On Thu, Aug 10, 2017 at 10:11:13AM +0100, Peter Maydell wrote:
> On 1 August 2017 at 10:41, Peter Maydell <peter.mayd...@linaro.org> wrote:
> > On 1 August 2017 at 10:35, KONRAD Frederic <frederic.kon...@adacore.com>
> > wrote:
> >> On 08/01/2017 11:30 AM, Edgar E. Iglesias wrote:
> >>> At this stage, perhaps we should just register the blocker when this dev
> >>> realizes.
> >>> If a request_ptr comes in during migration, the VM will fail either way...
> >> Yes but this will breaks migration for the spips device everytime
> >> and not only when mmio-execution is used?
> > This line of thought is why I ended up suggesting just disabling
> > the exec-in-place feature -- that way we just don't introduce
> > what would be a new-in-2.10 feature, rather than breaking something
> > that used to work in 2.9.
> OK, so what's the plan here? We have several options:
> * just disable exec-from-spips for 2.10 (I sent a patch for that)
> * disable exec-from-spips for 2.10 but with a device x-property
> to allow the user to turn it on again if they really want it
> * this patch or variants on it which try to only disable
> migration if exec-from-spips is actually used by the guest
> (I don't like these because of the awkward corner cases if
> migration and the guest using exec-from-spips happen at the
> same time)
> So my current view remains "we should just disable this feature
> for 2.10 and we can implement it properly with handling of
> migration for 2.11", unless somebody cares enough to implement
> the x-property thing within the next day or so.
I think the x-property sounds good.
Fred, would you like to send a patch for that?
Otherwise, I can do it later today.