Stephen Hahn wrote:
> [GPEC comment omitted]
>
> and the harder problem of
> determining which subset of the events should go to a non-global zone.
> Based on Erik's mail, I guess that may eventually be "all of the
> events associated with the networking endpoints associated with the
> zone", but it's hard in the interim.
I'm not sure we need to take that approach.
The hardware is still controlled by the global zone (just like Xen's
domain0 controls the hardware) even if the configuration is to grant
exclusive access to a name like "bge3" to a particular zone (or Xen domU).
Thus events indicating that the hardware is broken and needs to be
replaced might still belong to the global zone, since the global admin
is the only one that can repair it.
Events that effect the local zone (or Xen domUs) ability to communicate
should be delivered to the local zone. Thus all of
- the cable being unplugged (wired),
- loosing the access point (wireless), and
- the NIC chip dying (and the driver detecting it)
should be represented to the local zone as some 'link down' event.
(And the last one would also cause a hardware event in the global zone).
Erik
