On Thursday 10 June 2010 17:49:16 Avi Kivity wrote:
> On 06/10/2010 06:31 AM, Sheng Yang wrote:
> > +
> > +#define EXCEPTION_REGION_BEGIN(r, lb) \
> > + asm volatile("pushq $"lb"\n\t" \
> > + "mov $2, %0\n\t" \
> > + : "=r"(r))
> > +#define EXCEPTION_REGION_END(r, lb) \
> > + asm volatile("popq %%rdx\n\t" \
> > + "mov $0, %0\n\t" \
> > + lb":\n\t": \
> > + "=r"(r) :: "%rdx")
>
> These mess up the stack, no? So if any intervening code uses %rsp based
> addressing, it will get incorrect information.
Yes, but I meant to only include instruction directly in the “region”, and support
nobody should touch $rsp then. It's directly and simple enough.
>
> I suggest using a special data section like the kernel.
Kind of more complex...
>
> Also, most tests that use asm don't need an exception region, but will
> want to trap an exception on a single address. So need a macro designed
> to be placed in asm in front of the faulting instruction.
Use region here is due to we don't know the size of instruction(or I don't know
how to get it)... Use a label here seems the simplest way to get the address
right.
--
regards
Yang, Sheng
--
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
