x11/xfce ssh-agent once per logon for minimal (no gnome/kde) installation

Steve Franks bahamasfranks at gmail.com
Fri Jun 4 16:07:50 UTC 2010


On Thu, Jun 3, 2010 at 11:56 PM, Matthew Seaman
<m.seaman at infracaninophile.co.uk> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 04/06/2010 24:39:46, Steve Franks wrote:
>> I'm totally lost.  What I desire is to put in my passphrase for my
>> public key(s) when I logon to my box.  Since I usually install from
>> ports and use xfce, I have no infrastructure for this, and I'm getting
>> nowhere fast.  My Fedora box popped up a nice little "enter
>> passphrase" box the very first time I ssh'd to my server, and now it
>> 'just works'.  My FreeBSD boxes (which are many - everything *but* the
>> laptop with Fedora), 'just don't work'.  I've installed everything
>> with 'ssh'  and either 'key' or 'agent' in the name from
>> ports/security, and gone through the manpages, and tweaked countless
>> environment variables, but every time I ssh on a FreeBSD box, it
>> stubbornly locates the terminal I started any gui's from (i.e. meld +
>> bzr), and asks for the passphrase a great many times daily.  Add that
>> I've started my gui with meld& (so as to continue using said terminal
>> - don't love 'panels', 'choosers', 'menus', etc - guis are for word
>> processors and file managers, not desktops), I can't even type in the
>> passphrase there.  I  generally like using FreeBSD caveman style, but
>> this is starting to drive me nuts.  No meld/bzr==no work from home==no
>> happiness ;)
>>
>> Anyone have a 'standard' / FreeBSD-friendly best-practices for this?
>> I think I'm just cluttering up my system here.
>
> The problem you have is that you need to start ssh-agent(1) somewhere
> very early in your login process, so that your entire desktop
> environment can inherit all the necessary ssh-agent settings.
>
> Exactly what to do depends on how you get into X11.  If you run
> startx(1) manually when you want to switch from console to X11 then you
> need to edit your ~/.xinitrc
>
> Alternatively, if you use a display manager like xdm(1) -- ie. there's
> an X based graphical login -- then you have two choices: edit your
> ~/.xsession or tweak the pam configuration for your login manager.
>
> If you want to go the 'edit .xinitrc or .xsession' route then you need
> to do basically the same thing for either of those two files.  They're
> pretty much just scripts that start up the initial X applications for
> your login session: practically speaking, that means starting up your
> window manager.  It's possible you don't have either of those files
> explicitly in your account: in either case the system will run a
> standard default script if it can't find a user specific version.
>
> The .xinitrc or .xsession file should look something like this:
>
> #!/bin/sh
>
> # Import user environment settings
> . ${HOME}/.profile
>
> eval $( ssh-agent -s )
>
> # Eg. pop up an xterm so you can enter your ssh passphrase
> xterm -geometry 80x24-91+0 -e /usr/bin/ssh-add -c ${HOME}/.ssh/id_rsa &
>
> # Note: no '&' -- this should run in the foreground
> xfce
>
> eval $( ssh-agent -k )
> #
> # That's All Folks!
> #
>
> This is just a rough outline, which you should adapt to your own needs
> - -- in particular there are nicer apps you can use for entering a
> passphrase.  Also note that you can probably omit that bit from the
> .xinitrc or .xsession and have your window manager run it.  In any case,
> the important bits are the two 'eval' lines bracketing the
> window-manager startup.
>
> The other possibility -- which is only available if you are using a
> display manager like xdm(1) -- is tweaking the pam settings.  For xdm,
> edit the file /etc/pam.d/xdm and uncomment the two lines mentioning
> pam_ssh.  Now you will be able to log into your system via xdm using
> your ssh passphrase, and xdm will start up ssh-agent for your session
> and add your key to it.  Different display managers will have their own
> pam.d files (either in /etc/pam.d or in /usr/local/etc/pam.d) but you
> should be able to make equivalent changes there -- either uncomment or
> add pam_ssh lines in the auth or session sections.
>
>        Cheers,
>
>        Matthew
>
> - --
> Dr Matthew J Seaman MA, D.Phil.                   7 Priory Courtyard
>                                                  Flat 3
> PGP: http://www.infracaninophile.co.uk/pgpkey     Ramsgate
> JID: matthew at infracaninophile.co.uk               Kent, CT11 9PW
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG/MacGPG2 v2.0.14 (Darwin)
> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
>
> iEYEARECAAYFAkwIo58ACgkQ8Mjk52CukIxm/ACgjwPTgJjq8YjN/e1uwD9be2xj
> vBcAoIQ8aP+1pyV/050ooHCr9yUFjuXh
> =S7kV
> -----END PGP SIGNATURE-----
>


Ah, sometimes you're just a command away!  I'm surprised I couldn't
google this (too many ssh examples, all the x11+agent ones must be
hidden).

So for posterity, this is the relevant portion of .xinitrc file:
...
export SSH_ASKPASS=/usr/local/bin/x11-ssh-askpass ;export SSH_ASKPASS
eval $( ssh-agent -s )
ssh-add &
xfce4-session
eval $( ssh-agent -k )
....

I was using "exec xfce4-session" as in most of the examples for
.xinitrc files, which seemed to be mucking things up - ditching the
exec was one of the key steps on my system.

Now, when I login, xfce starts normally, and concurrently, a nice
little window pops up asking for my passphrase for my keys!  Now I can
ssh (and bzr) with impunity.

Thanks!
Steve


More information about the freebsd-questions mailing list