On 05/10/2013 02:06:53 PM, Caraman Mihai Claudiu-B02008 wrote:
> > > I didn't see Tiejun's patch... My goal was just to fix the
build
> > break
> > > without exposing problems, and to encourage a patch to fix it
> > properly
> > > to happen sooner rather than later. With Tiejun's patch, which
is
> > > similar to mine except that it doesn't disable e6500 support, a
user
> > > could BUG() the kernel by forcing an Altivec exception in a
guest.
> > I
> > > didn't want to go further down the road of adding reflectors for
> > those
> > > exceptions, which could make it look like the problem was dealt
with
> > > even though it's still not done.
> >
> > I agree it's quite annoying to hit a build breakage. Reflection
is not
> > a proper solution for this problem (though we will require it
later)
> > but program exception injection looks feasible as a simple fix.
>
> Program exception injection still doesn't deal with state
corruption.
Yes but it's not critical for this particular case since nobody is
able
to effectively use that state via altivec instructions. Leaking state
however can be a real issue.
Depending on guest behavior it could look like things are working even
though they aren't (e.g. a guest just enables MSR[VEC] and uses altivec
instructions, not relying on exceptions). This really isn't worth
spending a lot of time debating... Once Altivec is fixed properly (you
said that'd be soon, right?), we can add e6500 back to the list.
-Scott
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
