gtlssh-keygen [options] <command> [command options]
The gtlssh-keygen program is used for making key handling for gtlssh easier.
Generally, when you start using gtlssh on a system, you would run
and it would create keys for you. You should do the same thing on any target system you want to log into with gtlssh. Then copy the default.crt file in your $HOME/.gtlssh directory to the target's $HOME/.gtlssh/allowed_certs directory. Then run
on the target system to generate the hashes After that you should be able to log in without a password.
When you need to regenerate your keys, you run
again. It will prompt you for replacement. If you replace the keys, the old keys will be saved with a ".1" appended to the filename. Once you replace they keys, you need to push up new keys to all your target. You can run
gtlssh-keygen pushcert target1 [target2 [....]]
to update they keys on all those targets. It will use the old credentials (with the ".1" appended) to do this, so it should work easily.
Differences from SSH
Unlike ssh, ssl keys have lifetimes. By default gtlssh creates 1 year lifetimes on keys, though you can override this. Although this is a little annoying, it is a good idea to replace your keys periodically, so you could call this a good thing.
gtlssh lets you create keys for specific targets and use them automatically. Suppose, for instance, that you want to use a different key for logging into target abc.my.domain. You would do:
gtlssh-keygen keygen abc.my.domain
and it would create a key and certificate in the directory .gtlssh/keycerts with the names abc.my.domain.crt and abc.my.domain.key. You would need to copy that certificate (not the default.crt) to your remote target. gtlssh would see that those keys were there and use them automatically when you logged in to abc.my.domain. The pushcert command understands this, too, and will handle pushing the proper keys when you push to abc.my.domain.
You can also add ports to the key generation, and it will only use it if you connect to the specific target on the specific port. This could be useful for ser2net.
gtlsshd will work on Windows, and it will sort of work without a password, but certain things will not work as the logon has no stored credentials, and it is unable to create a linked token for admin logins so you can't do admin things.
To work around this issue and still allow certificate logins, you can use the storepw command of gtlssh-keygen to store your password in your .gtlssh directory. This is not ideal, but your private keys are there, anyway, so it's not a huge thing. Administrators on your system will be able to look at your password, so be warned.
- --keysize size
Create an RSA key with the given number of bits. Default is 2048. Usually 1024, 2048, or 4096.
- --keydays days
Create a key that expires in the given number of days. Default is 365.
- --basedir dir
Base directory for gtlssh. Default is $HOME/.gtlssh. Default keys go here.
- --keydir dir
Location to put the non-default generated keys. Default is $HOME/.gtlssh/keycerts for user certificates. For server certificates it is in sysconfdir (generally /etc) /gtlssh on Unix. On Windows it is the directory of the executable /../etc.
- --commonname name
Set the common name in the certificate. The default is your username for normal certificates and the fully qualified domain name (prefixed with the keyname if the keyname is given) for server certificates.
- --algorithm algname
Use the given algorithm for the key generation, one of rsa or ec. The default is rsa.
- --force | -f
Don't ask any questions, just force the operations. Be careful, this will overwrite data without asking.
Print the version number and exit.
- keygen [-p <port>] [hostname]
Setup of the base directory (if not already set up) and create keys. If nothing is specified, create the default key in <basedir>/default.key and <basedir>/default.crt. Otherwise create keys in the form <keydir>/<host>[,<port>].key and <keydir>/<host>[,<port>].crt.
Like keygen, but take no options and create the default certificate only.
- rehash [directory [directory [...]]]
The openssl library used by gtlssh requires that certificates in a library be hashed so it can find them. This command redoes the hashes in the given directories. If no directory is specified, it rehashes <basedir>/allowed_certs and <basedir>/server_certs.
Note that if you add keys to these directories, you must rehash them or they will not work.
rehash will automatically remove any certificates that have expired.
- addallow [-i] <hostname> <file>
Add the given file as an allowed public certificate for the given hostname. It will install this file in the directory in <basedir>/allowed_certs with the name "hostname.crt". It will also rehash the directory. If -i is specified, input comes from stdin and the file is not required or used. If the destination file already exists, it will rename it "hostname.crt.1.crt".
- pushcert [-n <name> ] [-p <port>] <hostname> [[-p <port>] <hostname> [...]]
Put the local certificate for the given host onto the remote host so it can be used for login. It uses old credentials (credentials with .1 appended to the name, per keygen) if they are there. This is useful if you have updated your certificate and need to send a new one to some remote hosts. It finds the certificate name as described in the keygen command. If old credentials exist, it will use those to connect with gtlssh and send the certificate. Otherwise it will use default credentials and hope for the best, probably only useful if passwords are accepcted. This only works one keygen back, if you have run the keygen command twice for the host, you will need to transfer the certificate manually. By default the credential on the remote host is named the output of 'hostname -f' on the local machine, -n overrides this.
Windows only, see "WINDOW HACKS" above for details. Prompts for your password to store in your .gtlssh directory.
Create keys for the server, generally in /etc/gtlssh/gtlsshd.key and /etc/gtlssh/gtlsshd.crt. You generally must be root to do this. Use for initial setup of gtlsshd.
If you specify a keyname, the common name used for the key will be "<keyname>.<hostname>" by default, unless you set it with --commonname. If you do not specify a keyname (defaulting it to gtlsshd), the common name is just the hostname. This avoids the following problem...
NOTE: If you run multiple servers on the same system, you must have a different common name for each one. Some version of openssl will look at another certificate with the same common name as part of the same certificate chain and you will get a failure:
authority and subject key identifier mismatch
So if, for instance, you are running ser2net and gtlsshd on the same system, they must have different common names.
Corey Minyard <firstname.lastname@example.org>
gtlssh(1), gtlsshd(8), ser2net.yaml(5).