Re: Privileged Resource Manager comments

On Nov 20, 2003, at 1:05 PM, Jim Pinkerton wrote:

>  
>
> In the minutes for the last IETF meeting, it states:
>
>  
>
> Q: Draft introduces Privileged Resource Manager as a mandatory
>
>       functional component.  How does this affect performance?
>
> A: The intent is that it not be involved in high-frequency
>
>       operations, but more work is needed to define the precise
>
>       functionality of the Privileged Resource Manager and check
>
>       if there are problems.
>
>  
>
>  
>
> In the latest draft, per feedback from the reflector, the following 
> text was added:
>
>  
>
> For example, the Privileged Resource Manager may be partially or 
> completely encapsulated in the Privileged Application. Regardless, it 
> is expected that the security analysis of the potential threats and 
> countermeasures still apply.
>
>  
>
> Thus I don’t believe the draft requires a distinct entity called the 
> Privileged Resource Manager. It can be encapsulated in an application 
> that has been trusted to access privileged resources directly. For 
> example, a particular implementation may run exactly one application, 
> which has physical access to all resources. In that case the security 
> issues described for the non-privileged application are unnecessary to 
> implement (because there is not one), and there is no need to separate 
> the Privileged Resource Manager from the privileged application.
>
>  
>

Agreed, the "Privileged Resource Manager" is merely a label that 
reflects the
fact that DDP Layer must be given permission to access user memory, and
that in most environments this fact must/should be validated.

But to be even more precise, the DDP Layer does not *have* to be given
access to "physical memory". The "RNIC" could be implemented *within*
a process, in which case the memory being accessed could be just
plain "normal" memory. The DDP Layer could also be operating over
a system bus with its own addressing (such as VME-64 or CPCI).