On Wed, Dec 19, 2012 at 10:59:28AM +0100, Alexander Graf wrote: > > On 19.12.2012, at 10:37, Gleb Natapov wrote: > > > On Wed, Dec 19, 2012 at 12:01:19AM +0100, Alexander Graf wrote: > >> > >> On 18.12.2012, at 23:54, Scott Wood wrote: > >> > >>> On 12/18/2012 06:38:41 AM, Alexander Graf wrote: > >>>> When we hit an emulation result that we didn't expect, that is an error, > >>>> but it's nothing that warrants a BUG(), because it can be guest triggered. > >>>> So instead, let's only WARN() the user that this happened. > >>>> Signed-off-by: Alexander Graf <agraf@xxxxxxx> > >>>> --- > >>>> arch/powerpc/kvm/powerpc.c | 3 ++- > >>>> 1 files changed, 2 insertions(+), 1 deletions(-) > >>>> diff --git a/arch/powerpc/kvm/powerpc.c b/arch/powerpc/kvm/powerpc.c > >>>> index be83fca..e2225e5 100644 > >>>> --- a/arch/powerpc/kvm/powerpc.c > >>>> +++ b/arch/powerpc/kvm/powerpc.c > >>>> @@ -237,7 +237,8 @@ int kvmppc_emulate_mmio(struct kvm_run *run, struct kvm_vcpu *vcpu) > >>>> r = RESUME_HOST; > >>>> break; > >>>> default: > >>>> - BUG(); > >>>> + WARN_ON(1); > >>>> + r = RESUME_GUEST; > >>> > >>> Do you have a specific way of a guest triggering this in mind, or is it just being cautious? The guest probably shouldn't be allowed to spam the kernel log with WARNs either. Is a traceback even useful here? > >> > >> For debugging, yes. But maybe we would be better off with a trace point. Anyway, a WARN is better than a BUG either way for now. > >> > >> I was able to provoke this by live patching an instruction without flushing the icache, so that the last_inst instruction fetch gets a different instruction from the instruction that resulted in the trap we're currently in. > >> > > If guest can trigger this it better be WARN_ON_ONCE(). Otherwise, as > > Scott said, guest will be able to spam host kernel log. > > I really think eventually we want a trace point and no WARN at all or all possible EMULATE targets handled, because a guest can legitimately not flush its icache and thus confuse our logic. > > Just consider this patch as a quick fix to make sure we enable people (me) to unload the module still after they hit this case ;). Real fix coming soon. > OK. -- Gleb. -- 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
