Re: trouble with pkinit
Nico Williams <[email protected]> Fri, 17 Apr 2026 17:56:12 -0500
| Newsgroups | gmane.comp.encryption.kerberos.devel |
|---|---|
| Message-ID | <aeK6jA+/fCa86CPJ@ubby> |
On Fri, Apr 17, 2026 at 06:05:20PM -0400, Geoffrey Thorpe wrote: Ok, so Heimdal client to Heimdal KDC works, but MIT client to Heimdal KDC gets: > root@alicia:~# kinit -V \ > -X X509_user_identity=FILE:/assets/pkinit-client-alicia.pem alicia > Using default cache: /tmp/krb5cc_0 > Using principal: [email protected] > PA Option X509_user_identity = FILE:/assets/pkinit-client-alicia.pem > kinit: Client '[email protected]' not found in Kerberos database while > getting initial credentials What does the KDC say in its log files? > * The (Heimdal-based) KDC is configured to use synthetic principals, so > indeed there is no "alicia" principal in the Kerberos database, the KDC is > configured to issue a TGT for whichever client principal is in the client > cert, whether or not that principal is in its database. This doesn't pose a > problem with the Heimdal kinit, so I don't know why it would be an issue > with the MIT one. (I.e. whether the principal is or isn't in the database is > a KDC-side consideration, so why is this error message showing up on the > client-side?) Oh! I wonder if the MIT client is trying an AS-REQ w/o pre-auth first to discover if PKINIT is supported for [email protected] -- if so that would prevent this working. Maybe it's trying to discover DH algorithms and parameters and what not? It would be better for MIT Kerberos' kinit to optimistically pick a suitable set of algorithms and parameters and jump rather than look-before-jumping. I could work around this in the Heimdal KDC by returning only PKINIT pre-auth metadata to an AS client that tries to get a ticket for a non-existent principal w/o pre-auth data. I should add a test of interop with MIT for this specific case. > * If I run the "kinit" command without the "-X" argument providing the > certificate, I see exactly the same error. That would be expected. > Any ideas? [...] See above. Nico -- _______________________________________________ krbdev mailing list [email protected] https://mailman.mit.edu/mailman/listinfo/krbdev