On 07/12/2016 04:19 PM, Simo Sorce wrote:
On Tue, 2016-07-12 at 15:46 +0200, Martin Babinsky wrote:On 07/12/2016 02:00 PM, Martin Babinsky wrote:On 07/12/2016 01:05 PM, Alexander Bokovoy wrote:On Mon, 11 Jul 2016, Martin Babinsky wrote:From 185bde00a76459430d95ff207bf1fb3fe31e811a Mon Sep 17 00:00:00 2001 From: Martin Babinsky <mbabi...@redhat.com> Date: Fri, 1 Jul 2016 18:09:04 +0200 Subject: [PATCH] Preserve user principal aliases during rename operation When a MODRDN is performed on the user entry, the MODRDN plugin resets both krbPrincipalName and krbCanonicalName to the value constructed from uid. In doing so, hovewer, any principal aliases added to the krbPrincipalName are wiped clean. In this patch old aliases are fetched before the MODRDN operation takes place and inserted back after it is performed. This also preserves previous user logins which can be used further for authentication as aliases. https://fedorahosted.org/freeipa/ticket/6028 --- ipaserver/plugins/baseuser.py | 46 +++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 46 insertions(+) diff --git a/ipaserver/plugins/baseuser.py b/ipaserver/plugins/baseuser.py index 0052e718afe639bcc1c0a698ded39ea8407a0551..e4288a5a131157815ffb2 452692a7edb342f6ac3 100644 --- a/ipaserver/plugins/baseuser.py +++ b/ipaserver/plugins/baseuser.py @@ -498,6 +498,50 @@ class baseuser_mod(LDAPUpdate): len = int(config.get('ipamaxusernamelength')) ) ) + + def preserve_krbprincipalname_pre(self, ldap, entry_attrs, *keys, **options): + """ + preserve user principal aliases during rename operation. This is the + pre-callback part of this. Another method called during post-callback + shall insert the principals back + """ + if options.get('rename', None) is None: + return + + try: + old_entry = ldap.get_entry( + entry_attrs.dn, attrs_list=( + 'krbprincipalname', 'krbcanonicalname')) + + if 'krbcanonicalname' not in old_entry: + return + except errors.NotFound: + self.obj.handle_not_found(*keys) + + self.context.krbprincipalname = old_entry.get( + 'krbprincipalname', ) + + def preserve_krbprincipalname_post(self, ldap, entry_attrs, **options): + """ + Insert the preserved aliases back to the user entry during rename + operation + """ + if options.get('rename', None) is None or not hasattr( + self.context, 'krbprincipalname'): + return + + obj_pkey = self.obj.get_primary_key_from_dn(entry_attrs.dn) + canonical_name = entry_attrs['krbcanonicalname'] + + principals_to_add = tuple(p for p in self.context.krbprincipalname if + p != canonical_name) + + if principals_to_add: + result = self.api.Command.user_add_principal( + obj_pkey, principals_to_add)['result'] + + entry_attrs['krbprincipalname'] = result.get('krbprincipalname', ) + def check_mail(self, entry_attrs): if 'mail' in entry_attrs: entry_attrs['mail'] = self.obj.normalize_and_validate_email(entry_attrs['mail']) @@ -557,9 +601,11 @@ class baseuser_mod(LDAPUpdate): self.check_objectclass(ldap, dn, entry_attrs) self.obj.convert_usercertificate_pre(entry_attrs) + self.preserve_krbprincipalname_pre(ldap, entry_attrs, *keys, **options) def post_common_callback(self, ldap, dn, entry_attrs, *keys, **options): assert isinstance(dn, DN) + self.preserve_krbprincipalname_post(ldap, entry_attrs, **options) if options.get('random', False): try: entry_attrs['randompassword'] = unicode(getattr(context, 'randompassword')) -- 2.5.5The approach looks good. For the record, we also support aliases for hosts and service kerberos principals but we don't support rename options for them, so there is no need to add similar logic there.That's right, I have updated the corresponding section of the design page  for future reference.  http://www.freeipa.org/page/V4/Kerberos_principal_aliases#Managemen t_frameworkAdding Simo to the loop since he is not convinced that this is the right behavior. As I see it, it seems to not be a security issue but more of a different expectations about the perceived correct behavior in this particular situation. I can see the use case in keeping the old aliases, e.g. keeping the old credentials after legal name change, but I can also see the available space for user principal names piling up and cluttering quickly in large organizations.after some thinking I think it is ok to keep by default and then drop as it avoid races if you do really want to keep the aliases. However the CLI/UI should probably offer a button/switch to allow to drop all aliases on rename, what it would do would be to modify the entry after the rename and drop the aliases. I am a bit uncertain what to do by default with the "original name". I can see it going both ways on whether to keep it by default or not. Simo.
So let me summarize the proposed behavior just to see if I understand it correctly:
1.) when the user is renamed, all his kerberos aliases are added back to the attribute after rename
2.) a new Flag will be added to *user-mod (e.g. --remove-aliases) which will suppress this behavior when True so that all previous aliases will be erased as is done now. The administrator will use this flag to override the default behavior in 1.)
We now have the following concerns: 3.) What to do with old canonical names:The proposed patch treats old canonical names as aliases which are re-added after rename (since krbPrincipalName before rename contains the alias equal to the original canonical name anyway, we get this for free). If we implement the alias-wiping flag from 2.), we have to decide whether the old canonical name should be wiped away as well, or implement an additional logic to preserve it.
4.) What to do with users coming from pre-4.4 FreeIPA versions. These have no krbCanonicalName attribute and have only one alias = canonical name, so the code does no special handling of them and behaves as before. If we want to keep his old principal after rename, we have to implement another path which constructs krbCanonicalName attribute for him during rename and adds old principal as an alias among his krbPrincipalNames.
I would say that decoupling the preservation of old canonical principals from the preservation of aliases is relatively easy to implement, but I am bit afraid that it will increase the complexity and potential fragility of the solution.
-- Martin^3 Babinsky -- Manage your subscription for the Freeipa-devel mailing list: https://www.redhat.com/mailman/listinfo/freeipa-devel Contribute to FreeIPA: http://www.freeipa.org/page/Contribute/Code