The file inetd.tar.gz contains source and binaries for the BSD4.3-Reno inetd
for A/UX The BSD4.3-reno inetd has several advantages over the A/UX inetd: it
supports rereading its setup files when sent SIGHUP; it has a number of small,
built-in services (echo, daytime, and others) and it can run daemons as any
user can pass arguments to them.
inetd has been further augmented to understand A/UX-format entries in
/etc/servers as well as BSD-format entries (thus, you can just use your
existing /etc/servers and update entries to BSD-format as required). It also
understands A/UX-format rpc entries, can log which daemons were spawned and,
thanks to the tcp_wrappers library, can also perform host-access control for
each daemon.
Due to the system-call fixes in A/UX 3.1, the provided 'inetd' binary will
only work under 3.1, but you can recompile it (using 'gcc') for older versions.
Files are:
Makefile :
A simple makefile to build and install inetd
README-inetd :
This file
hosts_access.5.Z :
Man page for tcp_wrapper hosts_access file format.
inetd :
Working inetd for A/UX 3.1 (must be recompiled with 'gcc'
for older versions).
inetd.1m.man :
BSD-format man page source (will not format with -man on A/UX)
inetd.1m.Z :
Formatted and compressed inetd man page
inetd.c :
Modified BSD inetd source
libwrap.a :
The tcp_wrapper libwrap archive. Required if you wish to
recompile inetd.
servers :
BSD-format conf file for A/UX (if you haven't modified your
/etc/servers, this is a drop-in replacement)
services :
Include just for the heck of it :)
The original port was done by John Coolidge (coolidge@apple.com). Since
then, Jim Jagielski (jim@jagubox.gsfc.nasa.gov) has "taken over" keeping
inetd up to date and fixing bugs. If you see any bugs or have any problems
or suggestion, please Email Jim.
VERSION HISTORY:
===Version 1.8===============================================================
1. Some minor code cleaning
2. Have '-p' option override '-a'... before, if 'inetd' was called
with '-p -a -l', the PARANOID option would be overruled by the
later "regular" access control option '-a'. This behavior "makes
sense", but we should know better. Now, '-p' has higher authority
than '-a'.
===Version 1.7===============================================================
1. Minor buglet: gethostbyname() test should have used 'while' not 'if'
2. Changed #1 to break out gethostby*() functions as 'if' loops.
3. Fixed PARANOID code
===Version 1.6===============================================================
2 minor (but #1 was important) changes.
1. Starting with v1.5, inetd was linked with '-lbsd' to get BSD-type
signals. However, this caused troubles with some ported shells (such as
'tcsh'). Now we just call 'set42sig()' to get _only_ what we want and
avoid having the child inherit compatibility stuff it shouldn't.
2. Better form of daemon()
===Version 1.5===============================================================
Not a lot of code changes (basically some modifications and code cleanup
so that compiling with '-Wall' looks cleaner) but some _major_ additional
functionality. Thanks to Wietse Venemas, author of tcp_wrapper (tcpd), who
allowed me to use some routines of tcpd, inetd 1.5 now includes host access
control. This means you can select which hosts can or cannot spawn which
daemons. Since inetd 1.5 uses the same routines, you can use the same access
control files (/etc/hosts.allow and /etc/hosts.deny) that tcpd does. In fact,
you won't even need to use tcpd anymore. Read the inetd man-page for more
info.
As expected, inetd 1.5 requires the tcp_wrappers distribution if you want to
perform a full recompile (my copy of the libwrap.a archive from tcpd 6.3 is
included).
===Version 1.4===============================================================
If /etc/servers doesn't exist, then inetd will try /etc/inetd.conf.
Changed ordering of daemon() code.
===Version 1.3===============================================================
Pretty minor little fixes. It's easier to determine the version of inetd
running. Some network related casts are made where needed. Some code
cleanup as well...
===Version 1.2===============================================================
I've fixed a few things... First of all, previous versions assumed that
the "<address>" magic cookie would always be the _only_ argument to the
daemon (and, of course, would _always_ be the 1st one). 'inetd'v1.2 no
longer assumes this.
... /usr/etc/in.telnetd in.telnetd -l <address>
will work correctly.
Also, I've changed the logging "procedure"; it's now logged at the
LOG_LOCAL2 facility (was LOG_DAEMON) and it's now semi-formatted to make
scanning the logs easier. All "daemons" are now logged and 'inetd' is
a little smarter in figuring-out the remote host's address and name.
The SIGHUP rereading-bug (which only happened occasionally) has been
fixed as well.
Finally, some semi-major code-shuffling was done. This was due to make
things a little more logical as well as to allow Apple-specific capabilities
to be handled in a similar way than non-Apple ones; for example, a failed
'bind' for an rpc daemon-socket will be retried in the same way the
normal steam socket-types were.
==Version 1.1================================================================
I've updated John's original port a bit. First of all, 'inetd' will correctly
handle tcp wrapper daemons, such as 'tcpd'. Secondly, I've added, for
future reference, the daemon() function to the program. Finally, I've
added logging capability... basically what this means is that if enabled
(by calling 'inetd' with the "-l" option), 'inetd' will log via syslogd
(LOG_DAEMON at LOG_INFO level) when a daemon is spawned and which host
made the request. This provides, internal to 'inetd' itself, some of the
functionality of 'tcpd' (although 'inetd' doesn't do access lists at all,
yet :) ).
--
#include <std/disclaimer.h>
Jim Jagielski | "That is no ordinary rabbit... 'tis the
jim@jagubox.gsfc.nasa.gov | most foul, cruel and bad-tempered
NASA/GSFC, Code 734.4 | rodent you ever set eyes on"
Greenbelt, MD 20771 | Tim the Enchanter