repassh 1.2.0

SSH wrapper to load private keys on first execution

repassh

SSH wrapper to load private keys via ssh-add when they are first needed, with port knocking support.

Heavily based on ssh-ident.

Usage

Use this script to start ssh-agents and load ssh keys on demand, when they are first needed.

All you have to do is modify your .bashrc to have:

alias ssh='/path/to/repassh'

or add a link to repassh from a directory in your PATH, for example:

ln -s /path/to/repassh ~/bin/ssh

If you use scp or rsync regularly, you should add a few more lines described below.

In any case, repassh:

• will start ssh-agent and load the keys you need the first time you actually need them, once. No matter how many terminals, ssh or login sessions you have, no matter if your home is shared via NFS.
• can prepare and use a different agent and different set of keys depending on the host you are connecting to, or the directory you are using ssh from. This allows for isolating keys when using agent forwarding with different sites (eg, university, work, home, secret evil internet identity, ...). It also allows to use multiple accounts on sites like github, unfuddle and gitorious easily.
• allows to specify different options for each set of keys. For example, you can provide a -t 60 to keep keys loaded for at most 60 seconds. Or -c to always ask for confirmation before using a key.
• performs port knocking before attempting to connect via ssh if configured.

Installation

All you need to run repassh is a standard installation of python >= 3.6.

To install it, run:

pip install repassh

Then you can use the repassh command just as you'd use ssh.

Alternatives

In .bashrc you can define an alias:

alias ssh=/path/to/repassh

then all you have to do is:

ssh somewhere

repassh will be called instead of ssh, and it will:

• check if ssh-agent is running. If not, it will start one.
• try to load all the keys in ~/.ssh, if not loaded.

If you use ssh again, repassh will reuse the same agent and the same keys.

About scp, rsync, and friends

scp, rsync, and most similar tools internally invoke ssh. If you don't tell them to use repassh instead, key loading won't work. There are a few ways to solve the problem:

Rename or link

Rename repassh to ssh or create a symlink ssh pointing to repassh in a directory in your PATH before /usr/bin or /bin.

For example:

ln -s /path/to/repassh ~/bin/ssh
export PATH="~/bin:$PATH"

Make sure echo $PATH shows ~/bin before /usr/bin or /bin. You can verify this is working as expected with which ssh, which should show ~/bin/ssh.

This works for rsync and git, among others, but not for scp and sftp, as these do not look for ssh in your PATH but use a hard-coded path to the binary.

If you want to use repassh with scp or sftp, you can simply create symlinks for them as well:

ln -s /path/to/repassh ~/bin/scp
ln -s /path/to/repassh ~/bin/sftp

More aliases

Add a few more aliases in your .bashrc file, for example:

alias scp='BINARY_SSH=scp /path/to/repassh'
alias rsync='BINARY_SSH=rsync /path/to/repassh'
...

The first alias will make the scp command invoke repassh instead, but tell repassh to invoke scp instead of the plain ssh command after loading the necessary agents and keys.

Note that aliases don't work from scripts - if you have any script that you expect to use with repassh, you may prefer the first method, or you will need to update the script accordingly.

Tell other programs to use repassh instead of ssh

Use command specific methods to force them to use repassh instead of ssh, for example:

rsync -e '/path/to/repassh' ...
scp -S '/path/to/repassh' ...

Config file with multiple identities

To have multiple identities:

1. create a $XDG_CONFIG_HOME/repassh/config.json file. In this file, you need to tell repassh which identities to use and when. The file should be a valid JSON (ignore/remove the lines starting with #, they are comments, but JSON does not have comments):

   {
   # Specifies which identity to use depending on the path I'm running ssh
   # from.
   # For example: ("mod-xslt", "personal") means that for any path that
   # contains the word "mod-xslt", the "personal" identity should be used.
   # This is optional - don't include any MATCH_PATH if you don't need it.
   
     "MATCH_PATH": [
       ["mod-xslt", "personal"],
       ["repassh", "personal"],
       ["opt/work", "work"],
       ["opt/private", "secret"]
     ],
   
   # If any of the ssh arguments have 'cweb' in it, the 'personal' identity
   # has to be used. For example: "ssh myhost.cweb.com" will have cweb in
   # argv, and the "personal" identity will be used.
   # This is optional - don't include any MATCH_ARGV if you don't
   # need it.
   
     "MATCH_ARGV": [
       ("cweb", "personal"),
       ("corp", "work")
     ],
   
   # KNOCK_PATH and KNOCK_ARGV work the same way as MATCH_PATH and MATCH_ARGV,
   # but instead of identity name you can provide port knocking configuration
   # executed before the ssh connection attempt.
   # Parameters:
   # - host: hostname
   # - ports: sequence of ports
   # - delay: time to wait between knocks (ms) [0]
   # - timeout: timeout for knocks (ms) [100]
   # - use_udp: set for UDP knocks, TCP is used by default
   
     "KNOCK_ARGV": [
       [
         "remotehost",
         {
           "host": "remotehost.domain.org",
           "ports": [123, 234, 345],
           "delay": 50
         }
       ]
     ],
   
   # Note that if no match is found, the DEFAULT_IDENTITY is used. This is
   # generally your loginname, no need to change it.
   # This is optional - don't include any DEFAULT_IDENTITY if you don't
   # need it.
   
   # "DEFAULT_IDENTITY": "foo",
   
   # Use running `ssh-agent`, true by default
   # If `SSH_AUTH_SOCK` and `SSH_AGENT_PID` environment variables are set
   # and the agent responds then it will be used instead of executing a new
   # one based on identity matching.
   # If the agent does not respond, a new one is started just like
   # `USE_RUNNING_AGENT` would be false.
   
   # "USE_RUNNING_AGENT": true,
   
   # This is optional - don't include any SSH_ADD_OPTIONS if you don't
   # need it.
   
     "SSH_ADD_OPTIONS": {
       # Regardless, ask for confirmation before using any of the
       # work keys.
       "work": "-c",
       # Forget about secret keys after ten minutes. repassh will
       # automatically ask you your passphrase again if they are needed.
       "secret": "-t 600"
     },
   
   # This is optional - don't include any SSH_OPTIONS if you don't
   # need it.
   # Otherwise, provides options to be passed to 'ssh' for specific
   # identities.
   
     "SSH_OPTIONS": {
       # Disable forwarding of the agent, but enable X forwarding,
       # when using the work profile.
       "work": "-Xa",
   
       # Always forward the agent when using the secret identity.
       "secret": "-A"
     },
   
   # Options to pass to ssh by default.
   # If you don't specify anything, UserRoaming=no is passed, due
   # to CVE-2016-0777. Leave it empty to disable this.
   
     "SSH_DEFAULT_OPTIONS": "-oUseRoaming=no",
   
   # Which options to use by default if no match with SSH_ADD_OPTIONS
   # was found. Note that repassh hard codes -t 7200 to prevent your
   # keys from remaining in memory for too long.
   
     "SSH_ADD_DEFAULT_OPTIONS": "-t 7200",
   
   # Output verbosity
   # valid values are:
   #   LOG_ERROR = 1, LOG_WARN = 2, LOG_INFO = 3, LOG_DEBUG = 4
   
     "VERBOSITY": 3
   }
   

2. Create the directory where all the identities and agents will be kept:

   mkdir -p ~/.ssh/identities; chmod u=rwX,go= -R ~/.ssh
   

3. Create a directory for each identity, for example:

   mkdir -p ~/.ssh/identities/personal
   mkdir -p ~/.ssh/identities/work
   mkdir -p ~/.ssh/identities/secret
   

4. Generate (or copy) keys for those identities:

   # Default keys are for my personal account
   $ cp ~/.ssh/id_rsa* ~/.ssh/identities/personal
   
   # Generate keys to be used for work only, rsa
   $ ssh-keygen -t rsa -b 4096 -f ~/.ssh/identities/work/id_rsa
   
   ...
   

Now if you run:

$ ssh corp.mywemployer.com

repassh will be invoked and:

1. checks ssh argv, determine that the work identity has to be used.
2. checks ~/.ssh/agents for a work agent loaded. If there is no agent, it will prepare one.
3. checks ~/.ssh/identities/work/ for a list of keys to load for this identity. It will try to load any key that is not already loaded in the agent.
4. finally run ssh with the environment setup such that it will have access only to the agent for the identity work, and the corresponding keys.

Note that repassh needs to access both your private and public keys. Note also that it identifies public keys by the .pub extension. All files in your identities subdirectories will be considered keys.

If you want to only load keys that have "key" in the name, you can add to your config.json:

PATTERN_KEYS = "key"

The default is:

PATTERN_KEYS = r"/(id_.*|identity.*|ssh[0-9]-.*)"

You can also redefine:

DIR_IDENTITIES = "$HOME/.ssh/identities"
DIR_AGENTS = "$HOME/.ssh/agents"

To point somewhere else if you so desire.