More forest trust related patches
abartlet at samba.org
Tue May 5 15:08:15 MDT 2015
On Tue, 2015-05-05 at 22:41 +0200, Stefan (metze) Metzmacher wrote:
> Hi Andrew,
> >> I moved a lot more stuff to the -ok branch (Note I also changed fixed some
> >> of the dsdb_trust_* helper functions compared to the last patchset!)
> >> It passed autobuild a few times and it's ready for master from my site.
> >> Note that samba-tool domain trust create needs to generate a true
> >> utf8 based password if --no-aes-keys is given, this is required
> >> because our kerberos client code can't handle random utf16munged passwords
> >> for arcfour-hmac-md5 pre-auth yet.
> >> However there're a few TODO's in the remaining patches.
> >> It's mainly related to bug #11130, where we should allow
> >> COMPUTERNAME at REALM and map it to COMPUTERNAME$@REALM.
> >> The same applies also for trust accounts (I guess it's just based on the
> >> '$').
> >> It's allowed as a client and also as a service principal.
> >> I added some tests for it and hacked a mostly working (but ugly
> >> implementation),
> >> Andrew maybe you can work out a better fix :-)
> >> Note that winbindd uses MYDOMAIN at OTHERREALM for kinit and generates some
> >> warnings
> >> without the fix for bug #11130, but it still work fine.
> >> Please review and push the -ok patches.
> > This is really, really good. The only concern I still have is around
> > testing. We need tests that
> > - walk over all the new samba-tool domain commands. That is important
> > because otherwise we won't even notice if we break them when trying
> > python3 upgrades, or other sweeping changes.
> I guess this is only possible for the non changing commands
> for the others we need two domains.
> > - specifically test for the referral shown by behaviour
> > HDB_ERR_WRONG_REALM. This is important because we will soon need to
> > update Heimdal, and folks like Debian combine Samba with untested
> > upstream versions.
> Ok, I'll see what I can do here.
We also need a test for "heimdal:kdc: generic support for 3part
servicePrincipalNames" for similar reasons.
> > - test for (the ban on) changing the trust password over LDAP
> > - test for listing local groups on the AD DC
> What do you want specifically here?
> I think we already test enumerating all groups including validation.
> > - test different KVNO values on trusts
> What do you mean here exactly?
> Changing the password a few times?
Both that, and simply asking for a ticket to an invalid kvno, and
checking which password it decrypted. I'm wanting this code tested in
The test probably belongs in the lsa.forest-trust test we discussed a
few weeks ago, using the approach of the new krb5.kdc tests. I realise
I'll probably need to help you out with some logistics here.
> > - test the new --local-dc (special_name) handling in Credentials
> > I realise that some of this is tested in integration tests, but I'm
> > starting to insist on unit tests (like the great work on the $ removal
> > stuff) for KDC changes. The other issue with the integration tests is
> > that a number of tests (validation, namespaces) are being done in the
> > environment creation, when these should be done as distinct unit tests.
Thank you so much for your understanding on this.
> > I do realise I'm asking for a lot of work, and I'm happy to help on
> > this, either between now and SambaXP, or at SambaXP, so we get this done
> > right.
> It would be cool if you could work on a proper fix for bug #11130.
I'll see what I can do.
Authentication Developer, Samba Team http://samba.org
Samba Developer, Catalyst IT http://catalyst.net.nz/services/samba
More information about the samba-technical