Re: [PATCH V6 05/10] audit: log creation and deletion of namespace instances

[ebiederm@xmission.com (Eric W. Biederman)]

Paul Moore <paul@paul-moore.com> writes:

> On Sat, May 16, 2015 at 5:46 AM, Daniel J Walsh <dwalsh@redhat.com> wrote:
>> On 05/15/2015 05:05 PM, Paul Moore wrote:
>>> On Thursday, May 14, 2015 11:23:09 PM Andy Lutomirski wrote:
>>>> On Thu, May 14, 2015 at 7:32 PM, Richard Guy Briggs <rgb@redhat.com> wrote:
>>>>> On 15/05/14, Paul Moore wrote:
>>>>>> * Look at our existing audit records to determine which records should
>>>>>> have
>>>>>> namespace and container ID tokens added.  We may only want to add the
>>>>>> additional fields in the case where the namespace/container ID tokens are
>>>>>> not the init namespace.
>>>>> If we have a record that ties a set of namespace IDs with a container
>>>>> ID, then I expect we only need to list the containerID along with auid
>>>>> and sessionID.
>>>> The problem here is that the kernel has no concept of a "container", and I
>>>> don't think it makes any sense to add one just for audit.  "Container" is a
>>>> marketing term used by some userspace tools.
>>>>
>>>> I can imagine that both audit could benefit from a concept of a
>>>> namespace *path* that understands nesting (e.g. root/2/5/1 or
>>>> something along those lines).  Mapping these to "containers" belongs
>>>> in userspace, I think.
>>> It might be helpful to climb up a few levels in this thread ...
>>>
>>> I think we all agree that containers are not a kernel construct.  I further
>>> believe that the kernel has no business generating container IDs, those should
>>> come from userspace and will likely be different depending on how you define
>>> "container".  However, what is less clear to me at this point is how the
>>> kernel should handle the setting, reporting, and general management of this
>>> container ID token.
>>>
>> Wouldn't the easiest thing be to just treat add a containerid to the
>> process context like auid.
>
> I believe so.  At least that was the point I was trying to get across
> when I first jumped into this thread.

It sounds nice but containers are not just a per process construct.
Sometimes you might know anamespace but not which process instigated
action to happen on that namespace.

>> Then make it a privileged operation to set it.  Then tools that care about
>> auditing like docker can set the ID
>> and remove the Capability from it sub processes if it cares.  All
>> processes adopt parent processes containerid.
>> Now containers can be audited and as long as userspace is written
>> correctly nested containers can either override the containerid or not
>> depending on what the audit rules are.
>
> This part I'm still less certain on.  I agree that setting the
> container ID should be privileged in some sense, but the kernel
> shouldn't *require* privilege to create a new container (however the
> user chooses to define it).  Simply requiring privilege to set the
> container ID and failing silently may be sufficient.

My hope is as things mature fewer and fewer container things will need
any special privilege to create.

I think it needs to start with a clear definition of what is wanted and
then working backwards through which messages in which contexts you want
to have your magic bits.

Eric