[Freeipa-devel] [PATCH] 0001 ipa_kdb add krbPrincipalAuthInd handling

[Alexander Bokovoy]

On Wed, 27 Apr 2016, Matt Rogers wrote:
>On 04/27, Sumit Bose wrote:
>> On Tue, Apr 26, 2016 at 02:02:04PM -0400, Matt Rogers wrote:
>> > On 04/26, Sumit Bose wrote:
>> > > On Thu, Apr 14, 2016 at 12:59:55PM -0400, Matt Rogers wrote:
>> > > >
>> > > >
>> > > > ----- Original Message -----
>> > > > > From: "Nathaniel McCallum" <npmccallum at redhat.com>
>> > > > > To: "Matt Rogers" <mrogers at redhat.com>, freeipa-devel at redhat.com
>> > > > > Sent: Thursday, April 14, 2016 10:32:15 AM
>> > > > > Subject: Re: [Freeipa-devel] [PATCH] 0001 ipa_kdb add krbPrincipalAuthInd handling
>> > > > >
>> > > > > On Mon, 2016-04-11 at 10:41 -0400, Matt Rogers wrote:
>> > > > > > Hi,
>> > > > > >
>> > > > > > The attached patch is a part of the authentication indicator
>> > > > > > enhancements,
>> > > > > > adding indicator value storage and retrieval for the KDB driver.
>> > > > > >
>> > > > > > https://fedorahosted.org/freeipa/ticket/5782
>> > > > >
>> > > > > Can you add some whitespace in next_attr()? The density of the code
>> > > > > there hurts readability.
>> > > > >
>> > > > Sure, I've attached the revised patch.
>> > >
>> > > Hi Matt,
>> > >
>> > > thank you for the patch. Currently I have the following question.
>> > >
>> > > You call krb5_dbe_set_string to remove the 'require_auth' data before
>> > > calling ipadb_get_ldap_mod_extra_data()
>> > >
>> >
>> > > > +        /* Delete authinds from tl_data so it is not included in krbExtraData. */
>> > > > +        kerr = krb5_dbe_set_string(kcontext, entry, "require_auth", NULL);
>> > > > +        if (kerr) {
>> > > > +            goto done;
>> > > > +        }
>> > > > +
>> > > >          kerr = ipadb_get_ldap_mod_extra_data(imods,
>> > > >                                               entry->tl_data,
>> > > >                                               mod_op);
>> > > >
>> > >
>> > > Why it is needed to filter this data again in
>> > > ipadb_get_ldap_mod_extra_data()?
>> > >
>> >
>> > Hi Sumit, thanks for looking. The above krb5_dbe_set_string() call I
>> > believe I left there in error - We decided to operate on a filtered copy
>> > of the tl_data (which filter_authind_str_attrs() handles) rather than
>> > removing it completely with krb5_dbe_set_string(). I had tested with the
>> > above code commented out but forgot to remove it with the submitted patch.
>>
>> ok, makes sense.
>>
>> Nevertheless, comments in kdb.h like:
>>
>> /* String attributes (currently stored inside tl-data) map C string keys to
>>  * values.  They can be set via kadmin and consumed by KDC plugins. */
>>
>> and
>>
>> /* String attributes may not always be represented in tl-data.  kadmin clients
>>  * must use the get_strings and set_string RPCs. */
>>
>> make me wonder if it is a good idea to operate on the tl-data of type
>> KRB5_TL_STRING_ATTRS directly? I know we do this in other places as well
>> so I'm not insisting to change it, I'm just wondering about the reasons.
>>
>> Would something like (error checks are missing)
>>
>>     kerr = krb5_dbe_get_string(kcontext, entry, "require_auth",
>>                                &require_auth_str);
>>
>>     if (require_auth_str != NULL) {
>>         kerr = krb5_dbe_set_string(kcontext, entry, "require_auth", NULL);
>>     }
>>
>>     kerr = ipadb_get_ldap_mod_extra_data(imods,
>>                                          entry->tl_data,
>>                                          mod_op);
>>
>>     if (require_auth_str != NULL) {
>>         kerr = krb5_dbe_set_string(kcontext, entry, "require_auth",
>>                                    require_auth_str);
>>     }
>>
>> have the same effect with only using the recommended API (and making the
>> patch smaller)?
>>
>
>I believe that would be the same effect, just less efficient. But
>sticking to the API is probably safer in the long run. I am okay
>with changing it if you would prefer.
I think from the maintenance perspective it would be better to use the
recommended API.
-- 
/ Alexander Bokovoy