the actual file data have no memory of which files are open, nor
even which clients have which partitions mounted. Each request
(readblock, writeblock, etc) contains all the information needed to
complete it (partition id provided by
mountd, inode number, block
number(s), read/write/etc, data). The HTTP protocol is similar in this
respect. An upside of statelessness if the server reboots, the clients
will notice only a brief period of interrupted access.
The Samba server
smbd provides file and print services (largely for
Windows clients). While it can be started from
inetd, it is
typically run as a stand alone daemon
nmbd is the netbios
nameserver (or WINS server). It too can be run from
inetd, but is
more typically run as a stand alone daemon
nmbd -D. Samba can
function as a server in a Windows WORKGROUP, as well as Primary Domain
The configuration file for both
/etc/samba/smb.conf. Copious configuration parameters are described
smb.conf man page. The
lmhosts file is used to map NetBios
names to IP addresses. It's format is similar to (but not identical
There are several excellent HOWTOs on the subject of Samba configuration as well as several books. This section touches on the basic ideas with pointers to more complete documentation.
smb.conf snippet allows users to access their
(local) home directories from remote Samba clients:
[homes] comment = Home Directories browseable = no
This is usually included in the default
Of the numerous Unix printing systems, CUPS is the least antiquated
and probably the currently most popular. Depending on your
distribution, CUPS may be enabled in the default
smb.conf. Here is a
simple example of a CUPS print share:
[global] load printers = yes printing = cups printcap name = cups [printers] comment = All Printers path = /var/spool/samba browseable = no public = yes guest ok = yes writable = no printable = yes printer admin = root [print$] comment = Printer Drivers path = /etc/samba/drivers browseable = yes guest ok = no read only = yes write list = root
CUPS can provide
ppd (Postscript printer description) files and
Windows drivers for clients, which, when setup properly, allow remote
users to take advantage of the full range of a printers features
(color versus black-and-white, resolution, paper tray select, double
vs single sided printing, etc). Traditional Unix printing systems are
quite cumbersome by comparison. Consult the
cupsaddsmb man page for
Samba (unlike NFS) requires individual users to authenticate with the
server. As with any network authentication service, care should be
taken to insure that passwords are never passed over the network
unencrypted. See the section on "encrypt passwords" in the
man page for details.
There are a variety of mechanisms Samba can use to authenticate remote users (clients). By their nature most of these are incompatible with the standard Unix password hash. The notable exception is when passwords are passed over the wire in the clear unencrypted, which is almost always a bad idea.
Assuming you encrypt passwords on the wire,
smbpasswd will usually
be used to setup users with an initial Samba password. The "Unix
password sync" option allows
smbpasswd to change Unix passwords
whenever users change their Samba password.
pam_smb module when configured can authenticate
Linux users using the Samba database directly. As if that's not enough
choices, LDAP can be used to authenticate Samba and/or Linux users.
When configuring a Samba server, the
testparm (also called
smbtestparm) command can be quite useful. It will parse the
smb.conf file and report any problems.
nmblookup command does for Samba what
nslookup does for DNS;
it queries the NetBios directory. See the
nmblookup man page for
smbclient command provides FTP-like access to a Samba file
share. Transparent access to SMB file shares is trickier; see the
smbmount man page or the
sharity package for more info.
FTP is an old and widely used network protocol. FTP is normally run over two separate ports, 20 and 21. Port 21 is used as a control stream (transmitting login information and commands) while port 20 is used as the data stream over which actual file content is transmitted.
Generally, FTP is not considered a very secure protocol in the sense that in its default mode of operation, control information--i.e. login passwords--are transmitted in the clear. For that matter, data streams are also unencrypted, but FTP shares that feature with NFS and Samba (for secure data channels, SSH/SCP is a better choice). It is possible to layer FTP's control port over SSH, hence protecting control information.
Traditional FTP clients provide their own shell environment over which to transmit control commands and configure connections. Sometimes GUI frontends are used to provide friendlier interfaces to FTP transfers. However, nowadays, many non-dedicated tools incorporate FTP--everything from file managers to text editors are often happy to work with files served by an FTP server.
For what FTP is most often used for, its security usually does not
matter. Probably most often, FTP servers are used for "anonymous
FTP"--that is, data that is available to the world at large, and hence
does not require security. By convention, a username of
configured to allow access, and an identifying password (often an
email address) is requested but not verified. Sometimes a
username/password is required, but such a combination is provided
without any deep user authentication (e.g. people who want to
volunteer for a project).
Most web browsers and many file managers and tools support FTP servers
transparently. Often these tools will use an FTP URL to request a
file (or also to upload a file to a server). For example, the
command line tool
wget will retreive files from FTP servers using
$ wget ftp://example.net/pub/somefile $ wget ftp://user:firstname.lastname@example.org/pub/somefile
File managers will often "mount" an FTP server in a manner that is
essentially identical to a local filesystem, or NFS or Samba drive
(this does not, however, use the
/etc/fstab system; and
such pseudo-partititions are usually named by their URL).
Given the age and ubiquity of FTP a bewildering number of implementations are available, and installed with various Linux distributions. Configuring the FTP server you decide to use will require a visit to the documentation accompanying the particular server.
Some popular Linux FTP servers include
BSD ftpd, and TUX FTP. There are many less used ones as well. In
most every case, the configuration of a server will live in a file
/etc/FOOftpd.conf (for an appropriate value of "FOO"). I am
vsftpd, which is both fast and avoids known security
glitches (the "vs" stands for "very secure").
Given the wealth of servers, configuration syntaxes will differ. But
a few concepts taken from
/etc/vsftpd.conf illustrate the types of
options other servers provide. For
vsftpd each option takes the
form "option=value", with the usual hash marks for comment lines.
Most other FTPd configuration files are similar.
anonymous_enable: Controls whether anonymous logins are permitted. anon_world_readable_only: When enabled, anonymous users will only be
allowed to download world-readable files.
* chroot_local_user: If enabled, local users will be placed in a
chroot() jail in their home directory after login.
* pasv_enable: Should the server use the "passive FTP" style in which
clients initiate ports (helps with firewalls at clients).
ssl_enable: If enabled, vsftpd will support SSL secure connections. tcp_wrappers: If enabled incoming connections will be fed through
access control (i.e.
In the simplest case, you may start an FTP server the same way you might launch any daemon, e.g.:
% sudo vsftpd
At this point the server will listen for incoming connections,
according the rules configured in its configuration file. You may
also launch an FTP server from an "network super-server" such as
xinetd. The LPI 202 tutorials will discuss these
Launching a daemon individually, even if in appropiate startup
scripts--either for a particular runlevel or in
you finer control over the behavior of an FTP server.