Linux Network Administrator's Guide, Second Edition

Serial Devices and the login: Prompt

It was once very common that a Unix installation involved one server machine and many “dumb” character mode terminals or dial-up modems. Today that sort of installation is less common, which is good news for many people interested in operating this way, because the “dumb” terminals are now very cheap to acquire. Dial-up modem configurations are no less common, but these days they would probably be used to support a SLIP or PPP login (discussed in Chapter 7 and Chapter 8) than to be used for a simple login. Nevertheless, each of these configurations can make use of a simple program called a getty program.

The term getty is probably a contraction of “get tty.” A getty program opens a serial device, configures it appropriately, optionally configures a modem, and waits for a connection to be made. An active connection on a serial device is usually indicated by the Data Carrier Detect (DCD) pin on the serial device being raised. When a connection is detected, the getty program issues a login: prompt, and then invokes the login program to handle the actual system login. Each of the virtual terminals (e.g., /dev/tty1) in Linux has a getty running against it.

There are a number of different getty implementations, each designed to suit some configurations better than others. The getty that we’ll describe here is called mgetty. It is quite popular because it has all sorts of features that make it especially modem-friendly, including support for automatic fax programs and voice modems. We’ll concentrate on configuring mgetty to answer conventional data calls and leave the rest for you to explore at your convenience.

Configuring the mgetty Daemon

The mgetty daemon is available in source form from ftp://alpha.greenie.net/pub/mgetty/source/,and is available in just about all Linux distributions in prepackaged form. The mgetty daemon differs from most other getty implementations in that it has been designed specifically for Hayes-compatible modems. It still supports direct terminal connections, but is best suited for dialup applications. Rather than using the DCD line to detect an incoming call, it listens for the RING message generated by modern modems when they detect an incoming call and are not configured for auto-answer.

The main executable program is called /usr/sbin/mgetty, and its main configuration file is called /etc/mgetty/mgetty.config. There are a number of other binary programs and configuration files that cover other mgetty features.

For most installations, configuration is a matter of editing the /etc/mgetty/ mgetty.config file and adding appropriate entries to the /etc/inittab file to execute mgetty automatically.

Example 4.6 shows a very simple mgetty configuration file. This example configures two serial devices. The first, /dev/ttyS0, supports a Hayes-compatible modem at 38,400 bps. The second, /dev/ttyS0, supports a directly connected VT100 terminal at 19,200 bps.

Example 4-6. Sample /etc/mgetty/mgetty.config File

#
# mgetty configuration file
#
# this is a sample configuration file, see mgetty.info for details
#
# comment lines start with a "#", empty lines are ignored
#
# ----- global section -----
#
# In this section, you put the global defaults, per-port stuff is below
#
# access the modem(s) with 38400 bps
speed 38400
#
# set the global debug level to "4" (default from policy.h)
debug 4
#

# ----- port specific section -----
# 
# Here you can put things that are valid only for one line, not the others
#
#
# Hayes modem connected to ttyS0: don't do fax, less logging
#
port ttyS0
  debug 3
  data-only y
#
# direct connection of a VT100 terminal which doesn't like DTR drops
#
port ttyS1
  direct y
  speed 19200
  toggle-dtr n
#

The configuration file supports global and port-specific options. In our example we used a global option to set the speed to 38,400 bps. This value is inherited by the ttyS0 port. Ports we apply mgetty to use this speed setting unless it is overwritten by a port-specific speed setting, as we have done in the ttyS1 configuration.

The debug keyword controls the verbosity of mgetty logging. The data-only keyword in the ttyS0 configuration causes mgetty to ignore any modem fax features, to operate just as a data modem. The direct keyword in the ttyS1 configuration instructs mgetty not to attempt any modem initialization on the port. Finally, the toggle-dtr keyword instructs mgetty not to attempt to hang up the line by dropping the DTR (Data Terminal Ready) pin on the serial interface; some terminals don’t like this to happen.

You can also choose to leave the mgetty.config file empty and use command-line arguments to specify most of the same parameters. The documentation accompanying the application includes a complete description of the mgetty configuration file parameters and command-line arguments. See the following example.

We need to add two entries to the /etc/inittab file to activate this configuration. The inittab file is the configuration file of the Unix System V init command. The init command is responsible for system initialization; it provides a means of automatically executing programs at boot time and re-executing them when they terminate. This is ideal for the goals of running a getty program.

T0:23:respawn:/sbin/mgetty ttyS0
T1:23:respawn:/sbin/mgetty ttyS1

Each line of the /etc/inittab file contains four fields, separated by colons. The first field is an identifier that uniquely labels an entry in the file; traditionally it is two characters, but modern versions allow four. The second field is the list of run levels at which this entry should be active. A run level is a means of providing alternate machine configurations and is implemented using trees of startup scripts stored in directories called /etc/rc1.d, /etc/rc2.d, etc. This feature is typically implemented very simply, and you should model your entries on others in the file or refer to your system documentation for more information. The third field describes when to take action. For the purposes of running a getty program, this field should be set to respawn, meaning that the command should be re-executed automatically when it dies. There are several other options, as well, but they are not useful for our purposes here. The fourth field is the actual command to execute; this is where we specify the mgetty command and any arguments we wish to pass it. In our simple example we’re starting and restarting mgetty whenever the system is operating at either of run levels two or three, and are supplying as an argument just the name of the device we wish it to use. The mgetty command assumes the /dev/, so we don’t need to supply it.

This chapter was a quick introduction to mgetty and how to offer login prompts to serial devices. You can find more extensive information in the Serial-HOWTO.

After you’ve edited the configuration files, you need to reload init to make the changes take effect. Simply send a hangup signal to the init process; it always has a process ID of one, so you can use the following command safely:

# kill -HUP 1