Re: [Xen-devel] [PATCH] xen/apic: implement io apic read with hypercall

[Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>]

On Tue, Apr 24, 2012 at 10:43:53PM +0800, Lin Ming wrote:
> On Mon, Apr 23, 2012 at 11:11 PM, Konrad Rzeszutek Wilk
> <konrad.wilk@oracle.com> wrote:
> >> >> > How about return -1 on error?
> >> >> > The calling function can check -1 for error.
> >> >>
> >> >> Isn't -1 potentially (at least theoretically) a valid value to read from
> >> >> one of these registers?
> >> >
> >> > Yeah, but then we are back to crashing at bootup (with dom0) :-)
> >> >
> >> > Perhaps the fallback is to emulate (so retain some of the original code)
> >> > as we have been since .. uh 3.0?
> >>
> >> Do you mean the return value of io_apic_read in 3.0?
> >
> > No. I meant that we would continue to emulate. The improvement
> > is that now we do:
> >
> >       if (reg == 0x1)
> >               return 0x00170020;
> >       else if (reg == 0x0)
> >               return apic << 24;
> >
> > instead of 0xfdfdfdfd.
> >
> >> It's 0xffffffff.
> >
> > Now it is 0xfdfdfdfd.
> >
> > I am suggesting that instead of BUG_ON(), we fallback to do returning
> > an emulatated IO_APIC values - like the ones that this original patch
> > cooked up..
> 
> But we still need to return some value if the register is not emulated.

Right. 0xfd;
> 
> How about below?


Almost perfect.
> 
> unsigned int xen_io_apic_read(unsigned apic, unsigned reg)
> {
>         struct physdev_apic apic_op;
>         int ret;
> 
>         apic_op.apic_physbase = mpc_ioapic_addr(apic);
>         apic_op.reg = reg;
>         ret = HYPERVISOR_physdev_op(PHYSDEVOP_apic_read, &apic_op);
>         if (!ret)
>                 return apic_op.value;
> 
>         /* emulate register */
>         if (reg == 0x1)
>                 return 0x00170020;
>         else if (reg == 0x0)
>                 return apic << 24;
>         else
>                 return -1;

	return 0xfd;
> }