[Freeipa-devel] [freeipa] #3668: CA-less install fails when intermediate CA is used

[Martin Kosek]

On 06/07/2013 02:04 PM, Dmitri Pal wrote:
> On 06/07/2013 03:47 AM, freeipa wrote:
>> #3668: CA-less install fails when intermediate CA is used
>> -------------------------------------+-------------------------------------
>>                Reporter:  jcholast   |             Owner:  jcholast
>>                    Type:  defect     |            Status:  assigned
>>                Priority:  major      |         Milestone:  2013 Month 06 -
>>               Component:             |  June (3.2.x bug fixing)
>>   Installation                       |           Version:
>>              Resolution:             |          Keywords:
>>              Blocked By:             |          Blocking:
>>           Tests Updated:  0          |       Affects DOC:  0
>> Patch posted for review:  0          |  Red Hat Bugzilla:
>>                  Source:             |       Effort Type:
>>        Targeted feature:             |       Design link:
>>           Design review:  0          |  Fedora test page:
>>                  Chosen:             |   Needs UI design:
>> -------------------------------------+-------------------------------------
>> Release Notes:
>>
>>
>> -------------------------------------+-------------------------------------
>> Changes (by mkosek):
>>
>>  * rhbz:  0 =>
>>
>>
>> Comment:
>>
>>  We not support intermediate CAs for external CA install or CA-less
>>  install. Thus, this ticket cannot be easily solved extensive changes to
>>  the installer. Related to #3274 (Pilsner milestone).
>>
>>  Moving back to triage to decide what to do about this ticket.
>>
> So you are saying that CA we chain to or get the certs from should
> always be a root CA?
> Why does it matter for our code whether the CA we deal with a Root CA or
> not?

No, this is a case when a CA you pass for FreeIPA is not a direct "parent" of
HTTP/DIRSRV certificates, i.e. there is an intermediate CA between the CA
passed to IPA and the actual certs.

It should not mean that the root CA you pass to IPA must be necessarily a root
CA of the entire chain. Jan, is this correct? Can you elaborate?

> If we are trying to say that we do not follow the chains of trust I can
> understand that. So the limitation is that we should always be given the
> certs from the same CA we chain to and not its sub CA's, right? That
> makes sense.
> 
> Can someone describe a use case where in real life the certs would come
> from one CA in the chain but we would be told to chain or use a sub CA
> of that CA. That seems strange. What am I missing?