Re: sasl/plain with hashed password not working

[btb@bitrate.net]

On Oct 2, 2013, at 09.44, Dan White <dwhite@olp.net> wrote:

> libsasl2, with default configuration, requires that the password be stored
> in cleartext, even for PLAIN.
> 
> To support {ssha} in this scenario, I recommend you configure your SASL
> slapd.conf file to authenticate against saslauthd, which in turn should be
> configured to perform ldap simple (non-sasl) authentication against slapd.
> Think of it as a two-level deep recursive authentication.
> 
> Create a slapd.conf SASL file (e.g. /usr/lib/sasl2/slapd.conf) with these
> contents:
> 
>  pwcheck_method: saslauthd
>  # Disallow shared secret mechanisms
>  mech_list: PLAIN LOGIN GSSAPI EXTERNAL

thanks, this worked.  though the slapd/sasl recursive auth was a bit of a brain teaser to get straight in my head, if i'm honest.  :)

a few follow up questions, if i may.

i'd like to, just for the sake of my own edification, understand the internal bits a little better.  having gone through this exercise, if i understand correctly now, when a client connects to slapd and uses sasl, slapd hands over [so to speak] the authentication portion to libsasl2 in almost the same way any other server software which might be using libsasl2 does - even though slapd is likely to ultimately be the source of the authentication data?  the one notable difference being that when the libsasl2 is used in its default state [e.g. pwcheck_method: auxprop], slapd knows to include/use a "special" internal auxprop module [auxprop_plugin: slapd - i see this referenced in doc/options.html].  this in turn allows slapd to, in essence, work "internally" to itself, but still within the constraints of libsasl2?

that came out a bit long winded, but i'm hoping i have the essence straight.

you mention "libsasl2, with default configuration, requires that the password be stored in cleartext, even for PLAIN.".  was that just referring to the default of pwcheck_method: auxprop, or is there some scenario in which pwcheck_method: auxprop could be made to somehow handle hashed password values?

regarding using saslauthd in the mix with slapd/libsasl2, it seems that perhaps olcauthzregexp does not really apply any longer in this scenario?  slapd does not need to itself canonicalize the username, as that work is done by the saslauthd ldap auth mechanism?  removing the olcauthzregexp elements from my config seems to confirm this, but i'm wondering if there are other effects i've not considered.

lastly, i'm curious about some logging i see going to the syslog auth facility.  when using pwcheck_method: saslauthd, i see the following logged when starting slapd:

Oct  7 21:26:12 aurora slapd[20228]: DIGEST-MD5 common mech free
Oct  7 21:26:14 aurora slapd[20240]: auxpropfunc error invalid parameter supplied
Oct  7 21:26:14 aurora slapd[20240]: _sasl_plugin_load failed on sasl_auxprop_plug_init for plugin: ldapdb
Oct  7 21:26:14 aurora slapd[20240]: canonuserfunc error -7
Oct  7 21:26:14 aurora slapd[20240]: _sasl_plugin_load failed on sasl_canonuser_init for plugin: ldapdb
Oct  7 21:26:14 aurora slapd[20240]: sql_select option missing
Oct  7 21:26:14 aurora slapd[20240]: auxpropfunc error no mechanism available
Oct  7 21:26:14 aurora slapd[20240]: _sasl_plugin_load failed on sasl_auxprop_plug_init for plugin: sql

i understand from reading various ml discussions that those messages are largely innocuous, but was curious nonetheless why messages that seem to be related to auxprop appear if i've explicitly specified saslauthd.  is auxprop still being used in some capacity?

similarly, before using saslauthd, when initially testing using pwcheck_method: auxprop, and setting auxprop_plugin: slapd, as per doc/options.html, i expected to no longer see the complaints about sql and ldapdd, yet they still appear.  are my expectations regarding this off-base?

-ben