conserver/INSTALL
2020-05-25 14:53:00 -07:00

278 lines
12 KiB
Plaintext

INSTALL
=======
Upgrading?
Whenever you upgrade I suggest you upgrade both the client and
server. Most times, however, you can get away without upgrading
the client (it's usually a fairly static piece of code). I'll
document any dependencies here, but check the CHANGES file for any
new features added to the client if you're considering *not*
upgrading.
Version 8.1.3
- The '^Ec;' sequence won't work correctly with 8.1.2 (where it
was introduced).
Version 8.1.2
- The 'devicesubst' and 'execsubst' formats have changed from
8.1.1. It's fairly simple to update your config file to the
new format...just check the conserver.cf manpage. Sorry for
having to change things, but it's for a good reason (I should
have though ahead when designing the original format).
Version 8.1.0
- The client/server protocol has changed to better protect 8-bit
data and to allow programs invoked with '^Ec|' not have to
worry about accidentally sending the escape sequence to the
server. Though it will look like things are mostly
backward-compatible, don't count on it and just upgrade.
Version 8.0.2
- I've added a '^Ec;' sequence to allow the client to signal the
server as to when it's ready to see console data. Without
this, verbose consoles will prevent clients from attaching
(the client sees unexpected data). An 8.0.2 client should be
compatible with an 8.0.1 server, but an 8.0.1 client is not
compatible with an 8.0.2 server.
Version 8.0.1
- There's a slight client/server protocol change to implement
the new 'initcmd' console option. If you use this
functionality with an 8.0.0 client, you'll run into a
compatibility problem while the 'initcmd' command is running.
Version 8.0.0
- The client/server protocol has been rearchitected. You *MUST*
use an 8.0.0 client with an 8.0.0 server. No combination of
client/server will work with pre-8.0.0 code.
- Upgrading from pre-8.0.0 code to 8.0.0 and beyond requires
you to change your conserver.cf and conserver.passwd files
because both of the file formats have changed.
The conserver.cf file changes are so major that there is a
convert program available in the conserver subdirectory. Just
run './conserver/convert <old-cf-file>' and it will attempt a
conversion to the new format, sending it to stdout. Any errors
will be printed to stderr. There are a couple of things
you might need to adjust. First are the user access lists.
If you are restricting users to certain consoles in your old
conserver.passwd file, you'll need to move those restrictions
into the new conserver.cf file. Restrictions are set with the
'ro' and 'rw' tags in the configuration file. Second are the
'access' blocks. What get produced by the convert program
will be functionally equivalent to the old behavior, but you
may be able to tune things to better suit your environment.
The conserver.passwd file's console restrictions have moved,
as described above. So to convert the conserver.passwd file,
all you really need to do is something like:
awk -F: '{print $1 ":" $2}' <old-passwd-file>
If you have comments or continuation lines in your file,
you'll have to do a bit more cleanup to strip out the third
field (which is what the awk command is intending to do).
- Conserver no longer trusts reverse DNS information by default.
If you use the --with-trustrevdns configure flag, you can
re-enable the use of gethostbyaddr() [I don't recommended it,
however]. If you are using domain names in access lists,
you'll either need to change those to use hostnames and/or ip
addresses/ranges or use the --with-trustrevdns flag. For
example, if you have (in the
8.0.0 format):
allowed conserver.com; # allow *.conserver.com
then you'll need to worry about this change. If you only use
full hostnames, you shouldn't have to do anything.
Version 7.2.4
- If SSL support is compiled into the code, older versions of
the client and server are, by default, incompatible because
encrypted connections are a requirement. Use of the -E flag in
the client and/or server can work around this (but I discourage
this - please upgrade the clients and servers instead).
Version 7.2.0
- The code related to broadcast messages in the client (-b) has
changed. If you want the username to come across properly in
the broadcast message, you'll need to make sure you upgrade
to the 7.2.0 client.
Version 7.1.1
- Both conserver.passwd and conserver.cf file parsing behaves
the same now. Both use leading whitespace as a continuation
line indicator - if you have leading whitespace on a line
(aside from comments) you probably should remove it.
Version 7.1.0
- The client/server protocol has changed. You *MUST* use a
7.1.0 client with a 7.1.0 and above server. A 7.1.0 client
is *not* backward compatible with a pre-7.1.0 server.
- Some of the flags in the client (-d, -D, and -r) and server
(-n) have been given new identities to make the client and
server flags more uniform.
- The conserver.passwd file now uses the first username match
to determine access rights - if you have multiple instances
of a username in an existing password file, they must be
combined into one to continue to work.
Quickie Instructions
- Download conserver (http://www.conserver.com/) and unpack
- Run './configure'
- Run 'make'
- Run 'make test'
- If all is well, run 'make install'
- Now set up config files, etc. (see below)
Detailed Instructions
- First thing to do is determine if you want different defaults. A
'./configure --help' will show you the basics. If you like all
the defaults shown, you're set. If not, here are the conserver
unique options:
--with-port=PORT Specify port number [conserver]
--with-base=PORT Base port for secondary channel [0]
--with-master=MASTER Specify master server hostname [console]
--with-ccffile=CFFILE Specify client config filename
[SYSCONFDIR/console.cf]
--with-cffile=CFFILE Specify config filename [SYSCONFDIR/conserver.cf]
--with-pwdfile=PWDFILE Specify password filename
[SYSCONFDIR/conserver.passwd]
--with-logfile=LOGFILE Specify log filename [/var/log/conserver]
--with-pidfile=PIDFILE Specify PID filepath [/var/run/conserver.pid]
--with-maxmemb=MAXMEMB Specify maximum consoles per process [16]
--with-timeout=TIMEOUT Specify connect() timeout in seconds [10]
--with-trustrevdns Trust reverse DNS information
--with-extmsgs Produce extended messages
--with-rpath Use -R as well as -L for libraries
--with-cycladests (deprecated - noop) Build for a Cyclades TS
--with-uds[=DIR] Use Unix domain sockets for client/server
communication [/tmp/conserver]
--with-trust-uds-cred Trust UDS credentials obtained via socket
--with-libwrap[=PATH] Compile in libwrap (tcp_wrappers) support
--with-openssl[=PATH] Compile in OpenSSL support
--with-req-server-cert Require server SSL certificate by client
--with-gssapi[=PATH] Compile in GSS-API support
--with-striprealm retry username without @REALM with gss-api
authentication
--with-freeipmi[=PATH] Compile in FreeIPMI support
--with-dmalloc[=PATH] Compile in dmalloc support
--with-pam Enable PAM support
--with-ipv6 (experimental) Use IPv6 for client/server
communication
Not surprisingly, some match the old conserver/cons.h items...here
they are for reference:
PORT or SERVICE - Socket used to communicate
HOST - Hostname of console server
CONFIG - Config file path
PASSWD_FILE - Password file path
MAXMEMB - Number of consoles per child process
A couple of notes. First, --with-libwrap will add tcp_wrappers
lookups to all socket connections in the server. --with-openssl
will add encryption between the client and server when you connect
to a console. --with-uds will cause the client and server to use
unix domain sockets for their communication, eliminating the
tcp communication they normally do (which means --with-master and
--with-port are not used). --with-dmalloc should only be used to
do memory allocation debugging and not used in production.
- Run './configure'. This will detect system specific
information. The --prefix option will redirect where things are
installed. Other options are available as well...try
'./configure --help'.
- Now run 'make'. Hopefully things will compile.
- To test your binaries, run 'make test'. If there are problems, it
should mean something is wrong, but check the output differences
to make sure it wasn't a temporary failure. I tried to make the
tests generic, but I may have missed something.
- Once things build, you can run 'make install'.
- If you'd like to build the autologin application, you'll need to
run 'make autologin'. If you'd like it installed, use 'make
autologin.install'.
- Now that the binaries are in place, we need to set up the
configuration files and such.
+ Does your conserver master hostname exist? This is the
hostname specified with the --with-master option. By default
the hostname is "console", so make sure it's in DNS, hosts
files, or whatever.
+ If you used a symbolic name for the --with-port option (by
default it uses "conserver", so the answer would be yes),
you'll need to enter a definition in your services file
(directly, via NIS, or whatever). Here's what we use:
console 782/tcp conserver # console server
If you used a number, you shouldn't have to worry about this
step.
+ Next, make sure conserver runs during boot. The init script we
use under Solaris is installed in
<DATADIR>/examples/conserver/conserver.rc. Use that or some
form of it for your own /etc/init.d script or an entry in
startup files (/etc/rc, /etc/rc.local, or whatever).
+ Now for the fun stuff. You need to create a conserver.cf and
conserver.passwd file. Those are defined with the
--with-cffile and --with-pwdfile settings. If you ever need to
know what values were compiled into conserver, run 'conserver
-V'. See the conserver.cf/INSTALL file for instructions on
setup of these files.
- That's it! Just start up the console server and enjoy!
Other Information And Gotchas
- Potential GCC bug
Adam Morris <AMorris@providence.org> reported a problem with
the following line in console/console.c:
if ((in_addr_t) (-1) == pPort->sin_addr.s_addr) {
This tickles a GCC bug under HP-UX 11.11 using GCC 3.0.2 in
64-bit mode with optimization enabled (-O). The bug could
possibly be provoked in other combinations as well. His fix is
to change the line to:
if ((in_addr_t) (-1) == inet_addr(pcToHost)) {
It's also reported that newer versions of the compiler fix the
issue, so if you happen to have problems with the client
connecting to servers, you might be tickling this bug and you
can upgrade the compiler, turn off the optimization, or apply
this code change.