Re: [Xen-devel] [PATCH] xen/arm: trap SMC instructions and inject an UND exception

[Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx>]

On Fri, 3 May 2013, Ian Campbell wrote:
> On Thu, 2013-05-02 at 17:08 +0100, Stefano Stabellini wrote:
> > On Thu, 2 May 2013, Ian Campbell wrote:
> > > On Thu, 2013-05-02 at 14:53 +0100, Stefano Stabellini wrote:
> > > > On Thu, 2 May 2013, Ian Campbell wrote:
> > > > > Currently only handles 32 bit guests. The 64-bit exception model is
> > > > > considerably different.
> > > > > 
> > > > > Signed-off-by: Ian Campbell <ian.campbell@xxxxxxxxxx>
> > > > >
> > > > > There are probably other places where injecting UND would be more 
> > > > > appropriate
> > > > > than killing the guest. It should also be reasonably easy to extend 
> > > > > this logic
> > > > > to inject DABT and PABT.
> > > > > ---
> > > > >  xen/arch/arm/traps.c                  |   56 
> > > > > ++++++++++++++++++++++++++++++++-
> > > > >  xen/include/asm-arm/arm32/processor.h |    4 ++-
> > > > >  xen/include/asm-arm/processor.h       |    8 +++++
> > > > >  xen/include/public/arch-arm.h         |    1 +
> > > > >  4 files changed, 67 insertions(+), 2 deletions(-)
> > > > > 
> > > > > diff --git a/xen/arch/arm/traps.c b/xen/arch/arm/traps.c
> > > > > index b7487b7..72dddb2 100644
> > > > > --- a/xen/arch/arm/traps.c
> > > > > +++ b/xen/arch/arm/traps.c
> > > > > @@ -59,7 +59,7 @@ void __cpuinit init_traps(void)
> > > > >      WRITE_SYSREG((vaddr_t)hyp_traps_vector, VBAR_EL2);
> > > > >  
> > > > >      /* Setup hypervisor traps */
> > > > > -    WRITE_SYSREG(HCR_PTW|HCR_BSU_OUTER|HCR_AMO|HCR_IMO|HCR_VM, 
> > > > > HCR_EL2);
> > > > > +    
> > > > > WRITE_SYSREG(HCR_PTW|HCR_BSU_OUTER|HCR_AMO|HCR_IMO|HCR_TSC|HCR_VM, 
> > > > > HCR_EL2);
> > > > >      isb();
> > > > >  }
> > > > 
> > > > My recent patch http://marc.info/?l=xen-devel&m=136734156810415 does
> > > > this
> > > 
> > > Right, I should have mentioned that. What do you do with the SMCs? I
> > > think we've concluded that undef is the right response?
> > 
> > I was going to support SMC as a conduit for PSCI calls.
> 
> I think we have agreed with the PSCI guys that HVC will be a supported
> conduit and since that is what we (will) advertise to guests I think
> that's all we should support -- there's no need to allow two ways to do
> it.

The patch were written before the discussion.
Now I agree that we can just support HVC and get away with it.
I'll update my series.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel