Re: linux-next: build failure

[Josh Boyer <jwboyer@linux.vnet.ibm.com>]

On Fri, 07 Nov 2008 11:24:26 -0600
Hollis Blanchard <hollisb@us.ibm.com> wrote:

> On Fri, 2008-11-07 at 12:05 -0500, Josh Boyer wrote:
> > On Fri, 07 Nov 2008 10:40:05 -0600
> > Hollis Blanchard <hollisb@us.ibm.com> wrote:
> > > 
> > > Hi Stephen, it looks like Josh accidentally set the following options in
> > > ppc44x_defconfig:
> > > +CONFIG_VIRTUALIZATION=y
> > > +CONFIG_KVM=y
> > > +CONFIG_KVM_BOOKE_HOST=y
> > 
> > That wasn't an accident.  I set them based on -rc2 Kconfig values, and
> > enabled it for better build coverage.  Seems it worked. 
> 
> But when you asked me if I thought it was a good idea to enable KVM in
> ppc44x defconfig, I said "no"...

Hm.  I recall you saying "no", and then we had more discussion and
apparently I took that to mean "maybe."

> I'm not sure if there's precedent for features marked EXPERIMENTAL to be
> in the defconfig, but simply put I have not done a thorough audit of the
> KVM 440 code for security or DoS issues. (For example, until yesterday,
> the guest could trivially flood the host console because I'd left a
> printk enabled.) It just seems like asking for trouble to enable it in
> the defconfig.

If you think it's really in rough enough shape that it's a concern for
.28, then I can change it back.  Let me know soon.

EXPERIMENTAL means very little these days because it's used everywhere
and gets forgotten about down the road.  That's more a comment on the
uselessness of EXPERIMENTAL rather than your code, but that's the state
we're at today.  Marking it BROKEN would have prevented me from
enabling it for sure (maybe).

> That said, I have no complaint about having it enabled for linux-next
> builds.

I don't have a separate defconfig for linux-next builds.  We could do
that but I don't see much value.  I want -next to be building (and
theoretically testing) what gets built for actual release kernels.

josh