Re: [PATCH] Revert "evtchn: refuse EVTCHNOP_status for Xen-bound event channels"

[Jan Beulich <jbeulich@suse.com>]

On 16.05.2024 21:15, Oleksii K. wrote:
> I am not deeply familiar with the technical details surrounding XSM,
> but if I understand Daniel's point correctly, the original change
> violates the access control framework. This suggests to me that the
> revert should be merged.
> 
> However, I have a question: if we merge this revert, does it mean that
> using channels a user ( domain ) will be able to get information about
> certain events such as EVTCHNSTAT_unbound, EVTCHNSTAT_interdomain,
> EVTCHNSTAT_pirq, EVTCHNSTAT_virq, and EVTCHNSTAT_IPI (based on the code
> of lseventch.c)? Is this information really so critical that it cannot
> be exposed for some time until a patch that changes the XSM policy
> consistently is provided and merged?
> 
> If this information is indeed critical and should not be exposed, I
> think we can consider Daniel's suggestion to add a check to the dummy
> XSM policy as a solution.

The question isn't whether it's critical. As pointed out elsewhere, my
view is that any exposure of information needs to come with a proof that
nothing undue can be derived from that information. I see Daniel has
responded there, so we'll continue the discussion there.

Jan