Re: [Xen-devel] [PATCH 2/2] arm: kill a guest which uses hvc with an immediate operand != XEN_HYPERCALL_TAG

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Tue, 2012-07-24 at 11:00 +0100, Stefano Stabellini wrote:
> On Mon, 23 Jul 2012, Ian Campbell wrote:
> > At best these guests are confused/broken and at worse they are malicious.
> > 
> > Signed-off-by: Ian Campbell <ian.campbell@xxxxxxxxxx>
> > ---
> >  xen/arch/arm/traps.c |    6 +-----
> >  1 files changed, 1 insertions(+), 5 deletions(-)
> > 
> > diff --git a/xen/arch/arm/traps.c b/xen/arch/arm/traps.c
> > index 52ab943..50b62c0 100644
> > --- a/xen/arch/arm/traps.c
> > +++ b/xen/arch/arm/traps.c
> > @@ -472,11 +472,7 @@ static void do_trap_hypercall(struct cpu_user_regs 
> > *regs, unsigned long iss)
> >      arm_hypercall_fn_t call = NULL;
> >  
> >      if ( iss != XEN_HYPERCALL_TAG )
> > -    {
> > -        printk("%s %d: received an alien hypercall iss=%lx\n", __func__ ,
> > -                __LINE__ , iss);
> > -        regs->r0 = -EINVAL;
> > -    }
> > +        domain_crash_synchronous();
> >  
> >      if ( regs->r12 > ARRAY_SIZE(arm_hypercall_table) )
> >      {
> 
> Are we sure that this is actually necessary?
> I think that the old behaviour of returning -EINVAL is sufficient.

We have no idea what semantics a guest using a tag other than ours is
expecting from the hypercall. We don't even know if they are expecting
negative errno style errors at all.

Worst case you've just answered the question "how much disk shall I
erase" with 0xffffffea (or whatever EINVAL is).

Even if they were expecting -errno we don't know that their idea of
EINVAL is the same as ours (the actual numbers are not standard, only
the names). (and in any case ENOSYS would be the correct one to use).

Ian.


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