[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH V6 4/4] x86/altp2m: fix display frozen when switching to a new view early

On 11/16/18 12:08 PM, Jan Beulich wrote:
> I've seen you mention such or alike a couple of times now while
> discussing this series, and I have to admit that I'm a little frightened:
> Making a change just based on the observation that nothing breaks
> is setting us up for breakage in some corner case. That is - seeing
> something break is a good indication that a change was wrong, but
> not seeing any breakage doesn't alone justify a change.
> In the particular case here I think whether the copying of the field
> is needed depends on what else is being copied (I'm sorry, I'm not
> that familiar with altp2m): If mappings get cloned from the host p2m,
> then this value ought to be cloned too. For any mappings that
> might be kept in sync between p2m-s, I think this field also would
> need to be updated in sync.

I agree, I was merely offering to test whether any breakage occurs if
not syncing the field, not suggesting that that's a comprehensive
analysis or a justification for change, in case that offers us more insight.

AFAIK, EPT restrictions are always propagated from the hostp2m to all
active altp2ms. I've tested this at some point in the past and it has
proven true - it's supposedly an altp2m feature although it can also be
a bit of a headache: if we would have been able to use the hostp2m
independently then we wouldn't have had this display freeze problem,
since we could have used the hostp2m as the default (restricted) EPT
view. As things stand now, we need to create a new view when starting
introspection and switch to that immediately, so changes to it will not
propagate into other views (in other words, so we can use it
independently of all other altp2ms).


Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.